2

Introduction to Colloid Science

Applications to sediment characterization

Claire Chassagne

PUBLISHER

3
Published by:
TU Delft OPEN
Delft University of Technology, Delft, The Netherlands

Cover design: Claire Chassagne

Images without an attribution are made by the author

https://books.open.tudelft.nl

This work is published under a Creative Commons License: CC-BY 4.0

by Claire Chassagne.
This work is the open access version of the work with the same title
published in 2019 by Delft Academic Press (DAP).

ISBN: 97890-6562-4376
DOI: https://doi.org/10.34641/mg.16

4
5
 Contents
Contents................................................................................................................ 6

Foreword ............................................................................................................ 11

Chapter 1 Introduction, general definitions......................................................... 13

Introduction ......................................................................................................... 14
Solutions, colloids and suspensions ..................................................................... 16
The difference between clay, silt and sand particle sizes .................................... 20
The definition of mud, silt and clay ...................................................................... 22
Clay minerals........................................................................................................ 25

Chapter 2 Settling, diffusion and stabilisation .................................................... 37

Stokes’ settling velocity........................................................................................ 38
Brownian motion ................................................................................................. 42
Fick’s laws ............................................................................................................ 47
Stable and unstable colloidal suspensions ........................................................... 52
Osmotic pressure ................................................................................................. 58

Chapter 3 DLVO forces ........................................................................................ 61

Van der Waals forces ........................................................................................... 63
Coulombic forces.................................................................................................. 67
DLVO theory ......................................................................................................... 71
Link between surface potential and surface charge ............................................ 78
Measurement of the zeta potential ..................................................................... 80
Ionic conductivity and diffusion coefficient.......................................................... 86

Chapter 4 Other colloids: polymers, surfactants, microorganisms … .................. 89

Polymers & polyelectrolytes................................................................................. 90
Polymers and clays .............................................................................................. 94
Depletion effect ................................................................................................... 96

6
 Stabilization ......................................................................................................... 97
Surfactants (amphiphilic molecules) .................................................................... 98
Surface tension .................................................................................................. 101
Plankton ............................................................................................................. 104
The marine cycle ................................................................................................ 109

Chapter 5 Floc formation and break-up ............................................................. 113

Cluster aggregation ........................................................................................... 114
Stable clay suspensions ...................................................................................... 115
Unstable clay suspensions: influence of salt ...................................................... 117
Swelling behaviour of clays ................................................................................ 120
Unstable clay suspensions: flocculation by polyelectrolytes .............................. 122
Shear rate influence on floc size ........................................................................ 122
Density of flocs and fractal dimensions ............................................................. 128

Chapter 6 Modelling and measuring the flocculation rate ............................... 135

Model of Smoluchowski: aggregation by Brownian motion .............................. 136
Orthokinetic aggregation: aggregation by shear .............................................. 142
Measuring the flocculation rate ........................................................................ 144
Settling velocity, optimal flocculation and zeta potential ................................. 152

Chapter 7 Rheological behaviour of colloidal suspensions ................................ 157

Viscosity and yield stress.................................................................................... 158
Viscoelastic fluids ............................................................................................... 163
Rheology of suspensions .................................................................................... 168
Phase transition ................................................................................................. 175
Shear thinning and shear thickening ................................................................. 182
Gels and hydrogels ............................................................................................. 186
A fifty cent rheometer for yield stress measurement ........................................ 191
Maximum yield and zeta potential .................................................................... 194

Chapter 8 Settling of (concentrated) suspensions............................................. 199

7
 Concentrated suspensions and fluid mud .......................................................... 200
Modelling concentrated suspensions................................................................. 201
Settling of concentrated suspensions ................................................................ 202
Chemical potential, osmotic pressure and thermodynamics ............................. 212
Settling profiles .................................................................................................. 219

Chapter 9 Permeability of slurries ..................................................................... 233

Link between settling and consolidation ........................................................... 238
A Darcy equation for settling particles .............................................................. 241
The Gibson equation .......................................................................................... 245
Finding the permeability using colloid science................................................... 256
Onsager relations............................................................................................... 260

Chapter 10 Modelling the consolidation of slurries ......................................... 267

Consolidation of slurries : the fractal approach ................................................. 268
Settling and consolidation of natural mud ........................................................ 289
Link with Chapter 8 : how to couple settling and consolidation? ...................... 292
Beyond self-weight consolidation ...................................................................... 299

Conclusion from colloid science to large-scale applications............................... 311

References for colloid science books ................................................................. 315

8
 “Who do not
honour the small things
is not worth the big ones”

[Dutch saying]
 Foreword
This book is meant as an introduction to the field of colloid science, i.e. the study of
the behaviour of micrometric particles in a fluid (or a gas). The book was written with
a special emphasis on sediment particles. Sediment particles are complex colloidal
particles due to their composition, shape and interaction with their environment.
Characterization of the colloidal fraction of sediment is done by recording, among
others, the particles’ size, shape and electric surface charge and evaluating their
interactions. These properties are important for civil engineering applications, even
though the size range of these particles and their interactions is microscopic. Large-
scale sediment transport models for example require as input the settling velocity of
individual particles. In concentrated areas, this velocity becomes a function of the
particles’ concentration and particle-particle interactions lead to the creation of
larger particles, called flocs. These flocs can settle and, when reaching the bed,
consolidate in time. All these aspects, and related models, are treated in the present
book.

The book was written for students having no special beforehand knowledge in
colloid science, and a limited knowledge in physics and chemistry, the two main
disciplines relevant to this branch of science. The mathematics are also kept
minimal. They are given when the derivations present no particular problem. The
more lengthy derivations, often very elegant (but sometimes rather tedious) can be
found in the references given as footnotes in the different chapters.

Key words are highlighted in the text, which will enable the reader to find further
information by searching for them in a web browser.

I take the opportunity to thank the colleagues that reviewed this book, and helped
with their comments to improve it. To the present reader, I would like to say: do not
hesitate to report any mistake or unclarity to me.

Claire Chassagne
Delft, spring 2019
 Chapter 1
Introduction, general definitions
and some properties of
clayey material
 Introduction to Colloid Science

Introduction
Mud and clayey systems have from origin been studied in the field of soil science 1,
i.e. the branch of science concerned with the formation, classification and fertility
properties of soils. In soil science, the classification of soils is of prime importance,
as the properties of a soil are used by farmers to decide which types of crops,
livestock and soil management are best for their piece of land. The information
obtained by a soil survey can also be used by an architect or a builder to determine
whether a given soil is suitable for a specific type of construction.

In agriculture, soils are primary viewed for their

chemical properties, i.e. their ability to bring
nutrients to plants. At the beginning of the 19th
century chemical studies were extensively
conducted which confirmed for instance the
role of phosphate, nitrogen and potassium in
plant growth, and commercial fertilizers were
developed. The composition of a soil is also
determinant for the type of plants that are
growing on it. The first report that clearly
related the soil properties to the different
1350 BC, Egypt : thanks to the fertile climates, vegetation and parent rocks from
soil of the river Nile, Egypt became a which the soils were formed was written by
powerful and long-lasting civilisation Dokuchaiev in 1883 2. Many of the chemical
changes occurring in a soil are related to the
action of bacteria. At the end of 19th century it was for instance discovered that the
bacteria which live in nodules on the roots of legume plants absorb nitrogen from
the air and convert it to a form which plants can utilize. The differences in soil
characteristics are connected not only to the (bio)chemistry but also to the
mineralogical and physical properties of soils. These properties are important for
construction, but also to the farmer as the porosity of a soil governs gas exchange
with the atmosphere, water penetration and root growth. The porosity of a soil is
strongly connected to the particle size distribution of the grains forming the soil and
the soil’s state of consolidation. The mineralogical composition of a clay, which can
be determined from X-ray crystallography and electron microscopy tests, will tell if
a clay soil is prone to weathering, swelling and retain or exchange cations 3. This last
property is pH-dependent. This shows that both chemical and physical properties

1 We will focus on natural soils. Another important research concerns the use of clays in
ceramics, cosmetics and drugs which will not be discussed here.
2 Dokuchaiev, V.V. (1883) Russian chernozem, Monograph, Sankt-Peterburg. See also

Greenland, Dennis James, and Michael Hilary Bermingham Hayes. The chemistry of soil
constituents. Wiley and sons (1997).
3 a cation is a positively charged ion; an anion is a negatively charged ion.

14
 Chapter 1 Introduction, general definitions

are of importance to understand the diffusion and retentions of ions within the soil
fabric. Similarly, the consolidation, permeability, strength and aging of soils are
related to chemo-physical processes: by oxidation (chemical process) the
permeability of a soil (physical property) can for instance been altered.

In the context of sediment dynamics, i.e. the study of sediment transport, deposition
and erosion, the first studies have been initiated starting from the work done on
sand. We emphasize that we are here dealing with water-saturated bodies and not
Aeolian transport. The studies we refer to are therefore relevant for fresh and salt
water systems such as rivers, lakes, seas and estuaries. Aquatic sand transport,
deposition and erosion has been monitored and modelled successfully in the past
decades. The models have been tried on clays and clayey systems, but it was soon
recognized that these systems were much more complex and the models and
theories should be adapted 4.

In this first chapter (Chapter 1), clayey soils and suspensions are introduced from
their classical definitions, derived primarily from soil science, in terms of particle size,
clay mineralogy and physical properties (fluid, solid). We will start by defining
colloidal sized particles. Colloids (= colloidal suspensions) will be the main topic of
interest in the following chapters.

There are several differences between clay and sand particles. One of them is their
size, which affects their diffusion behaviour in the water body, as detailed in Chapter
2. A major difference between sand and (colloidal) clay particles is that clay particles
have the ability to aggregate. the reason behind this property is discussed in Chapter
3. The clay particles can aggregate between themselves, adsorb (poly)ions, stick to
or interact with other particles (some of them are reviewed in Chapter 4). In Chapter
5, the different types of aggregation (also called flocculation) and the different types
of aggregates (flocs) produced are reviewed. The rate of flocculation is modelled in
Chapter 6.

From the concepts introduced in Chapters 2 – 6, it will become clear that the main
reason sand models are not adapted to clayey systems is that because of their size,
clay particles are extremely dependent on surface forces (interaction with other
particles in suspension), whereas the dynamics of sand particles are mainly
controlled by a volume force (gravity force). Sand particles can be assumed to have
a constant Stokes’ settling velocity 5, which can be calculated from their density and
volume whereas clayey particles have a time-dependent density and volume.
Sediment transport models have as input the concentration of particles of a certain
size and their Stokes’ settling velocity, it is therefore clear that some adjustment is

4 Leussen, W. van, Estuarine macroflocs and their role in fine-grained sediment transport, PhD
thesis, RIKZ (1994)
5 Stokes’ settling velocity is defined in Chapter 2

15
 Introduction to Colloid Science

required to the models as aggregation implies that particles change both their size
and their settling velocity.

The density and composition of the clay suspensions influence their rheology
(Chapter 7) and the settling of their constitutive particles (Chapter 8). Once settled,
mud suspensions will form a soil with a time-dependent permeability (Chapter 9).
Given a specific relation between permeability and density, the early stage of
consolidation can be studied (Chapter 10). This so-called early stage of
consolidation, when mud is freshly deposited at the bottom of a water body, is
important to study in connection to sediment dynamics: the strength of a freshly
consolidated mud layer is a key parameter for sediment transport models as it will
determine the amount of mud eroded from the bottom at a given water velocity.
The link between Chapters 8 and 10 is discussed. In the last part of Chapter 10, we
will briefly review the later stage of consolidation, when forces at the contact points
between particles become predominant. This stage is extensively studied in soil
science and the reader is referred to books in that field of science for further reading.

Solutions, colloids and suspensions

For the remainder of the book, we will be concerned with what is technically defined
as a two-phase system: water and particles. Already we have to be more precise:
water itself does contain particles. First of all water molecules of course, but also
dissolved ions, which (see Chapter 3), can play a key role in the aggregation of clay
particles. Even fresh water contains ions, for example HCO3− and Ca2+ ions which
are the predominant ions found in river waters, and originate from the weathering
of limestone or feldspar (carbonate rocks). These type of rocks are typically found in
Europe, North America and Asia. The weathering of silicate rocks will produce
different types of ions. In chemistry, one does therefore speak of a solution (water +
ions) and not of water (as is done in civil engineering). In a large branch of colloid
science, solutions are seen as a continuum. This means that the solution is a fluid
with given properties and that no distinction is made for the individual ionic particles
or water molecules in the water. This fluid has bulk properties such as density and
ionic concentration. In this book, water will be seen as a continuum.

The particles we refer to are predominantly clay particles, but, as we will see in
Chapters 4 and 5 mud is composed of numerous other particles. These particles can
either be seen as individual particles which interact with each other and the gravity
field or as a continuum with a given (time-dependent) density. This continuum
approach will in particular be used in Chapters 9 and 10 but it will become clear that
all models presented in this book depend in a way or another on the properties of
the microscopic constitutive particles.

16
 Chapter 1 Introduction, general definitions

Mainly depending on the size of the particles in suspension, three different types of
fluids can be distinguished: solutions, colloids and suspensions. They are
represented underneath:

A solution is a homogeneous mixture that appears clear, such as salt water in a glass. In a
colloid, such as milk, the particles are much larger but remain dispersed and do not settle. A
suspension, such as mud, is a heterogeneous mixture of suspended particles that can either
remain in suspension (for the colloidal fraction) or can settle (for the coarser fraction).

A solution is a homogeneous mixture resulting from the dissolution of a solute (ex:

salt) in a solvent (ex: water). In the case of salt in water, one speaks of electrolyte
solutions (or electrolytes), as dissolved salts are charged and electrically conducting.
It can happen that a solution is saturated in one of its solutes, in which case it cannot
dissociate this solute anymore. This happens for example when too much salt is
added in a glass of water: some salt crystals will remain at the bottom of the glass.
Note that due to gravity, when the height of the water column is important, like in
the sea, some stratification can be observed in the solution, and its density may vary
as function of height.

A colloid (or colloidal suspension) is composed of particles or droplets (in which case
one speaks of an emulsion – milk is for instance an emulsion) that are much larger
than ions or molecules, but that are small enough not to settle out. These colloidal
particles are dispersed in a solvent. We will only consider water or an electrolyte
solution as solvent in the remaining of the book, as these are the most common
solvents for sediment particles. If the colloidal particles or droplets are dispersed in
a gas one refers to an aerosol (smoke and mist are aerosols).

17
 Introduction to Colloid Science

To be classified as colloidal, a material must have one or more of its dimensions

(length, width, or thickness) in the approximate range of 1-1000 nm. The particles in
a colloid are large enough to scatter light, a phenomenon called the Tyndall effect.
This can make colloidal suspensions appear cloudy or opaque. The Tyndall effect is
an easy way of determining whether a fluid contains a significant amount of colloidal
particles. When light is shined through a true solution, the light passes cleanly
through the solution (ex: an electrolyte solution), however when light is passed
through a colloidal solution, the substance in the dispersed phases scatters the light
in all directions, making it readily seen (ex: milk).

John Tyndall (1820 – 1893) was a prominent 19th-

century physicist. His initial scientific fame arose in
the 1850s from his study of diamagnetism. Later he
made discoveries in the realms of infrared radiation
and the physical properties of air. Tyndall also
published more than a dozen science books which
brought state-of-the-art 19th century experimental
physics to a wide audience. From 1853 to 1887 he
was professor of physics at the Royal Institution of
Great Britain in London.

extract from the book The Glaciers of the Alps, by John Tyndall 6

Expedition of 1857: the Lake of Geneva. Blueness of the water

“On Thursday, the 9th of July, 1857, I found myself upon the Lake of Geneva,
proceeding towards Vevey. I had long wished to see the waters of this renowned
inland sea, the colour of which is perhaps more interesting to the man of science
than to the poets who have sung about it. Long ago its depth of blue excited
attention, but no systematic examination of the subject has, so far as I know, been
attempted. It may be that the lake simply exhibits the colour of pure water. Ice is
blue, and it is reasonable to suppose that the liquid obtained from the fusion of ice
is of the same colour; but still the question presses—"Is the blue of the Lake of
Geneva to be entirely accounted for in this way?" The attempts which have been
made to explain it otherwise show that at least a doubt exists as to the sufficiency
of the above explanation.

It is only in its deeper portions that the colour of the lake is properly seen. Where
the bottom comes into view the pure effect of the water is disturbed; but where the
water is deep the colour is deep: between Rolle and Nyon for example, the blue is
superb. Where the blue was deepest, however, it gave me the impression of
turbidity rather than of deep transparency. At the upper portion of the lake the

6 http://www.gutenberg.org/files/34192/34192-h/34192-h.htm#Page_33

18
 Chapter 1 Introduction, general definitions

water through which the steamer passed was of a blue green. Wishing to see the
place where the Rhone enters the lake, I walked on the morning of the 10th from
Villeneuve to Novelle, and thence through the woods to the river side. Proceeding
along an embankment, raised to defend the adjacent land from the incursions of the
river, an hour brought me to the place where it empties itself into the lake. The
contrast between the two waters was very great: the river was almost white with
the finely divided matter which it held in suspension; while the lake at some distance
was of a deep ultramarine.

The lake in fact forms a reservoir where the particles held in suspension by the river
have time to subside, and its waters to become pure. The subsidence of course takes
place most copiously at the head of the lake; and here the deposit continues to form
new land, adding year by year to the thousands of acres which it has already left
behind it, and invading more and more the space occupied by the water.
Innumerable plates of mica spangled the fine sand which the river brought down,
and these, mixing with the water, and flashing like minute mirrors as the sun's rays
fell upon them, gave the otherwise muddy stream a silvery appearance.”

The Tyndall effect can be linked to theories developed by Rayleigh and Mie, in the
frame of optics: when the particle is smaller than the wavelength, Rayleigh theory is
used, which predicts that for this type of particles, short-wavelength light is
scattered more than longer wavelengths. The blue colour (short wavelength) is
therefore more scattered than the red (long wavelength): this is why the sky appears
blue to us, as the gas molecules of the atmosphere are much more effective in
scattering the blue wavelength. Some suspensions, containing nanoparticles,
similarly have a blueish tinge. Clouds, on the other hand, are formed by water
droplets that are much larger than gas molecules and scatter in the same way all the
parts of the light spectrum: this is why clouds appear white (like milk does).

A solution is clearly scattering much less light than a colloidal suspension, and this is
why the way a fluid is scattering light can be used to determine if this fluid contains
a substantial amount of colloidal-sized particles.

Note that the fact that a true solution might be coloured is not due to the scattering
of light, but on the contrary to the adsorption of light by the atoms and molecules
that compose this solution. What we see is not the colour absorbed, but its
complementary colour, originating from the removal of the absorbed wavelengths.
For example, beta-carotene, an organic pigment abundant in plants and fruits, has a
maximum absorption at 454 nm (blue light) and consequently what we see appears
orange (the complementary colour for that type of blue). (Beta-carotene takes its
name from the carrot (daucus carota) from which it can be extracted.)

A suspension (not necessarily colloidal) is a fluid composed of a solvent and particles

in suspension. At low or at no shear rate, these fluids are usually containing colloidal
particles to be classified as “suspensions”. If the particles constituting the suspension

19
 Introduction to Colloid Science

would be all much larger than a few microns the time the particles would remain in
suspension would be very limited at low shear: sand particles for example, settle out
very rapidly and large gas bubbles usually move rapidly towards the water/air
interface. In natural systems, however, it is possible to keep fine particles in
suspension by turbulent mixing. This mixing insures that sand can be transported
over some considerable distance during storms for example.

Typical orders of magnitude for ions, colloidal and sand particles:

Ions colloidal particles sand

typical particle size 10-10 m 10-6 m 10-4 m
number of particles
in a spoon (15 mL) 1025 1013 107
number of sand
equivalent in 10000 times the number of
particles in the
number to population of Earth inhabitants of Paris
Sahara

The unit to measure the number of ions in a litre is mol/L. A mole (symbol: mol)
corresponds to 6.02x1023 particles, which is the number of 12C particles contained in
12 g. One speaks of Avogadro’s number (symbol: NA) to express the fact that there
are 6.02x1023 particles in one mole (NA = 6.02x1023 mol-1). A solution of 1 mol/L
therefore contains 6.02x1023 particles per litre. The molar mass of a product enables
to determine how many grams of product is contained in one mol. The molar mass
of NaCl salt (kitchen salt) is 58 g/mol. There are therefore 6.02x1023 salt particles in
58 g of NaCl, which will give, when dissociated in water, 6.02x1023 Na+ particles and
6.02x1023 Cl- particles. Comparatively, 58 g of clay particles which have a density of
2.6 kg/L (2600 kg/m3) and a size of 1 micron will give approximatively 1013 clay
particles, i.e. 1010 times less particles, or, in other words, if 58 g of clay is mixed with
58 g of salt in one litre of water, for each clay particle, there will be 1010 ionic
particles. The typical concentration of salt ions in natural systems ranges from less
than mM (millimol/L or 10-3 mol/L) for fresh water system to hundreds of mM for
sea water. Clay concentrations are usual of the order of a few mg/L at sea to
hundreds of mg/L and beyond in estuaries.

The difference between clay, silt and sand particle sizes

Sand is a naturally occurring granular material composed of finely divided rock and
mineral particles. It is defined by size, being finer than gravel and coarser than silt.
Here we differentiate sand, silt and clay based on size. Depending on the context,
silt can refer to a type of mud or particles with a specific grain size. A similar
confusing double definition exists for clays, see the table below. Particles that have
a “clay” size may not necessarily be clay minerals (organic matter for instance can
have a clay size but is not composed of clay minerals).

20
 Chapter 1 Introduction, general definitions

Sand feels gritty when rubbed between the fingers. Silt, by comparison, feels like
flour. Even though the exact definitions, in terms of size, of sand, silt and clay are
depending on countries, one assume in general that they can be defined as follows:

Traditionally, sieves could not be made for particle sizes smaller than 63 µm. This is
why the lower limit of 63 µm was adopted for sand. The size of silt particles is
measured traditionally by hydrometer test. Nowadays, Static Light Scattering (laser
diffraction) devices enable to measure the whole range of particle sizes (from 1 nm
to 1 mm) with a single device, see Chapter 2.

The most common constituent of sand, in inland continental settings and non-
tropical coastal settings, is silica (silicon dioxide, or SiO2), usually in the form of
quartz, which, because of its chemical inertness and considerable hardness, is the
most common mineral resistant to weathering. In contact with water or electrolyte,
the presence of hydroxyl (silanol) groups have however been observed on the
surface of silicas and silicates, leading to the creation of a surface charge.

An easy way to find out roughly the composition of a sediment is to use a jar test:
Fill the jar halfway with sediment. Add water to fill up the jar, close the lid, shake,

21
 Introduction to Colloid Science

and set on the table. The sand will settle out quickly, in a matter of hours (like “Sand”
in the example underneath). The middle jar (“Loam”) represents the mixture of clay
and silt, which takes up to 24 hours to settle out. The third “Clay” jar represents how
long it takes clay content to settle out, in this example up to three days.

Loam is soil composed mostly of sand (particle size > 63 µm), silt (particle size > 2
µm), and a smaller amount of clay (particle size < 2 µm). Loam soils generally contain
more nutrients, moisture, and humus than sandy soils, have better drainage and
infiltration of water and air than silty soils, and are easier to till than clay soils. Loam,
(combined or not with straw), has been used in construction since ancient times.

The definition of mud, silt and clay

Mud is composed, in part, of clay mineral and water. For the other parts, most types
of mud are include ions and organic material. Particles of different sizes, such as clay,
silt and sand particles can also be incorporated in mud. The most common
definitions of mud, silt and clay in English, French and Dutch are given in the table
below 7.

Clay particles are colloidal particles when defined by their size (< 2µm). The lower
size range of silt particles can also be seen as colloidal, but in most cases silt particles
(when composed of mineral clay) are non-cohesive, which implies that their
dynamics in the water column are dominated by the gravity force and
hydrodynamics similarly to sand particles.

Clay particles can be found in all water bodies (sea, lake, river), but their relative
ratio compared to silt and sand varies of course from site to site. A coastal wetland
where clays and silts can be found are mudflats. Mudflats are formed when mud is
deposited by tides (tidal flat) or rivers. Mudflats (French: vasière; Dutch: wad) can
be seen as exposed layers of (wet) mud. A tidal flat is submerged approximately
twice daily enabling estuarine silts, clays and organic matter to be deposited. On the
Baltic Sea coast of Germany in places, mudflats are exposed not by tidal action, but
by wind-action driving water away from the shallows into the sea. These wind-
affected mudflats are called windwatts in German. Mudflat hiking (Dutch:
Wadlopen; German: Wattwandern; Danish: Vadehavsvandring) is a recreation
enjoyed by Dutch, Germans and Danes in the Netherlands, northwest Germany and
in Denmark. Mudflat hikers are people who, with the aid of a tide table, use a period
of low water to walk and wade on the watershed of the mudflats. The study of the

7 See in particular for the French definitions: “ La vase” by J. Bourcart and C. Francis-Boeuf
(1942)

22
 Chapter 1 Introduction, general definitions

morphology of mudflats is an on-going topic of research (see middle figure below

where students fix instruments to a rig).

English French Dutch

mud boue modder
liquid or semi-liquid from old Celtic word “baw” Mud and modder are
mixture of principally water (cf gall. baw, “dirt” ): etymologically related to
and clay minerals creamy substance that is Middle Low German and
found, after rainfalls, on Dutch mudde and modde
roads . denoting something wet or
dirty
vase slib
from old Saxon word probably derived from old
“wase”: wet mud that is to Dutch slik/slijk: deposited
be found under water (in wet mud (on mudflats)
lakes, rivers or seas) or
deposited on mudflats
silt limon silt
a - mud carried by running a - from latin “limo - limus” a - from Middle Dutch silte,
water and deposited as a (“mud”): mud that sulte “salt marsh brine”.
sediment, especially in a accumulates due to water Contrary to slib, silt is found
channel or harbour. flows on river sides. in open air and can
therefore oxidize.

b - particles with a size > 2 b - particles with a size > 2 b - particles with a size > 2
µm and < 63 µm µm and < 63 µm µm and < 63 µm
clay 8 argile klei
a - mud that can be a - from latin “argila”: mud a - potklei: clay used in
moulded when wet, used having plastic properties, pottery
in bricks, pottery, and used in bricks, pottery, and
ceramics ceramics b – klei is composed in
majority of lutum (from
b - particles b - particles latin “lut(um)”: clay) i.e.
with a size < 2 µm with a size < 2µm clay particles
with a size < 2µm
c - clay (minerals) are c – argiles (minéraux
hydrous aluminium argileux) are hydrous c – klei(mineralen) are
phyllosilicates aluminium phyllosilicates hydrous aluminium
phyllosilicates

8
Geologists and soil scientists call “clays” particles less than 2 μm. The Dutch klei (clay) is
composed in majority of lutum particles, which are by definition particles smaller than 2 μm.
Sedimentologists often use the limit 4–5 μm, and colloid chemists use 1 μm.

23
 Introduction to Colloid Science

Tidal flat

Wadden Sea in the Netherlands 9

9
http://www.werelderfgoed.nl/werelderfgoed/waddenzee

24
 Chapter 1 Introduction, general definitions

Clay minerals
Clay minerals are common weathering products and low-temperature hydrothermal
alteration products. Clay minerals are very common in soils, in fine-grained
sedimentary rocks such as shale, mudstone, and siltstone and in fine-grained
metamorphic slate and phyllite. Depending on the soil's content in which it is found,
clay can appear in various colours from white to dull grey or brown to deep orange-
red.

Clay minerals are hydrous aluminium phyllosilicates. Phyllosilicates are minerals

formed by parallel sheets of silicate tetrahedra, sometimes with variable amounts
of iron, magnesium, alkali metals, alkaline earths, and other cations found on or near
some planetary surfaces. Depending on the academic source, there are three or four
main groups of clays: kaolinite, montmorillonite-smectite, illite, and chlorite.
Chlorites are not always considered a clay, sometimes being classified as a separate
group within the phyllosilicates. There are approximately 30 different types of
"pure" clays in these categories, but most "natural" clay deposits are mixtures of
these different types, along with other weathered minerals.
Single mineral clay particles are in the clay (colloidal) size range. The identification
of the type and structure of clay minerals requires special analytical techniques as
x-ray and electron diffraction methods (XRD, XRF and SEM in short).

Clay mineralogy

The crystalline structure of clay minerals is built up from different types of sheets or
layers. The fundamental building blocks of these sheets are the tetrahedron and the
octahedron units. The tetrahedron is composed of either a central silicon or
aluminium surrounded by four oxygen ions in a tetrahedral coordination. The
octahedron is composed usually of a central polyvalent cation surrounded by six
oxygen (O) or hydroxyl (OH) ions in an octahedral coordination. Whether a cation
forms tetrahedral or octahedral coordination with oxygen depends on the relative
size of the cations and anions involved.

25
 Introduction to Colloid Science

Tetrahedral layer: All the tetrahedrons in a tetrahedral layer are in the same plane
and therefore their tips point in the same direction. The unit cell formula of the Si
tetrahedron is (Si4O10)4-. Electrical neutrality is obtained by replacement of four
oxygens by hydroxyls (-OH) or by union with a sheet of different composition that is
positively charged.

Octahedral layer: This sheet structure is composed of magnesium or aluminium in

octahedral coordination with oxygens or hydroxyls. If the cation is trivalent, then
normally only two-thirds of the possible cationic spaces are filled and the structure
is termed dioctahedral. In the case of aluminium the composition is Al(OH)3. This
composition and structure form the mineral gibbsite. If the octaedrally coordinated
cation is divalent then normally all possible cation sites are occupied and the
structure is trioctahedral. In the case of magnesium the composition is Mg(OH)2,
which forms the mineral brucite.

Combining the tetrahedral and octahedral layers, the following minerals are formed:

Each platelet has a very large length to width ratio. For commodity, the symbols used
above for the clay minerals have a much smaller ratio.

The chemical structure of minerals is usually far from ideal, as the actual composition
of minerals is frequently altered by isomorphous substitution, i.e. the substitution
of ions within the structure. Weathering allows Si4+, Al3+ and Mg2+ to be substituted
with cations with comparable ionic radii in their respective tetrahedral and
octahedral sheets.

In each of these minerals, the layers combine differently. For kaolinite, we get:

26
 Chapter 1 Introduction, general definitions

A kaolinite particle usually consist of a pile (stack) of platelets (tetrahedral +

octahedral layers forming one platelet). We have here only represented a pile of
two, but the number of platelets forming a single particle can vary significantly:

Scanning Electron Microscopy (SEM) picture of kaolinite platelets

27
 Introduction to Colloid Science

The different classes of Phyllosilicates

Serpentine group

Antigorite – Mg3Si2O5(OH)4
Chrysotile – Mg3Si2O5(OH)4
Lizardite – Mg3Si2O5(OH)4 Lizardite
Clay mineral group

Halloysite – Al2Si2O5(OH)4
Kaolinite – Al2Si2O5(OH)4
Illite – (K,H3O)(Al,Mg,Fe)2(Si,Al)4O10[(OH)2,(H2O)]
Montmorillonite – (Na,Ca)0.33(Al,Mg)2Si4O10(OH)2·nH2O
Vermiculite – (MgFe,Al)3(Al,Si)4O10(OH)2·4H2O
Talc – Mg3Si4O10(OH)2
Sepiolite – Mg4Si6O15(OH)2·6H2O
Palygorskite (or attapulgite) – (Mg,Al)2Si4O10(OH)·4(H2O)
Pyrophyllite – Al2Si4O10(OH)2

Mica group

Biotite – K(Mg,Fe)3(AlSi3)O10(OH)2
Muscovite – KAl2(AlSi3)O10(OH)2
Phlogopite – KMg3(AlSi3)O10(OH)2 Margarite
Lepidolite – K(Li,Al)2–3(AlSi3)O10(OH)2
Margarite – CaAl2(Al2Si2)O10(OH)2
Glauconite – (K,Na)(Al,Mg,Fe)2(Si,Al)4O10(OH)2

Chlorite group

Chlorite – (Mg,Fe)3(Si,Al)4O10(OH)2·(Mg,Fe)3(OH)6

The surface charge of clay minerals

As we will see in the forthcoming Chapters 3 and 5, the surface charge of mineral
clays play a very important role in understanding their aggregation behaviour. The
surface of a clay mineral particle is defined as the parts of the clay exposed to the
bulk water (in contrast to the interlayer water, see figure above).

The surface charge of clays is very dependent on pH. In particular the hydroxyls (OH)
are exposed on the surfaces and edges of the particles and they dissociate in water
under the influence of pH:

28
 Chapter 1 Introduction, general definitions

𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 + 𝐻𝐻2 𝑂𝑂 ⇋ 𝑆𝑆𝑆𝑆𝑂𝑂 − + 𝐻𝐻3 𝑂𝑂+

The higher the pH, the greater is the tendency to dissociate the hydroxyl 𝑂𝑂𝑂𝑂 into
𝐻𝐻3 𝑂𝑂+ (often noted 𝐻𝐻 + for simplification) and 𝑆𝑆𝑆𝑆𝑂𝑂− .

For kaolinite, the following pH dependence is observed:

At low pH, the edges of a platelet are positively charged, whereas the faces are not
(or very weakly negatively) charged. At medium pH, the edges are uncharged, and
the faces are not (or weakly negatively) charged. At high pH, both edges and faces
are negatively charged.

The way clay particles are interacting with each other will be studied in Chapter 5.

The different states of clay

If the water content of a sample made predominantly of water, clay and/or silt is
varied, different states can be observed, for example:

In soil science, criterions have been established to quantify these states. These are
traditionally based on the measurements of the Atterberg limits of the sample. Four
states are defined: solid, semi-solid, plastic and liquid. The boundary between each
state is defined on a change of the soil’s behaviour. From the extended studies done
over the years, it is now possible to distinguish between different types of silts and
clays, based solely on the Atterberg limits of the sample. This is in particular possible

29
 Introduction to Colloid Science

for “standard” soils, i.e. as commonly found in nature. Nonetheless, caution is

required when the soil composition is complex (including a lot of organic material
for example), as these non-mineral components, even in low amounts, can greatly
influence the properties of the soil.

Atterberg and the Atterberg limits

Albert Mauritz Atterberg (1846–1916) was a Swedish

chemist and agricultural scientist who created the
Atterberg limits. He received his Ph.D. in chemistry from
Uppsala University in 1872. At the age of 54, while
continuing his work on chemistry, he began to focus his
efforts on the classification and plasticity of soils.
Atterberg was apparently the first to suggest the limit
<0.002 mm as a classification for clay particles.
Atterberg found plasticity to be a particular
characteristic of clay and as a result of his investigations
arrived at the consistency limits which bear his name
today. He also conducted studies aiming to identify the specific minerals that give a
clayey soil its plastic nature.
The importance of Atterberg’s work has never been fully realized in his own field of
agricultural science. Its introduction to the field of geotechnical engineering was due
to Karl Terzaghi, who came to realise its importance at a relatively early stage of his
research. Terzaghi’s assistant, Arthur Casagrande, standardized the tests in his
paper in 1932 and the procedures have been followed worldwide ever since.

The Atterberg limits are defined as follows:

The liquid state and liquid limit

In the liquid state, the clay sample cannot withstand any type of loading, it is said to
have no strength. When poured, It will flow like
a fluid. The viscosity of the fluid can however
change with clay concentration, this will be
detailed in Chapter 7. The Liquid Limit (LL) is
defined as the moisture content which soil
begins to behave as a liquid material and begins
to flow.

The liquid limit (LL) is obtained by putting a

paste-like sample into a brass cup (Casagrande
cup). A groove is cut in the paste using a
standard tool. The cup is then bumped a
standard number of times using a crank-
operated mechanism. The LL is the moisture content at which the shear strength of

30
 Chapter 1 Introduction, general definitions

the sample is so small that the soil “ flows” to close the groove. The LL can also been
obtained using a cone penetrometer, where a pointy cone is made to free-fall on a
sample from a certain height. If the cone penetrate a given distance (20 mm) inside
the sample the LL is reached.

The plastic state and plastic limit

In the plastic state, as in the example of the potter’s clay, the sample is in a state
where it is possible to make shapes out of it. If the sample is left to dry out for a
short time (so that it is not actually completely dry) it will lose its plasticity. If we
then try to shape it, many cracks will appear, indicating that the clay sample is in its
semi-solid state.

The Plastic Limit (PL) is defined as the moisture content at which the soil begins to
behave as a plastic material.

The plastic limit (PL) is obtained by rolling

down a dough-like sample into threads of 3
mm in diameter on a glass plate. Being able to
roll such a thread is an indication that the PL is
reached. If the water content is too high, the
thread will be too soft and impossible to roll, if
it is too low, some crumbling will take place,
indicating that the semi-solid state is reached.

The shrinkage limit and the solid state

The Shrinkage Limit (SL) is defined as the moisture content at which no further
volume change occurs with further reduction in moisture content.

The shrinkage limit (SL) is obtained when a semi-solid sample does not shrink
anymore by losing moisture content.

In the solid state, the sample cannot be shaped, and becomes brittle.

Moisture content

In the representation of the Atterberg limits hereunder, 𝑤𝑤 represents the moisture

content (also called water content) usually defined as:
𝑚𝑚water 𝑚𝑚wet sample − 𝑚𝑚dry sample
𝑤𝑤 (%) = 100 × = 100 ×
𝑚𝑚dry sample 𝑚𝑚dry sample

where 𝑚𝑚water represents the mass of water in the sample, 𝑚𝑚wet sample the mass of
the wet sample and 𝑚𝑚dry sample the mass of the sample after drying it at a
temperature not exceeding 115°C.

31
 Introduction to Colloid Science

Atterberg limits: depending on water content, several states of a sample made of

predominantly water, clay and/or silt can be observed. The transition between one state and
another is given as a specific water content (in %). SL: Shrinkage Limit, PL: Plastic Limit and
LL: Liquid Limit. The soil volume increases with moisture content once passed the solid state.

As a classification of fine-grained soils, the plasticity index (PI) is introduced:

𝑃𝑃𝑃𝑃 = 𝐿𝐿𝐿𝐿 − 𝑃𝑃𝑃𝑃

The water content of a soil can be related to the soil’s stress-strain response 10:

By increasing the water content (w) of a soil, the amount of stress to exert on it to
get a given strain is reduced.

The stress-strain is obtained by measuring the amount of deformation (strain) at

distinct intervals of tensile or compressive loading (stress). More about stress and

10See for instance An introduction to Geotechnical Engineering, by Holtz & Kovacs, Prentice-
Hall inc.

32
 Chapter 1 Introduction, general definitions

strain is to be found in Chapters 7 and 10, where the rheology and consolidation of
clays is discussed.

Mud found in water systems is mainly liquid, but sampled at the bottom of a water
body its consistency can be plastic. Semi-solid and solid mud is found on land, as it
requires the dissipation of a certain amount of water, which can be obtained by
evaporation. To study and characterize the physical properties of liquid and plastic
mud rheological tests are usually performed, see Chapter 7. These tests are to be
preferred to Atterberg limits tests as they can give quantitative estimations for the
stress/strain (or shear rate) responses. For more consolidated materials, other tests
such as the oedometer tests are performed (see Chapter 10). Attenberg limit tests
have therefore only an engineering value and are usually used in the preliminary
stages of designing any structure to ensure that the soil will have the correct amount
of shear strength. We nonetheless like to present them for their historical value.

Cassagrande’s plasticity chart

Cassagrande (1932) studied the relationship of the plasticity index to the liquid limit
of a variety of natural soils and proposed a plasticity chart, given here:

Casagrande’s plasticity chart, showing several representative soil types

33
 Introduction to Colloid Science

Two lines are defined in the chart:

The A-line which separates the inorganic clays (above the A-line) from the inorganic
silts (below the A-line).

The U-line which is approximately the upper limit of the relationship of the plasticity
index to the liquid limit for any currently known soil. Above the U-line the soils are
therefore assumed to be cohesionless.

“Activity” of clays

It has been demonstrated that the plasticity index to the clay fraction content (<
2µm) is in good approximation constant 11:

This led to the definition of “colloidal activity”:

plasticity index
activity =
clay fraction

In simple words, the clay (colloidal) content of a soil is directly proportional to its
cohesiveness. We will see in Chapter 5 that montmorillonite clay has special
properties that explains the fact it is more “active” than other clays. The presence or
not of salt within the sample also greatly influences the values of the plasticity
index 12. This can be understood from colloidal interactions, which strongly depend
on salinity, as will be seen in Chapter 3.

11Skempton, A. W. "The colloidal activity of clays." Selected papers on soil mechanics (1953):
106-118.
12 Bjerrum, Laurits. "Geotechnical properties of Norwegian marine clays." Geotechnique 4.2
(1954): 49-69.

34
 Chapter 1 Introduction, general definitions

Illustrations
the tomb –chapel of Nebamun : Nebamun hunting in the marshes, adapted from a wall
painting currently to be seen at the British Museum; ca. 1350 BC (public domain)
https://commons.wikimedia.org/wiki/File:TombofNebamun-2.jpg

John Tyndall (public domain)

https://en.wikipedia.org/wiki/John_Tyndall

Mudflat (creative commons license)

https://en.wikipedia.org/wiki/Mudflat

lizardite photograph (creative commons license)

https://upload.wikimedia.org/wikipedia/commons/7/71/Lizardite%2C_Chrysotile-
288581.jpg

margarite photograph (creative commons license)

https://en.wikipedia.org/wiki/Margarite

Albert Atterberg (public domain)

http://geotecnia-sor.blogspot.nl/2010/11/consistencia-del-suelo-limites-de_17.html

Casagrande cup (creative commons license)

http://labmodules.soilweb.ca/soil-compaction-atterberg-limits/

35
 Chapter 2
Settling, diffusion and
stabilisation
 Introduction to Colloid Science

Fine (colloidal) particles can remain in suspension for an extremely long (sometimes
infinite) time. In this chapter, we are going to explain why and when these colloidal
particles remain in suspension and how they diffuse and settle. We will primarily
address dilute suspensions i.e. suspensions containing a limited amount of colloidal
particles. The settling behaviour of concentred suspensions will be discussed in
Chapter 8.

Stokes’ settling velocity

In Chapter 1, we have already stated that in colloid science the typical length scale
for a particle is 1 µm. By this, we mean of the order of 1 µm which means a particle
somewhere in the range between 0.01 µm and 10 µm.

Let us now consider a particle of any size > 1 nm in water. Three forces are exerted
on this particle: the force of gravity, the force of Archimedes and the force of friction.
We furthermore assume that

1) the water far from the particle is at rest

2) we have reached the regime where the particle’s velocity is constant (the initial
acceleration is not considered)
3) the velocity is small enough for the fluid to be in the laminar regime.

The fact that a flow is laminar or not can be evaluated by the estimation of the
Reynolds number 13 Re which represents the ratio of inertial forces (which create
turbulence) to viscous forces (which create friction):

𝑣𝑣𝑣𝑣𝑣𝑣
𝑅𝑅𝑅𝑅 =
𝜂𝜂

where 𝑣𝑣 is the fluid’s velocity, 𝐿𝐿 is a characteristic length (in our case the size of
studied particle, as the particle is setting the fluid in motion) and the kinematic
viscosity is given by 𝜂𝜂/𝜌𝜌 where 𝜂𝜂 is the viscosity and 𝜌𝜌 the density of the fluid. The
laminar regime is defined by a low Reynolds number (𝑅𝑅𝑅𝑅 < 10 for a sphere),
implying that the friction force is dominating, in accordance with our initial
assumption. The balance of forces gives:

𝑚𝑚𝒈𝒈 = 6𝜋𝜋𝜋𝜋𝜋𝜋𝒗𝒗

where 𝑚𝑚 is the mass of the particle, compensated for Archimedes, 𝑔𝑔 the gravitation
constant, 𝜂𝜂 the viscosity of the water, 𝑎𝑎 the radius of the particle and 𝑣𝑣 its velocity.
The fact that the force of friction might be expressed as 6𝜋𝜋𝜋𝜋𝜋𝜋𝒗𝒗 for a spherical
particle in a laminar flow in due to the work of Georges Stokes (1819-1903). We have

13 The number has been invented by Stokes but is named after Osborne Reynolds (1842-1912)

who popularised its use.

38
 Chapter 2 Settling, diffusion and stabilisation

expressed the vectors in bold letters, implying that the vector 𝒗𝒗 is directed along 𝒈𝒈.
The mass of the particle, compensated for Archimedes can be expressed as:

4 3
𝑚𝑚 = 𝜋𝜋𝑎𝑎 �𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤 �
3
where 𝜌𝜌𝑝𝑝 is the density of the particle and 𝜌𝜌𝑤𝑤 the density of the water. From the
balance of forces we get:

2 2 𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤
𝒗𝒗 = 𝑎𝑎 𝒈𝒈
9 𝜂𝜂

From this expression which is Stokes’ settling velocity, we can verify that:

(a) the velocity scales as the size squared: a particle twice as big will settle 4 times as
fast.
(b) if the density of the water and the particle are the same the velocity is zero, the
particle remains in suspension. If 𝜌𝜌𝑝𝑝 < 𝜌𝜌𝑤𝑤 the particle will float (this happens for
certain types of algae for instance)

Several instruments make use of Stokes’ settling velocity to assess the size of (sub-)
colloidal particles, like the sedigraph, the hydrometer and the sediment balance.
These techniques are explained below.

Hydrometer and sedigraph methods are based on recording the density evolution in
a settling column as function of time.

39
 Introduction to Colloid Science

Determining particle sizes experimentally

The hydrometer test

For particles that pass the smallest sieve (63 μm in general), their particle size can
be assessed by the use of a hydrometer. The hydrometer consists of a
cylindrical stem with a precise scale on it and a bulb weighted with lead to
make it float upright. The suspension to test is poured into a column, and
the hydrometer is gently lowered into the liquid until it floats freely. The
hydrometer indicates the ratio of the density of the tested fluid to the
density of water (this ratio is called the specific gravity). After calibration
of the hydrometer, the height 𝐻𝐻𝑒𝑒 (see figure above) can be related to the
density of the suspension. The height 𝐻𝐻𝑒𝑒 is increasing in time as the fluid
becomes less and less dense. From Stokes’ law it is possible to estimate the
size of particles that, for a given 𝑡𝑡, will be above 𝐻𝐻𝑒𝑒 . These are the particles
with a radius smaller than:

9𝜂𝜂𝐻𝐻𝑒𝑒
𝑎𝑎 = �
2�𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤 �𝑔𝑔𝑔𝑔

The density evaluated from the hydrometer at time 𝑡𝑡 is given by: 𝜌𝜌(𝑡𝑡) = 𝜙𝜙(𝑡𝑡)𝜌𝜌𝑝𝑝 +
�1 − 𝜙𝜙(𝑡𝑡)�𝜌𝜌𝑤𝑤 where 𝜙𝜙(𝑡𝑡) = ∑ 𝑁𝑁𝑖𝑖 (𝑡𝑡) 𝑉𝑉𝑖𝑖 /𝑉𝑉 is the total volume fraction of suspended
particles; 𝑁𝑁𝑖𝑖 (𝑡𝑡) is the number of particles with volume 𝑉𝑉𝑖𝑖 in suspension at time t and
𝑉𝑉 is the total volume of fluid in the column. Assuming that all particles have the same
density (𝜌𝜌𝑝𝑝 = 𝑚𝑚𝑖𝑖 /𝑉𝑉𝑖𝑖 where 𝑚𝑚𝑖𝑖 is the mass of particle i) we see that 𝜌𝜌𝑝𝑝 𝜙𝜙(𝑡𝑡)
represents the total mass in suspension at time 𝑡𝑡 per unit volume. The percentage
in mass of particles remaining in suspension (so with a radius smaller than 𝑎𝑎) can
therefore be estimated from:

𝜌𝜌𝑝𝑝 (𝜌𝜌(𝑡𝑡) − 𝜌𝜌𝑤𝑤 )

% smaller than 𝑎𝑎 = 100
𝜌𝜌0 �𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤 �

where 𝜌𝜌0 = 𝑚𝑚/𝑉𝑉 ; 𝑚𝑚 is the mass of dry soil in the column.

The sedigraph

The sedigraph uses a paralleled X-ray beam to detect changes in suspended

sediment concentration: the X-ray attenuation is proportional to mass
concentration, following the so-called Beer-Lambert law. The beam can be
positioned at different vertical positions in the analysis cell, and records the changes
at different times. In essence, the principle is very similar to the hydrometer (see
above) except that the concentration is recorded through X-ray attenuation instead
of the estimation of the specific gravity.

40
 Chapter 2 Settling, diffusion and stabilisation

The sediment balance

The sediment balance records the mass 𝑀𝑀 of particles falling on a balance as a

function of time: a long rod with a disk attached to it is plunged in a
settling column containing a suspension. The rod is attached to a
laboratory scale. The assumptions made when performing a
sediment balance measurement are that 1) the suspension is
homogeneous at t = 0 (start). The suspension has been mixed just
prior the measurements and 2) the particles settle according to
Stokes’ law. This implies that particles with a diameter 𝑑𝑑 = 2𝑎𝑎 will
all have arrived on the balance at the time: 𝑡𝑡 = 18ℎ𝜂𝜂/[(𝜌𝜌 −
𝜌𝜌𝑊𝑊 )𝑔𝑔𝑑𝑑 2 ] where ℎ is the height of the liquid in the column. From t =
0 (start) until t these particles have been falling continuously on the
balance, in a linear way, since the suspension is considered to be
properly mixed. It is therefore possible to estimate the percentage of
particles that at a time t have a settling time smaller than (or equal
to) t, i.e. are larger than or equal to d. The relation is given by:

𝑀𝑀 − 𝑡𝑡 ∙ 𝑑𝑑𝑑𝑑/𝑑𝑑𝑑𝑑
% particles larger than d = 100 ×
𝑀𝑀𝑒𝑒𝑒𝑒𝑒𝑒

where 𝑀𝑀 is the mass on the balance at time t, 𝑀𝑀𝑒𝑒𝑒𝑒𝑒𝑒 is the final mass on the balance
(after one day, typically) and 𝑑𝑑𝑑𝑑/𝑑𝑑𝑑𝑑 represents the slope of the curve M(t) at time
t. The formula is called the Oden formula, as it has originally been derived by Oden
in 1916 14. Note that 𝑡𝑡 ∙ 𝑑𝑑𝑑𝑑/𝑑𝑑𝑑𝑑 represents the amount of particles at time t with a
settling velocity smaller than or equal to h/t whereas 𝑀𝑀 represents the amount of
particles with a settling velocity smaller than or equal to h/t between t = 0 and t. If
there would be only particle of one size, then 𝑀𝑀 − 𝑡𝑡 ∙ 𝑑𝑑𝑑𝑑/𝑑𝑑𝑑𝑑 = 0 until all the
particles have fallen on the balance. When this happens, 𝑑𝑑𝑑𝑑/𝑑𝑑𝑑𝑑 = 0 and we find
that :

% particles larger than d = 100

For each of the techniques (hydrometer, sedigraph and sediment balance), it is

crucial to know the proper density 𝜌𝜌 of the particles. In the case of aggregated clay
particles, the density can vary significantly from the mineral clay density, in which
case the measurements are not reliable. In most protocols it is therefore important
to deflocculate the particles prior measurements. This is done by chemical agents or
ultrasonication.

A colloidal mineral clay particle, exposed to gravity, will always settle according to
the expression for Stokes’ settling since clay minerals have a density of the order of

14 Odén, Sven. "Eine neue methode zur bestimmung der körnerverteilung in

Suspensionen." Colloid & Polymer Science18.2 (1916): 33-48.

41
 Introduction to Colloid Science

2.6 kg/L (the density of water is of the order of 1 kg/L). As colloidal clay particles can
also remain in suspension without settling, it implies that we missed important
forces while setting-up the force balance. One is related to so-called Brownian
motion. Another is related to the repulsion that can exist between colloidal particles.
These forces and their principles are described in the next sections.

120 100

100 80

80 60

% larger than
mass (g)

60 40

40 20

20 0

0 -20
0 200 400 600 800 1000 0 42 55 95 150
time(s) diameter (microns)

Sediment balance principle. Left: The coloured curves represent the mass on the balance as
function of time for suspensions made of: (red): 35 g of particles with a diameter of 95
microns, (magenta): 25 g of particles with a diameter of 55 microns, (green): 50 g of particles
with a diameter of 42 microns and (blue) a suspension made by the mixture of the three
types of particles. After using Oden’s formula on the left blue curve, one obtains the blue
curve on the right figure, which gives the amount of particles larger than a certain size in %.

Brownian motion
This transport phenomenon is named after the botanist
Robert Brown (1773-1858). In 1827, while looking
through a microscope at particles trapped in the water
of the interior of pollen grains, he noted that the
particles moved through the water but he was not able
to determine the mechanisms that caused this motion.
It is Albert Einstein in a paper in 1905 that explained in
precise detail how the motion that Brown had observed
was a result of the particles being moved by individual
water molecules. Atoms and molecules had long been
theorized as the constituents of matter and therefore Einstein’s explanation of
Brownian motion served as convincing evidence that atoms and molecules exist. It
was verified experimentally by Jean Perrin in 1908. Perrin was awarded the Nobel
Prize in Physics in 1926 "for his work on the discontinuous structure of matter"
(Einstein had received the award five years earlier "for his services to theoretical
physics" with specific citations of different researches).

The fact that atoms or molecules (like water molecules) are moving randomly is
associated to temperature: temperature is an indirect measure of the microscopic

42
 Chapter 2 Settling, diffusion and stabilisation

movement of these particles. A whole branch of science (i.e. thermodynamics) is

based on the phenomena associated to temperature and temperature fluctuations.
From thermodynamic principles, one has defined a proper unit to measure
temperature: the Kelvin, which is linked to Celsius degrees by:

𝐾𝐾 = ℃ + 273.15

The temperature corresponding to zero Kelvin is called the “absolute zero” , a

temperature never to be reached as it would correspond to a state where molecules
would have a zero velocity, which is in contradiction with Heisenberg’s uncertainty
principle in quantum mechanics 15. In 1908, the Dutch scientist Kamerlingh Onnes 16
and his research group in Leiden were the first ones to be able to measure a
temperature of 0.9 K, which made Leiden “the coldest place on Earth”. This led to a
Nobel prize in Physics in 1913 for K. Onnes.

Brownian movement: due to the thermal agitation of the water molecules, colloidal particles
experience a random motion (the trajectory of such a particle is given on the right panel)

The direction of the total force (red arrow in the figure above) on the particles due
to the bombardment by water molecules is random, and therefore constantly
changing, which makes it impossible to derive such a force. At different times the
particle is hit more on one side than another, leading to the stochastic (random)
nature of the motion of the particle. This type of motion is called Brownian motion
or random walk. The trajectory of a particle experiencing a random walk is given in
the figure on the right.

Statistically, if one would add all the velocities of all the particles in the water at a
given time, the resulting velocity would be zero: there is no global movement
(contrary to particles in a flow, where clearly when adding all the velocities, the total
velocity will be in the direction of the flow). If one would follow a given particle

15 The uncertainty principle states that it is impossible to know exactly at the same time both
the momentum and the position of quantum (i.e. very very small) particles. If the velocity (and
momentum) of particles would be zero, then it is obvious that their position should exactly be
known.
16 https://www.lorentz.leidenuniv.nl/history/cold/cold.html

43
 Introduction to Colloid Science

submitted to Brownian motion as function of time, one would observe that this
particle gets further and further away from its starting point, but that no direction is
privileged. It is a talented mathematician, Louis Bachelier 17, who, in a thesis
presented in 1900, demonstrated that what characterized the movement of such a
particle was not the arithmetic mean of its displacement 〈𝑋𝑋〉 but in fact its root
mean square 〈𝑋𝑋 2 〉:

1 𝑡𝑡
〈𝑋𝑋 2 (𝑡𝑡)〉 = � 𝑥𝑥 2 (𝜏𝜏)𝑑𝑑𝑑𝑑
𝑡𝑡 0

where 𝑥𝑥 is the position of particle and 𝑡𝑡 the time. From this relation it can be shown
that the root mean square displacement is proportional to time:

〈𝑋𝑋 2 (𝑡𝑡)〉 = 2𝑑𝑑𝑑𝑑𝑑𝑑

where 𝑑𝑑 is the dimension of the movement 18 (linear d=1, planar d=2 or spatial d=3),
𝐷𝐷 is the diffusion coefficient and 𝑡𝑡 the time. Note that if the particle would undergo
a regular translation of the type 𝑥𝑥(𝑡𝑡) = 𝑣𝑣0 𝑡𝑡 where 𝑣𝑣0 is a constant, one finds:

1 𝑡𝑡 𝑣𝑣0 𝑡𝑡
〈𝑋𝑋(𝑡𝑡)〉 = � 𝑣𝑣0 𝜏𝜏𝜏𝜏𝜏𝜏 =
𝑡𝑡 0 2

and in that case, it would be the arithmetic mean of the displacement 〈𝑋𝑋(𝑡𝑡)〉 which
would be proportional to time (at a given time 𝑡𝑡, the distance between the particle
and its origin at 𝑡𝑡 = 0 is proportional to time). For a particle under a 3D Brownian
motion, one has 𝑑𝑑 = 3, hence:

�〈𝑋𝑋 2 (𝑡𝑡)〉 = Δ𝑟𝑟 = √6𝐷𝐷𝐷𝐷

where Δ𝑟𝑟 = �〈𝑋𝑋 2 (𝑡𝑡)〉 symbolises the averaged position relative to the original
position (at 𝑡𝑡 = 0) of the particle (see illustration underneath). In this case, the
(averaged) distance between the particle and its origin at 𝑡𝑡 = 0 is proportional to
the square root of time.

One unknown in the previous equation remains the diffusion coefficient 𝐷𝐷. It is
Albert Einstein in 1905 and independently Marjan Smoluchowski in 1906, who
derived an expression for this diffusion coefficient, based on kinetic theory. In fact
the so-called Einstein-Smoluchowski relation is an early example of the fluctuation-
dissipation relation, which is a powerful theorem in statistical physics:

17https://en.wikipedia.org/wiki/Louis_Bachelier
18In Chapter 10 the dimension 𝑑𝑑 will be represented by D. In the present chapter it could be
confused with the diffusion coefficient D, so we adapted the notation.

44
 Chapter 2 Settling, diffusion and stabilisation

𝐷𝐷 = 𝜇𝜇𝑘𝑘𝐵𝐵 𝑇𝑇

where 𝑘𝑘𝐵𝐵 is Boltzmann’s constant (𝑘𝑘𝐵𝐵 = 1.38 × 10−23 𝐽𝐽/𝐾𝐾) and 𝑇𝑇 the temperature
in Kelvin.

Illustration of a 2D Brownian displacement of a spherical particle. At 𝑡𝑡 = 0 the particle is at

origin, and 𝛥𝛥𝛥𝛥 = 0. For a given time 𝑡𝑡 > 0 one can infer that the particle will be, on average,
at a distance 𝛥𝛥𝛥𝛥 = √4𝐷𝐷𝐷𝐷 from its origin. The direction of the particle relative to its origin is
unknown.

The product 𝑘𝑘𝐵𝐵 𝑇𝑇 is called the thermal energy. The mobility 𝜇𝜇 of the particle is
defined as the ratio between the particle’s terminal velocity 𝑣𝑣 to the applied force
𝐹𝐹:
𝑣𝑣
𝜇𝜇 =
𝐹𝐹
Thanks to Newton’s equation of motion (that we used earlier to establish Stokes’
settling velocity), and using Stokes’ s friction force, we have:

𝑭𝑭 = 6𝜋𝜋𝜋𝜋𝜋𝜋𝒗𝒗

We have already said that Stokes’ frictional force (also called drag force) exerted on
a spherical particle is only valid at very small Reynolds numbers (at small Reynolds
numbers, the fluid flow round the particle is laminar, i.e. non turbulent. This happens
for small flow velocities). The force 𝑭𝑭 is due to the bombardment of water molecules
on the particles. Without having to know this force (which is of the order of 𝑘𝑘𝐵𝐵 𝑇𝑇/𝑎𝑎),
we easily get 𝜇𝜇 = 1/(6𝜋𝜋𝜋𝜋𝜋𝜋) and we obtain the Stokes-Einstein relation:

𝑘𝑘𝐵𝐵 𝑇𝑇
𝐷𝐷 =
6𝜋𝜋𝜂𝜂𝑎𝑎

This is the theoretical expression for the diffusion coefficient of a sphere subjected
to Brownian motion. It is from this relation that one has experimentally determined

45
 Introduction to Colloid Science

the radius of ions and small colloidal particles. The radius thus obtained is called
hydrodynamic radius (or Stokes radius) as it is known that moving particles always
carry some molecules of water with them. The hydrodynamic radius of particles can
be assessed by Dynamic Light Scattering 19.

Static and Dynamic Light Scattering

Static and Dynamic Light Scattering (SLS and DLS) are two different techniques that
both enable to assess the size of particles. Dynamic Light Scattering can be used for
particles in the range [ 1nm – 1 μm] whereas Static Light Scattering can be used for
the range of particle size, besides gravel: [1 nm – 1 mm]. DLS works on the following
principle: when light hits particles, it scatters in all direction provided that the
particles are smaller than the wavelength (about 600 nm for a red laser). The total
scattering intensity of the light that is collected in a photodetector fluctuates over
time, owing to the Brownian motion of the particles that produce interferences. This
fluctuation can be linked to a characteristic time which is a function of the diffusion
coefficient D. From D, using the Stokes-Einstein relation, the size of the particles can
be obtained.

Principle of Dynamic Light Scattering (DLS). The scattered light is measured at an

certain angle relative to the incident laser beam direction

Like DLS, Static light scattering (SLS) also measures the scattered light. But, instead
of measuring the time-dependent fluctuations in the scattering intensity, SLS makes
use of the time-averaged intensity of scattered light. For particle size measurements,
SLS makes use of multiple photodetectors, positioned at various angles relative to
the incident beam: depending on the size of the particles, the scattering intensity
will be different in the different detectors (Rayleigh scattering occurs for extremely

19 for particles larger than a few nm. For ions (of size about 0.1 nm), the hydrodynamic radius

is determined using conductivity measurements and can nowadays easily to be found in

Handbooks. See also the end of Chapter 3.

46
 Chapter 2 Settling, diffusion and stabilisation

small particles, whereas Mie scattering occurs for particles larger than the
wavelength) :

In the Rayleigh regime, the intensity of the scattered light is proportional to

1 + cos 2 (𝜃𝜃)
𝐼𝐼 ~
𝜆𝜆4
where 𝜃𝜃 is the angle between the incident beam and the observer and 𝜆𝜆 is the
wavelength of the incident beam. It can be verified that

𝐼𝐼(𝜃𝜃 = 0)
𝜋𝜋 = 2
𝐼𝐼 �𝜃𝜃 = �
2
which is illustrated on the figure above: the scattering at right angles is half the
forward intensity. The dependence of the intensity of the fourth power of the wave
length leads to the fact that short wave lengths are more scattered than long wave
lengths. As already discussed in Chapter 1, this is the reason why the blue colour
(short wavelength) is more scattered than the red (long wavelength) for small
particles.

Fick’s laws
Fick’s first law

The diffusion coefficient 𝐷𝐷 expressed above is very important as it is used in

fundamental relations, such as, for example Fick’s law of diffusion (Fick’s first law):

𝜕𝜕𝜕𝜕
𝑱𝑱 = −𝐷𝐷 𝒆𝒆
𝜕𝜕𝜕𝜕 𝒙𝒙
where 𝑱𝑱 is the diffusion flux, 𝑛𝑛 is the concentration of particles (ex: colloidal clay
particles) and 𝒆𝒆𝒙𝒙 the unit vector in the x direction. The units of 𝑱𝑱 are in number of
particles per square meters per second. The law has here been written for a 1d

47
 Introduction to Colloid Science

diffusion, it is easy to generalize to more dimensions (using the general relation 𝑱𝑱 =

−𝐷𝐷𝛁𝛁𝑛𝑛).

Fick’s law expresses the fact that given a gradient of concentration, the colloidal
particles will move towards the region of low particle concentration (hence the
minus sign). For example, if a suspension of colloidal particles (or an electrolyte
solution) is pipetted into a jar of water the particles will diffuse into the whole jar,
and after some time no gradient in concentration will be observed anymore: the
particle concentration in the jar will be everywhere the same:

Colloidal particles obeying Fick’s law and diffusing to regions of low particle concentration.
When the particle concentration is everywhere the same, the particles will still move due to
Brownian motion, but the particle concentration will not change anymore and remain
uniform in the jar. (the action of gravity is neglected in this example)

Fick’s second law

Conversation of matter dictates that for a given volume, during a small time 𝑑𝑑𝑑𝑑, the
difference between the fluxes of particles entering and going out of this volume
must be equal to the variation of concentration of the same particles. In
mathematical notations this implies that:

𝜕𝜕𝜕𝜕
[𝐽𝐽(𝑥𝑥) − 𝐽𝐽(𝑥𝑥 + 𝑑𝑑𝑑𝑑)] = 𝑑𝑑𝑑𝑑
𝜕𝜕𝜕𝜕
and therefore:

𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
− =
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
Using Fick’s first law, one gets Fick’s second law:

𝜕𝜕 2 𝑛𝑛 𝜕𝜕𝜕𝜕
𝐷𝐷 =
𝜕𝜕𝑥𝑥 2 𝜕𝜕𝜕𝜕

48
 Chapter 2 Settling, diffusion and stabilisation

Fick’s laws – and similar type of relations - are found in many branches of science,
where diffusion plays a role (these equations can also be used to describe the
transport of heat or momentum).

Fick’s second law enables to give a good example of the connection between
Brownian motion and diffusion processes. We consider the case where the diffusion
of particles can be schematically visualized as follows:

The initial conditions of the experiment are that all the particles are located at 𝑥𝑥 =
0 in an infinitely thin layer Δ𝑙𝑙. We have: 𝑛𝑛(0,0) = 𝑛𝑛0 Δ𝑙𝑙. To account for the fact that
the layer is infinitely small, one expresses usually that: 𝑛𝑛(𝑥𝑥, 0) = 𝑛𝑛0 Δ𝑙𝑙𝑙𝑙(𝑥𝑥) where
𝛿𝛿(𝑥𝑥) is called the Dirac distribution which is a function such 𝛿𝛿(𝑥𝑥 = 0) = 1 and
𝛿𝛿(𝑥𝑥 ≠ 0) = 0. Using this Dirac function insures that one has:
∞
� 𝑛𝑛(𝑥𝑥, 𝑡𝑡) 𝑑𝑑𝑑𝑑 = 𝑛𝑛0 Δ𝑙𝑙
0

for any time t (conservation of mass principle). One can verify that the solution of
Fick’s second law, in the simple case considered here (i.e. the diffusion is only along
𝑥𝑥) is:

𝑛𝑛0 Δ𝑙𝑙 −𝑥𝑥 2

𝑛𝑛(𝑥𝑥, 𝑡𝑡) = exp � �
√𝜋𝜋𝜋𝜋𝜋𝜋 4𝐷𝐷𝐷𝐷

The characteristic timescale associated to the diffusion is:

𝑥𝑥 2
𝜏𝜏 =
4𝐷𝐷
Note that the characteristic time 𝜏𝜏 can be linked to the measure of how far the
substance has spread in a given time 𝑡𝑡 = 𝜏𝜏:

〈𝑋𝑋 2 (𝑡𝑡)〉 = 2𝑑𝑑𝑑𝑑𝑑𝑑

49
 Introduction to Colloid Science

where 𝑑𝑑 = 1 as we have here a linear movement (dimension 1). This is the

expression we have seen to be related to Brownian motion.

The functions plotted underneath were obtained numerically from the expression of
𝑛𝑛(𝑥𝑥, 𝑡𝑡) given above, and each function was normalized by dividing it by 𝑛𝑛(𝑥𝑥 = 0, 𝑡𝑡):

1
Dt = 0.05
0.9 Dt = 0.1
0.8 Dt = 0.25
Dt = 1
0.7
n(x,t)/n(x=0,t)

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.5 1 1.5 2
x

At large times (𝑡𝑡 ≫ 𝜏𝜏), one will obtain: 𝑛𝑛/𝑛𝑛(𝑥𝑥 = 0, 𝑡𝑡) = 1 as there will be no
gradient in concentration anymore.

Important: It should be emphasized that the diffusion that has been discussed here
is not the one defined in hydrodynamics: in that case, the diffusion term, still
𝜕𝜕𝜕𝜕
expressed as 𝐷𝐷 as in Fick’s first law, refers to a diffusion coefficient 𝐷𝐷 that is
𝜕𝜕𝜕𝜕
orders of magnitude larger than the diffusion coefficient discussed above. In colloid
science, the diffusion coefficient is associated to the thermal energy 𝑘𝑘𝐵𝐵 𝑇𝑇 whereas in
hydrodynamics the diffusion is related to the water flow (usually in the context of
turbulent mixing). The typical diffusion coefficient in colloidal science is of the order
of 10-9 m2/s whereas it is 10-1 m2/s in hydrodynamics: in short, it is more efficient to
stir a cup of coffee than to wait for the sugar molecules to diffuse according to
Brownian motion!

This does not imply that Brownian motion should be neglected when dealing with
turbulent mixing: it plays an important role for predicting the flocculation of colloids.

50
 Chapter 2 Settling, diffusion and stabilisation

Application of Fick’s law : cleaning colloidal suspensions

In order to study colloidal suspensions, it is often required to “clean” them, i.e.

remove as much dissolved salt or small dissolved molecules as possible from the
suspension. Doing so, it is then possible to perform accurate measurements as
function of a chosen added salt for example and avoid that the background
electrolyte concentration contaminates the measurements.

A well-established technique to clean suspensions is to perform a dialysis. The

suspension is put in a flexible tube (Visking tube), made of a membrane that only
allow the passage of small molecules. The principle of the membrane is based on
diffusion: the tube containing the suspension is placed in a jar containing ultra-pure
water (of extremely low conductivity). By diffusion the small molecules and ions will
then tend to pass the membrane of the tube and invade the ultra-pure water. As the
colloidal particles are too big, they cannot pass the membrane and remain in the
tube. After an equilibrium state is reached, the tube is removed and the
contaminated water is replaced by new ultra-pure water. This procedure is repeated
several times, until the conductivity of the water at equilibrium is close to the one
of pure water.

Dialysis of a colloidal suspension

The cleaning can also be accelerated by using electrodialysis, where the application
of an electric field speeds up the migration of ions:

51
 Introduction to Colloid Science

The ions move by electrophoresis to the oppositely charged electrodes, a

phenomenon that will be discussed in Chapter 3.

Stable and unstable colloidal suspensions

An important condition to keep a colloidal particle suspended is that it cannot “glue”
to another particle (and another, and
another…) to form a larger entity that might
eventually settle under its own weight. One
says of such isotropic suspensions in which
particles do not glue or settle that they are
“stable”. In the next chapter, we are going
to review in detail the mechanisms leading
to unstable suspensions. For the moment,
we simply note that stable suspensions are composed of:

 particles with a surface charge of same sign, since these particles are
repelling each other (whereas particles with opposite charges are attracting
each other)

 particles in special mixtures (defined in the Chapter 4), or particles coated

with a special type of molecules can also create stable suspensions,
irrespective of the surface charge of the particles. (The mechanisms leading
to stability are then not linked to surface charge).

In order to be stable, the particles should also be able to overcome gravity. This can
be done when the particles are colloidal by Brownian motion, or, in special cases by
electrostatic forces (discussed below). We will distinguish four different types of
suspensions:

52
 Chapter 2 Settling, diffusion and stabilisation

Stable suspensions: case (A) and case (B)

A stable isotropic dilute suspension of particles can be

achieved when the particles are undergoing Brownian
motion. Gravity then usually plays a negligible role (this is
the case for a small jar or settling column – for some meters
high suspended sediment, on the other hand, a gradient in
concentration can appear). The particles should also repel
each other, otherwise they would glue at some point , form
a larger entity and settle. Ions in water, even though ions are
usually smaller than the typical colloid size 20, satisfy all these
conditions. Small colloidal particles, with a significant
surface charge, also satisfy these conditions. A typical example is the colloidal gold
suspension (see picture on the right) prepared by Faraday (1791-1867) that have
remained stable over more than a century.

The difference between case (A) and case (B) lays in the fact that because of the
high concentration the particles are “ trapped” in their positions in case (B) and do
not “ walk around” easily (thanks to Brownian motion) as in case (A). An important
consequence of this trapping is that the relative distance between the particles is
rather constant, similar to what one has in crystals. This is why this type of
suspension is called a liquid crystal. A lot of work in the 1960-70s has been done on
colloidal crystals, as they were used as models for atomic systems that cannot easily
been investigated. The research continues on this type of systems, especially on non-
spherical colloids that can arrange themselves in very different ways, depending on
parameters such as temperature, pressure, and concentration:

20 Some ions can be of colloidal size: an example are ions called polyoxometalate (abbreviated

POM), composed of metal and oxide; the size of these ions can reach 10 nm – the lower end
of what is called a colloidal particle.

53
 Introduction to Colloid Science

Liquid crystals made a great break-through in the 1970s as they became

commercially available and were used in digital calculators as Liquid-Crystal Displays
(LCDs). The principle of a LCD is similar to the one used in research to investigate the
birefringence of complex suspensions. In birefringence studies, an electric field is
applied to a monodisperse suspension, such that all particles will align with the
electric field thanks to electric dipole-dipole interactions. When the electric field is
switched off, the particles will relax because of their Brownian motion. From
evaluating the time needed for the suspension to relax completely (i.e. become fully
isotropic again), information about the particle’s size can be deduced. The
observation is done by analysing the amount of light going through a photodetector
as function of time. The time for the particles to relax is extremely fast and the
transition cannot be observed by the eye, which implies that for LCDs what one sees
is a pixel switching directly from black to white (and vice-versa):

Liquid crystal principle: a cholesteric nematic liquid crystal is placed between two
transparent electrodes (in pink). Cholesteric liquid crystals have the property that they
organize themselves in layers and that each layer has a director axis, represented here by the
direction of a blue rod. The distance between electrodes is tuned in such a way that the light
exiting the cell can pass through the filter and illuminate a pixel. When an electric field is
applied, the ordering of the crystal changes and no light can pass through the filter.

54
 Chapter 2 Settling, diffusion and stabilisation

Liquid crystals made of clay minerals are an on-going topic of research 21. The
complexity of the clay mineral structures (see Chapter 1), explains why for years
researchers have privileged simpler colloidal particles than clays. However, the
understanding of the transition of one clay liquid crystal phase to another will enable
a better predictability of the rheology of these liquids, and their consolidation
behaviour. The applications in Civil Engineering are enormous: drilling fluids can be
seen as examples of clay liquid crystals (they are a mixture of bentonite clays and
polymers 22), and so can cohesive sediments (a mixture of clay, silt and organic
matter), found in many delta regions, rivers, river shores and lakes.

Unstable suspensions: case (C) and Case (D)

When the colloidal particles are noticeably sensitive to gravity in the jar or settling
column, they will settle. We still only consider the case that the particles are not
gluing to each other, meaning that they just settle because of their own weight.

Mud pool after a rainfall: the colloidal

particles are in suspension and will slowly
settle down in time.

Now an interesting question arises:

Let’s take silica particles as an

example. We consider two types of
particles: small (colloidal) silica
particles and large (sand) silica
particles. The only difference
between the particles is their size, which means they have the same surface charge
(in C/m2) and the same density. We suppose that the particles are highly charged.
The question is: is it possible to define a size of particles such that the particles are
so repellent that they remain in suspension like in case (B)? Intuitively, we know
that this is not possible for sand particles: sand particles will always settle.

The answer lays in the comparison of the forces exerted on the particles. The
dominant forces are gravity and the electric repulsion. We have already seen that
the gravity force scales with a3 (the radius of the particle to the power 3). The electric
force is linked to the surface charge of the particles. This force is thus proportional
to a2. The ratio of these forces is therefore proportional to the radius of the particles
a. Of course, there are more factors in this ratio besides the size but the general idea
is that for a large particle (large a) volume forces will always be large than surface

21 van der Beek, David, and Henk NW Lekkerkerker. "Liquid crystal phases of charged colloidal

platelets." Langmuir 20.20 (2004): 8582-8586.

22 Polymers will be defined in the next chapters. They are a special type of colloidal particles.

55
 Introduction to Colloid Science

forces. A criterion 23 for determining if a particle can or cannot settle is therefore its
size: the typical particle size for which the two forces balance is around 1 µm, i.e. the
size of colloidal particles.

In Case (C), we have displayed the case where, even though the particles settle under
their own weight, there is a small region above the bed where particles are not
touching the bed, because of the strong repulsion between the particles and the
gradient in concentration that tends to push them vertically up. This phenomenon
has been studied in 2003 in an article called “Defying gravity with entropy and
electrostatics: sedimentation of charged colloids” 24. A particular case of this general
theory is when only gravity and the gradient in concentration plays the dominant
role (the repulsion between particles just prevent them from gluing). Supposing that
the particles do not hinder each other, they all settle according to Stokes’ settling
velocity:

2 2 𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤
𝑣𝑣 = 𝑎𝑎 𝑔𝑔
9 𝜂𝜂

The corresponding flux of particle is given by 𝐽𝐽𝑔𝑔 = 𝐶𝐶𝑣𝑣 where 𝐶𝐶 is the concentration
of particles (in mol/L or number/m3 depending on one’s choice). By settling, a
concentration gradient will establish, and this will lead to a flux, according to Fick’s
first law (we take z as the coordinate along the vertical axis such that 𝒈𝒈 = −𝑔𝑔𝒆𝒆𝑧𝑧 ):

𝑘𝑘𝐵𝐵 𝑇𝑇 𝑑𝑑𝑑𝑑
𝐽𝐽𝐶𝐶 = −
6𝜋𝜋𝜋𝜋𝜋𝜋 𝑑𝑑𝑑𝑑

At equilibrium (steady state), the two fluxes must be equal: 𝐽𝐽𝐶𝐶 = 𝐽𝐽𝑔𝑔 from which we
obtain:

𝑑𝑑𝑑𝑑 4 𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤

= − 𝜋𝜋𝑎𝑎3 � � 𝑔𝑔𝑔𝑔𝑔𝑔
𝑛𝑛 3 𝑘𝑘𝐵𝐵 𝑇𝑇

Note that this equation is mathematically similar to the one obtained for a Rouse
profile, where the balance is between the downwards settling (due to gravity, like
here) and the upward turbulent diffusion, instead of the thermal diffusion. We have
already stressed above that the two diffusion coefficients are extremely different in
magnitude, implying that a Rouse profile can easily be observed in-situ in the case
of clays, and the one derived here generally not.

The equation can be solved using 𝑛𝑛(𝑧𝑧 = 0) = 𝑛𝑛0 and leads to:

23 It is of course not the only criterion, surface charge and the ability to undergo Brownian
motion are others.
24 van Roij, R. (2003). Journal of Physics: Condensed Matter, 15(48), S3569.

56
 Chapter 2 Settling, diffusion and stabilisation

4 𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤
𝑛𝑛(𝑧𝑧) = 𝑛𝑛0 exp �− 𝜋𝜋𝑎𝑎3 � � 𝑔𝑔𝑔𝑔�
3 𝑘𝑘𝐵𝐵 𝑇𝑇

A case where this equation leads to an easily observable profile is the case of gas.
Assuming an ideal gas, the relation between particle concentration and pressure is:

𝑃𝑃 = 𝑛𝑛𝑘𝑘𝐵𝐵 𝑇𝑇

where 𝑃𝑃 is the pressure and 𝑛𝑛 the number of gas particles per unit of volume.
4
Furthermore, defining the volume of one gas molecule as 𝑉𝑉𝑝𝑝 = 𝜋𝜋𝑎𝑎3 and realizing
3
that there is no Archimedes force in this case, one gets:

𝑉𝑉𝑝𝑝 𝜌𝜌𝑝𝑝
𝑃𝑃(𝑧𝑧) = 𝑃𝑃0 𝑒𝑒𝑒𝑒𝑒𝑒 �− 𝑔𝑔𝑔𝑔�
𝑘𝑘𝐵𝐵 𝑇𝑇

The mass of one gas particle 𝑚𝑚𝑝𝑝 = 𝑉𝑉𝑝𝑝 𝜌𝜌𝑝𝑝 can be substituted by using the molar mass
𝑀𝑀 of the gas, and one obtains:

𝑀𝑀𝑔𝑔
𝑃𝑃(𝑧𝑧) = 𝑃𝑃0 𝑒𝑒𝑒𝑒𝑒𝑒 �− 𝑧𝑧�
𝑅𝑅𝑅𝑅
where 𝑅𝑅 = 𝑁𝑁𝐴𝐴 𝑘𝑘𝐵𝐵 = 8.314 J ∙ K −1 ∙ mol−1 is called the gas constant. This expression
for the pressure with altitude is called the barometric formula. It is valid in the limit
of ideal gases, for temperatures that do not vary with height. For air ( 𝑀𝑀 =
0.02896 kg/mol) at T = 288 K (15 ℃), one gets that 𝑃𝑃(𝑧𝑧) ≈ 0 for 𝑧𝑧 = 8.4 km –
clearly an observable length.

Another case where this equation leads to an observable length is for particles with
a density close to the one of water for which (𝜌𝜌𝑝𝑝 −
𝜌𝜌𝑤𝑤 ) becomes very small. This case is illustrated on the
right-hand-side where latex colloids are dispersed in
ultra-pure water. The latex particles were deposited
gently with a pipette on the bottom of the tube, see
picture (t = 0) and left unstirred. The picture (t = 40h)
was taken 40 hours after the sample was prepared. A
fuzzy region of a few cm high can be observed above
the bed.

In Case (D), the particles are all settled on the bed. The
structure of the bed, and in particular its porosity, will
be extremely dependent on the way the particles have
settled down: did they settle rapidly and just stick to the bed? Did they have time to
rearrange their position relative to the particles already settled? Did the particles
aggregate (flocculate) before they reached the bed? The evolution of a bed formed
by settled particles will be discussed in Chapters 9 and 10.

57
 Introduction to Colloid Science

Osmotic pressure
When a suspension is brought into contact with a semi-permeable membrane,
another effect takes place to ensure that the suspension reaches a stable state. This
effect is linked to a transport of water (not of colloidal particles) through the
membrane, and is driven by what is called osmotic pressure.

We have derived above the barometric pressure for gases where we used the ideal
gas equation of state:

𝑃𝑃 = 𝑛𝑛𝑘𝑘𝐵𝐵 𝑇𝑇

where 𝑛𝑛 is the number of gas atoms per unit of volume. An analogue of this relation,
in the context of colloids, is the osmotic pressure of the suspension, defined for a
dilute suspension as:

Π = 𝑛𝑛𝑘𝑘𝐵𝐵 𝑇𝑇

where 𝑛𝑛 is the number of colloidal particles per unit of volume. This equation is
demonstrated in Chapter 8. Thus, a dilute suspension of colloidal particles behave
thermodynamically like a collection of “giant atoms”: they have the same equation
of state and show the same sedimentation equilibrium as a classical ideal atomic gas
(even though not on the same length scale as discussed above).

The osmotic pressure can be seen as the pressure that must be applied to a
suspension in contact with a bath of pure solvent across a semi-permeable
membrane (which the particles cannot cross) in order to stop the flow of solvent
from the bath to the suspension. It can therefore easily be measured:

At the initial state, a suspension is set in contact with a semipermeable membrane that only
allows the flow of water. In time, water will flow to the compartment with the suspension. At
equilibrium, the osmotic pressure can be determined from the height difference between the
water and the suspension or by applying a counter-pressure so as to have the same fluid
levels.

58
 Chapter 2 Settling, diffusion and stabilisation

The relation between height and osmotic pressure is given by:

Π = 𝜌𝜌𝜌𝜌ℎ

where 𝜌𝜌 is the density of the solvent (water) and the measured height difference at
equilibrium.

The concept of osmotic pressure plays an important role in the stability of

suspensions. This will be discussed further in the Chapters 3 and 4.

Applications

Biology

An important domain of application for osmotic pressure is biology and medicine. As

cells in biological bodies can be seen as fluids encapsulated in semi-permeable
membranes, the properties of the solutions in which they are found determine their
structure: when the composition of the cell’s fluid and the suspending solution are
such that water flows out of the cell, the cell will shrink, and it will swell when water
will flow into the cell. Three situations are defined:

Hypertonicity: there is a greater concentration of solutes outside the cell than inside:
water will get out the cell. Some organisms have evolved intricate methods of
circumventing hypertonicity. For example, saltwater is hypertonic to the fish that
live in it. They need a large surface area in their gills in contact with seawater for gas
exchange, thus they lose water osmotically to the sea from gill cells. They respond
to the loss by drinking large amounts of saltwater, and actively excreting the excess
salt. This process is called osmoregulation.

Isotonicity: there is the same concentration of solutes outside and inside the cell. In
this case the cell neither swells
nor shrinks. Water molecules
diffuse through the plasma
membrane in both directions,
and as the rate of water diffusion
is the same in each direction that
cell will neither gain nor lose
water. Isotonic sport drinks
contain similar concentrations of
salt and sugar as in the human
body. These drinks are marketed
as soft drinks (they contain
approximately 15 grams of sugar
per 250 ml). They are supposed
to help athletes replace “water,

59
 Introduction to Colloid Science

electrolytes, and energy” after training or competition but their efficacy is still not
proven.

Hypotonicity: there is a greater concentration of solutes inside the cell than outside:
water will get in the cell. In the case of plants, the cell walls are rigid enough to
contain the internal osmotic pressure and limit the cell expansion (for animal cells,
like blood cells, the cells will eventually burst). It is hypotonicity that enables
herbaceous plants to stand upright.

Water purification

Osmotic pressure is the basis of filtering ("reverse osmosis"), a process commonly

used in water purification. The “reverse osmosis” compares to the case where a
counter pressure is applied to equilibrate the liquid heights in the tube, as depicted
above. The water to be purified is placed in a chamber and put under an amount of
pressure greater than the osmotic pressure exerted by the water and the solutes
dissolved in it. Part of the chamber opens to a differentially permeable membrane
that lets water molecules through, but not the solute particles. The osmotic pressure
of ocean water is about 27 atm (1 atm = 105 Pa). Reverse osmosis desalinates fresh
water from ocean salt water.

Illustrations
Robert Brown (public domain)
https://en.wikipedia.org/wiki/Robert_Brown_(botanist,_born_1773)

Faraday gold
https://www.rigb.org/

Tonicity (creative commons license)

https://en.wikipedia.org/wiki/Tonicity
https://commons.wikimedia.org/wiki/File:Turgor_pressure_on_plant_cells_diagram.svg

60
 Chapter 3
DLVO forces
 Introduction to Colloid Science

Given certain conditions, colloidal particles can “glue 25” (aggregate) to each other,
hereby creating a bigger entity (an aggregate or floc) that may be able to settle
down. Its settling velocity is then different from the settling velocity of the individual
particles that created it. This, in turn, has important consequences in terms of
sediment transport as Stokes’ settling velocity is for example one of the parameters
used in large numerical codes that predicts large scale sediment transport.

(a): stable suspension with non-aggregated (primary) particles; (b): unstable suspension with
large flocs consisting of aggregated (flocculated) particles. In (a) the solvent properties are
different than in (b), but the primary particles are the same in (a) as in (b). The flocs in (b) are
made by aggregation of primary particles. The primary particles in (a) have a smaller settling
velocity than the flocs in (b) 26

In the present chapter, the conditions leading to the (non) aggregation of particles
are reviewed. These conditions originates from electrical interactions between
particles. Some of these interactions occur between molecules and atoms
constituting each particle. The sum of all the interactions between these atoms and
molecules leads to forces between particles. This is detailed below in the section
“Van der Waals forces”. The van der Waals forces are usually attractive 27. Other
forces (“Coulombic forces”) originate from the interaction between the surface
charges of the particles. When the particles have surface charges of same sign, the
forces are repulsive. It is the relative strength between the Coulombic and the van
der Waals forces that determines the stability of a suspension: if van der Waals
dominates, the particles will aggregate and the suspension is said unstable as its
composition changes over time. If the repulsive Coulombic forces dominate, the
particles will stay away from each other and the suspension is said to be stable as its
composition will not change over time. This is provided of course that gravity does

25 The word “colloid” comes from the greek word glue.

26 From Shih et al., Aggregation of Colloidal Particles with a Finite Interparticle Attraction
Energy; Journal of Statistical Physics, VoL 62, Nos. 5/6, 1991
27 The van der Waals force between two identical bodies in a medium is always attractive,

while that different bodies in a medium can be attractive or repulsive.

62
 Chapter 3 DLVO forces

not play a significant role, in which case the particles will settle and the suspension
is again said unstable as its composition is changing over time.

Van der Waals forces

Microscopic observations of colloidal particles have led
scientists in the 19th century to discover that colloidal
particles have the tendency to form persistent
aggregates through collisions induced by Brownian
motion. This led them to conclude that there should be
attractive forces between the particles. It is the Dutch
theoretical scientist and thermodynamist Johannes
Diderik van der Waals (1837-1923) who was the first to
quantify these forces that nowadays carry his name.

Van der Waals is famous for another discovery: the van der Waals equation of state
that describes the behaviour of gases and their condensation to the liquid phase.
This equation was published in his PhD thesis entitled Over de continuiteit van den
gas- en vloeistoftoestand (On the continuity of the gaseous and liquid state). The
original PhD thesis can still be found in the library of the University of Leiden where
he studied. He got the Nobel prize in Physics in 1910.

The van der Waals forces originate from weak electrical interactions between atoms
and molecules. These forces have 3 origins, and are all related to electric dipole
moments.

Electric dipole moments are created by the fact that two (or more) charges of
opposite sign are in the close vicinity of each other. Water molecules, for instance,
have dipole moments as they have a slight positive charge on H-sides and a slight
negative charge on the O-side (even though the molecule H2O is not charged):

The water molecule, H2O, is uncharged but possesses a dipolar moment 𝑃𝑃 due to the
asymmetry of the shared electrons’ distribution in the covalent OH bonds. This results in the
oxygen atom having a slight negative charge −2𝛿𝛿𝛿𝛿 (where 𝑒𝑒 is the electron charge) whereas
each hydrogen atom has a slight positive charge +𝛿𝛿𝛿𝛿.

63
 Introduction to Colloid Science

The dipole moment 𝑃𝑃 of two charges ( +𝑞𝑞 and –𝑞𝑞) separated by a distance 𝑑𝑑 is given
by

𝑷𝑷 = 𝑞𝑞𝒅𝒅

The bold notation indicates vectors, and 𝒅𝒅 is oriented from −𝑞𝑞 to +𝑞𝑞. In the case of
water, illustrated above, we first find the location of the barycentre of the positive
charges which leads to 𝑞𝑞 = 2𝛿𝛿𝛿𝛿 (illustrated above). The dipole moment 𝑃𝑃 can then
be evaluated. For water one finds it is equal to 1.85 D (1 D = 3.335 64 × 10−30 C m).

Dipole-dipole interactions can be either attractive or repulsive:

It is due to this polar nature of water that water molecules make hydrogen bonds,
which are defined in Chapter 4.

The three forces that form the van der Waals forces are due to:

1 – the positive (or negative) interaction between permanent dipoles of the

atoms/molecules. These forces are also called Keesom forces, in honour of the
Dutch physicist who derived them mathematically.

Schematic representation of dipole-dipole interactions in a solid and a liquid

2 – the positive attraction between a permanent dipole and an induced dipole. An

induced dipole is created when there is a temporary (extremely brief) deformation
of the electronic atmosphere around an atom: as the nucleus of an atom is positively
charged and the electrons are negatively charged, a temporary dipole is created.
This deformation is induced by the proximity of another dipole or an ion. These

64
 Chapter 3 DLVO forces

forces are called Debye forces, in honour of the Dutch physicist Peter Debye, who
made significant contributions to colloid science, and hence his name is also
associated with an equation and a length scale that will be introduced later.

3 – the positive attraction between two induced dipoles. These forces are also called
London forces, in honour of the German physicist who derived them
mathematically.

The van der Waals forces were primarily derived for modelling the interaction
between two molecules. However, it was soon discovered that it was possible to
extend them to larger objects. This was done in particular by the Dutch scientist
Hugo Christiaan Hamaker (1905-1993), who gave his name to the Hamaker
constant used to describe the van der Waals forces between macroscopic objects
(macroscopic in the sense that they are larger than molecules). We will only give the
examples of the van der Waals forces exerted in the cases of the two following
geometries:

For the two semi-infinite plates separated by a distance h, the van der Waals
potential per area of interface is given by:

Φ −𝐴𝐴
=
𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 12𝜋𝜋ℎ2
where 𝐴𝐴 is the Hamaker constant. For equal spheres of radius a at center-to-center
separation r it is given by:

−𝐴𝐴 2𝑎𝑎2 2𝑎𝑎2 𝑟𝑟 2 − 4𝑎𝑎2

Φ= � 2 + + 𝑙𝑙𝑙𝑙 � ��
6 𝑟𝑟 − 4𝑎𝑎2 𝑟𝑟 2 𝑟𝑟 2

The van der Waals force F𝑣𝑣𝑣𝑣𝑣𝑣 can be estimated using the relation:

𝑑𝑑Φ
F𝑣𝑣𝑣𝑣𝑣𝑣 (𝑟𝑟) = −
𝑑𝑑𝑑𝑑
The van der Waals interaction potentials can be calculated from the adding-up all
the inter-atomic dipole contributions and the Hamaker constants are then obtained

65
 Introduction to Colloid Science

by identification. These Hamaker constants are then compared with the ones found
experimentally, by performing flocculation experiments (see Chapter 5). When the
system is too complex, the Hamaker constant can only be determined
experimentally. The Hamaker constants are often noted 𝐴𝐴123 symbolizing the fact
that it is the Hamaker constant for media 1 and 3 interacting across medium 2. For
instance 𝐴𝐴121 could represent a clay particle (medium 1) interacting with another
clay particle (also medium 1) across water (medium 2). For typical clays, across water
one finds 28:

kaolinite 𝐴𝐴121 = 3.1 ∙ 10−20 J

illite 𝐴𝐴121 = 2.5 ∙ 10−20 J
montmorrillonite 𝐴𝐴121 = 2.2 ∙ 10−20 J

To give an idea, Hamaker constants are about 10-19 J for interactions across vacuum.
Typical values are in between 10-19-10-21 J. These values can decrease slightly with
increasing salt concentration.

Peter Joseph William Debije (1884 - 1966) studied in the

Aachen University of Technology under the supervision of
the theoretical physicist Arnold Sommerfeld, who later
claimed that his most important discovery was Peter
Debye. Debije’ s name is usually written Debye as the
digraph of the letter i and j (ij is considered a letter in
itself) is only known to the Dutch language.

Debye made major contributions to the field of physics

and physical chemistry (including colloid science). He
applied the concept of dipole moment to the charge
distribution in asymmetric molecules in 1912 and developed equations relating
dipole moments to temperature and dielectric constant. The units of dipole
moments are termed Debye (D) in his honour (1 D = 3.335 64 × 10−30 C m). His name
is also associated with the Debye frequency, the frequency above which an electric
double layer does not polarize anymore and the Debye length that we are going to
define in the next section. In 1923, together with his assistant Erich Hückel, he
developed an improvement of Svante Arrhenius' theory of electrical conductivity in
electrolyte solutions. This work resulted in the Debye-Hückel equation. In 1936 he
got the Nobel prize in chemistry “for his contributions to our knowledge of molecular
structure through his investigations on dipole moments and on the diffraction of X-
rays and electrons in gases”.

28 Novich, B. E., & Ring, T. A. (1984). Colloid stability of clays using photon correlation

spectroscopy. Clays Clay Miner., 32(5), 400. [Note that photon correlation spectroscopy = DLS,
see previous chapter]

66
 Chapter 3 DLVO forces

Coulombic forces
The first given reason for the stability of suspensions is associated to surface charge
of the particles. These charges lead to Coulombic forces that are repulsive when two
particles have charges of same sign (and are attractive otherwise).

In order to calculate these Coulomb forces, one needs first to know the electric
potential around one particle, without any interaction with another.

Electric potential distribution around a charged sphere in an electrolyte solution

When a spherical particle

of radius a, say negatively
charged, is in an
electrolyte solution, there
will be a so-called electric
double layer around the
particle. This layer is
composed of the (fixed)
surface charges (the first
layer) and its surrounding
cloud of ions (the second
layer). This cloud is mainly
composed of counter-ions,
i.e. ions with charge + (but
not only: there are also co-
ions, i.e. ions with charges -). The distribution of ions around the particle is given by
the Boltzmann distribution:

−𝑞𝑞𝑖𝑖 𝜓𝜓(𝑟𝑟)
n𝑖𝑖 (𝑟𝑟) = n𝑖𝑖 (∞)exp � �
𝑘𝑘𝐵𝐵 𝑇𝑇

where n𝑖𝑖 (𝑟𝑟) is the concentration of ions i(= +, −) in number / m3 as function of the
distance from the centre of the sphere. The other parameters are: 𝑞𝑞𝑖𝑖 the electric
charge of ion i and 𝜓𝜓(𝑟𝑟) is the electric potential around the sphere. The electric
potential 𝜓𝜓(𝑎𝑎) is the potential at the surface of the particle. Note the presence of
the thermal energy 𝑘𝑘𝐵𝐵 𝑇𝑇 in the expression: it is linked to the Brownian motion of the
ions in the cloud.

The expression for 𝜓𝜓(𝑟𝑟) is found using the Poisson equation, derived from
electrostatics, given here in spherical coordinates:

1 𝑑𝑑 𝑑𝑑𝑑𝑑(𝑟𝑟) −1
Δ𝜓𝜓 = �𝑟𝑟 2 �= (𝑞𝑞 n + 𝑞𝑞− n− )
𝑟𝑟 2 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 𝜀𝜀0 𝜀𝜀𝑟𝑟 + +

67
 Introduction to Colloid Science

By combining the Poisson and Boltzmann equations one obtains the non-linear
Poisson-Boltzmann equation for 𝜓𝜓(𝑟𝑟). Using some mathematics, it is then possible
to derive the electric potential around one particle. Many expressions for this
potential are available, depending on the approximations made for the
mathematical derivations. The exact solution of the Poisson-Boltzmann can also be
obtained numerically 29.

At low surface potentials: At low surface potential 𝜓𝜓(𝑎𝑎) it is possible to find a simple
analytical solution by using the fact that 𝑒𝑒𝑥𝑥𝑥𝑥(𝑥𝑥) = 1 + 𝑥𝑥 when 𝑥𝑥 is small. One then
obtains:

1 𝑑𝑑 𝑑𝑑𝑑𝑑(𝑟𝑟)
2
�𝑟𝑟 2 � = 𝜅𝜅 2 𝜓𝜓(𝑟𝑟)
𝑟𝑟 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑

which can be solved into, using the convention 𝜓𝜓(∞) = 0:

𝑎𝑎
𝜓𝜓(𝑟𝑟) = 𝜓𝜓(𝑎𝑎) exp�−𝜅𝜅(𝑟𝑟 − 𝑎𝑎)�
𝑟𝑟
where the parameter 𝜅𝜅 −1 is called the Debye length (or double layer thickness) and
can be evaluated from:

∑ n𝑖𝑖 (∞)𝑞𝑞𝑖𝑖 2
𝜅𝜅 2 =
𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇

where 𝜀𝜀0 𝜀𝜀𝑟𝑟 is the permittivity of the medium (𝜀𝜀0 = 8.85 × 10−12 F/m is the
permittivity of vacuum and 𝜀𝜀𝑟𝑟 = 80 the relative permittivity of water). The electric
potential around a charged particle decays therefore to zero over a typical length
equal to the Debye length. Beyond the double layer the ionic concentration is equal
to the bulk concentration n𝑖𝑖 (∞). The double layer thickness does not depend on the
size of the particle, it is only depending on the salt concentration. Therefore, for a
given salt concentration, 𝜅𝜅 −1 is the same for a 1 nm or a 1 mm particle.

At high ionic strength: When the ionic strength is large (a few tenth of mM to give
an order of idea), one usually satisfy the condition 𝜅𝜅𝜅𝜅 ≫ 1, independently of the size
of 𝑎𝑎. For a sphere with a thin double layer, i.e. for which 𝜅𝜅𝜅𝜅 ≫ 1, one can show that
the electric potential around that sphere, for any surface potential 𝜓𝜓(𝑎𝑎), is given
by:

29Chassagne, C., and D. Bedeaux. "The dielectric response of a colloidal spheroid." Journal of
colloid and interface science 326.1 (2008): 240-253. See also Chassagne, C. "Dielectric
response of a charged prolate spheroid in an electrolyte solution." International Journal of
Thermophysics 34.7 (2013): 1239-1254.

68
 Chapter 3 DLVO forces

4𝑘𝑘𝐵𝐵 𝑇𝑇 𝑞𝑞𝑞𝑞(𝑎𝑎)
𝜓𝜓(𝑟𝑟) = tanh � � exp�−𝜅𝜅(𝑟𝑟 − 𝑎𝑎)�
𝑞𝑞 4𝑘𝑘𝐵𝐵 𝑇𝑇

where the ion charge is usually given as 𝑞𝑞 = 𝑧𝑧𝑧𝑧 where 𝑧𝑧 is the valence of the ion,
and 𝑒𝑒 the elementary particle charge (𝑒𝑒 = 1.6 × 10−19 C).
We will here only consider a symmetric electrolytes for which 𝑧𝑧 = 𝑧𝑧+ = −𝑧𝑧− . For
monovalent salts (ex: for NaCl, KCl) one has 𝑧𝑧 = 1. For small 𝑥𝑥 one has tanh(𝑥𝑥) = 𝑥𝑥
which implies that for small 𝜓𝜓(𝑎𝑎) the above expression reduces to:

𝜓𝜓(𝑟𝑟) = 𝜓𝜓(𝑎𝑎)exp�−𝜅𝜅(𝑟𝑟 − 𝑎𝑎)�

We have previously found that for a spherical particle with small 𝜓𝜓(𝑎𝑎) (and any 𝜅𝜅𝜅𝜅)
the potential takes the form:
𝑎𝑎
𝜓𝜓(𝑟𝑟) = 𝜓𝜓(𝑎𝑎) exp�−𝜅𝜅(𝑟𝑟 − 𝑎𝑎)�
𝑟𝑟
For small 𝜅𝜅𝜅𝜅 one has:
𝑎𝑎 𝜅𝜅𝜅𝜅 𝜅𝜅𝜅𝜅
= ≈ ≈1
𝑟𝑟 𝜅𝜅𝜅𝜅 + 𝜅𝜅(𝑟𝑟 − 𝑎𝑎) 𝜅𝜅𝜅𝜅 + 1

since the characteristic distance over which the electric potential is non-zero is
(𝑟𝑟 − 𝑎𝑎) ≈ 𝜅𝜅 −1 . This implies that we recover the previous expression, valid for small
𝜓𝜓(𝑎𝑎) and large 𝜅𝜅𝜅𝜅 i.e.

𝜓𝜓(𝑟𝑟) = 𝜓𝜓(𝑎𝑎)exp�−𝜅𝜅(𝑟𝑟 − 𝑎𝑎)�

This expression is in fact also valid for two planar surfaces (for any 𝜓𝜓(𝑎𝑎) this time
and large 𝜅𝜅𝜅𝜅). For thin double layers (𝜅𝜅𝜅𝜅 ≫ 1) the curvature of the interfaces does
not influence the profile of 𝜓𝜓(𝑟𝑟):

For thin double layers (the thickness of which is

given by the double arrow), the electric
potential as function of the distance from the
(here grey) surface is the same for planar or
curved surfaces

69
 Introduction to Colloid Science

Particles approaching one another

Beyond the double layer, the ionic concentration is the same as if there would be no
colloidal particles in the solution. This implies that the ion cloud screens the
particle’s charge: from a distance, the colloidal particle surrounded by its double
layer appears uncharged (reflected by the fact that the electric potential is then zero
(𝜓𝜓(∞) = 0). Two colloidal particles can therefore approach each other without
“feeling” each other. They start to interact only when their double layers start to
overlap, i.e. when their 𝜓𝜓(𝑟𝑟) overlap.

If the two particles would be in vacuum and therefore have no double layers, they
would then simply undergo an electrostatic repulsion. It is now important to realize
that in the present situation, the overlap of the double layers results in an increased
ion concentration between the two particles. This implies that the local osmotic
pressure is higher between the two particles, and it is this pressure that pushes the
particles apart until their double layers are not overlapping anymore.

Slipping plane and Stern layer

When a particle is moving, it is possible to define a slip(ping) plane (also called

surface of shear). This is the surface behind which everything moves with the same
velocity as the particle. The electric potential at this plane is defined as the zeta
potential 𝜁𝜁 . In ideal situations 𝜁𝜁 = 𝜓𝜓(𝑎𝑎) but quite often some water and ions are
tightly bound to the particle and the slipping plane is of the order of 0.1 nm away
from the particle’s surface: a small distance, but due to exponential behaviour of the
electric potential one then finds that 𝜁𝜁 ≠ 𝜓𝜓(𝑎𝑎). The region between the slip plane
and the particle is called the Stern layer region.

70
 Chapter 3 DLVO forces

Schematic representation of the arrangement of ions close to a charged surface. Note that in
some Stern layer models two regions are distinguished: the inner one where ions are strongly
attached to the surface, with loss of part of their hydration shell and one where they are
tightly bound but still possess their entire hydration shell. Between the charged surface and
the shear plane the ions are not free to move. Beyond the shear plane it is assumed that
Poisson-Boltzmann is valid.

Coulombic repulsion between spheres

The expression for the Coulomb repulsion potential between a couple of spheres
with thin double layers, i.e. particles for which 𝜅𝜅𝜅𝜅 ≫ 1 is given by:

𝑘𝑘𝐵𝐵 𝑇𝑇 2 𝑞𝑞𝑞𝑞(𝑎𝑎)
Φ = 32π𝜀𝜀0 𝜀𝜀𝑟𝑟 � � 𝑎𝑎 tanh2 � � exp�−𝜅𝜅(𝑟𝑟 − 2𝑎𝑎)�
𝑞𝑞 4𝑘𝑘𝐵𝐵 𝑇𝑇

The Coulomb repulsion force F𝑟𝑟𝑟𝑟𝑟𝑟 can be estimated using the relation:

𝑑𝑑Φ
F𝑟𝑟𝑟𝑟𝑟𝑟 (𝑟𝑟) = −
𝑑𝑑𝑑𝑑

DLVO theory
This theory is named after the scientists Derjaguin, Landau, Verwey and Overbeek.
Both Derjaguin and Landau, and independently Verwey and Overbeek proposed to
explain the stability of colloidal suspensions by adding the van der Waals potential
and the Coulomb repulsion potential, so as to give the interaction potential:

Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 = Φ𝑣𝑣𝑣𝑣𝑣𝑣 + Φ𝑟𝑟𝑟𝑟𝑟𝑟

71
 Introduction to Colloid Science

One formulation for the DLVO potential between two spheres of equal size a, for
any 𝜓𝜓(𝑎𝑎) and κa ≫ 1 is obtained from the expressions given above:

−𝐴𝐴 2𝑎𝑎2 2𝑎𝑎2 𝑟𝑟 2 − 4𝑎𝑎2

Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 = � 2 2
+ 2 + 𝑙𝑙𝑙𝑙 � ��
6 𝑟𝑟 − 4𝑎𝑎 𝑟𝑟 𝑟𝑟 2
𝑘𝑘𝐵𝐵 𝑇𝑇 2 𝑞𝑞𝑞𝑞(𝑎𝑎)
+ 32π𝜀𝜀0 𝜀𝜀𝑟𝑟 � � 𝑎𝑎 tanh2 � � exp�−𝜅𝜅(𝑟𝑟 − 2𝑎𝑎)�
𝑞𝑞 4𝑘𝑘𝐵𝐵 𝑇𝑇

By evaluating Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 for a given suspension, it is possible to say if the suspension is

stable (i.e. the particles will not aggregate) or unstable (i.e. they will aggregate):
when the potential energy is below (approximately) 10 𝑘𝑘𝐵𝐵 𝑇𝑇 then aggregation is
possible, and the suspension is unstable.

Typical examples for Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 are given in the following examples:

50
repulsion
40
van der Waals
30 DLVO

20

10
Φ/kBT

-10

-20

-30

-40

-50
0 0.02 0.04 0.06 0.08 0.1
h/a=(r-2a)/a

Interaction energy between two spheres of radius 250 nm, using A = 5.10-21 J and 𝜓𝜓(𝑎𝑎) =
12.5 mV. The spheres are immersed in an electrolyte solution made of KCl at a salt
concentration of 5 mM.

Note that the distance at which the particles start to interact is very small: it is less
than 1/50 of the radius of the particles. The Debye length in this case is less than
1/200 of the radius.

72
 Chapter 3 DLVO forces

250
10 mM
200 50 mM
100 mM
150 200 mM Interaction energy between two
kaolinite particles of radius 370
100
nm, using A = 3.10-20 J, and
Φ/kBT

50 𝜓𝜓(𝑎𝑎) = 40 mV. The spheres are

immersed in an KCl electrolyte
0
solution. The concentrations are
-50 indicated in the figure.
-100

-150
0 0.02 0.04 0.06 0.08 0.1
h/a=(r-2a)/a

The values used for this example correspond to the ones given by Novich, B. E., &
Ring 30. These authors have indeed found the fastest aggregation for 200 mM of
added KCl.

Between 50 and 100 mM of added KCl (where Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 is of the order of some 10 𝑘𝑘𝐵𝐵 𝑇𝑇)
aggregation is still possible, but will take a longer time, as the particles have to
overcome the energy barrier (the peak in the picture). This was also found
experimentally by the authors.

By looking closely at the interaction potential curves, one can distinguish the
following features:

At long-range separation, the particles experience a weak attraction, due to the

long-range action of the van der Waals force: the van der Waals force acts on a
longer range than the repulsion 5
force, which becomes faster 4.5
zero. This is best seen when 4
plotting |Φ𝑣𝑣𝑣𝑣𝑣𝑣 |/�Φ𝑟𝑟𝑟𝑟𝑟𝑟 �, see 3.5
figure. It is clear that both for 3
|Φ vdw|/| Φ rep|

long-range (h/a > 0.06) and

2.5
small-range (h/a < 0.001) the
2
van der Waals potential is
1.5
larger in magnitude than the
1
repulsion potential (|Φ𝑣𝑣𝑣𝑣𝑣𝑣 | >
�Φ𝑟𝑟𝑟𝑟𝑟𝑟 �) . At mid-distance, the 0.5

repulsion potential dominates 0

0 0.02 0.04 0.06 0.08 0.1
h/a=(r-2a)/a
(|Φ𝑣𝑣𝑣𝑣𝑣𝑣 | < �Φ𝑟𝑟𝑒𝑒𝑒𝑒 �).

30 See earlier footnote

73
 Introduction to Colloid Science

At mid-range separation, a minimum is appearing in the curve, more or less visible,

depending on the value of |Φ𝑣𝑣𝑣𝑣𝑣𝑣 |/�Φ𝑟𝑟𝑟𝑟𝑟𝑟 �: if |Φ𝑣𝑣𝑣𝑣𝑣𝑣 |/�Φ𝑟𝑟𝑟𝑟𝑟𝑟 � ≪ 1 then the repulsion
is so strong that the minimum is barely visible. In that case, the particles will simply
be “pushed back” by the energy barrier. If, on the other hand, |Φ𝑣𝑣𝑣𝑣𝑣𝑣 |/�Φ𝑟𝑟𝑟𝑟𝑟𝑟 � ≤ 1 ,
then the minimum is visible (see the curves for 50 mM and 100 mM on the picture
of the previous page). If the minimum is sufficiently deep, the particles will be
trapped inside. For the 50 mM case, the minimum is located at about 3.7 nm (so
1/100 of the particle’s radius). The two particles will therefore, after having
approached each other, stay located 3.7 nm apart. On the scale at which
measurements are performed, this means that the detecting device will “see” the
two particles as aggregated. This type of aggregation is reversible. By lowering the
salt concentration (by diluting the sample with water for example), the energy
barrier will increase again (|Φ𝑣𝑣𝑣𝑣𝑣𝑣 |/�Φ𝑟𝑟𝑒𝑒𝑒𝑒 � ≪ 1), the two particles will repel each
other and no longer form an aggregate. By shearing or sonicating the sample, the
particles will also be able to escape the minimum, but when the sample is then left
to stay unstirred, the particles will slowly come back into the minimum, and
aggregation will start again.

This minimum is called secondary minimum by opposition to the primary minimum

that always exists at very small separations and that we are now going to discuss.

At small-range separation, a very deep minimum (the primary minimum) is always

present, indicating the attractive action of the van der Waals forces at very low
separations. When two particles are trapped in this minimum, they are nearly
irreversibly aggregated, as the energy to get them out of the minimum has to be
extremely large. The separation between the particles is then close to 0 (the particles
are touching). The DLVO model does not account for what exactly occurs when
particles are touching. Depending on the surface properties of the touching
particles, the depth of the primary minimum and the height of the energy barrier, it
is sometimes possible to de-aggregate (= disperse) the particles by sonication or
using a deflocculating agent (also called a dispersing or destabilizing agent). This
agent is a chemical product. For instance, in soil science, sodium
hexametaphosphate 31 (NaPO3)6 is used in combination with sodium carbonate 32
Na2CO3 to disperse soil particles.

Even if the energy barrier is high (of some order of 10 𝑘𝑘𝐵𝐵 𝑇𝑇 – but not too high i.e. not
of the order of 100 𝑘𝑘𝐵𝐵 𝑇𝑇), it is still possible for the particles to “ jump” the barrier
thanks to thermal fluctuation and get into the primary minimum. Particles in the
secondary minimum are therefore called “ kinetically” stable, implying that over

31 It is also used as an active ingredient in toothpastes as an anti-staining and tartar prevention

ingredient.
32also known as washing soda

74
 Chapter 3 DLVO forces

time (days, weeks, years) (some) particles inside the suspension could get over the
energy barrier and aggregate.

The scientists behind the DLVO theory

Boris Vladimirovich Derjaguin (or Deryagin; Russian: Бори́ с Влади́ мирович

Деря́гин) (1902– 1994) was a Soviet chemist. He helped laying the foundation of the
modern science of colloids and surfaces. He is at the origin of the Derjaguin
approximation widely used in order to approximate the interaction between curved
surfaces from the knowledge of interacting planar ones.

Lev Davidovitch Landau (Russian: Ландау, Лев Давидович) (1908 – 1968) was a
Soviet physicist who made fundamental contributions to many areas of theoretical
physics. He received the 1962 Nobel Prize in Physics. Together with Derjaguin, they
published what would become known as the DLVO theory in an article entitled
“Theory of the stability of strongly charged lyophobic sols and of the adhesion of
strongly charged particles in solutions of electrolytes.” by Derjaguin, B. V., & Landau,
L. (1941). Acta physicochim. URSS, 14(6), 633-662.

Evert Johannes Willem Verwey (Verweij) (1905 – 1981) was a Dutch chemist, who
also did research in physical chemistry. He obtained his PhD in 1934 under the
guidance of Hugo Rudolph Kruyt (who is one of the pioneer in colloid science). In
1934 he moved to the Philips Laboratories (NatLab) in Eindhoven. He continued
work on colloids, which was also the topic of his dissertation, and on oxides. The
Verwey transition in magnetite is named after him.

Jan Theodoor Gerard Overbeek (1911 - 2007) was a Dutch professor of physical
chemistry at the university of Utrecht. Like Verwey, he obtained his PhD under the
guidance of Kruyt, and his PhD (obtained in 1941) was entitled Theorie der
Electrophorese, het Relaxatie-effect. (Theory of electrophoresis, the relaxation
effect). He then went to work for Philips, where Evert Verwey was his direct boss.
The van der Waals forces had already been studied at Philips in 1936-37 by Hamaker
and De Boer. On the basis of thermodynamic concepts, they were able to calculate

75
 Introduction to Colloid Science

the free energies that would enable to derive the interaction potential. Their work,
which led independently to the DLVO theory, was published in some key articles in
1946, and soon they wrote a book: “Theory of the stability of Lyophobic colloids” by
Verwey and Overbeek, with the collaboration of K. van Nes (Elsevier, 1948).

It is clear that the DLVO theory has much to thank to the Philips Natuurkundig
Laboratorium (Philips Physics Laboratory) or NatLab in short which was the Philips
research department located in Eindhoven. Famous physicists like Hendrik Casimir
and Balthasar van der Pol from Utrecht University also worked for the NatLab on
experimental physics.

The Schulze-Hardy rule

The Schulze-Hardy rule is a classical well-known empirical relation that states that
the critical coagulation 33 concentration (c.c.c.) of a suspension varies with the
valence z of the counter-ions as:

1
c. c. c. ∼
𝑧𝑧 𝑛𝑛
where n is an exponent which is generally observed to be between 2 and 6. The
relation was found originally by Schulze in 1882 34. In article of 1925, Weiser 35 states:

“From an investigation of the coagulation by electrolytes of negative arsenious

sulfide and antimony trisulfide sols, Schulze concluded that the coagulating power of
electrolytes is greater the higher the valence of the ion having a charge opposite to
that on the colloidal particles. This conclusion has been supported by the later work
of Prost, Linder and Picton, Hardy, Freundlich, and others and has come to be known
as Schulze’s Law.”

One of the remarkable features of the DLVO theory is that it is in agreement with
the Schulze-Hardy rule. The c.c.c. corresponds to the situation at which the
maximum in the DLVO potential energy curve just touches the horizontal axis (which
is the case in the example shown above for kaolinite particles at 100 mM of salt).
The point on the curve where this happens is mathematically defined by:

𝑑𝑑Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷
Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 = =0
𝑑𝑑𝑑𝑑

33 when the aggregation of particles is due to ions, as discussed until here in the present

chapter, it is called coagulation.

34 H. Schulze, J. Prakt. Chem., 1882, 25, 431; 1883, 27, 320; 1885, 32, 390.
35 Weiser, H. B. (1925). Adsorption and Schulze's Law. The Journal of Physical Chemistry, 29(8),

955-965.

76
 Chapter 3 DLVO forces

The expression we have used so far for the DLVO potential is rather complicated to
differentiate. To illustrate the fact that the Schulze-Hardy rule can be found from the
DLVO theory, we will use another expression for Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 which is valid for spheres
with low 𝜓𝜓(𝑎𝑎) and at large separations:

−𝐴𝐴𝐴𝐴 2
Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 = + 2π𝜀𝜀0 𝜀𝜀𝑟𝑟 �𝜓𝜓(𝑎𝑎)� 𝑎𝑎 exp�−𝜅𝜅(𝑟𝑟 − 2𝑎𝑎)�
12(𝑟𝑟 − 2𝑎𝑎)
𝑑𝑑Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷
One then finds from = 0 that:
𝑑𝑑𝑑𝑑

2 𝐴𝐴𝐴𝐴
2π𝜀𝜀0 𝜀𝜀𝑟𝑟 �𝜓𝜓(𝑎𝑎)� 𝜅𝜅𝜅𝜅 exp�−𝜅𝜅(𝑟𝑟 − 2𝑎𝑎)� =
12(𝑟𝑟 − 2𝑎𝑎)2

By substituting this expression in the expression for Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 = 0 one finds that

−𝐴𝐴𝐴𝐴 𝐴𝐴𝑎𝑎
+ =0
12(𝑟𝑟 − 2𝑎𝑎) 12𝜅𝜅(𝑟𝑟 − 2𝑎𝑎)2

which is satisfied when 𝑟𝑟 = 2𝑎𝑎 + 𝜅𝜅 −1 . Using this value in the first equation of the
page and setting Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 = 0 one gets that:

2 2
24π𝜀𝜀0 𝜀𝜀𝑟𝑟 �𝜓𝜓(𝑎𝑎)�
2
𝜅𝜅 = � � exp(−2)
𝐴𝐴

where the inverse of the Debye length squared for a symmetric salt of valence z is
given by:

2n∞ (𝑒𝑒𝑒𝑒)2
𝜅𝜅 2 =
𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇

and where n∞ = n+ (∞) = n− (∞). The number of ions per volume 36 and the
concentration of salt C are related by:

n∞ (number/m3 ) = 𝑁𝑁𝐴𝐴 (mol−1 ) × 𝐶𝐶(mM)

One obtains:

2 2
𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇 24π𝜀𝜀0 𝜀𝜀𝑟𝑟 �𝜓𝜓(𝑎𝑎)�
𝐶𝐶(mM) = � � exp(−2)
2𝑁𝑁𝐴𝐴 (𝑒𝑒𝑒𝑒)2 𝐴𝐴

36 In physics the unit of volume is the cubic meter (m3 ) whereas it is litre (L) in chemistry. In

the formula for 𝜅𝜅 2 , n∞ should be in number/m3 .

77
 Introduction to Colloid Science

This relation clearly indicates that the salt concentration required to obtain the c.c.c.
is proportional to 1/𝑧𝑧 2 . From another expressions of the DLVO theory, based on
other approximations 37, it is possible to show that the salt concentration required to
obtain the c.c.c. is then proportional to 1/𝑧𝑧 6 or 1/𝑧𝑧 𝑛𝑛 with n values between 2 and
6.

Note that Schulze-Hardy rule in the simple case we have illustrated above originates
from the fact that 𝜅𝜅 2 ~𝑧𝑧 2 . This can be linked to an intuitive understanding:

For ions with a higher valence, the Debye length will be smaller for the same salt
concentrations. As the Debye length indicate the distance over which the electric
repulsive potential is going to zero, one can say that multivalent ions screen better
the particle’s surface charge than monovalent ions (for the same amount of salt):

250
monovalent
200 trivalent

150

100
Φ/kBT

50

-50

-100

-150
0 0.02 0.04 0.06 0.08 0.1
h/a=(r-2a)/a

Interaction energy between two kaolinite particles of radius 370 nm, using A = 3.10-20 J, and
𝜓𝜓(𝑎𝑎) = 40 mV. The spheres are immersed in an electrolyte solution made of 10 mM
monovalent salt (z = 1 → 𝜅𝜅 −1 = 3 𝑛𝑛𝑛𝑛) and 10 mM trivalent salt (z = 3 → 𝜅𝜅 −1 = 1 𝑛𝑛𝑛𝑛).

Link between surface potential and surface charge

We have until now expressed the fact that the particle is charged by using its electric
surface potential 𝜓𝜓(𝑎𝑎) in the equation. Usually, however, it is the surface charge of
a particle that is known (the surface charge can be determined by potentiometric
titration). The relation between 𝜓𝜓(𝑎𝑎) and the particle surface charge 𝜎𝜎𝑆𝑆 (𝐶𝐶/𝑚𝑚2 )
cannot be derived analytically, but numerical solutions are available. Empirical

37 E. J. W. Verwey and J. Th. G. Overbeek, The Theory of the Stability of Lyophobic Colloids
(Elsevier, Amsterdam, 1948) and I. M. Metcalfe and T. W. Healy, Charge-regulation Modelling
of the Schulze-Hardy Rule and related Coagulation Effects, Faraday Discuss. Chem. SOC.,
1990,90, 335-344

78
 Chapter 3 DLVO forces

relations, based on the numerical results, can also be found. A very useful one 38,
valid for all 𝜁𝜁 = 𝜓𝜓(𝑎𝑎) (implying the shear plane is at the surface of the particle) and
𝜅𝜅𝜅𝜅 ≥ 0.5 is:

𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑒𝑒𝑒𝑒𝑒𝑒 4 𝑒𝑒𝑒𝑒𝑒𝑒

𝜎𝜎𝑆𝑆 = 𝜅𝜅 �2 sinh � �+ tanh � ��
𝑒𝑒𝑒𝑒 2𝑘𝑘𝐵𝐵 𝑇𝑇 𝜅𝜅𝜅𝜅 4𝑘𝑘𝐵𝐵 𝑇𝑇

At low surface potential, one finds, using sinh(𝑥𝑥) = 𝑥𝑥 and tanh(𝑥𝑥) = 𝑥𝑥 for low 𝑥𝑥 :

1
𝜎𝜎𝑆𝑆 = 𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜅𝜅𝜅𝜅 �1 + �
𝜅𝜅𝜅𝜅
and we find a linear relationship between surface charge and surface electric
potential. (This linear relation obviously breaks down for higher potentials). The
linear equation can be related to the definition of a capacitance 39 C (as found in
electrostatics), namely:

𝑄𝑄 𝜎𝜎𝑆𝑆 𝑆𝑆
𝐶𝐶 = =
𝑈𝑈 𝑈𝑈
where 𝑄𝑄 (in Coulomb, symbol C) is the charge on one condensator plate of surface
S (it is – 𝑄𝑄 on the other plate), U is the electric potential difference between the two
condensator plates. In the present case, one can identify 𝐶𝐶/𝑆𝑆 = 𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜅𝜅[1 + 1/(𝜅𝜅𝜅𝜅)]
and 𝑈𝑈 = 𝜁𝜁 − 0 = 𝜁𝜁 (the difference between the surface electric potential and the
electric potential in the bulk (i.e. far from the particle).

As we have already discussed, often, there is a Stern layer in between the surface of
the particle and the slip plane. This layer has an extremely small thickness d.
Therefore, it is usually assumed that the relation between the surface electric
potential 𝜓𝜓(𝑎𝑎) and the zeta potential 𝜁𝜁 (i.e. the potential at the slip plane) is given
by a similar linear relationship:

𝜎𝜎𝑆𝑆 𝑆𝑆 𝜀𝜀0 𝜀𝜀𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆

𝐶𝐶𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 = =
𝜓𝜓(𝑎𝑎) − 𝜁𝜁 𝑑𝑑

Where the permittivity 𝜀𝜀𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 of the Stern layer is not necessarily the same as water:
it could be lower, because the water molecule dipoles in that region are strongly

38 Loeb, A. L., Overbeek, J. T. G., Wiersema, P. H., & King, C. V. (1961). The electrical double

layer around a spherical colloid particle. Journal of The Electrochemical Society, 108(12),
269C-269C.
39 The capacitance, symbolised by C has as unit F (Farad, named after Michael Faraday). Not

to be confused with C (Coulomb), the unit of charge, nor with C (in mol/L) the ionic
concentration.

79
 Introduction to Colloid Science

oriented and bound to the surface. Usually, 𝐶𝐶𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 is used as an adjustable parameter
in the models.

Measurement of the zeta potential

We will see in Chapter 6 how the zeta potential can be related to the optimal
flocculation rate. Measuring the zeta potential of a particle (as function of solvent
property, like the pH, salinity or added polyelectrolyte concentration) therefore
enables to predict the range (of pH, salinity, etc..) where flocculation is most likely
to occur: this happens when the zeta potential is close to zero. A very well
established method to assess the zeta potential of colloidal particles in suspension
is the use of the electrophoresis technique. Electrophoresis occurs when a colloidal
suspension is under influence of an electric field and each charged particle is moving
to the electrode which has a charge of opposite sign. Usually the electric field is a
fast alternating (AC) field, to prevent the electrodes from corroding. The particles
then follow the reversal of electrode charges by oscillating at the same frequency as
the electric field (for the low frequencies that are typically used in the
measurements). The velocity of the colloidal particles in the electric field is recorded.
This can be done with laser Doppler velocimetry where the phase shift in the
scattered laser light can be related to the particle’s velocity, or simply by video
microscopy. In video microscopy, a movie of the moving particles is analysed to get
their velocity. Laser Doppler velocimetry is suited for particles in the 1 nm - 1μm
range, whereas video microscopy can be used for larger particles, in the range 0.5μm
– 10 μm.

From the measure of the particle’s velocity, the zeta potential can be deduced. Charged
colloidal particles will move to the oppositely charged electrodes, whereas uncharged
particles have no electrophoretic velocity and hence their zeta potential is zero. During
electrophoresis, the electric double layer around a particle is deformed, and an electric dipole
is created.

A simple relation between particle velocity and zeta potential has been derived by
Smoluchowski. This relation is:

80
 Chapter 3 DLVO forces

𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜁𝜁
𝑣𝑣 = 𝐸𝐸
𝜂𝜂

where 𝑣𝑣 is the velocity of a colloidal particle, 𝜂𝜂 is the viscosity of the solvent (water)
and 𝐸𝐸 the applied electric field. Usually the electrophoretic mobility 𝜇𝜇𝐸𝐸 of the
particle is given instead of its velocity. The electrophoretic mobility 𝜇𝜇𝐸𝐸 is defined by:
𝑣𝑣
𝜇𝜇𝐸𝐸 =
𝐸𝐸
The Smoluchowski formula is valid when 𝜅𝜅𝜅𝜅 is very large – in fact so large that it
corresponds to salt concentrations so high that they are usually not measured in
practise. However, it is often used in articles to convert the electrophoretic mobility
(which has units of m2 V-1 s-1) into zeta potential units (V), which are more appealing:
an electrophoretic mobility of 1.7 × 10−8 m2 V-1 s-1 corresponds to a zeta potential
of 25 mV for example. One can therefore defined a measured “Smoluchowski” zeta
potential by:
𝜂𝜂
𝜁𝜁𝜇𝜇𝐸𝐸 = 𝜇𝜇
𝜀𝜀0 𝜀𝜀𝑟𝑟 𝐸𝐸

and plot 𝜁𝜁𝜇𝜇𝐸𝐸 instead of 𝜇𝜇𝐸𝐸 .

There exists nowadays advanced numerical and analytical models that enable to
convert the electrophoretic mobility into zeta potentials, for any 𝜅𝜅𝜅𝜅. A lot of work
on this topic has been performed by Hiroyuki Ohshima professor at the University of
Tokyo, who derived convenient, easy to implement, analytical formulas 40.

First, let us look at the simple case where we assume that the Smoluchowski relation
would be correct. We will assume the following surface potential/surface charge
relation:

𝑘𝑘𝐵𝐵 𝑇𝑇 𝑒𝑒𝜎𝜎𝑆𝑆
𝜓𝜓(𝑎𝑎) = 𝜁𝜁 = 2 arcsinh � �
𝑒𝑒 2𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇𝑇𝑇

This potential is in fact the zeta potential / surface charge relation for a flat surface
(one can see, by comparing it with the relation given above for spherical particles,
that many elements are similar). This relation is valid for spheres provided that 𝜅𝜅𝜅𝜅 is
large.

40 See the books: Ohshima, H. (2006). Theory of colloid and interfacial electric phenomena

(Vol. 12). Academic Press, and: Ohshima, H. (2011). Biophysical chemistry of biointerfaces.
John Wiley & Sons.

81
 Introduction to Colloid Science

In the figure below, we have plotted 𝜁𝜁 as a function of ionic strength (salinity) for a
constant surface charge given by 𝜎𝜎𝑆𝑆 = 3.9 × 10−2 C/m2.

Variation of the surface potential assuming constant charge density as a function of ionic
strength. We took 𝐷𝐷𝑘𝑘 = 2 × 10−9 𝑚𝑚2 /𝑠𝑠 which is representative for the diffusion coefficients
of K+ ions. The particle’s radius was taken to be 300 nm.

One can clearly see that the surface potential (equal to the zeta potential as the
shear plane is at the particle’s surface ) is varying as a function of salinity, whereas
the surface charge is not. This is due to the fact that, by adding ions, the double layer
gets compressed and that the potential close to the particle’s surface is changing
accordingly. The relation between the particle surface charge and the electric
potential is given by Gauss’s law:

𝑑𝑑𝑑𝑑
𝜎𝜎𝑆𝑆 = −𝜀𝜀0 𝜀𝜀𝑟𝑟 � �
𝑑𝑑𝑑𝑑 𝑟𝑟=𝑎𝑎

This is a fundamental relation, from which all the surface charge / potential relations
given above are derived. This equation tells us in particular that if 𝜎𝜎𝑆𝑆 is constant,
𝑑𝑑𝑑𝑑/𝑑𝑑𝑑𝑑 should be constant. For the sake of argument, we can make the assumption
that:

𝑑𝑑𝑑𝑑 𝜓𝜓(𝑎𝑎 + 𝜅𝜅 −1 ) − 𝜓𝜓(𝑎𝑎) −𝜓𝜓(𝑎𝑎)

� � ~ ~ −1
𝑑𝑑𝑑𝑑 𝑟𝑟=𝑎𝑎 (𝑎𝑎 + 𝜅𝜅 −1 ) − 𝑎𝑎 𝜅𝜅

as 𝜓𝜓(𝑎𝑎 + 𝜅𝜅 −1 ) ~ 0 . Therefore 𝜓𝜓(𝑎𝑎) should vary if 𝜅𝜅 −1 is changing.

If the Smoluchowski relation would be correct, this implies that we should have
𝜁𝜁𝜇𝜇𝐸𝐸 = 𝜁𝜁 and the measured electrophoretic mobility would be equal to:

82
 Chapter 3 DLVO forces

𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑒𝑒𝑒𝑒

𝜇𝜇𝐸𝐸 = 𝜁𝜁 = 2 arcsinh � �
𝜂𝜂 𝑒𝑒𝜂𝜂 2𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇𝑇𝑇

The “expected to be measured” 𝜁𝜁𝜇𝜇𝐸𝐸 should look like the dashed red curve given
above. Note that for moderate to low salinities the expected zeta potential gets
unrealistic values (in general |𝜁𝜁| < 150 mV).

A formula, derived by Ohshima, valid for 𝜅𝜅𝜅𝜅 > 10 and monovalent electrolytes (z =
1), relates the particle’s surface charge (assuming that the shear plane is at the
particle’s surface and no Stern layer exists) to the electrophoretic mobility or 𝜁𝜁𝜇𝜇𝐸𝐸 .
We give it here (see the book given in footnote to find the full derivations), as it will
enable us to understand an important feature of the mobility ( 𝜇𝜇𝐸𝐸 or 𝜁𝜁𝜇𝜇𝐸𝐸 ) versus
salinity curve.

The formula found by Ohshima is given by:

2𝐹𝐹 𝑘𝑘𝐵𝐵 𝑇𝑇 1 𝑒𝑒𝑒𝑒

𝜁𝜁𝜇𝜇𝐸𝐸 = 𝜁𝜁 − ln � �1 + exp � ���
1 + 𝐹𝐹 𝑒𝑒 2 2𝑘𝑘𝐵𝐵 𝑇𝑇

with

2 2𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇 2 𝑒𝑒𝑒𝑒

𝐹𝐹 = �1 + � � � �exp � � − 1�
𝜅𝜅𝜅𝜅 𝜂𝜂𝐷𝐷𝑘𝑘 𝑒𝑒 2𝑘𝑘𝐵𝐵 𝑇𝑇

where 𝐷𝐷𝑘𝑘 is the ionic diffusion coefficient of the counterion.

In the following figure, we have

plotted the measured 𝜁𝜁𝜇𝜇𝐸𝐸
according to the formula of
Ohshima, the formula of
Smoluchowski and the full
numerical solution, using the
same values as given above.

𝜁𝜁𝜇𝜇𝐸𝐸 according to the formula of

Ohshima, Smoluchowski and the full
numerical solution as function of
ionic strength C (mM) and 𝜅𝜅𝜅𝜅.
Particles have a radius of 300 nm and
a constant charge 41.

One can now see that the

behaviour of the curve is complex. This is due to the fact that the double layer, when

41 Chassagne, C., & Ibanez, M. (2012). Electrophoretic mobility of latex nanospheres in

electrolytes: Experimental challenges. Pure and Applied Chemistry, 85(1), 41-51.

83
 Introduction to Colloid Science

the electric field is applied, is deforming. This deformation is then changing the local
electric field “felt” by the colloidal particle and hence the colloidal particle adapts its
velocity. At high ionic strengths, the double layer is so compressed that it hardly
deforms and the conditions used by Smoluchowski to derive his equation applies.
This happens for salinities larger than 100 mM (larger than the salinity of sea water)
which are seldom encountered in experiments.

The full numerical solution behaviour has been confirmed by experiments 42:

Electrophoretic mobility of 156 nm

latex particles in NaCl solution: the
counterions of the particles are
nearly the only ions in solution. At
very low ionic strength, the data
follows the theoretical prediction
and the mobility increases
dramatically.

These experiments were done in

very “clean” water.

In the following figure, NUM

represents the full numerical solution and OHSHIMA is the curve according to the
formula given above. No
adjustable parameters
have been used. The
experiments were done
using spherical,
homogeneously charged,
latex spheres. The
symbols correspond to
measurements. The
values of the variables are
the ones given in the
previous theoretical
figures. It would now
appear that the formula of
Ohshima works better
than the full numerical
solution at low ionic strength: this is an artefact. In the experiments performed here,
below 0.1 mM of added salt, the ionic strength remained constant, due to the

42 M. R. Gittings, D. A. Saville , Electrophoretic Mobility and Dielectric Response

Measurements on Electrokinetically Ideal Polystyrene Latex Particles, Langmuir, 1995, 11 (3),

798-800

84
 Chapter 3 DLVO forces

presence of background ions. At very low concentrations of added salt, the added
salt concentration is then not representative of the true electrolyte concentration.
These background ions were released when the particles were added to the demi-
water 43. To remove these background ions, dialysis (see Chapter 2) can be
performed. This was not done in the present experiments but was performed in the
experiments performed by Gittings et al. given above. In most experiments on clays,
dialysis is not often done, as it could trigger some unwanted effects, such as the
delamination of the primary clay particles (see Chapter 5). This implies, in the
present case, that the measured 𝜁𝜁𝜇𝜇𝐸𝐸 remains constant below 0.1 mM of added salt.
The difference between white and black symbols are related to the experimental
protocol. From this study, it was concluded that the high > low protocol was the most
adapted, since it corresponds best to the expected theoretical values.

From the measured 𝜁𝜁𝜇𝜇𝐸𝐸 it is also possible to have an idea of the surface charge of
the particles. One relation between surface charge and electric potential some pages
above for 𝜅𝜅𝜅𝜅 ≥ 0.5. One can show that in most cases encountered in practice, this
formula simplifies into:

2𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑒𝑒𝑒𝑒𝑒𝑒

𝜎𝜎𝑆𝑆 = 𝜅𝜅sinh � �
𝑒𝑒𝑒𝑒 2𝑘𝑘𝐵𝐵 𝑇𝑇

which is the transposed of the relation given at the beginning of this section, where
it was given as zeta potential / surface charge relation for a flat surface. If we plot:

2𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑒𝑒𝑒𝑒𝜁𝜁𝜇𝜇𝐸𝐸

𝜎𝜎𝜇𝜇𝐸𝐸 = 𝜅𝜅sinh � �
𝑒𝑒𝑒𝑒 2𝑘𝑘𝐵𝐵 𝑇𝑇

We get:

where the black dotted line represents the surface charge of the particle. For small
C, 𝜎𝜎𝜇𝜇𝐸𝐸 goes to zero (not shown) as 𝜅𝜅 becomes very small, but for C > 10 mM one has

43 Demi-water is water with an extremely low conductivity.

85
 Introduction to Colloid Science

in good approximation 𝜎𝜎𝜇𝜇𝐸𝐸 ~ 𝜎𝜎𝑆𝑆 . This is due to the fact that in that range of
concentrations one has 𝜁𝜁𝜇𝜇𝐸𝐸 ~ 𝜁𝜁 as discussed above.

Ionic conductivity and diffusion coefficient

When an electric field is applied to a solution, ions are moving like colloidal particles
do via electrophoresis. The electrophoretic mobility of ions cannot be measured by
the techniques described above, as they are too small to be seen, even by laser
Doppler velocimetry (ions’ sizes are below 1 nm). An alternative way to assess the
electrophoretic mobility of ions is to measure conductivity. As ions are charged,
their displacement can be related to an electric current which in turn can be related
to the applied electric field E by:

𝐽𝐽 = 𝐾𝐾𝑒𝑒 𝐸𝐸

where 𝐾𝐾𝑒𝑒 is the conductivity (in S/m) of the solution. The symbol S (1 S = 1 Ω-1 )
stands for “Siemens” (in honour of Werner von Siemens, founder of the electrical
and telecommunications company Siemens). Note that in the previous relation the
electric current 𝐽𝐽 is expressed in A/m2. The electric intensity 𝐼𝐼 is in A (Ampère 44) in
the standard relation 𝑈𝑈 = 𝑅𝑅𝑅𝑅, where 𝑈𝑈 is the applied voltage difference and 𝑅𝑅 the
resistance.

In Chapter 2, we have seen that mobility of a particle of radius a (independently of

the applied force) is given by 𝜇𝜇 = 1/(6𝜋𝜋𝜋𝜋𝜋𝜋) if Stokes’ s friction force applies. This
expression gives the right order of magnitude for the electrophoretic mobility of a
colloidal particle, however it is imperfect, as the application of an electric field
induces corrections to the friction force on the colloidal particle. For ions, Stokes’ s
friction force is a good approximation. For example, for a typical ion radius of 0.1
nm, one finds that the ionic diffusion coefficient (see Chapter 2) defined by 𝐷𝐷𝑘𝑘 =
𝑘𝑘𝐵𝐵 𝑇𝑇/(6𝜋𝜋𝜋𝜋𝜋𝜋) gives 2.1 × 10−9 𝑚𝑚2 /𝑠𝑠 which is very close to the diffusion coefficient
of K+ and Cl- for example, which have radii of the order of 0.1 nm.

The conductivity of a solution is linked to the ionic diffusion coefficients of its ionic
constituents by:

𝑒𝑒 2 𝑧𝑧𝑘𝑘 2 n∞
𝐾𝐾𝑒𝑒 = � 𝐷𝐷𝑘𝑘
𝑘𝑘𝐵𝐵 𝑇𝑇
𝑘𝑘

where we recall that: n∞ (number/m3 )

= 𝑁𝑁𝐴𝐴 (mol−1 ) × 𝐶𝐶(mM). The ionic diffusion
coefficients can be found in handbooks, and are usually given as limiting ionic
conductivities Λ∞
𝑘𝑘 (S m mol ):
2 -1

In honour of André-Marie Ampère (1775-1836) a French physicist and mathematician who

44

was one of the founders of classical electromagnetism.

86
 Chapter 3 DLVO forces

𝑒𝑒 2
Λ∞
𝑘𝑘 = |𝑧𝑧𝑘𝑘 |𝐷𝐷𝑘𝑘 𝑁𝑁𝐴𝐴
𝑘𝑘𝐵𝐵 𝑇𝑇

A convenient relation between the Debye length and the conductivity is given by:

𝑧𝑧+ 𝐷𝐷+ − 𝑧𝑧− 𝐷𝐷−

𝐾𝐾𝑒𝑒 = 𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜅𝜅 2
𝑧𝑧+ − 𝑧𝑧−

Most electrophoretic measuring devices also measure the conductivity of the

suspension. Knowing the type of salt present in the suspension, 𝜅𝜅 2 can then be
estimated. Note that electrophoretic measurement are always done for small
concentrations of colloidal particles (to avoid particle-particle interactions), and that
the contribution of colloidal particles to conductivity is negligible compared to the
one of ions: in Chapter 1, we have seen that on average for 1 colloidal particle there
is 1010 ions in solution.

Illustrations
van der Waals
https://en.wikipedia.org/wiki/Johannes_Diderik_van_der_Waals

Derjaguin
https://link.springer.com/article/10.1023%2FA%3A1020686631909

Landau
https://en.wikipedia.org/wiki/Lev_Landau

Verwey
https://chg.kncv.nl/geschiedenis/biografieen/v/verweij,-e.j.w.

Overbeek
http://www.ecis-web.eu/overbeek.htm

87
 Chapter 4
Other colloids:
polymers, surfactants,
microorganisms …
 Introduction to Colloid Science

So far, we have only discussed electrostatic stabilization, induced by charges (the

charges on the particle’s surface) and destabilization triggered by the ions in
solution. There are however other mechanisms leading to (de)stabilization. These
are related to the presence of special molecules in the colloidal suspension. These
molecules are called polymers and surfactants. These molecules have also a
colloidal size and combine with clay particles to create new colloidal particles (flocs,
micelles…). Flocculation (the formation of flocs) will be the topic of Chapters 5 and
6. In the present chapter, we will review the properties of polymers and surfactants.
These molecules can have an industrial origin, but can also originate from plants and
animals. Polymers and other type of organic matter are produced or altered by
microorganisms which are living in the mud. Some of these microorganisms,
presented here, have also a colloidal size and can be part of flocs.

Polymers & polyelectrolytes

Polymers come in various size, shape and charges. They can be found in nature or
made in the laboratory. On some of the next pages, we give an overview of some
polymers. Polymers that are electrically charged are called polyelectrolytes.
Polymers are colloidal particles, as one of their characteristic size (usually their
apparent radius or their width) lays within the colloidal range.

Illustration of a polyelectrolyte in water. The polyelectrolyte is seen as a long flexible chain

made of repeating units i.e. the monomers (circles) connected by springs. The counterions (in
this case cations) are in vicinity of the negatively charged units

Polyelectrolytes have been studied extensively over the years 45 as their behaviour
can be quite complex. As the charged colloidal particles studied in Chapter 3, they
are subjected to electrostatic interactions, but the fact that they are flexible chains
adds a conformation behaviour. Depending on the affinity of the polyelectrolyte
with the solvent and the ionic concentration, the polyelectrolyte conformation can
be different.

45Dobrynin, Andrey V., and Michael Rubinstein. "Theory of polyelectrolytes in solutions and
at surfaces." Progress in Polymer Science 30.11 (2005): 1049-1118 and references within
(especially the books)

90
 Chapter 4 Other colloids: polymers, surfactants, microorganisms …

Very simple illustrations of polymers in solutions are given here:

random coil expanded conformation coiled conformation

the uncharged the charges on the polymer in the presence of salt,

polymer configuration are repelling each other the charges on the
is close to a self- through Coulomb forces polymer will be
avoiding (=two leading to an extension of screened (see Chapter
monomers cannot the chain in the water 3), and it is
occupy the same (at low salinity). The energetically
position at the same extension of the polymer favourable for the
time) random walk depends on its degree of polymer to coil.
(see Chapter 2) dissociation .

The conformation of polymers in suspension can be studied theoretically by

numerical simulations. As in the picture above, one then represents usually the
monomers by spheres that interact through Coulomb and Lennard-Jones potentials
(see Chapter 3 for Coulomb and Chapter 7 for Lennard-Jones potentials). The
numerical models are compared to different types of experimental data (light
scattering, fluorescence or Raman spectroscopy, Nuclear Magnetic Resonance
(NMR), etc...).

The behaviour of polymers and polyelectrolytes is also extensively studied from

thermodynamic concepts, similar to the ones introduced at the end of Chapter 8.

Branched and crosslinked polymers

We have represented polyelectrolytes as long chains, but there exist also branched
and crosslinked polyelectrolytes/polymers.

left: branched polymer; right: crosslinked polymer

91
 Introduction to Colloid Science

As can be seen underneath, natural polymers are often branched or crosslinked.

Some examples of polymers, polyelectrolytes and (de)flocculating agents

By definition a polymer is a macromolecule, which consists of many repeated

subunits (for example n repeated ethylene units –(CH2)n–). Natural polymeric
material include fibres like cellulose (found in wood, and used in paper making),
proteins, gums, DNA, etc… Synthetic polymeric material include plastics (latex –
natural latex also exists), paints and glue.

Polyacrylamide (-CH2CHCONH2-) : It forms a soft gel when hydrated, and is used in

applications such as gel electrophoresis. One of the largest
uses for polyacrylamide is to flocculate solids in a liquid.
Polyacrylamide can be supplied in copolymer 46 forms of
acrylamide combined with other chemical species to form an
acrylic acid or a polyelectrolyte. Common uses of
polyacrylamide and its derivatives are in Enhanced Oil Recovery, in water treatment
(for flocculation), and processes like paper making and screen printing.

Xanthan gum (-C35H49O29 -) : This natural polymer is a polysaccharide secreted by the

bacterium Xanthomonas campestris. It is composed of
repeat units of glucose, mannose (sugars). In foods,
xanthan gum is most often found in salad dressings and
sauces. In the oil industry, xanthan gum is used in large
quantities, usually to thicken drilling mud. In cosmetics,
xanthan gum is used to prepare water gels, usually in
conjunction with bentonite clays. It is also used in oil-in-water
emulsions 47 to help stabilize the oil droplets against coalescence.
Natural polymers secreted by microorganisms are in general called Extracellular
Polymeric Substances (EPS) – generally
referred to as slime. They are composed of
polysaccharides, and include other macro-
molecules such as DNA, lipids and humic
substances. In nature, they are important
for the formation of biofilms. Biofilms (see
picture) are communities of symbiotic
micro-organisms (bacteria, fungus, algae,
protozoa) which can stick to each other and
are embedded within a self-produced matrix of EPS. Biofilms are an important food
resource for invertebrates, and can be found at flood on mudflats. They can also help
to stabilize a soil against erosion.

46 A copolymer is a polymer having more than one type of monomer.

47 See “surfactant” , below.

92
 Chapter 4 Other colloids: polymers, surfactants, microorganisms …

Humic acids : They are a major organic constituent of soil, and produced by the
biodegradation of dead organic
material. Each humic acid is a complex
mixture of different acids containing
carboxyl and phenolate groups. They can
form complexes with ions commonly
found in the environment creating
humic colloids. Fulvic acids for instance
(a special type of humic acid) are
colloidal polyelectrolytes. Humic acids
are able to interact with each other and
create higher order complexes. They can also form complexes with metal ions, and
hereby regulate their bioavailability. Modern investigations have found that humic
acid is released from straw when mixed with mud and increases clay's plasticity. The
bricks made with mixtures of straw and mud are stronger (they are less likely to
break or lose their shape) than mud bricks 48. The adsorption behaviors of humic
acids and fulvic acids onto clay minerals in an on-going topic of research.

Biodegradation of polymers

The polymers used in industry are often synthetic and made from polyolefins. These
are for example polyethylene and polystyrene. These polymers are produced from
fossil fuels, and are considered to be undegradable. In recent years there has been
a gain in interest in biodegradable polymers 49. These polymers are defined as those
that undergo microbially induced chain scission leading to their mineralization. The
biodegradation is strongly dependent on environmental conditions, like pH,
humidity, oxygenation and the presence of some metals. Biodegradable polymers
are made from corn, wood cellulose or are synthetized by bacteria from small
molecules like butyric acid or valeric acid that give polyhydroxybutyrate and
polyhydroxyvalerate.

The degradation of polymers in soil is an on-going topic of research, and is strongly

linked to the biochemical and physical processes occurring within the soil. The
degradation of a polymer within a clayey fabric has for instance consequences for
the consolidation, permeability and strength of this soil.

48 Lucas, A.; Harris, J.R. (1998). Ancient Egyptian Materials and Industries. New York: Dover

Publications. p. 49. ISBN 0-486-40446-3

49 Ray, Suprakas Sinha, and Mosto Bousmina. "Biodegradable polymers and their layered

silicate nanocomposites: in greening the 21st century materials world." Progress in materials
science 50.8 (2005): 962-1079.

93
 Introduction to Colloid Science

Polymers and clays

Polymers can interact in different ways with clay colloidal particles. A polymer can
adsorb on the surface of a colloidal particle as a result of a Coulombic (charge-
charge) interaction, dipole-dipole interaction, hydrogen bonding, van der Waals
forces (see Chapter 3), or any combination of these mechanisms.

A hydrogen bond is the electrostatic attraction between polar groups that occurs
when a hydrogen (H) atom bound to a highly electronegative atom such as nitrogen
(N), oxygen (O) or fluorine (F) experiences attraction to some other nearby highly
electronegative atom (see Chapter 3). These hydrogen-bond attractions can occur
between molecules (intermolecular) or within different parts of a single molecule
(intramolecular). Depending on geometry and environmental conditions, the
hydrogen bond typically has between 5 and 30 kJ/mole in thermodynamic terms.
This makes it stronger than a van der Waals interaction, but weaker than charge-
charge (covalent) bonds. This type of bond can occur in inorganic molecules such as
water and in organic molecules like DNA and proteins.

Water molecules have both hydrogen bonds and covalent bonds

As the polymer is a very long snake-like molecule, the way it sticks to a particle
depends not only of its affinity with the particle’s surface, but also its affinity with
the solvent. The usual result is that the polymer sticks to certain points of the surface
(as trains), separated from one another by loops and for much of its length it is able
to trail out into the solvent (as tails).

Aggregation by the addition of polyelectrolytes is a method widely used in industry

in the treatment of mineral ores and in the purification of water.

94
 Chapter 4 Other colloids: polymers, surfactants, microorganisms …

Tailings ponds in Alberta (Canada): these are by-products of the oil sand industry. After
extracting the bitumen, the remaining fine particles (tailings) are flocculated by polymers,
deposited in large ponds and left to settle.

When two particles, coated with polymers, are getting close to each other, this can
lead to two different aggregation mechanisms 50:

Bridging aggregation 51: the loops and tails of one polymer of one particle will be
able to attach to the other particle. This process is facilitated when the amount of
polymers adsorbed on the particles is not too high. Usually, one finds that the
optimum polymer concentration to achieve flocculation corresponds to half surface
coverage for the polymer. Polymeric bridges are changing as function of shear (see
Chapter 5). Bridging aggregation can even occur with polyelectrolytes having surface
charges of same sign as the ones of the particles. In that case, aggregation is enabled
by the presence of oppositely charged ions in the water, that will act as “binders”
between particle and flocculant.

Left and middle: bridging flocculation. The

anionic polyelectrolyte on the left needs a
cation (in red) to bridge to the clay; right:
patching flocculation

Patching aggregation: patching aggregation occurs usually when polyelectrolytes

have a charge that is opposite in sign to the one of the particle. These polyelectrolyte
then strongly bind to the particles, and their tails do not extend much into the

50Bergaya, Faïza, and Gerhard Lagaly. Handbook of clay science. Vol. 5. Newnes, 2013.
51When particles are aggregated through polymers, one usually speaks of flocculation (as
oppose to coagulation, i.e. salt-induced aggregation).

95
 Introduction to Colloid Science

solvent. Aggregation is then made possible between one polymer patch of one
particle and the bare surface area of another particle.

Interlayer interaction

Until now we have represented clay particles are spheres, but as we have seen in
Chapter 1, clay particles are not spherical. Many types of clay particles have the form
of stacks of platelets, and the interaction between clay and polyelectrolyte can also
influence the stacking of platelets:

Clay interlayer with a cationic polyelectrolyte

Cationic polymers can strongly interact with clay minerals and penetrate between
the layers if they are small enough (in particular montmorillonite due to its swelling
properties, see Chapter 5).

Another mechanism is induced by the strong interaction between a polyelectrolyte

coated face of a clay particle with the bare face of another: one silicate layer of one
or each clay particle can then be peeled-of so that one (or two) new bare faces are
exposed and able to adsorb polymers:

Interaction between polyelectrolyte and clay platelets. Two silicate layers are peeled-of so
that one polymer can adsorb on one of the bare surfaces. Subsequently another stack of
platelets can adsorb on the polyelectrolyte, and again a silicate layer is peeled-of. This leads
to the intercalation of polyelectrolytes between the platelets of the clay.

Depletion effect
When small (or coiled) polymers are added to a suspension and do not stick to the
suspended particles they can lead to their destabilization through a depletion effect.
This effect occurs when two colloidal particles are close to each other, and that there
is a region (indicated by the black arrow below) where the polymers cannot
penetrate because they are too big to get inside the volume. The resulting gradient
in polymer concentration gives rise to a lower osmotic pressure in that region,
leading to aggregation.

96
 Chapter 4 Other colloids: polymers, surfactants, microorganisms …

Even though they were not aware of its cause, early makers of inks and paints used
to add natural gums or other polymers
to their pigments to promote the
binding between the pigments via
depletion effects 52.

Art work made of clay mineral suspensions,

probably using saliva as a binder. Saliva is
composed of 99.5% water (electrolyte)
with several natural colloidal and
polymeric agents. (Grotte de Lascaux,
17.000 BC)

Stabilization
In some cases, the presence of polymers in the suspension can enhance its
stabilization. This occurs in particular when:

1 – the particles are fully coated by polyelectrolytes of opposite charge. The particles
then get an effective charge of same sign as the polyelectrolyte they are coated with
and start repelling each other, much like standard particles of same charge do:

Charge repulsion between fully polyelectrolyte-coated particles

2 – the particles are coated with (uncharged) polymers. In that case, aggregation will
be prevented by steric repulsion. Steric effects originate from the osmotic pressure
in the region where the polymers overlap, due to the crowding of the polymer

52Lambourne, R., & Strivens, T. A. (Eds.). (1999). Paint and surface coatings: theory and
practice. Elsevier.

97
 Introduction to Colloid Science

chains, much like what we have seen for overlapping double layers (in that last case,
it was due to the crowding of ions):

Like for DLVO theory, theories have been developed to account for steric effects,
from which stability criteria are derived 53.

Surfactants (amphiphilic molecules)

Colloids that has not yet been reviewed and are of interest are surfactants (their
name come from surface active agents, see surface tension below) .

The surfactant called sodium stearate (sodium octadecanoate, CH3(CH2)16CO2Na) is found in

most soap. Soaps represent about 50% of commercial surfactants. In water, sodium stearate
dissociates to form an ionic (carboxylate) group and a counterion (Na+)

Surfactants can be seen as a special class of polymers. Their main property is that
they possess both a hydrophilic part (their “head” ) and a hydrophobic part (their
“tail”, which is an hydrocarbon chain).

53Stuart, MA Cohen, et al. "Adsorption of ions, polyelectrolytes and proteins." Advances in

colloid and interface science 34 (1991): 477-535.

98
 Chapter 4 Other colloids: polymers, surfactants, microorganisms …

Surfactant classification according to the composition of their head

Surfactants can also be found in natural systems, usually in the group of lipids,
molecules that include fats, waxes, glycerides, vitamins… Lipids can be either
hydrophobic or amphiphilic (amphiphile is another word for surfactant). An
abundant amphiphilic lipid is the phospholipid (note that this surfactant has two tails
(in yellow) – there are surfactants which have more):

Thanks to their dual hydrophilic/hydrophobic properties, surfactants are able to

form particular structures called micelles:

From left to right: adsorption of surfactant on an oil droplet

and formation of a micelle (right figure)

99
 Introduction to Colloid Science

We consider here a drop of oil adsorbed on a surface. Surfactant molecules are

added into the water. Energetically, it is favourable for the molecules to go to the
oil/water interface and position themselves such that their hydrophilic (from the
greek “water-loving”) head sits in the water, and their hydrophobic (from the greek
“water-hating”) tail in the oil. It is the need for the surfactants to sit at the interface
that leads to the rapid removal of the oil droplet from the surface, so that more
surfactant can adsorb on it. The entity such formed (the droplet coated with a layer
of surfactant molecules) is called a micelle. This micelle will remain in suspension
and not stick to the surface, as the hydrophilic heads prefer to remain in water and
have no affinity for the surface. This is how soap is used to clean oily dishes or
clothes. Note that shearing (using a brush or rubbing clothes) helps to detach the
micelles from the surface.

Micelles do not have to be spherical. Their shape is dependent on the shape of the
surfactant, its charge and its environment (pH, salinity, temperature…). For example,
another extensively used surfactant in cleaning and hygiene products is Sodium
Dodecyl Sulfate (SDS), which undergoes a transition from spherical micelle to rod-
shape micelles upon increase of ionic strength:

Structural (sphere-to-rod) transition in charged micelles (charges not shown) induced by high
ionic strength. The structural transition takes place in charged micelles at concentrations
well above the Critical Micelle Concentration (CMC, defined below). Note that the packing of
the surfactant molecules is different in the spherical ‘end caps’ (darker shade) and the
cylindrical central part (lighter shade). The spacing is reduced in the cylindrical part of the
rod-shaped micelle due to attenuation of charge interactions at high ionic strength 54.

Surfactants in nature

In nature, phospholipids are extremely important amphiphilic molecules that form,

in particular, the cell membrane. They can be part of the natural organic matter that
is found in sediments. These molecules can arrange themselves in micelles,
liposomes or bilayer sheets.

54 from: A. Chaudhuri et al. / Chemistry and Physics of Lipids 165 (2012) 497– 504

100
 Chapter 4 Other colloids: polymers, surfactants, microorganisms …

A vesicle is a large structure consisting of liquid

enclosed by a lipid bilayer. A liposome is a
spherical vesicle. It can be created by the
disruption of biological membranes (which form
bilayer sheets) by sonification. To maximize the
contact of its hydrophilic head and water, a
vesicle is then formed. Liposomes can be as big as
1000 nm (with a standard size between 20 and
1000 nm). Only some small molecules can pass
the bilayers of liposomes, and larger ones can be
trapped inside. Liposomes are for example used
to carry drugs which they subsequently deliver by
fusing with the cell membranes.

Surface tension
Let us consider an interface between a liquid (water) and a gas (air).The cohesive
forces between the water molecules are at the origin of the existing surface tension:
in the bulk, each molecule is surrounded by
water molecules but the molecules at the
water/air interface do not have the same
molecules on all sides. The molecules in the
layer in contact with air experience other
interactions than the molecules in the bulk,
which results in a higher surface energy
(J/m2). This energy can also be expressed in a
surface tension (N/m) by realising that 1 J = 1
N.m. A system will always try to minimize its surface energy, resulting in the fact
that it will try to minimize its interface with the other phase (here one phase is water,
the other air). This is why one finds many spherical, often colloidal, drops of liquid in
gas (or gas in liquids, like the bubbles depicted here) in nature: a sphere is the
volume that offers the smallest interface. This also explains why when two drops get
into contact, they tend to form a larger drop (this process is called coalescence):
again this is the smallest interface for a given volume:

101
 Introduction to Colloid Science

Measuring surface tension

Principle of the Wilhelmy plate: the force needed to pull the plate out of the liquid is
recorded; the perimeter of the plate is known : the surface tension can be estimated.

It is possible to measure a surface tension. One of the method is to use a Wilhelmy

plate, which is a thin plate of a few square centimetres in area. The plate should be
made in a material that is – obviously – able to be wetted by the liquid. The plate is
pulled from the liquid, and the force is then measured (usually of the order of mN).
The perimeter of the plate L is known: 𝐿𝐿 = 2𝑑𝑑 + 2𝑏𝑏 where 𝑑𝑑 is the length of the
plate and 𝑏𝑏 its width. The surface tension 𝛾𝛾 is given by:

𝐹𝐹
𝛾𝛾 =
𝐿𝐿cos(𝜃𝜃)

In general the contact angle 𝜃𝜃 is unknown 55, but assumed to be close to zero, as
there should be complete wetting. One then simply gets 𝛾𝛾 = 𝐹𝐹/𝐿𝐿 : the surface
tension can be seen as the force, perpendicular to the interface, needed to deform
this interface per unit of length.

Lowering the surface tension

When a surfactant is added in water in a clean

jar (which therefore contains nothing else
than water), the surfactant molecules tend to
go to the water/air interface, where their
hydrophobic tail can stick out of the water.
This lowers the surface tension of the water,
hence their name of surfactants. On the
picture, an example of the evolution of the
surface tension as function of concentration
surfactant is given. Below a critical

55 Nowadays, if needed, cameras can record the angle with good accuracy.

102
 Chapter 4 Other colloids: polymers, surfactants, microorganisms …

concentration called Critical Micelle Concentration (CMC), the surfactant populate

the interface. At the CMC, the interface is full, and micelles start to form: the surface
tension does not vary much anymore, even though still some surfactant molecules
will be able to pop in-between those who are already at the interface.

Both polymers and surfactants are important parameters to account for when clay
aggregation is studied, but they can be used in other engineering applications as
well. In petroleum engineering for example the oil production is increased when
water containing surfactant is flushed in reservoir pores. By lowering the interfacial
tension the oil mobility is increased thus allowing a better displacement of the oil by
injected water. The addition of polymers on the other hand increases the viscosity
of water injected into the oil reservoir enabling it to exert more pressure on the oil.

Biosurfactants

Besides phospholipids, there exist many more natural surfactants (“biosurfactants”)

produced by microorganisms from various substrates like sugars, oils and wastes 56.
These surfactants are synthesized as metabolic by-products , and their composition
depends on pH, nutrient composition, substrate and temperature. These
biosurfactants are usually either anionic or neutral. Only a few are cationic such as
those containing amine groups. Most biosurfactant-producing organisms are
aerobic, but a few anaerobic producers exist.

Biosurfactants are used for bioremediation of contaminated soils. For instance

rhamnolipid surfactants were found to efficiently remove hydrocarbons from a
sandy loam soil, and this was applied to the beaches in Alaska after the Exxon Valdez
tanker spill. Standard surfactants, like sodium dodecyl sulfate (SDS), were found to
be less effective than biosurfactants in removing hydrocarbons. Biosurfactants were
also shown to better able than SDS to enhance the solubilisation of polycyclic
aromatic carbons (PAH’s). PAH’s are uncharged, non-polar molecules found in coal
and tar deposits. They are also produced by thermal decomposition of organic
matter. The dominant source of PAH’s in the environment comes from human
activity by the combustion of biofuels and fossil fuels. Emissions from vehicles such
as cars and trucks are a source of PAHs in particulate air pollution. Industrial activity
such as aluminum, iron, and steel manufacturing, coal gasification and production of
coke, tar distillation, shale oil extraction, road paving and asphalt manufacturing,
rubber tire production can produce and distribute PAHs. Soil and river sediment near
industrial sites can be highly contaminated with PAHs. PAHs have a strong affinity
for organic carbon, and thus highly organic sediments in rivers, lakes, and the ocean
can be a substantial sink for PAHs.

56 Mulligan, C. N., R. N. Yong, and B. F. Gibbs. "Surfactant-enhanced remediation of

contaminated soil: a review." Engineering geology 60.1-4 (2001): 371-380.

103
 Introduction to Colloid Science

Plankton
Any water in a sea or lake that is neither close to the bottom nor near the shore can
be said to be in the pelagic zone. The pelagic zone can be thought of in terms of an
imaginary cylinder or water column that goes from the surface of the sea almost to
the bottom. The pelagic zone can be contrasted with the benthic and demersal
zones at the bottom of the sea. The benthic zone is the ecological region at the very
bottom of the sea. It includes the sediment surface and some subsurface layers.
Marine organisms living in this zone, such as clams and crabs, are called benthos.
The demersal zone is just above the benthic zone. It can be significantly affected by
the seabed and the life that lives there.

In the marine environment, polymers, polyelectrolytes, surfactants and humic acids

are originating from microorganisms and plants. We here review some of the
microorganisms present in the water body, which go under the generic name of
“plankton”. Some of these microorganisms have a colloidal size and can therefore
experience colloidal interactions: algae for instance are known to aggregate
between themselves.

Plankton (singular plankter) are a diverse group of organisms that live in the water
column of large bodies of water and that cannot swim against a current. They
provide a crucial source of food to many
large aquatic organisms, such as fish and
whales.

These organisms include drifting or

floating bacteria, fungi, archaea, algae,
protozoa and animals that inhabit, for
example, the pelagic zone of oceans,
seas, or bodies of fresh water. Though
many planktonic species are microscopic
(colloidal) in size, plankton includes
organisms covering a wide range of sizes, including large organisms such as jellyfish.

Biologists have a peculiar way to define the size of plankton, as can be seen in the
table underneath:

name size example

picoplankton < 2µm cyanobacteria
nanoplankton 2µm - 20µm diatoms
microplankton 20µm - 200µm dinoflagellates, diatoms
mesoplankton 200µm – 20mm small molluscs
macroplankton 20mm – 20 cm jellyfish, snails, krill
megaplankton > 20 cm jellyfish

104
 Chapter 4 Other colloids: polymers, surfactants, microorganisms …

Picoplankton and nanoplankton can be considered as colloidal particles, but it is

important to remember that not only colloidal plankton but also larger planktons
are able to secrete or decompose into smaller organic parts.

Phytoplankton

Phytoplankton are photosynthesizing microscopic organisms that inhabit the upper

sunlit layer of almost all oceans and bodies of fresh water. They use solar energy to
synthesize complex organic molecules from simpler inorganic compounds such as
carbon dioxide (CO2) and water (H2O). A simple photosynthetic reaction is:

CO2 + H2O + light → CH2O + O2

By this type of reaction carbon-based polymers (CH2O)n are formed, typically

molecules such as glucose or other sugars. These relatively simple molecules may be
then used to further synthesise more complicated molecules, including proteins,
complex carbohydrates, lipids, and nucleic acids (DNA, RNA). This happens when
carbon-based molecules are eaten by living organisms (by small microorganisms, like
bacteria, which in turn are eaten by larger organisms, or directly by larger organism,
like fishes) that cannot themselves synthetize carbon 57.

The most important groups of phytoplankton include the diatoms, cyanobacteria

and dinoflagellates.

Diatoms

Diatoms are algae with distinctive, transparent cell walls made of silicon dioxide
hydrated with a small amount of water (Si02 + H20). Diatoms are abundant in nearly
every habitat where water is found – oceans, lakes, streams, mosses, soils, even the
bark of trees. These algae form part of the base of aquatic food webs in marine and
freshwater habitats. Assemblages of diatom species are often specific to particular
habitats and can be used to characterize those habitats. Nearly all diatoms are
microscopic - cells range in size from about 2 microns to about 500 microns.

Planktonic diatoms in freshwater and marine environments typically exhibit a "boom

and bust" (or "bloom and bust") lifestyle. When conditions in the upper mixed layer
(nutrients and light) are favourable (as at the spring), their competitive edge and
rapid growth rate enables them to dominate phytoplankton communities ("boom"
or "bloom"). They can then also produce large quantities of polymeric slime (the
freshwater diatom Didymosphenia geminate, for example, produces large quantities
of a brown jelly-like material called "brown snot" or "rock snot").

57 this type of organisms include humans, who rely on eating fruit, vegetables and meat to get

their necessary carbon-based molecules.

105
 Introduction to Colloid Science

Several species of fresh-water diatoms

When conditions turn unfavourable, usually upon depletion of nutrients, diatom

cells typically increase in sinking rate and exit the upper mixed layer ("bust"). This
sinking can for example be induced by a loss of buoyancy control or the synthesis of
mucilage (= polymeric slime) that sticks diatoms cells together. Cells reaching deeper
water or the shallow seafloor can then rest until conditions become more favourable
again. In the open ocean, many sinking cells are lost to the deep, but refuge
populations can persist near the thermocline 58.

In the open ocean, the diatom (spring) bloom is typically ended by a shortage of
silicon. Unlike other minerals, the requirement for silicon is unique to diatoms and
it is not regenerated in the plankton ecosystem as efficiently as, for instance,
nitrogen or phosphorus nutrients. Because of this bloom-and-bust cycle, diatoms are
believed to play a disproportionately important role in the export of carbon from
oceanic surface waters. Significantly, they also play a key role in the regulation of the
biogeochemical cycle of silicon in the modern ocean.

58The thermocline divides the warmer upper layer from the cooler (and calm) deep water
below.

106
 Chapter 4 Other colloids: polymers, surfactants, microorganisms …

Cyanobacteria

Cyanobacteria are a group of photosynthetic, nitrogen fixing bacteria that live in a

wide variety of habitats such as moist soils and in water. They may be free-living or
form symbiotic relationships with plants or with lichen-forming fungi. They range
from unicellular to filamentous and include colonial species. Colonies may form
filaments, sheets, or even hollow balls.

a typical cyanobacteria cell

Aquatic cyanobacteria are known for their extensive and highly visible blooms that
can form in both freshwater and marine environments. The blooms can have the
appearance of blue-green paint or scum. These blooms can be toxic, and frequently
lead to the closure of recreational waters when spotted. Marine bacteriophages
(which are viruses) are significant parasites of unicellular marine cyanobacteria.

Dinoflagellates

The dinoflagellates are marine plankton but can be found in freshwater habitats as
well. Their populations are distributed depending on temperature, salinity or depth.
Many dinoflagellates are known to be photosynthetic, but a large fraction combine
photosynthesis with ingestion of prey. Dinoflagellates are unicellular and possess
two dissimilar flagella arising from the ventral cell side. The flagellar movement
produces forward propulsion and also a turning force.

A bloom of dinoflagellates can result in a visible coloration of the water colloquially

known as red tide, which can cause shellfish poisoning if humans consume
contaminated shellfish.

107
 Introduction to Colloid Science

Left: photographs of dinoflagellates; Right: algal bloom

Zooplankton

Zooplankton are organisms drifting in water (oceans, seas or fresh water) that are
usually microscopic, but some (such as jellyfish) are larger and visible with the naked
eye. Many zooplankton have locomotion, used to avoid predators or to increase prey
encounter rate. Zooplankton feed on bacterioplankton (the bacterial component of
the plankton), phytoplankton, other zooplankton or dead organic material (detritus).
Zooplankton are therefore primarily found in surface waters where food is
abundant.

Some zooplankton have a symbiotic relationship with

algae. Paramecium bursaria for instance has such a
relation with the green algae called Zoochorella. The
algae lives inside the cytoplasm of Paramecium and
provides it with food while Paramecium provides the
algae with movement and protection. Paramecium is
larger than a colloidal particle, being 80-150 µm long.
Other zooplankton can be much smaller.

108
 Chapter 4 Other colloids: polymers, surfactants, microorganisms …

The marine cycle

When considering colloids in natural environment, it is important to realize the
realm of important (variable) parameters that has to be considered in order to
correctly model their behaviour. For instance, predicting suspended sediment
behaviour requires not only to be able to know physical parameters such as shear
stresses, turbulent mixing conditions, erosion parameters, or chemical parameters
(salinity, pH, presence of polymeric substances), but also to understand – at some
extend - the role of biology and bio-chemistry: When and why does EPS become
available? How does organic decomposition change the properties of the flocs? Is
plankton able to bind to sediment particles? Can minerals, like silica, have a
biological origin?

The oceanic carbon cycle (or marine carbon cycle) is composed of processes that exchange
carbon between various pools within the ocean and between the atmosphere. There are four
distinct carbon pools (POC, DIC, PIC and DOC). Phytoplankton are responsible for most of the
transfer of carbon dioxide from the atmosphere to the ocean. Carbon dioxide is consumed
during photosynthesis, and the carbon is incorporated in the phytoplankton. Most of the
carbon is returned to near-surface waters when phytoplankton are eaten or decompose, but
some falls into the ocean depths.

A part of the sediment (silica) found in the marine environment (as part or not of
marine snow) comes from the decomposition of microorganisms:

109
 Introduction to Colloid Science

Cycling of silica in the marine environment: Silicon commonly occurs in nature as silicon
dioxide (SiO2), also called silica. It cycles through the marine environment, entering primarily
through riverine runoff. Silica is removed from the ocean by organisms such as diatoms and
radiolarians (a type of zooplankton) that use an amorphous form of silica in their cell walls.
After they die, their skeletons settle through the water column and the silica redissolves. A
small number reach the ocean floor, where they either remain, forming a silaceous ooze, or
dissolve and are returned to the water column. Opal is a hydrated amorphous form of silica
(SiO2nH2O)

Many suspended particles in the marine environment are aggregates of particles of

different origin. Marine snow is a type of aggregate made up of a variety of organic
matter, including dead and living phytoplankton, bacteria, fecal matter, sand, and
other inorganic dust.

A majority of marine snow composition is actually made up of aggregates of smaller

particles held together by a sugary mucus, transparent extracellular polysaccharides
(TEP). These are natural polymers exuded as waste products mostly by
phytoplankton and bacteria, but also by zooplankton. These aggregates grow over
time and may reach several centimetres in diameter, traveling for weeks before
reaching the ocean floor. Phytoplankton, microorganisms and bacteria live attached
to aggregate surfaces and are involved in rapid nutrient recycling.

The composition of the flocs is also depending on the seasonal variations, linked to
the natural lifecycles of the microorganisms involved. In general, most detritus
linked to mineral particles are found in estuaries or higher up in the river stream. In
winter, they form the most abundant type of aggregates in the whole system (from
the sea to up in the river).

110
 Chapter 4 Other colloids: polymers, surfactants, microorganisms …

marine snow (bottom row), faecal pellets (centre row) and “others” (top row). The “others”
category includes all recognizable planktonic organisms (alive and carcasses) and optically
dense debris that does not classify as marine snow or faecal pellets. For each image, the size
(μm) and depth sampled (m) are given 59.

Illustrations
Biofilm (public domain)
https://en.wikipedia.org/wiki/Biofilm#/media/File:Staphylococcus_aureus_biofilm_01.jpg

Lascaux (public domain)

https://fr.wikipedia.org/wiki/Grotte_de_Lascaux#/media/Fichier:Lascaux2.jpg

Phospholipides (public domain)

https://commons.wikimedia.org/wiki/File:Phospholipids_aqueous_solution_structures.svg

Micelles (public domain)

https://en.wikipedia.org/wiki/Micelle

Plankton (public domain)

https://en.wikipedia.org/wiki/Plankton

Diatoms (public domain)

https://en.wikipedia.org/wiki/Diatom
https://en.wikipedia.org/wiki/Diatom#/media/File:Diatomeas_w.jpg

Cyanobacteria (creative commons)

https://en.wikipedia.org/wiki/Cyanobacteria

59 Bochdansky, Alexander B., Melissa A. Clouse, and Gerhard J. Herndl. "Dragon kings of the

deep sea: marine particles deviate markedly from the common number-size spectrum."
Scientific reports 6 (2016): 22633.

111
 Introduction to Colloid Science

Dinoflagellates (creative commons)

https://en.wikipedia.org/wiki/Dinoflagellate

Paramecium_bursaria (creative commons)

https://en.wikipedia.org/wiki/Paramecium_bursaria

112
 Chapter 5
Floc formation and break-up
 Introduction to Colloid Science

Given certain conditions (see for instance DLVO theory, Chapter 3), colloidal
particles can “glue” (aggregate) to each other, hereby creating a bigger entity called
“floc” that is able to settle down. The settling velocity and size of flocs is recorded
in-situ or in the lab, from which their density is evaluated using Stokes’ settling
velocity (defined in Chapter 2). In the present chapter we are going to see that the
density of flocs is a complex function of their size, related to their structures and
composition. We discuss why these structures are found. We also discuss the effect
of shear on flocs as well as delamination and swelling, which are processes that
break the structures of mineral clay aggregates.

Cluster aggregation
Before forming large clusters, particles first have to form small aggregates (mainly
doublets at early stages), then they grow larger and larger:

Two important aggregation mechanisms are linked to the sticking probability of two
particles:

DLCA (Diffusion Limited Cluster Aggregation): the particles stick at first contact (ex:
can they immediately get in the “primary minimum” defined by the DLVO theory 60,
as there is a very low energy barrier)

RLCA (Reaction Limited Cluster Aggregation): the particles need to position

themselves in a comfortable way to stay attached (ex: there is a substantial energy
barrier to be overcome first)

The results of these two types of aggregation lead to different floc structure:

60
See Chapter 3

114
 Chapter 5 Floc formation and break-up

From the figures it is evident that the density of these two types of flocs is different:
RLCA flocs are denser (i.e. they contain less water) than DLCA flocs. We will see later
that this can be quantified in terms of a (pseudo) fractal dimension.

In the previous pictures, spherical particles are shown, as this geometry is the easiest
to investigate from a theoretical point of view. In nature, most of colloidal particles
are not spherical. Clay minerals in particular are highly anisotropic: they can have
the shape of platelets (used in the examples underneath), but some are also
cylindrical-shaped:

Depending on the clay mineral type, clay particles will disperse in different way in
water – and also flocculate in a different way.

Stable clay suspensions

Here we show two examples, one of a non-swelling and one of a swelling clay,
kaolinite and montmorillonite respectively, as they are the most encountered in
engineering, and natural environments. Kaolinite is used in ceramics (it is the main

115
 Introduction to Colloid Science

component of porcelain), in toothpaste, in paint, as adsorbents in water and

wastewater treatment. Montmorillonite is a component of drilling mud, used in the
oil-drilling industry, as it makes the mud slurry viscous. Its swelling property makes
montmorillonite-containing bentonite useful as a seal for water wells and as a
protective liner for landfills. Montmorillonite has also been used in cosmetics and
sodium montmorillonite is used as the base of some cat litter products, due to its
adsorbing and clumping properties.

Montmorillonite clays (part of the so-called swelling clays) can usually disperse
easily in water as their layers are only weakly bound together. They then form a
suspension of very small platelets (usually 1 nm thick and 100 nm long). In the
illustration, it is said that the layers are held together by van der Waals forces, but
this point is controversial. It is more generally assumed that electrostatic
attraction is the dominant force.

Kaolinite clays (part of the so-called non-swelling clays) on the other hand also
disperse in water, but as their layers are tightly bound together by hydrogen
bonds, they disperse in the form of stacks of platelets. These stacks are usually
not so anisotropic in shape, and can be quite large compared to montmorillonite
clays: something between 1 and 10 micrometers in diameter.

Until now, we illustrated the flocs by represented clay particles bound together. This
type of flocs can be produced in an electrolyte solution. Another type of flocculation
is possible which involves the presence of polyelectrolytes or polymer and
microorganisms, see Chapter 4. In an electrolyte, different effects can occur, called

116
 Chapter 5 Floc formation and break-up

delamination and swelling, which lead to the breakage of the aggregates. Aggregates
can also break and re-conform due to shear as discussed below.

Unstable clay suspensions: influence of salt

The flocculation by salt (coagulation) of kaolinite and montmorillonite has been
studied for decades, given their importance in practical applications. Flocculation of
mixtures of clays have also been studied 61. As we have seen in Chapter 3, the
amount of ions present in the water is of extreme importance: at moderate ionic
strength 62, the particles can undergo a “secondary minimum” aggregation, whereas
at high ionic strength the particles can be strongly aggregated, thanks to van der
Waals forces.

In the case of montmorillonite clays, there is a complication in predicting the role of

ions. As montmorillonite particles can swell, when ions are added to the suspension,
some will penetrate the montmorillonite interlayers. An important question is then:
how does these ions alter the structure of the montmorillonite particle? Will the
particle remain a (swollen) stack of platelets, will the stack be delaminated?

Delamination of a stack of montmorillonite

The answer is: usually the stack will delaminate for small amount of added ions, and
this is why most montmorillonite particles are usually found as single platelets in
water. The delamination process is related to the fact that by adding a small amount
of ions, these ions (and related water) will penetrate the interlayer space. The
platelets forming the stack will therefore be “pushed” away from the secondary
minimum where they were residing, and undergo Coulombic repulsion. Note that
this does not happen for non-swelling types of clays (like kaolinite): there, the
platelets are more strongly bound thanks to hydrogen bonds. By adding substantially
more ions in the system, a second important effect will take place: the screening of

61 For extensive details, we refer to Bergaya, Faïza, and Gerhard Lagaly. Handbook of clay
science. Vol. 5. Newnes, 2013.
62 ionic strength is the term used in physical chemistry. In civil engineering one usually speaks

of salinity or salt concentration.

117
 Introduction to Colloid Science

the surface charge of each platelet. This will lead to a new aggregation (in the
secondary or primary minimum, depending on ionic strength):

Band-type aggregation: small amount of calcium added to a suspension of delaminated

montmorillonite leads to face-face aggregation. Usually the aggregates are then found in
band-type.

An important parameter that should not be forgotten while studying this type of
flocculation is pH (i.e. the concentration of H+ or OH- ions present in the water).
Contrary to most ions, H+ and OH- ions can interact chemically with the clay’s surface
and hereby change its surface charge. An example has been given in Chapter 1 in the
case of kaolinite. When the electric charge of the faces of the clay particles is of
different sign than their edges a strong Coulombic attraction between the positive
edge of one particle and the negative face of another will occur. The pH, for most
clays, where a change in surface charge sign occurs, is between 4 and 6. Depending
on ionic strength and pH, different aggregate structures can therefore be obtained.

Some examples of clay aggregates

118
 Chapter 5 Floc formation and break-up

These TEM images have been obtained on samples for which the pH and salinity has
not been specified. Several structures can usually be identified, depending on pH
and ionic strength 63:

Delaminated particles with

moderate/large addition of salt,
at pH > 8: each particle is purely
negatively charged. This charge is
screened sufficiently to induced
flocculation. A preferred order
(face/face) is required, to maximize the
van der Waals forces between particles.

Delaminated particles with

low/moderate addition of salt,
at pH < 4: the positive edge with one
particle will bind to the negative face to
another. If the salt concentration would
be high, face/face aggregation would also
be possible, and the structure would be a
mixture between this one and the one
shown above (and hence be more
compact).

Non-delaminated particles with

low/moderate addition of salt,
at pH < 4: the positive edge with one
particle will bind to the negative face to
another. If the salt concentration would
be high, face/face aggregation would also
be possible, leading to a more compact
structure.

63 O’Brien, Neal R. "Fabric of kaolinite and illite floccules." Clays and Clay Minerals 19.6 (1971):

353-359.

119
 Introduction to Colloid Science

Swelling behaviour of clays

Swelling clays (like montmorillonite) do not always delaminate. One important
reason is related to the concentration of particles (or in other word: how much water
is added to the clay). If the water content is high, the montmorillonite becomes a
thixotropic 64 gel, and at even higher water content it becomes a suspension (also
called a sol), where the particles can delaminate. Here we are concerned with a
water content such that the clay is a very thick paste, obtained from putting the dry
clay in water. Na-montmorillonite (i.e. montmorillonite where the interlayer cations
are Na+), when placed in water, can take up
10 g of water per g of clay, and its volume can
increase by about 20 times. In Chapter 2, we
have already addressed the concept of
osmotic pressure. This plays an important
role is the swelling of clays.

In montmorillonite and vermiculite the

aluminosilicate sheets are separated by
water layers whose thickness varies with the
concentration and type of electrolyte. In the
crystalline state, these extremely thin,
negatively charged sheets are held together
by electrostatic forces between alternate
layers of bridging cations (like Na+, K+, Ca2+).
When the clay is placed in water or a dilute
electrolyte suspension, some of the
adsorbed counterions tend to take up water
and may dissociate from the clay surface,
creating a diffuse double layer. It is the repulsive interaction between these double
layers that is at the origin of the swelling behaviour. On the schematic
representation on the left, innercrystalline swelling of Na-montmorillonite is
displayed. The layer distances and the maximum number of water molecules per
sodium ion are given. The first water to enter the interlayer positions is the result of
hydration of the ions, after which the water forms distinct layers which increase in
number to four. The water molecules, of at least the first layers, are probably
arranged in a hexagonal network whose order is determined by hydrogen bonding
to the surface oxygens of the clay 65.

64See Chapter 7
65Norrish, K. (1954). The swelling of montmorillonite. Discussions of the Faraday society, 18,
120-134; Madsen, Fritz T., and Max Müller-Vonmoos. "The swelling behaviour of clays."
Applied Clay Science 4.2 (1989): 143-156.

120
 Chapter 5 Floc formation and break-up

The swelling discussed so far is commonly referred to as crystalline swelling. In this

range, the adsorbed water increases to about 0.5 g water per g of clay while the
interlayer spacing increases from 0.96 nm, for dry material, to about 2 nm
corresponding to 4 layers of water. This water content is the one that is adsorbed at
99% relative humidity and at this stage the montmorillonite has shown little physical
swelling.

When this Na-montmorillonite is placed in contact with water, and increased about
20 times its original volume, one speaks of regular (osmotic) swelling of the clay (see
last picture of the sketch above). Unlike innercrystalline swelling, which acts over
small distances (up to 2 nm), osmotic swelling is based on the repulsion between
overlapping electric double layers, and can act over much larger distances. (In Na-
montmorillonite it can result in the complete separation of the layers, leading to
delamination, if the water content allows it). The expansion is associated with the
formation of these diffuse double layers, as more water penetrates the interlayer
space: the electrostatic attractive force that existed between the cations and the
negative surfaces of the interlayers is now changed into a (osmotic) repulsive one.

Before osmotic swelling, cations and their hydration shells (white balls) are electrostatically
bound to the clay platelets. After water is added to the clay, double layers start to form with
both anions and cations (white and black balls), leading rapidly to an electrostatic repulsion
between the clay platelets, since the ion concentration C1 between the layers is much higher
than the ion concentration C2 in the bulk water. An equilibration of the concentration can
only be reached through the penetration of water into the space between clay layers
(osmotic swelling). If the water content allows, this will eventually lead to delamination.

The repulsive force between the overlapping double layers of the clay particles also
exists inside some geological formations. This force is in equilibrium with the
overburden pressure coming from the mass above the double layers. If the load on
them is removed and if water is available, a new equilibrium is sought: the water
intrudes between the clay layers and pushes them apart. The swelling continues
until the new balance between this inner force and outside, resisting forces is

121
 Introduction to Colloid Science

reached. This phenomenon results in landslides and is further discussed in Chapter

7.

Norrish (1954) demonstrated that, when small crystals of montmorillonite were

arranged in a parallel orientation and allowed to equilibrate with a dilute electrolyte
suspension (10 mM < C < 250 mM), then the separation between the sheets (𝑑𝑑) was
given by:

1.14
𝑑𝑑 (𝑛𝑛𝑛𝑛) = 2.1 +
�𝐶𝐶(𝑀𝑀)

[for 10 mM the separation is 13.5 nm and it is 4.38 nm for 250 mM]. It is logical that
the separation decreases with increasing ionic strength, as the double layers are
then more compressed and the concentration in the interlayer space, C1, is then not
too different from the concentration in the bulk, C2.

Unstable clay suspensions: flocculation by polyelectrolytes

In the previous chapter, we have defined polyelectrolytes. Polyelectrolytes
(flocculants) can attach themselves to clay particles and form large, elastic flocs. The
structure of sediment-polyelectrolyte flocs is very complex and is still an on-going
topic of research. In applied research, the open questions are related to the
efficiency of polyelectrolytes as binders of colloidal particles, and to the strength of
the obtained flocs. In particular, the floc size and density evolution as function of
shear rate are important parameters to investigate. This is discussed in the next
section.

Shear rate influence on floc size

The way the clay particles flocculate is also strongly dependent on the shear rate: at
low shears, flocculation is usually promoted, whereas at high shear, flocs can be
broken (or prevented to grow further). In Chapter 6, we are going to show how to
model the growth and break-up of flocs with shear, and we will concentrate on the
case where only aggregation is occurring. Here we simply give and discuss some
measurements results, and we will pinpoint the limitations of the current models
and measurement techniques.

The influence of shear rate on flocculation can best be studied using static light
scattering (see Chapter 2). The set-up is schematized here:

122
 Chapter 5 Floc formation and break-up

The set-up consists of a 1L jar connected to two tubes through which the suspension is
pumped in and out the static light scattering device. To keep the suspension suspended it is
gently stirred by a rotating impeller. The particle size distribution (PSD) is measured.

From the commercial instrument software, the raw data is converted into equivalent
volumes, using the assumption that the particles are spherical. For each size (there
are usually 100 given sizes, ranging logarithmically between 0.1 and 1000 µm in most
commercial devices), a volume is evaluated that is equal to the volumes of all
particles of that size added up. Each Particle Size Distribution (PSD in short) curves is
normalized such that the integral over each curve gives 100%. This means that the
volume corresponding to one size is a percentage volume. In other words, PSD
measurements give the relative ratio between volumes, and changing the
concentration of particles (adding more of the same) will not change the PSD. The
PSD will however change if flocculation or break-up occur in the suspension as
function of time, as each measured PSD will then not have the same volume ratios.

Underneath we give an example obtained from measuring the change in PSD over
time of a suspension of clay in presence of a cationic polyelectrolyte.

At t = 30 s, 0.7 g/L of clay is mixed with cationic flocculant. The ratio flocculant to clay is 0.5
mg/g. The full PSD is recorded every 30 s by static light scattering. The change in PSD over
time is given in the left figure. The D10, D50 and D90 of the distributions are given in the
right figure.

In the example given above we start at t = 0 with a clay suspension, which average
mean size D50 is 6 µm. This mean size D50 is defined as the size for which 50% of

123
 Introduction to Colloid Science

the particles are smaller (here in volume) than this D50 size. Similarly one defines
the sizes D10 and D90 as the sizes for which respectively 10% and 90% of particles
are smaller than these sizes.

At t = 30 s, cationic flocculant is added to the clay suspension. The PSD is evolving in

time: as can be seen on the left figure, a new peak in size is appearing, corresponding
to the creation of aggregates, and this peak is shifting to higher sizes over time. This
is reflected in the changes in D50 over time (right figure). After 2000 s, an equilibrium
size seems to have been reached. However, one can observed that after 2000 s there
is a slight decrease in D50 over time. At the same time there is a larger decrease in
D90 and increase in D10. This implies that the largest particles are eroding over time,
creating small particles. This erosion of particles over time is usually not accounted
for in traditional flocculation models (see Chapter 6).

It is important to note that there are large differences between flocs measured in-
situ and the ones created in the jar. First of all, the shear rates generated in the jar
and especially in the tubes leading to the measurement device are much higher
generally than the ones encountered in-situ. As the tubes are about a few mm in
diameter, this limits the growth of flocs and will also be a cause for erosion.

Clay in presence of salt

In the example given below 66, a large amount of NaCl salt has been added to a
suspension of polystyrene latex particles (10 µm in diameter and with a density of
1055 g/L) and a suspension of silica particles (10 µm in diameter and with a density
of 2000 g/L) to ensure that each suspension is unstable and the particles are
aggregating. For each suspension and each shear, the mean equilibrium size of the
suspension, D50, is plotted. This mean size is obtained after the PSD does not evolve
any more in time.

The shear rate is varied by changing the rotational speed of the impeller. For this
experiment it was verified that the flocs were not significantly broken/eroded in the
tubes. The pumping speed was chosen as low as possible to prevent flocs settling in
the tubes. The Kolmogorov microscale (defined in Chapter 6), reflecting the size at
which the turbulence generated in the jar could disrupt a floc, is also plotted in the
figure.

66 Mietta, F., C. Chassagne, and J. C. Winterwerp. "Shear-induced flocculation of a suspension

of kaolinite as function of pH and salt concentration." Journal of colloid and interface science
336.1 (2009): 134-141.

124
 Chapter 5 Floc formation and break-up

Mean floc size as function of shear rate, for two types of particles. Note that in in-situ typical
shear rates are less than 10 s-1. Larger shear rates are however encountered in industry.

We see that this size is an important parameter, as the polystyrene particles follow
this microscale, indicating that they cannot grow much larger than this size. Below
the Kolmogorov microscale, the floc is simply transported (advected) with the flow
whereas above the Kolmogorov microscale shear forces are tearing the floc apart.
Shear forces are discussed in Chapter 7.

The silica particles behave differently. At low shear rate the silica particles seem to
grow smaller than at moderate shear rates. This is however only an artefact. At low
shears, large (heavy) silica particles are settling in the measuring jar, and are
therefore not recorded by the static light scattering device used to measure their
size: only the smallest particles remain in suspension. Polystyrene particles, having
a density close to the one of water, are always fully suspended independently of
their size.

Each suspension is gently stirred in the jar, and the sample is pumped through the static light
scattering device, where the D50 is recorded; at low shear the largest flocs made of silica
particles are setting in the jar and are therefore not recorded.

125
 Introduction to Colloid Science

Clay in presence of polyelectrolyte

Contrary to flocs created by salt, flocs created by polyelectrolytes can follow or not
the Kolmogorov microscale. In the example underneath, the suspensions were made
in demi-water (water containing a very low amount of ions). It was verified (not
shown) that flocs created by anionic flocculant are always much smaller than flocs
created by cationic flocculant, for all flocculant to clay ratios.

Equilibrium D50 for a suspension of clay mixed with (a) 10 mg/g Zetag 7587 (cationic
flocculant) to clay ratio and (b) 0.5 mg/g Zetag 4110 (anionic flocculant) to clay ratio. The
arrows indicate that each suspension was created at low shear, the shear was then increased
and lowered again.

This is due to the fact that anionic flocculant needs cations to bridge with the clay
(see Chapter 4) and there are not sufficient cations in the water to have an optimal
flocculation. Flocs produced by anionic flocculant follow the Kolmogorov microscale
quite well, for all shears, and the flocs that have been broken at high shear regrow
to the size they had before the high shearing for all shear rates.

126
 Chapter 5 Floc formation and break-up

The interaction between the polymer and the clay is dependent on the bridging
cation. Under the action of shear, when the Kolmogorov length is reached by the
floc, the cation will not be able to bridge them anymore, and clay and polyelectrolyte
will move independently. At sufficient low shear, the floc will be reformed.

Flocs produced by cationic flocculant, on the other hand are created thanks to
attractive Coulombic interactions between the negatively charged clay and the
positively charged polymer. Due to their low density, the large flocs (the cationic
flocs grow nearly as large as 1 mm) remain in suspension even at low shears. When
the shear is increasing the D50 is decreasing, but the D50 remain well above the
Kolmogorov microscale. When the shear is lowered again, it is observed that the D50
barely increases in size. The kinetics of floc formation (at 150 s-1) when the
suspension is made at this low shear is very different from the small change in size
when the shear is decreased from 300 s-1 to 150 s-1 after that the floc has been
disrupted at higher shear:

Change of D50 size over time at a shear rate of 150 s-1. (D50 inc): formation of large flocs
when clay and polyelectrolyte are mixed. The time to reach an equilibrium D50 is about 150-
200 s. (D50 dec): flocs formed when the shear is decreased from the previous shear step at
300 s-1. Only a small change in size is observed, and the change in D50 occurs is less than
100 s.

One see that flocs created at low shear will never regrow fully to their original large
size after having experienced a higher shear. There can be two reasons for this: (a)
the flocs broken at high shear are weakly positively charged and will experience a
mutual repulsion and (b) due to the high shear the loops and tails (see Chapter 4) of
the polyelectrolyte will be able to collapse and attach to any remaining exposed clay
particle in their floc matrix. This will reduce the floc size and increase its density.

127
 Introduction to Colloid Science

Only at low shear will there be a small re-flocculation, which might be due to a steric
entanglement of flocs whose ends are able to decoil at low shears.

Density of flocs and fractal dimensions

Since flocs are usually non-spherical, permeable to water and can be composed of
particles of different refractive indexes, this poses a problem for the in-situ recording
of their size by light scattering (see Chapter 2), as the conversion of the raw data
depend on a model which assumes that particles are spherical and have a given
refractive index. Thanks to the development of camera’s in the past years, it is
nowadays possible to estimate the size of flocs and their settling velocity by video
recording. From Stokes, it has then been found that the density of a floc is a
decreasing function of its size, and this has led scientists to conveniently model them
as if they were fractal objects. As flocs, especially the ones created in-situ have a
very heterogeneous composition they are not “real” fractals as their structure is not
self-similar.

Fractals and self-similarity

At the beginning of the chapter, we have evoked the term of fractal. This term stems
from its inventor, B.B. Mandelbrot who introduced it in the late 1970s. In 1967 he
wrote a famous article entitled “How Long Is the Coast of Britain?” 67 where he noted
that the length of the coast depends on the scale at which it is measured. In fact, it
was empirically found that the measured length of a coast could be estimated by:

𝐿𝐿(𝜆𝜆) = 𝑀𝑀𝜆𝜆1−𝐷𝐷

where 𝜆𝜆 is the measurement scale, L(𝜆𝜆) the length of the coast, M a positive constant
and D another constant, greater than or equal to 1. If D = 1, the length is independent
of the measurement scale (𝐿𝐿 = 𝑀𝑀). It has been found that D is ranging from 1.02 for
the coastline of South Africa to 1.25 for the West coast of Britain. Mandelbrot in the
paper introduces the concept of statistical self-similarity and fractional dimension
that will enable him to develop his concept of fractals in later publications.

Self-Similarity

Here we give an example showing how to construct a real self-similar fractal object.
The one given here is called Koch snowflake:

67 Mandelbrot, B. B. (1967). How long is the coast of Britain. Science, 156(3775), 636-638.

128
 Chapter 5 Floc formation and break-up

If one zoom-in on such a flake, one will realize that the structure is self-similar: at
various scales, the shape of the interface will be the same, and indeed the perimeter
of the fractal object will depend on the measurement scale used (the first triangle
has obviously a smaller perimeter than the floc on the right).

L-system and the growth of floc-like structures

So-called L-systems were introduced and developed in 1968 by Aristid Lindenmayer,

a Hungarian theoretical biologist and botanist at the University of Utrecht. A simple
example of L-system is as follows, and was introduced to understand the growth of
algae:

variables: A, B
rules: (A →AB), (B→A)

Starting with variable A and applying the rules, one obtains:

n=0:A

n = 1 : AB [as A →AB]

n = 2 : ABA [as A →AB and B→A]

n = 3 : ABAAB [as A →AB , B→A and A →AB ]

n = 4 : ABAABABA

n = 5 : ABAABABAABAAB

n = 6 : ABAABABAABAABABAABABA

n = 7 : ABAABABAABAABABAABABAABAABABAABAAB

By using a computer it is nowadays easy to generate beautiful 2 or 3D images of

plants grown using similar type of L-systems:

129
 Introduction to Colloid Science

2D plant-like structures

Note from the examples that depending on the “grammar” used, the general
symmetry can be broken. This also happens in natural systems.

Fractal dimension

Let us consider a floc consisting of N particles of radius a:

floc of radius RN consisting of N particles (red balls) of radius a.

The number of particles inside the floc is given by:

𝑅𝑅𝑁𝑁 𝐷𝐷
𝑁𝑁 = � �
𝑎𝑎
where D is called the fractal dimension. If the floc would be non-fractal (imagine
that all the red balls have fused to make a giant red floc), one can estimate that the
volume of the giant floc is N times the volume of a red ball, i.e.:

4 4
𝜋𝜋𝑅𝑅𝑁𝑁 3 = 𝑁𝑁 𝜋𝜋𝑎𝑎3
3 3
from which follows that

130
 Chapter 5 Floc formation and break-up

𝑅𝑅𝑁𝑁 3
𝑁𝑁 = � �
𝑎𝑎
implying D = 3. For natural flocs, one finds experimentally that 2 < D < 3, which is
equivalent to say that not all space within a floc is occupied by clay particles.

When the fractal dimension of DLCA and RLCA flocs is analysed, it is found that the
DLCA flocs have a fractal dimension that is lower than the RLCA flocs, in accordance
with the fact that DLCA flocs have a loose structure compared to RLCA flocs. The
fractal dimension is also depending on the environment of the flocs, in particular the
shear stresses, that affects non only the aggregation but also the re-conformation of
the floc:

Re-conformation of a floc due to shear: the (pseudo) fractal dimension increases; on the
right: flocs observed in the North Sea, close to the Port of Rotterdam, at different stages of
coiling. The scale of each floc is approximately 125 µm.

Fractal dimension of flocs

A powerful technique to determine the fractal dimension of flocs is the use of video
microscopy 68. As stated above, flocs are not fractal objects as they are not self-
similar but nonetheless a (pseudo) fractal dimension can be derived knowing their
size and settling velocity.

The flocs are pipetted into a settling column and their settling is recorded. From the
analysis of the video, both particle size 𝑅𝑅𝑁𝑁 and velocity 𝑣𝑣𝑁𝑁 can be determined. Here
the settling velocity is plotted as function of the diameter 2𝑅𝑅𝑁𝑁 . Each blue cross
represents a measurement. The flocs were obtained by mixing river clay and cationic
polyelectrolyte.

68 Manning, A.J. and Dyer, K.R. (2002). The use of optics for the in-situ determination of

flocculated mud characteristics. J. Optics A: Pure and Applied Optics, Institute of Physics
Publishing, 4, S71-S81

131
 Introduction to Colloid Science

Note that for a given size, for example 100 microns, the velocity 𝑣𝑣𝑁𝑁 can vary by a
factor 100, implying that the flocs have very different densities. Using Stokes’ s law
(see Chapter 2) the floc density 𝜌𝜌𝑁𝑁 can then easily be determined by

𝜌𝜌𝑁𝑁 (𝑅𝑅𝑁𝑁 ) = 𝜌𝜌𝑤𝑤 + 9𝜂𝜂𝜂𝜂𝑁𝑁 /�2𝑅𝑅𝑁𝑁 2 𝑔𝑔�

The mean density of flocs can be estimated by averaging the density for each size.
Below the mean effective density (𝜌𝜌𝑁𝑁 − 𝜌𝜌𝑤𝑤 ) is plotted as function of the floc
diameters in red. The fractal dimension can then be calculated realising that

𝜌𝜌𝑁𝑁 − 𝜌𝜌𝑤𝑤 𝜙𝜙𝑠𝑠 �𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤 �

=
𝜌𝜌𝑝𝑝 𝜌𝜌𝑝𝑝

where 𝜙𝜙𝑠𝑠 is the volume fraction of clay inside a floc, 𝜌𝜌𝑝𝑝/𝑤𝑤 the density of the
clay/water and that

𝑁𝑁 ∙ 𝑎𝑎3 𝑅𝑅𝑁𝑁 𝐷𝐷−3

𝜙𝜙𝑠𝑠 = =� �
𝑅𝑅𝑁𝑁 3 𝑎𝑎

Combining these equations leads to:

𝑅𝑅𝑁𝑁 𝐷𝐷−3
𝜌𝜌𝑁𝑁 − 𝜌𝜌𝑤𝑤 = �𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤 � � �
𝑎𝑎
and by fitting the data (dashed line) a fractal dimension of D = 2.39 was obtained.

132
 Chapter 5 Floc formation and break-up

Illustrations
TEM of minerals
Image reproduced from the ‘Images of Clay Archive’ of the Mineralogical Society of Great
Britain & Ireland and The Clay Minerals Society (https://www.minersoc.org/images-of-
clay.html)

Fractal (creative commons)

https://en.wikipedia.org/wiki/Koch_snowflake

Fractal tree
Generated using MuPad (Matlab)

133
 Chapter 6
Modelling and measuring
the flocculation rate
 Introduction to Colloid Science

From the examples given in the previous chapters, it is clear that modelling (and
predicting) the floc structure is challenging, owing to the different modes of
aggregation and the role of the shape of the particles. It is however easier to say
something about the rate of flocculation as this depends mainly on the interactions
between particles. This is what we are going to show in the present chapter.

The simplest model for predicting

the flocculation rate has been
developed by the Polish scientist
Marian von Smoluchowski (1875 -
1917). Smoluchowski moved to
Kraków (Poland) in 1913, to take
over the chair in Experimental
Physics Department, and died there
in 1917, a victim of a dysentery
epidemic, aged 45. His scientific
output is however large: in 1904 he
was the first who noted the
existence of density fluctuations in
the gas phase and in 1908 he
became the first physicist to ascribe the phenomenon of critical opalescence to large
density fluctuations. His investigations also concerned the blue colour of the sky as
a consequence of light dispersion on fluctuations in the atmosphere. In 1906,
independently of Albert Einstein, he described Brownian motion. Smoluchowski
presented an equation which became an important basis of the theory of stochastic
processes. In 1916, he proposed the equation of diffusion in an external potential
field. This equation bears his name. In 1916 he also published the article in which he
describes the rate of flocculation of colloidal particles 69.

Model of Smoluchowski: aggregation by Brownian motion

We consider the evolution of the number of particles of one class, say a class
numbered k as function of time. A class is the ensemble of all the particles 𝑁𝑁𝑘𝑘 which
have as characteristic to be formed of k primary particles (particles that cannot
break, i.e. the smallest in the system). For instance:

69 Marian Smoluchowski, « Drei Vorträge über Diffusion, Brownsche Molekularbewegung und

Koagulation von Kolloidteilchen », Physik. Zeit., vol. 17, 1916, p. 557–571, 585–599

136
 Chapter 6 Modelling and measuring the flocculation rate

We assume that all primary particles are the same. Particles arrive in class k because:

 they are formed by aggregation of particles of smaller size. We call this gain
by aggregation (GA)
 they are formed by the destruction of larger particles. We call this gain by
break-up (GB)

Particles leave class k because:

 they are lost by aggregation of a particle in class k with another particle (of
any class). We call this loss by aggregation (LA)
 they are lost by break-up of a particle in class k. We call this loss by break-
up (LB)

Examples of GA, GB, LA and LB

The general formulation of the rate of change of the number of particle in class k is
given by:

𝑑𝑑𝑁𝑁𝑘𝑘
= 𝐺𝐺𝐺𝐺 + 𝐺𝐺𝐺𝐺 − 𝐿𝐿𝐿𝐿 − 𝐿𝐿𝐿𝐿
𝑑𝑑𝑑𝑑
To simplify the model, we assume that there is no break-up in our system (all
particles grow in time). The formulation then reduces to:

𝑑𝑑𝑁𝑁𝑘𝑘
= 𝐺𝐺𝐺𝐺 − 𝐿𝐿𝐿𝐿
𝑑𝑑𝑑𝑑
where:
𝑘𝑘−1
1
𝐺𝐺𝐺𝐺 = � 𝛼𝛼𝑖𝑖,𝑘𝑘−𝑖𝑖 𝛽𝛽𝑖𝑖,𝑘𝑘−𝑖𝑖 𝑁𝑁𝑖𝑖 𝑁𝑁𝑘𝑘−𝑖𝑖
2
𝑖𝑖=1

137
 Introduction to Colloid Science

This term represents the rate of formation of a class k particle by aggregation of any
particles from lower classes (class 1 until class (k-1)), such that 𝑁𝑁𝑖𝑖 + 𝑁𝑁𝑘𝑘−𝑖𝑖 ⇌ 𝑁𝑁𝑘𝑘 . The
½ accounts for the fact that the aggregates are counted twice: 𝑁𝑁𝑘𝑘−𝑖𝑖 + 𝑁𝑁𝑖𝑖 ⇌ 𝑁𝑁𝑘𝑘 is
the same as 𝑁𝑁𝑖𝑖 + 𝑁𝑁𝑘𝑘−𝑖𝑖 ⇌ 𝑁𝑁𝑘𝑘 . the coefficient 𝛼𝛼𝑖𝑖,𝑗𝑗 is called the collision efficiency
(between particle of class i and particle of class j), and 𝛽𝛽𝑖𝑖,𝑗𝑗 is the collision frequency.
We also have:
𝑛𝑛−𝑘𝑘

𝐿𝐿𝐿𝐿 = 𝑁𝑁𝑘𝑘 � 𝛼𝛼𝑘𝑘,𝑖𝑖 𝛽𝛽𝑘𝑘,𝑖𝑖 𝑁𝑁𝑖𝑖

𝑖𝑖=1

This term represents the rate at which a class k particle disappears from class k by
aggregation with any other particle. The summation is limited to (𝑛𝑛 − 𝑘𝑘) where n is
the total number of classes considered (usually taken to be extremely large). This
means that we cannot make particles larger that class n particles. This is mainly done
for numerical procedures reasons. In doing the numerical simulations, one always
insures that class n is never populated. This is equivalent to say that the summation
can be extended to infinity. Usually one defines:

𝐾𝐾𝑘𝑘,𝑖𝑖 = 𝛼𝛼𝑘𝑘,𝑖𝑖 𝛽𝛽𝑘𝑘,𝑖𝑖

In the simplest models, as the one derived by Smoluchowski, on usually assumes

that 𝛼𝛼𝑘𝑘,𝑖𝑖 = 1 implying that the particles always stick when they touch. In more
realistic models, 𝛼𝛼𝑘𝑘,𝑖𝑖 is a value between 0 and 1. The collision frequency is linked to
the way two particles approach each other. Smoluchowski considered the case
where particles were moving because of Brownian motion, which we will discuss
now. We have already introduced Fick’s first law in a previous chapter. From this
law, we may evaluate the flux of particles 𝑛𝑛𝑘𝑘 (the number of particles of class k per
unit of volume 𝑉𝑉, for example the volume of the settling column in which the
experiment is performed) that enters a sphere of radius r centred in a given particle
of radius 𝑎𝑎𝑖𝑖 . It is given by:

𝜕𝜕𝑛𝑛𝑘𝑘
𝐽𝐽𝑘𝑘,𝑖𝑖 = 4𝜋𝜋𝑟𝑟 2 𝐷𝐷
𝜕𝜕𝜕𝜕
The units of 𝐽𝐽𝑘𝑘,𝑖𝑖 are in number of particles (of class) k entering the sphere per second.
Assuming that 𝐽𝐽𝑘𝑘,𝑖𝑖 is constant (steady-state condition), it is easy to integrate this
equation, and one obtains:

𝐽𝐽𝑘𝑘,𝑖𝑖
𝑛𝑛𝑘𝑘 = 𝑛𝑛𝑘𝑘,∞ −
4𝜋𝜋𝜋𝜋𝜋𝜋
where 𝑛𝑛𝑘𝑘,∞ = 𝑁𝑁𝑘𝑘 /𝑉𝑉 is the (bulk) particle concentration far from the sphere. We now
assume that when a particle k reaches 𝑟𝑟 = 𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 it “disappears” from class k: it is

138
 Chapter 6 Modelling and measuring the flocculation rate

glued to the particle i which centre is located at 𝑟𝑟 = 0. Thus 𝑛𝑛𝑘𝑘 (𝑟𝑟 = 𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 ) = 0
from which follows that:

𝐽𝐽𝑘𝑘,𝑖𝑖 = 4𝜋𝜋𝜋𝜋(𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 )𝑛𝑛𝑘𝑘,∞

This flux is now the rate (number/second) at which particles k meet particle i. The
particle i which centre is located at 𝑟𝑟 = 0 is not immobile: it also diffuses according
to Brownian motion. This implies that the diffusion coefficient 𝐷𝐷 is not the diffusion
coefficient of a particle, but should be the mutual diffusion coefficient 𝐷𝐷𝑘𝑘,𝑖𝑖 that
accounts for the fact that both particle k and particle i are experiencing Brownian
motion:

𝑘𝑘𝐵𝐵 𝑇𝑇 𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖

𝐷𝐷 = 𝐷𝐷𝑘𝑘,𝑖𝑖 = 𝐷𝐷𝑘𝑘 + 𝐷𝐷𝑖𝑖 =
6𝜋𝜋𝜋𝜋 𝑎𝑎𝑘𝑘 𝑎𝑎𝑖𝑖

The number of particles of class k that are leaving the class due to aggregation is
then given by the rate at which particles k meet particle i multiplied by the amount
of particles i and summed over all types of particles i :
𝑛𝑛−𝑘𝑘 𝑛𝑛−𝑘𝑘
𝑑𝑑𝑁𝑁𝑘𝑘 𝑁𝑁𝑘𝑘 2𝑘𝑘𝐵𝐵 𝑇𝑇 (𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 )2
= � 𝑁𝑁𝑖𝑖 𝐽𝐽𝑘𝑘,𝑖𝑖 = � 𝑁𝑁𝑖𝑖
𝑑𝑑𝑑𝑑 𝑉𝑉 3𝜂𝜂 𝑎𝑎𝑘𝑘 𝑎𝑎𝑖𝑖
𝑖𝑖=1 𝑖𝑖=1

By dividing this equation on both sides by 𝑉𝑉 one can express it as a relation between
concentrations:
𝑛𝑛−𝑘𝑘
𝑑𝑑𝑛𝑛𝑘𝑘 2𝑘𝑘𝐵𝐵 𝑇𝑇 (𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 )2
= 𝑛𝑛𝑘𝑘 � 𝑛𝑛𝑖𝑖
𝑑𝑑𝑑𝑑 3𝜂𝜂 𝑎𝑎𝑘𝑘 𝑎𝑎𝑖𝑖
𝑖𝑖=1

and define a corresponding rate constant 𝑘𝑘𝑘𝑘,𝑖𝑖 (m3/s) such that:

𝐾𝐾𝑘𝑘,𝑖𝑖 2𝑘𝑘𝐵𝐵 𝑇𝑇 (𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 )2

= 𝑘𝑘𝑘𝑘,𝑖𝑖 =
𝑉𝑉 3𝜂𝜂 𝑎𝑎𝑘𝑘 𝑎𝑎𝑖𝑖

We then find that:

2𝑘𝑘𝐵𝐵 𝑇𝑇 (2𝑎𝑎1 )2 8𝑘𝑘𝐵𝐵 𝑇𝑇

𝑘𝑘1,1 = =
3𝜂𝜂 𝑎𝑎1 2 3𝜂𝜂

Note that this rate constant is also equal to the rate 𝑘𝑘𝑘𝑘,𝑘𝑘 for all colliding particles of
equal sizes. (For particles of different sizes, one can easily verify that the collision
rate will always be smaller than for equal particles).

The rate of change of the number of particle in class 1 at the early stage of
aggregation is given by:

139
 Introduction to Colloid Science

𝑑𝑑𝑛𝑛1
= −𝑘𝑘1,1 𝑛𝑛12
𝑑𝑑𝑑𝑑
Primary particles cannot be formed, therefore their gain by aggregation (GA) is zero.
At the early stage of aggregation, only doublets are formed from the aggregation of
primary particles, so we do not have to consider the aggregation of class 1 particles
with other particles than themselves.

One can solve the equation using the fact that 𝑛𝑛1 (𝑡𝑡 = 0) = 𝑛𝑛1,0 and one obtains:
𝑛𝑛1,0
𝑛𝑛1 (𝑡𝑡) =
1 + 𝑘𝑘1,1 𝑛𝑛1,0 𝑡𝑡

If we now consider the aggregation process at any given time, one should find the
full equations:

𝑑𝑑𝑛𝑛1
= −𝑘𝑘1,1 𝑛𝑛12 − 𝑘𝑘1,2 𝑛𝑛1 𝑛𝑛2 − 𝑘𝑘1,3 𝑛𝑛1 𝑛𝑛3 ∙∙∙
𝑑𝑑𝑑𝑑
𝑑𝑑𝑛𝑛2 1
= 𝑘𝑘1,1 𝑛𝑛12 − 𝑘𝑘2,1 𝑛𝑛2 𝑛𝑛1 − 𝑘𝑘2,2 𝑛𝑛2 𝑛𝑛2 ∙∙∙
𝑑𝑑𝑑𝑑 2
𝑑𝑑𝑛𝑛3 1 1
= 𝑘𝑘1,2 𝑛𝑛1 𝑛𝑛2 + 𝑘𝑘2,1 𝑛𝑛2 𝑛𝑛1 − 𝑘𝑘3,1 𝑛𝑛3 𝑛𝑛1 ∙∙∙
𝑑𝑑𝑑𝑑 2 2
By assuming that all the rates 𝑘𝑘𝑖𝑖,𝑗𝑗 are the same (which is reasonable if the particles
are not too different in size) and using 𝑘𝑘𝑎𝑎 = 𝑘𝑘𝑖𝑖,𝑗𝑗 for all i and j, one obtains:

𝑑𝑑(𝑛𝑛1 + 𝑛𝑛2 + 𝑛𝑛3 ∙∙∙) 𝑘𝑘𝑎𝑎

= − (𝑛𝑛1 + 𝑛𝑛2 + 𝑛𝑛3 ∙∙∙)2
𝑑𝑑𝑑𝑑 2
the factor ½ can be understood easily when one considers again the early stage of
aggregation, when only doublets are formed. Then we have:

𝑑𝑑(𝑛𝑛1 + 𝑛𝑛2 ) 𝑘𝑘𝑎𝑎

= − (𝑛𝑛1 + 𝑛𝑛2 )2
𝑑𝑑𝑑𝑑 2
because:

𝑑𝑑𝑛𝑛1
= −𝑘𝑘1,1 𝑛𝑛12
𝑑𝑑𝑑𝑑
𝑑𝑑𝑛𝑛2 1
= 𝑘𝑘1,1 𝑛𝑛12
𝑑𝑑𝑑𝑑 2
From this, we realize that the aggregation (loss) of two primary particles leads to the
creation of one doublet.

140
 Chapter 6 Modelling and measuring the flocculation rate

We define the total amount of particles in the system by:

𝑛𝑛 𝑇𝑇 = 𝑛𝑛1 + 𝑛𝑛2 + 𝑛𝑛3 ∙∙∙

and the characteristic time:

2
𝜏𝜏 =
𝑘𝑘𝑎𝑎 𝑛𝑛1,0

The solutions to the full equations can be shown to be:

𝑛𝑛1,0
𝑛𝑛1 (𝑡𝑡) =
(1 + 𝑡𝑡/𝜏𝜏)2

𝑛𝑛1,0 𝑡𝑡/𝜏𝜏
𝑛𝑛2 (𝑡𝑡) =
(1 + 𝑡𝑡/𝜏𝜏)3

𝑛𝑛1,0 (𝑡𝑡/𝜏𝜏)𝑘𝑘−1
𝑛𝑛𝑘𝑘 (𝑡𝑡) =
(1 + 𝑡𝑡/𝜏𝜏)𝑘𝑘+1
𝑛𝑛1,0
𝑛𝑛 𝑇𝑇 (𝑡𝑡) =
1 + 𝑡𝑡/𝜏𝜏

Note that for 𝑡𝑡 ≪ 𝜏𝜏 one finds as expected:

𝑛𝑛1,0
𝑛𝑛1 (𝑡𝑡) =
1 + 𝑘𝑘𝑎𝑎 𝑛𝑛1,0 𝑡𝑡

𝑛𝑛𝑘𝑘>1 (𝑡𝑡) = 0

The solutions are plotted here:

1
n1
0.9
n2
0.8
n3
0.7 nT
0.6
n(t)/n1,0

0.5

0.4

0.3

0.2

0.1

0
0 0.5 1 1.5 2 2.5 3
t/τ

From the figure it can be seen that the concentration of primary particles (n1) is
decreasing as function of time, as these particles are aggregating. The other classes

141
 Introduction to Colloid Science

start from a concentration equal to zero (we have assume that at t = 0 there are only
primary particles), then their concentration increases (n2 is formed by aggregation
of n1), and eventually decrease (n2 aggregates with n1 to form n3 and aggregates with
n2 to form n4 etc…).

The Brownian motion aggregation we have been reviewing is called perikinetic

aggregation in textbooks. This type of aggregation does not lead to the rapid
formation of very large aggregates, especially in dilute suspension. We can for
instance estimate

8𝑘𝑘𝐵𝐵 𝑇𝑇
𝑘𝑘𝑎𝑎 = ~1.23 × 10−17 𝑚𝑚3 /𝑠𝑠
3𝜂𝜂

Using an initial concentration of 𝑛𝑛1,0 = 58 𝑔𝑔/𝐿𝐿 = 1016 particles/𝑚𝑚3 (particles of 1

μm)

2 2
𝜏𝜏 = ~ = 16 𝑠𝑠
𝑘𝑘𝑎𝑎 𝑛𝑛1,0 1.23 × 10−17 1016

Concentrations in natural environments (estuaries, rivers) are more of the order of

mg/L leading to 𝜏𝜏~1.6 × 104 𝑠𝑠 (approximatively 5 hours). Aggregation is natural
environments, or in flocculation tanks, as the ones used in sanitary engineering are
usually triggered by fluid motion. This type of aggregation is called orthokinetic
aggregation.

Orthokinetic aggregation: aggregation by shear

As for the case of Brownian motion, we evaluate the flux of particles k that collide
with a particle i, but because of the water shear rate 𝐺𝐺(𝑠𝑠 −1 )70, the particle k arrives
laterally:

70
In Chapter 7 the shear rate will be defined by the symbol 𝛾𝛾̇ . Both symbols are used in
literature.

142
 Chapter 6 Modelling and measuring the flocculation rate

The flux of particles k arriving on i is given by:

𝑎𝑎𝑘𝑘 +𝑎𝑎𝑖𝑖
𝐽𝐽𝑘𝑘,𝑖𝑖 = 4𝑛𝑛𝑘𝑘 � (𝐺𝐺𝐺𝐺)𝑥𝑥𝑥𝑥𝑥𝑥
0

where 𝐺𝐺𝐺𝐺 is the fluid velocity (m/s) at position 𝑧𝑧 (the velocity is along the x axis, as
the blue arrow indicates in the illustration and depends on z only). There is contact
between the two spheres when 𝑟𝑟 = 𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 and therefore 𝑥𝑥 = �(𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 )2 − 𝑧𝑧 2 .
The 4 comes from the fact that one wants to know the number of particles whose
center pass through the capture cross-section 𝜋𝜋(𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 )2 accounting for the fact
that the particles can come from any direction and thus the surface are of contact is
represented by a sphere of area 4𝜋𝜋(𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 )2 . The units of 𝐽𝐽𝑘𝑘,𝑖𝑖 are in number of
particles (of class) k entering the sphere per second. It is easy to find that:

4
𝐽𝐽𝑘𝑘,𝑖𝑖 = 𝐺𝐺𝐺𝐺𝑘𝑘 (𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 )3
3
Similarly to what we have done before, we can construct:
𝑛𝑛−𝑘𝑘
𝑑𝑑𝑛𝑛𝑘𝑘
= � 𝑛𝑛𝑖𝑖 𝐽𝐽𝑘𝑘,𝑖𝑖
𝑑𝑑𝑑𝑑
𝑖𝑖=1

From which we can evaluate the rate constant 𝑘𝑘𝑘𝑘,𝑖𝑖 (m3/s) :

4
𝑘𝑘𝑘𝑘,𝑖𝑖 = 𝐺𝐺(𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 )3
3
Comparing this rate of aggregation with the one obtained from Brownian motion,
one observe that the rate we have found now depends significantly on the size of
the particles (in fact it scales as the volume of the particles): a large particle will
aggregate a lot of other particles. In the case of Brownian motion, a large particle
would not capture so many other particles, because its (Brownian) motion would be
very limited: large particles have a low Brownian diffusion coefficient.

In turbulent conditions, a mean shear stress can be obtained from:

𝜖𝜖
𝐺𝐺 = �
𝜈𝜈

where 𝜖𝜖 is the power input per unit of mass of fluid and 𝜈𝜈 the kinematic viscosity
(the ratio between the viscosity and the density of the fluid). The Kolmogorov
microscale, which separates the inertial range (where the energy is transferred with
very little dissipation) from the viscous subrange (where the energy is dissipated as
heat) is given by:

143
 Introduction to Colloid Science

𝜈𝜈
𝐿𝐿 = �
𝐺𝐺

For typical values of shear rates (50-100 s-1) in aqueous dispersions, L is of the order
of 100-150 μm, meaning that flocs will be prevented to grow larger than these
values, as shown in Chapter 5. If large flocs come in a region of high shear, break-
up (GB and LB) should not be neglected. The general formulation of the rate of
change of the number of particle is then generally calculated numerically for all the
classes. These models are found under the name “Population Balance Equation” in
literature.

Measuring the flocculation rate

The rate at which particles aggregate can be estimated by light scattering
measurements (see Chapter 2), by evaluating the (mean) particle size as a function
of time. In Chapter 5, we have introduced the concept of DLCA and RLCA. These can
be linked to the mean aggregation diameter, which can be derived from the size of
floc in the corresponding class. One finds the following behaviours 71:

By adapting the Population Balance Equation, so as, in particular, to include the

(pseudo) fractal size of the aggregates, it is possible to reasonably well fit measured
data. The RLCA mode of aggregation is in particular verified for salt-induced
aggregation (see Runkana et al. given in footnote). At high ionic strength however,
it would seem that the process goes through a transition between RLCA (short times)
and DLCA (longer times). This change in flocculation regime may be due to aggregate
restructuring during aggregation.

Stability ratio

Despite the fact that shear-induced aggregation is probably the dominant

mechanism for aggregation in engineering, Brownian-induced aggregation is quite

71Runkana, V., Somasundaran, P., & Kapur, P. C. (2005). Reaction‐limited aggregation in

presence of short‐range structural forces. AIChE journal, 51(4), 1233-1245.

144
 Chapter 6 Modelling and measuring the flocculation rate

important to determine the stability of a suspension. When discussing the DLVO

theory in an earlier chapter, we already made the link between zeta potential and
stability. We are now going to explain how this link can be understood.

The force on a particle due to the energy barrier is given by F𝑟𝑟𝑟𝑟𝑟𝑟 = −𝑑𝑑Φ/𝑑𝑑𝑑𝑑 and
the particle’s velocity can be estimated from the relation derived in Chapter 2:

𝐷𝐷𝑘𝑘
𝑣𝑣𝑘𝑘 = 𝐹𝐹
𝑘𝑘𝐵𝐵 𝑇𝑇

The flux of particles produced by the force field can be estimated to be:
𝑟𝑟𝑟𝑟𝑟𝑟
𝐽𝐽𝑘𝑘,𝑖𝑖 = 4𝜋𝜋𝑟𝑟 2 𝑛𝑛𝑘𝑘 (2𝑣𝑣𝑘𝑘 )

The 2 comes from the fact that each particle experience the energy barrier (similarly
to what we did when we introduced the mutual diffusion coefficient). We have here,
for simplicity, assumed that all the particles have the same size. This implies that
𝐷𝐷 = 2𝐷𝐷𝑘𝑘 . The total number of particles that hits the central particle i per second is
given by:

𝜕𝜕𝑛𝑛𝑘𝑘 𝑛𝑛𝑘𝑘 𝑑𝑑Φ

𝐽𝐽𝑘𝑘,𝑖𝑖 = 8𝜋𝜋𝑟𝑟 2 𝐷𝐷𝑘𝑘 � + �
𝜕𝜕𝜕𝜕 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑑𝑑𝑑𝑑

Assuming a steady-state, one has 𝐽𝐽𝑘𝑘,𝑖𝑖 constant, and therefore:

𝐽𝐽𝑘𝑘,𝑖𝑖 𝜕𝜕𝑛𝑛𝑘𝑘 𝑛𝑛𝑘𝑘 𝑑𝑑Φ

2
= + = 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐
8𝜋𝜋𝑟𝑟 𝐷𝐷𝑘𝑘 𝜕𝜕𝜕𝜕 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑑𝑑𝑑𝑑

In our case both 𝑛𝑛𝑘𝑘 and Φ depend on 𝑟𝑟 only (we can therefore replace the 𝜕𝜕 by 𝑑𝑑 in
the expression above). Using the relation:

𝑑𝑑 Φ 𝑑𝑑𝑛𝑛𝑘𝑘 Φ 𝑛𝑛𝑘𝑘 𝑑𝑑Φ Φ

�𝑛𝑛𝑘𝑘 exp � �� = exp � �+ exp � �
𝑑𝑑𝑑𝑑 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑑𝑑𝑑𝑑 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑑𝑑𝑑𝑑 𝑘𝑘𝐵𝐵 𝑇𝑇

one gets:

𝐽𝐽𝑘𝑘,𝑖𝑖 Φ 𝑑𝑑 Φ
2
exp � �= �𝑛𝑛𝑘𝑘 exp � ��
8𝜋𝜋𝑟𝑟 𝐷𝐷𝑘𝑘 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑑𝑑𝑑𝑑 𝑘𝑘𝐵𝐵 𝑇𝑇

which can be integrated from 𝑟𝑟 to ∞ (where Φ = 0 and 𝑛𝑛𝑘𝑘 = 𝑛𝑛𝑘𝑘,∞ = 𝑁𝑁𝑘𝑘 /𝑉𝑉 is the
(bulk) particle concentration far from the sphere )

We thus get:
∞
Φ 𝐽𝐽𝑘𝑘,𝑖𝑖 1 Φ
𝑛𝑛𝑘𝑘,∞ − 𝑛𝑛𝑘𝑘 exp � �= � 2 exp � � 𝑑𝑑𝑑𝑑
𝑘𝑘𝐵𝐵 𝑇𝑇 8𝜋𝜋𝐷𝐷𝑘𝑘 𝑟𝑟 𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇

145
 Introduction to Colloid Science

which can be rewritten as:

∞
−Φ 𝐽𝐽𝑘𝑘,𝑖𝑖 −Φ 1 Φ
𝑛𝑛𝑘𝑘 = 𝑛𝑛𝑘𝑘,∞ exp � �− exp � � � 2 exp � � 𝑑𝑑𝑑𝑑
𝑘𝑘𝐵𝐵 𝑇𝑇 8𝜋𝜋𝐷𝐷𝑘𝑘 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑟𝑟 𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇

When 𝑟𝑟 = 2𝑎𝑎 we have aggregation and therefore 𝑛𝑛𝑘𝑘 = 0 leading to:

8𝜋𝜋𝐷𝐷𝑘𝑘 𝑛𝑛𝑘𝑘,∞
𝐽𝐽𝑘𝑘,𝑖𝑖 =
∞ 1 Φ
∫2𝑎𝑎 2 exp �𝑘𝑘 𝑇𝑇 � 𝑑𝑑𝑑𝑑
𝑟𝑟 𝐵𝐵

In the limiting case when there is no potential between the particles (Φ = 0), except
for an infinitely strong van der Waals attraction when they make contact, we get the
so-called fast aggregation rate:

𝐽𝐽𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 = 𝐽𝐽𝑘𝑘,𝑖𝑖 (Φ = 0) = 𝐽𝐽𝑘𝑘,𝑖𝑖 = 16𝜋𝜋𝜋𝜋𝐷𝐷𝑘𝑘 𝑛𝑛𝑘𝑘,∞

which is indeed the one we found for Brownian motion (= 4𝜋𝜋𝜋𝜋(𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 )𝑛𝑛𝑘𝑘,∞ ).

The stability ratio is defined by:

∞ ∞
𝐽𝐽𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 1 Φ 1 Φ
𝑊𝑊 = = 2𝑎𝑎 � 2 exp � � 𝑑𝑑𝑑𝑑 = 2 � 2 exp � � 𝑑𝑑𝑑𝑑
𝐽𝐽𝑘𝑘,𝑖𝑖 2𝑎𝑎 𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇 2 𝑠𝑠 𝑘𝑘𝐵𝐵 𝑇𝑇

where 𝑠𝑠 = 𝑟𝑟/𝑎𝑎. This integral can be evaluated numerically, and Verwey and
Overbeek showed in 1948 that 𝑊𝑊 was determined almost entirely by the value of Φ
at its maximum (Φ𝑚𝑚𝑚𝑚𝑚𝑚 ). A simple approximation, due to Reerink and Overbeek 72 in
1954, for unequal spheres, is:

1 Φ𝑚𝑚𝑚𝑚𝑚𝑚
𝑊𝑊 ~ exp � �
𝜅𝜅(𝑎𝑎𝑘𝑘 + 𝑎𝑎𝑖𝑖 ) 𝑘𝑘𝐵𝐵 𝑇𝑇

For 1 μm particles, in 100 mM of a 1-1 electrolyte, 𝜅𝜅𝜅𝜅~500. For a barrier of 10𝑘𝑘𝐵𝐵 𝑇𝑇

one would get 𝑊𝑊~45, whereas for 20𝑘𝑘𝐵𝐵 𝑇𝑇 one would get 𝑊𝑊~106 i.e. only one in
every million collisions would occur between particles having sufficient energy to
overcome the barrier.

A small change in electrolyte concentration or in particle’s surface potential can have

dramatic effects on the aggregation rate. In Chapter 3, when we derived the Schulze-
Hardy rule, we have given an different form for Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 . We will now use the following
one:

72 Reerink, H., & Overbeek, J. T. G. (1954). The rate of coagulation as a measure of the stability

of silver iodide sols. Discussions of the Faraday Society, 18, 74-84.

146
 Chapter 6 Modelling and measuring the flocculation rate

−𝐴𝐴𝐴𝐴 𝑘𝑘𝐵𝐵 𝑇𝑇 2 𝑞𝑞𝑞𝑞(𝑎𝑎)

Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 = + 32π𝜀𝜀0 𝜀𝜀𝑟𝑟 � � 𝑎𝑎 tanh2 � � exp�−𝜅𝜅(𝑟𝑟 − 2𝑎𝑎)�
12(𝑟𝑟 − 2𝑎𝑎) 𝑞𝑞 4𝑘𝑘𝐵𝐵 𝑇𝑇

which is valid for any 𝜓𝜓(𝑎𝑎) and κa ≫ 1 (as one can see, we have here simplified the
expression for the van der Waals force by taking the one derived for plates, i.e. which
is valid for spheres as well when κa ≫ 1 ). The position of the energy barrier when
Φ𝑚𝑚𝑚𝑚𝑚𝑚 = 0 is represented by :

𝑑𝑑Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷
Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 = =0
𝑑𝑑𝑑𝑑
From evaluating 𝜅𝜅Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 + 𝑑𝑑Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 /𝑑𝑑𝑑𝑑 = 0 one obtains:

−𝐴𝐴𝐴𝐴𝐴𝐴 𝐴𝐴𝐴𝐴
+ =0
12(𝑟𝑟 − 2𝑎𝑎) 12(𝑟𝑟 − 2𝑎𝑎)2

and therefore:

𝜅𝜅(𝑟𝑟 − 2𝑎𝑎) = 1

Using this expression in Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 = 0, and the definition of 𝜅𝜅 for a z-z electrolyte (𝑞𝑞 =
𝑧𝑧𝑧𝑧):

2(𝑒𝑒𝑒𝑒)2 𝑁𝑁𝐴𝐴
𝜅𝜅 2 = 𝐶𝐶(mM)
𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇

it follows that the critical concentration at which fast aggregation will occur is:

𝜀𝜀0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇 384π𝜀𝜀0 𝜀𝜀𝑟𝑟 2 𝑘𝑘𝐵𝐵 𝑇𝑇 4 𝑒𝑒𝑒𝑒𝑒𝑒(𝑎𝑎)

𝐶𝐶fast (mM) = 2
� � � � tanh4 � � exp(−2)
2(𝑒𝑒𝑒𝑒) 𝑁𝑁𝐴𝐴 𝐴𝐴 𝑒𝑒𝑒𝑒 4𝑘𝑘𝐵𝐵 𝑇𝑇

It is convenient to introduce here the Bjerrum length (after Danish chemist Niels
Bjerrum) which is the separation at which the electrostatic interaction between two
elementary charges is comparable in magnitude to the thermal energy scale, 𝑘𝑘𝐵𝐵 𝑇𝑇.
The Bjerrum length reads:

𝑒𝑒 2
𝑙𝑙𝐵𝐵 =
4𝜋𝜋𝜋𝜋0 𝜀𝜀𝑟𝑟 𝑘𝑘𝐵𝐵 𝑇𝑇

For water at room temperature (𝑇𝑇 = 300 𝐾𝐾), 𝜀𝜀𝑟𝑟 ~ 80 and 𝑙𝑙𝐵𝐵 ~ 0.7 𝑛𝑛𝑛𝑛.

We obtain:

962 exp(−2) 𝑘𝑘𝐵𝐵 𝑇𝑇 2 1 𝑒𝑒𝑒𝑒𝑒𝑒(𝑎𝑎)

𝐶𝐶fast (mM) = � � 3 6 tanh4 � �
8𝜋𝜋𝜋𝜋𝐴𝐴 𝐴𝐴 𝑙𝑙𝐵𝐵 𝑧𝑧 4𝑘𝑘𝐵𝐵 𝑇𝑇

147
 Introduction to Colloid Science

This expression scales with 1/𝑧𝑧 6 for large surface potentials but when when
𝑒𝑒𝑒𝑒𝑒𝑒(𝑎𝑎) ≪ 4𝑘𝑘𝐵𝐵 𝑇𝑇 it scales with 1/𝑧𝑧 2 as we have found while deriving the expression
for the Schulze-Hardy rule in Chapter 3.

We can also evaluate Φ𝑚𝑚𝑚𝑚𝑚𝑚 : Φ𝑚𝑚𝑚𝑚𝑚𝑚 occurs when 𝑑𝑑Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 /𝑑𝑑𝑑𝑑 = 0 i.e. when

𝐴𝐴𝐴𝐴 𝑘𝑘𝐵𝐵 𝑇𝑇 2 𝑞𝑞𝑞𝑞(𝑎𝑎)

= 32π𝜀𝜀0 𝜀𝜀𝑟𝑟 � � 𝑎𝑎 tanh2 � � exp�−𝜅𝜅(𝑟𝑟𝑚𝑚𝑚𝑚𝑚𝑚 − 2𝑎𝑎)�
12𝜅𝜅(𝑟𝑟𝑚𝑚𝑚𝑚𝑚𝑚 − 2𝑎𝑎)2 𝑞𝑞 4𝑘𝑘𝐵𝐵 𝑇𝑇

For which we get:

𝐴𝐴𝐴𝐴 1
Φ𝑚𝑚𝑚𝑚𝑚𝑚 = � − 1�
12(𝑟𝑟𝑚𝑚𝑚𝑚𝑚𝑚 − 2𝑎𝑎) 𝜅𝜅(𝑟𝑟𝑚𝑚𝑚𝑚𝑚𝑚 − 2𝑎𝑎)

Around the fast aggregation concentration, it is expected that 𝜅𝜅(𝑟𝑟𝑚𝑚𝑚𝑚𝑚𝑚 − 2𝑎𝑎) should
be close to one (see derivation just above). One may therefore write, using the fact
that ln(𝑥𝑥) ~ 1 + 𝑥𝑥 for small 𝑥𝑥 :

−𝐴𝐴𝐴𝐴
Φ𝑚𝑚𝑚𝑚𝑚𝑚 ~ ln[𝜅𝜅(𝑟𝑟𝑚𝑚𝑚𝑚𝑚𝑚 − 2𝑎𝑎)]
12(𝑟𝑟𝑚𝑚𝑎𝑎𝑥𝑥 − 2𝑎𝑎)

This implies that:

Φ𝑚𝑚𝑚𝑚𝑚𝑚 −𝐴𝐴𝐴𝐴
ln(𝑊𝑊) ~ ~ �ln(𝜅𝜅) + ln[(𝑟𝑟𝑚𝑚𝑚𝑚𝑚𝑚 − 2𝑎𝑎)]�
𝑘𝑘𝐵𝐵 𝑇𝑇 12𝑘𝑘𝐵𝐵 𝑇𝑇(𝑟𝑟𝑚𝑚𝑚𝑚𝑚𝑚 − 2𝑎𝑎)

In view of the relation 𝜅𝜅 ~ 𝐶𝐶 1/2 one then gets:

−𝐴𝐴𝐴𝐴
ln(𝑊𝑊) ~ ln(𝐶𝐶) + 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐
24𝑘𝑘𝐵𝐵 𝑇𝑇(𝑟𝑟𝑚𝑚𝑚𝑚𝑚𝑚 − 2𝑎𝑎)

Using (see above on this page):

𝐴𝐴𝐴𝐴 𝑎𝑎 𝑧𝑧𝑧𝑧𝑧𝑧(𝑎𝑎)
~ 2 tanh2 � �
𝜅𝜅(𝑟𝑟𝑚𝑚𝑚𝑚𝑚𝑚 − 2𝑎𝑎)2 𝑧𝑧 4𝑘𝑘𝐵𝐵 𝑇𝑇

One also gets:

−𝑎𝑎 𝑧𝑧𝑧𝑧𝑧𝑧(𝑎𝑎)
ln(𝑊𝑊) ~ tanh2 � � ln(𝐶𝐶) + 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐
𝑧𝑧 2 4𝑘𝑘𝐵𝐵 𝑇𝑇

From this expression it is clear that when 𝜓𝜓(𝑎𝑎) ~ 0 (and thus the zeta potential 𝜁𝜁
close to zero), one has ln(𝑊𝑊) ~ 0 and thus 𝑊𝑊 ~ 1 (fast aggregation). This is why, as
a rule of thumb, one says that flocculation is predicted to happen at small zeta

148
 Chapter 6 Modelling and measuring the flocculation rate

potentials: typically fast flocculation occurs when 𝑒𝑒𝑒𝑒/𝑘𝑘𝑘𝑘 is of the order of 1 i.e. at
room temperature, when 𝜁𝜁 is of the order of 25 mV.

Niels Janniksen Bjerrum (1879 – 1958) had a sister Dr.

Kirstine Meyer who was a prominent physicist 73. His early
interest was natural history, but this soon changed to
chemistry. He carried out research stimulated by S. M.
Jörgensen, one of the pioneers on coordination complexes,
obtaining a Master’s degree in 1902 and a Doctor’s degree
in 1908. He visited and worked with several of the great
physical chemists of the time: in 1905 Luther in Leipzig, in
I907 Werner in Zürich, in 1910 Perrin in Paris and in 1911
Nernst in Berlin. In 1912 he became docent at the
University of Copenhagen. In 1914 he succeeded O.T. Christensen as Professor of
chemistry at the Royal Agricultural College, Copenhagen where he remained until he
retired in 1949. In the period 1906-1908 Bjerrum was determining the structure of
various chromium complexes and measuring hydrogen ion concentrations before
the expression pH had yet been invented by Sörensen. He also did fundamental work
on the factors determining the acidity of soil.

The Bjerrum plot is named after him. A Bjerrum plot is a graph of the concentrations
(or ratio of concentrations) of the different species of a polyprotic acid (= which are
able to donate more than one proton per acid molecule) in a solution, as functions
of the solution's pH, when the solution is at equilibrium. Most often, the carbonate
system is plotted, where the polyprotic acid is carbonic acid (a diprotic acid), and the
different species are carbonic acid, carbon dioxide, bicarbonate, and carbonate. In

73 She received her PhD in physics from the University of Copenhagen in 1909, becoming the
first Danish woman to earn a doctorate in natural sciences. Her dissertation,
Temperaturbegrebets Udvikling gennem Tiderne (The Development of the Temperature
Concept through Time), was an in-depth treatment of the history of the concept of
temperature. In 1902, Meyer founded Fysisk Tidsskrift, the Danish journal of physics. She was
its editor until 1913.

149
 Introduction to Colloid Science

acidic conditions, the dominant form is CO2; in basic (alkalinic) conditions, the
dominant form is CO32−; and in between, the dominant form is HCO3−. At every pH,
the concentration of carbonic acid (H2CO3) is assumed to be negligible compared to
the concentration of CO2, and so is often omitted from Bjerrum plots. These plots
are typically used in ocean chemistry to track the response of an ocean to changes
in both pH and of inputs in carbonate and CO2.

Experimental verification

The stability ratio is not only depending on the

zeta potential 𝜁𝜁, it is also highly dependent on
the ionic strength, as we have seen in Chapter
3 that 𝜁𝜁 is highly depending on it. Fast
flocculation will occur when 𝜁𝜁 is in good
approximation zero. This fact is verified
experimentally in most cases.

There are however still open questions

related to the prediction of the stability ratio.
Even though the DLVO theory can correctly
predict the fast flocculation regime (based on
zeta potential measurements in particular), it
fails to describe correctly the slope between
ln(𝑊𝑊) and ln(𝐶𝐶): from the DLVO theory, the
slope should be much less steep than it is in reality for most systems. Sometimes,
also, the fast flocculation regime is not exactly predicted from DLVO:

Left: Forces between pairs of latex amidine spheres (size 1μm) versus the surface. Right:
measured stability ratios (symbols) with calculated ones based on the force measurements
(solid lines). The results for DLVO theory are shown for comparison (dashed line)

150
 Chapter 6 Modelling and measuring the flocculation rate

From the force/separation graphs 74, the Hamaker constant could be estimated
using the data at high salt concentration (where van der Waals force dominates).
An additional attractive non-DLVO force could be estimated from the other curves.
The result of DLVO theory only is shown for comparison (dashed lines). The non-
DLVO behaviour was attributed to surface charge heterogeneities.

Measuring forces between particles

In order to estimate the forces between colloidal particles (as in the previous figure,
on the left), one makes use of an Atomic Force Microscope (AFM) which is a very-
high-resolution type of scanning probe microscopy with a resolution for the
distances on the order of fractions of a nanometer (more than 1000 times better
than a regular optical microscope) and for the forces of the order of 10-8 N. The AFM
was invented by IBM Scientists in 1982. The precursor to the AFM, the scanning
tunneling microscope (STM), was developed by Gerd Binnig and Heinrich Rohrer in
the early 1980s at IBM Research in Zurich, a development that earned them the
Nobel Prize for Physics in 1986. The first commercially available atomic-force
microscope was introduced in 1989. Another
technique to measure interparticle forces is the
Surface Force Apparatus (SFA), developed in the
early 1970s at Cambridge 75. The SFA is more
ideally suited than AFM to measuring surface-
surface interactions, and can measure much
longer-range forces more accurately. The SFA
technique is however quite demanding and
hence, only a handful of labs worldwide have
these instruments.

Both AFM and SFA techniques work on the

principle of approaching two surfaces and
measuring their interaction. Here we present the AFM only. To measure the force
between two colloidal particles, a colloidal particle is glued to the tip of the
cantilever and brought closer to a colloidal particle that is glued to the XYZ scanner.
Depending on the interaction between the particles, the tip is attracted to or
repelled from the other particle, and this deviation is measured through the
deviation of the laser beam. This deviation can be linked to the force between the
particles.

74 Figures taken from Montes Ruiz-Cabello, F. Javier, et al. "Interaction forces and aggregation

rates of colloidal latex particles in the presence of monovalent counterions." The Journal of
Physical Chemistry B 119.25 (2015): 8184-8193.
75 See the book by Israelachvili, Jacob N. (1992). Intermolecular and surface forces. Boston:

Academic Press

151
 Introduction to Colloid Science

Another AFM use is the scanning mode, where the tip is dragged along the surface
and the change in elevation is recorded. 3D pictures can then be obtained from the
scanned surface 76:

Typical AFM images of a) bentonite, b) montmorillonite, c) kaolin, d) halloysite, e) silica and

f) graphene oxide

Settling velocity, optimal flocculation and zeta potential

Instead of measuring particle sizes via light scattering the optimal flocculation rate
can also be determined by measuring settling velocities: at optimal settling rate, the
zeta potential is minimum 77.

In this last example we will illustrate this fact and we will briefly show the
experimental problems to overcome when one wants to study the flocculation of a
natural clay with a commercial cationic polyelectrolyte 78:

76 Kryuchkova, Marina, et al. "Evaluation of toxicity of nanoclays and graphene oxide in vivo:

a Paramecium caudatum study." Environmental Science: Nano 3.2 (2016): 442-452.

77 Yu, X., & Somasundaran, P. (1996). Role of polymer conformation in interparticle-bridging

dominated flocculation. Journal of Colloid and Interface Science, 177(2), 283-287.

78 unpublished results

152
 Chapter 6 Modelling and measuring the flocculation rate

End of settling for a suspension made of 8.7 g/L clay in presence of various amount
of cationic flocculant Zetag 7587; flocculant to clay ratios are indicated above each
column.

From recording the suspension/water interface over time, the initial settling velocity
(estimated from the first 30 s of settling) can be estimated:

The settling velocities given here were measured in 250 mL columns, and each
column was mixed by rotating the column upside down 10 times. Several
parameters, like the volume and diameter of the column and the way to mix the
suspension is found to influence the settling velocity values. There are several
reasons for this: (1) the aggregation and break-up of clay and polyelectrolyte
particles is highly dependent on the shear rate and the residence time in the column,

153
 Introduction to Colloid Science

(2) the return flow of liquid in the column is influencing the settling of the flocs and
(3) the flocs at higher concentrations of flocculant become positively charged and
display an interaction with the wall. The walls of the settling columns are made of
glass, which is slightly negatively charged and therefore flocs remain electrostatically
attached to them.

From the protocol used to estimate the settling velocities it was found that the
fastest settling velocity is obtained for 6 mg/g flocculant to clay ratio. Additional
studies on the same system proved that the way of mixing the suspension did not
change the ratio for which the fastest settling velocity is observed. The quantitative
value of this velocity is however protocol-dependent, as some protocols lead to
larger flocs (and larger settling velocities) than others.

The same system was studied by static light scattering (see Chapters 2 and 5) and
electrophoresis (see Chapter 3). For these techniques, lower clay concentrations are
required, but it was shown that the results nicely compare to the ones obtained for
the settling tests.

Left figure: mean particle (D50) measured by static light scattering as function of
polyelectrolyte to clay ratio, for two times, indicated in the legend. The clay concentration is
0.7 g/L. Right figure: zeta potential estimated from the Smoluchowski formula and derived
from electrophoretic mobility results. The zeta potential is given as function of
polyelectrolyte to clay ratio for different clay concentrations, indicated in the legend.

As can be seen in the left figure, the largest flocs are obtained for a flocculant to clay
ratio of about 5-25 mg/g when the flocs are measured 120 s after mixing. One can
see that it corresponds to the ratio for which the zeta potential is zero (right figure).
Zeta potential measurements are performed extremely rapidly after mixing, to
prevent flocculation in the cell.

After 600 s, the largest sizes are found at ratios about 10-50 mg/g, and for higher
ratio’s the D50 does not decrease as fast as for flocs obtained at 120 s, as flocs keep
growing slowly in time, since they are continuously stirred in the jar. For these high
flocculant to clay ratio’s the zeta potential of flocs is highly positive, implying the

154
 Chapter 6 Modelling and measuring the flocculation rate

aggregation mechanism between flocs is due to steric interaction and/or hydrogen

bonding, but not to electrostatic attraction.

Interpretation of the electrophoretic mobility data: at low flocculant to clay ratio, the clay
particle is not fully covered by polyelectrolyte and its charge (and zeta potential) is negative.
At a specific flocculant to clay ratio the zeta potential of the system clay + polyectrolyte is
close to zero and rapid flocculation is expected. For higher flocculant to clay ratio, the system
becomes positively charged

Illustrations
Smoluchowski (public domain)
https://en.wikipedia.org/wiki/Marian_Smoluchowski

Niels Janniksen Bjerrum (public domain)

https://pubs.rsc.org/en/content/articlelanding/1959/tf/tf959550x001/unauth#!div
Abstract

155
 Chapter 7
Rheological behaviour
of colloidal suspensions
 Introduction to Colloid Science

Rheology is the study of the flow of a liquid in response to an applied force. The word
rheology has been introduced in 1928 by Eugene Bingham, professor at the
university Lehigh in the United States, after a suggestion of a colleague refereeing to
a famous quotation of Heraclitus: Panta rhei “everything flows” [the word rhei
means in Greek “to stream”]. In this Chapter we are going to see how the flow of a
colloidal suspension is different from that of a simple fluid (water for example).

Viscosity and yield stress

Viscosity

Intuitively, we know what viscosity is: it is a measure of “how well” a liquid flows.
Honey is a highly viscous liquid, and water is flowing rather well and is therefore a
low viscous fluid.

Different clay suspensions exhibiting different flow behaviours; (a) low viscous liquid (clay
particles have settled and a clear water layer can be seen); (d) fluid mud; (e) high viscosity
mud. The other photographs display mud is various stages of consolidation.

In order to quantify the way liquids flow one has first to define the conditions for the
flow: is it a laminar flow or a turbulent flow?

Laminar flow occurs in layers without mixing. Viscosity causes drag between layers
as well as with the fixed surface. Laminar flows occur at low water velocities, or low
Reynolds numbers. An obstruction and high water velocities generate turbulence
which mixes the fluid.

In rheological measurements in the laboratory mainly laminar flows are studied, as

in laminar flows the shear rate (defined underneath) can properly be quantified. This
is more difficult in turbulent flows, even though turbulent flows are the most
encountered in the natural environment. The Kolmogorov microscale (see Chapter
6) becomes an important parameter in turbulent mixing.

158
 Chapter 7 Rheological behaviour of colloidal suspensions

Let us consider a simple flow in the laminar regime, generated under controlled
conditions in the lab. An upper plate is moved parallel with a velocity v with respect
to another plate which is fixed (its velocity is zero in the frame of the laboratory).
The distance between the two plates is L. The force applied to slide the upper plate
is F and the cross-sectional area of the plate is A. The fluid will flow everywhere
parallel to the plates, since there is no turbulence and the velocity is assumed to vary
linearly across the gap. In most cases, the liquid layers near each plates have the
same velocity as that plate (“no-slip” condition).

schematic representation of a laminar shear flow between a sliding and a fixed plate
separated by a distance L; slices of liquid in blue and variation of the velocity in red

The fact that the fluid velocity varies linearly across the gap will now be discussed.
One first estimates the force on one of the water slices. We call (2) the slice of water
in contact with the upper plate:

The force on the slice (2) per unit area is given by

159
 Introduction to Colloid Science

𝐹𝐹
𝜎𝜎 =
𝐴𝐴
where 𝜎𝜎 (Pa) is called the shear stress acting on the surface (also often denoted 𝜏𝜏).
Due to stress materials deform. The deformation can be measured by the angle α in
the top figure. The shear strain 𝛾𝛾 due to this stress is defined as the displacement
𝑑𝑑𝑑𝑑 of the top surface with respect to the bottom plate relative to the thickness of
layer (2) which is 𝑑𝑑𝑑𝑑 :

𝑑𝑑𝑑𝑑
𝛾𝛾 = = tan(𝛼𝛼)
𝑑𝑑𝑑𝑑

In the case we would consider a purely elastic material (like a rubber eraser for
example) the deformation is reversible, and the material assumes its original shape
once the force is removed. As the material we consider is a fluid, we cannot apply a
static force, as in the case of a rubber eraser: because of its composition, the rubber
will oppose a resistance to the deformation, and a balance of forces will establish
between the imposed stress and the rubber resistance to the stress. A pure fluid
with no elastic behaviour (like water) has an irreversible deformation, so one has to
continuously shear the fluid to be able to measure its response. For fluids, one
therefore defines the shear rate 𝛾𝛾̇ (s-1) :

𝑑𝑑tan(𝛼𝛼) 1 𝑑𝑑𝑑𝑑 𝑑𝑑𝑣𝑣𝑥𝑥

𝛾𝛾̇ = = =
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑

For a constant shear rate, if the properties of the fluid do not change with height y,
one finds by integration that the velocity indeed varies linearly across the gap as:

𝑣𝑣𝑥𝑥 = 𝛾𝛾̇ 𝑦𝑦

where 𝑦𝑦 = 0 is defined at the bottom plate. The velocity 𝑣𝑣𝑥𝑥 = 𝛾𝛾̇ 𝐿𝐿 is the velocity of
the upper plate. The relation between the applied shear stress and the induced shear
rate (or the induced shear stress and measured shear rate – both are possible with
the modern rheometers) is given by:

𝑑𝑑𝑣𝑣𝑥𝑥
𝜎𝜎 = 𝜂𝜂𝛾𝛾̇ = 𝜂𝜂
𝑑𝑑𝑑𝑑

where 𝜂𝜂 is the shear (dynamic) viscosity (Pa ∙ s) of the fluid. One can therefore
estimate that the force per unit area exerted by slice (2) on slice (1) and in general
slice (k+1) on slice (k) is given by:

𝑑𝑑𝑣𝑣𝑥𝑥
𝜎𝜎𝑘𝑘+1→𝑘𝑘 = 𝜂𝜂
𝑑𝑑𝑑𝑑

160
 Chapter 7 Rheological behaviour of colloidal suspensions

This relation is true for so-called Newtonian fluids, the simplest fluids, which display
a linear relation between shear stress and shear rate. Most concentrated
suspensions, however, are non-Newtonian fluids. Here are the most common non-
Newtonian behaviours:

Modelling the shear stress as function of shear rate

For fluids with a zero yield stress (𝜎𝜎(𝛾𝛾̇ = 0)), which are the shear-thinning, shear-
thickening and Newtonian fluids, the relation between shear stress and shear rate
can be expressed as:

𝜎𝜎 = 𝑘𝑘𝛾𝛾̇ 𝑛𝑛

where n = 1 for a Newtonian fluid (and then 𝑘𝑘 has the dimension of a viscosity: 𝑘𝑘 =
𝜂𝜂), n < 1 for shear-thinning and n > 1 for shear-thickening. Note that the dimension
of 𝑘𝑘 depends on n.

For Bingham fluids, one uses:

𝜎𝜎 = 𝜎𝜎0 + 𝜂𝜂𝛾𝛾̇

where 𝜎𝜎0 is the yield stress (also often noted 𝜏𝜏𝑐𝑐 and defined underneath) . The non-
Newtonian plastic fluids can be modelled with the so-called Herschel-Bulkley model:

𝜎𝜎 = 𝜎𝜎0 + 𝑘𝑘𝛾𝛾̇ 𝑛𝑛

Many more models exist, as obviously the rheological behaviours of suspensions can
be complicated. A model used for the flow of blood is the Casson equation, given by:

161
 Introduction to Colloid Science

𝜎𝜎 𝑛𝑛 = 𝜎𝜎0 𝑛𝑛 + 𝑘𝑘𝛾𝛾̇ 𝑛𝑛

with 𝑛𝑛 = 0.5.

Yield stress

The yield stress (Pa) can be, in a simplistic way, defined as the stress at which a
material begins to flow. In the figure above, it is given by the point 𝜎𝜎(𝛾𝛾̇ = 0). For
Newtonian fluids, as soon as a shear rate is imposed, the material starts to respond,
therefore the yield stress is zero for a Newtonian fluid. Other types of liquids (see
ideal Bingham fluid) needs a certain amount of stress in order to start to flow. One
can define their yield stress by incrementally increasing the shear stress until their
shear rate is changing to a non-zero value. One can also, in the case of a Bingham
fluid, extrapolate the 𝜎𝜎(𝛾𝛾̇ ) line to find 𝜎𝜎(𝛾𝛾̇ = 0).

A question that arises is how to properly define the yield stress, as obviously it will
depend on several factors:

1 – the shear history: has the sample been stirred before? One can imagine for
example a mud that was undisturbed for a long time. Its yield stress will be higher
than a mud that has just been stirred.

2 – the rate at which the shear stress is applied: if the shear stress increments are
done slowly, the liquid would have time to adapt and its yield stress would be
different than when the increments are done fast.

When yield stress experiments are performed, both these points should therefore
be addressed. The change in fluid properties, like shear thinning and shear
thickening, will be discussed later, as function of the colloidal interactions occurring
in the suspensions.

In the case of suspensions which exhibit a yield stress for moderate volume fractions,
one can use this yield stress 𝜎𝜎0 to estimate the “amount” of colloidal forces within
the suspension:

𝜎𝜎0 ~ 𝑁𝑁 × 𝐹𝐹𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏

where 𝑁𝑁 is the number of bonds per unit area between the particles and 𝐹𝐹𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 the
mean force of a bond. We make the hypothesis that the network breakage requires
the rupture of almost every bond. The amount of hydrodynamic effects can be
estimated through the stress of the equivalent suspension of non-interacting
particles:

𝜎𝜎(𝜙𝜙𝑆𝑆 ) = 𝜂𝜂(𝜙𝜙𝑆𝑆 )𝛾𝛾̇

162
 Chapter 7 Rheological behaviour of colloidal suspensions

where 𝜙𝜙𝑆𝑆 is the volume fraction of particles, 𝜂𝜂(𝜙𝜙𝑆𝑆 ) is the viscosity of a force-free
particle suspension and 𝛾𝛾̇ is the shear rate. Examples of 𝜂𝜂(𝜙𝜙𝑆𝑆 ) are given in the next
sections. The ratio of the hydrodynamic to colloidal stresses is given by:

𝜂𝜂𝛾𝛾̇
Γ=
𝜎𝜎0

Formally Γ −1 is called a Bingham number. In general 𝜎𝜎0 rapidly increases with 𝜙𝜙𝑆𝑆 :
for concentrated clay-water mixtures 𝜎𝜎0 is found to follow an exponential law :

Yield stress of mud suspensions as a function of the volume fraction of particles 79

For fine mud suspensions the range of Γ for the transition from a colloidal regime
to a hydrodynamic one is found to be [0.3 - 50].

We will see in the last section how 𝜎𝜎0 can be linked to DLVO forces, which were
introduced in Chapter 3.

Viscoelastic fluids
Viscoelastic fluids combine the properties of elastic solids and those of viscous fluids.
This is in particular true for densely flocculated suspensions. Ideal plastic material
always return to their non-deformed state when the stresses are released. These
materials are said to have a perfect “memory” of their non-deformed reference
configuration. On the other hand, liquids have no memory at all, and when the stress
is released, they remain in their last position. Energetically, the work done in an
elastic deformation is stored in the material as potential energy and can be totally
recovered when the material returns to its non-deformed state.

79Coussot, P. (1995). Structural similarity and transition from Newtonian to non-Newtonian

behavior for clay-water suspensions. Physical review letters, 74(20), 3971.

163
 Introduction to Colloid Science

In order to study the viscoelastic properties of fluids, oscillatory shear flows are used.
If, instead of having a constant speed, the upper plate executes a sinusoidal motion,
this will induce a sinusoidal deformation or strain in the sample. Let the oscillatory
motion of the upper plate be:

𝑥𝑥(𝑡𝑡) = 𝑥𝑥0 cos(𝜔𝜔𝜔𝜔)

where 𝑥𝑥0 is the peak displacement and 𝜔𝜔 the frequency of the oscillation. The
deformation (strain) of the sample can then be expressed as:

𝛾𝛾(𝑡𝑡) = 𝛾𝛾0 cos(𝜔𝜔𝜔𝜔)

with 𝛾𝛾0 = 𝑥𝑥0 /𝐿𝐿 where 𝐿𝐿 is the distance between the two plates. In the case of an
ideal elastic sample, the shear stress should follow the shear strain deformation in
agreement with Hooke’s law:

𝜎𝜎(𝑡𝑡) = 𝐺𝐺𝐺𝐺(𝑡𝑡)

where the shear modulus G (Pa) is defined as the ratio between shear stress to shear
strain. Stress 𝜎𝜎 and strain 𝛾𝛾 are then in phase. The stress for a viscous liquid, on the
other hand, depends, as we have seen, on the shear rate 𝛾𝛾̇ . We then have:

𝛾𝛾̇ (𝑡𝑡) = −𝜔𝜔𝜔𝜔0 sin(𝜔𝜔𝜔𝜔) = 𝜔𝜔𝜔𝜔0 cos(𝜔𝜔𝜔𝜔 + 𝜋𝜋/2)

If we assume that the liquid is Newtonian we have 𝜎𝜎 = 𝜂𝜂𝛾𝛾̇ and therefore stress 𝜎𝜎
and strain 𝛾𝛾 are 𝜋𝜋/2 out of phase. Viscoelastic fluids have a phase shift between
stress 𝜎𝜎 and strain 𝛾𝛾 that is between 0 and 𝜋𝜋/2. A generalized Hooke’s law can be
defined for these kind of fluids:

𝜎𝜎(𝑡𝑡) = 𝛾𝛾0 [𝐺𝐺 ′ cos(𝜔𝜔𝜔𝜔) − 𝐺𝐺 ′′ sin(𝜔𝜔𝜔𝜔)]

The in-phase (real part) 𝐺𝐺 ′ describes the elastic component of the stress. It is called
the storage modulus. The out-of-phase (imaginary part) 𝐺𝐺 ′′ represents the viscous
part. It is called the loss modulus. Clearly, for pure elastic samples, 𝐺𝐺 ′′ = 0 and 𝐺𝐺 ′ =
𝐺𝐺 and the stress 𝜎𝜎 is given by 𝜎𝜎 = 𝐺𝐺𝐺𝐺. For pure Newtonian viscous samples 𝐺𝐺 ′ = 0
and 𝜎𝜎 = −𝛾𝛾0 𝐺𝐺 ′′ sin(𝜔𝜔𝜔𝜔)=−𝜂𝜂𝜔𝜔𝛾𝛾0 sin(𝜔𝜔𝜔𝜔), from which we deduce that 𝐺𝐺 ′′ = 𝜂𝜂𝜂𝜂. The
phase angle between stress and strain determines how much mechanical energy will
be dissipated in heat and is referred to as loss angle 𝛿𝛿:

tan 𝛿𝛿 = 𝐺𝐺 ′′ /𝐺𝐺 ′

164
 Chapter 7 Rheological behaviour of colloidal suspensions

Maxwell model

A viscoelastic fluid having a Newtonian loss modulus and a Hooke storage modulus
can be represented by a pure viscous damper (symbol 𝜂𝜂) and a purely elastic spring
(symbol 𝐺𝐺) connected in series 80. This model is called after Maxwell, who proposed
it in 1867:

The total stress and total strain can be defined as:

𝜎𝜎 = 𝜎𝜎𝜂𝜂 = 𝜎𝜎𝐺𝐺

𝛾𝛾 = 𝛾𝛾𝜂𝜂 + 𝛾𝛾𝐺𝐺

[note the analogy with electronic (equivalent) circuits where 𝜎𝜎 would represent a
current and 𝛾𝛾 a voltage]. One obtains:

𝜎𝜎 1 𝑑𝑑𝑑𝑑 𝜎𝜎 1
𝛾𝛾̇ = + = + 𝜎𝜎̇
𝜂𝜂 𝐺𝐺 𝑑𝑑𝑑𝑑 𝜂𝜂 𝐺𝐺

where we used 𝜎𝜎𝜂𝜂 = 𝜂𝜂𝛾𝛾̇𝜂𝜂 as for Newton fluids for the pure viscous damper, and 𝜎𝜎𝐺𝐺 =
𝐺𝐺𝛾𝛾𝐺𝐺 for a purely elastic spring. From that equation a characteristic relaxation time
can be estimated: 𝜏𝜏 = 𝜂𝜂/𝐺𝐺. If we adopt the complex representation such that 𝛾𝛾 ∗ =
𝛾𝛾 + 𝑖𝑖 Im(𝛾𝛾 ∗ ) = 𝛾𝛾0 exp(−𝑖𝑖𝑖𝑖𝑖𝑖) and 𝜎𝜎 ∗ = 𝜎𝜎 + 𝑖𝑖 Im(𝜎𝜎 ∗ ) = 𝜎𝜎0 exp(−𝑖𝑖𝑖𝑖𝑖𝑖), one can
rewrite the previous equation as:

𝜎𝜎 ∗ 1 ∗
𝛾𝛾̇ ∗ = + 𝜎𝜎̇
𝜂𝜂 𝐺𝐺

Taking the real part of both sides of this equation will give back the original equation.
By substitution we find:

1 𝑖𝑖𝑖𝑖 1 − 𝑖𝑖𝑖𝑖𝜏𝜏 ∗
𝛾𝛾̇ ∗ = −𝑖𝑖𝑖𝑖𝛾𝛾 ∗ = � − � 𝜎𝜎 ∗ = 𝜎𝜎
𝜂𝜂 𝐺𝐺 𝜂𝜂

From which we deduce that:

80There exist also a model where the damper and the spring are in parallel. That model is
called the Kelvin–Voigt model. This model is used to describe a simple viscoelastic solid.

165
 Introduction to Colloid Science

−𝑖𝑖𝑖𝑖𝑖𝑖 ∗
𝜎𝜎 ∗ = 𝛾𝛾
1 − 𝑖𝑖𝑖𝑖𝑖𝑖
A complex shear modulus 𝐺𝐺 ∗ can be defined such that:

𝜎𝜎 ∗ = 𝐺𝐺 ∗ 𝛾𝛾 ∗

with

𝐺𝐺 ∗ = 𝐺𝐺 ′ − 𝑖𝑖𝐺𝐺 ′′

We find from the previous relations:

𝜔𝜔2 𝜂𝜂𝜂𝜂 𝜔𝜔𝜔𝜔

𝐺𝐺 ′ − 𝑖𝑖𝐺𝐺 ′′ = 2 2
− 𝑖𝑖
1 + 𝜔𝜔 𝜏𝜏 1 + 𝜔𝜔 2 𝜏𝜏 2
Often, a complex viscosity 𝜂𝜂 ∗ is also defined such that:

𝐺𝐺 ′′ 𝐺𝐺 ′
𝜂𝜂 ∗ = 𝜂𝜂 ′ + 𝑖𝑖𝜂𝜂 ′′ = + 𝑖𝑖
𝜔𝜔 𝜔𝜔
which leads to:
𝜂𝜂
𝜂𝜂 ′ =
1 + 𝜔𝜔 2 𝜏𝜏 2
𝜂𝜂𝜂𝜂𝜂𝜂
𝜂𝜂 ′′ =
1 + 𝜔𝜔 2 𝜏𝜏 2
From which we can deduce that

𝐺𝐺 ′′ = 𝜂𝜂 ′ 𝜔𝜔

For a pure viscous liquid 𝜏𝜏 = 𝜂𝜂/𝐺𝐺 → 0 and 𝐺𝐺 ′′ = 𝜂𝜂𝜂𝜂 as discussed above. We have
now found for the complex stress/strain relation that

𝜔𝜔2 𝜂𝜂𝜂𝜂 𝜔𝜔𝜔𝜔

𝜎𝜎 ∗ = � 2 2
− 𝑖𝑖 � 𝛾𝛾 ∗
1 + 𝜔𝜔 𝜏𝜏 1 + 𝜔𝜔 2 𝜏𝜏 2

Taking the real part gives:

𝜔𝜔2 𝜂𝜂𝜂𝜂 𝜔𝜔𝜔𝜔

𝜎𝜎(𝑡𝑡) = 𝛾𝛾 cos(𝜔𝜔𝜔𝜔) − 𝛾𝛾 sin(𝜔𝜔𝜔𝜔)
1 + 𝜔𝜔 2 𝜏𝜏 2 0 1 + 𝜔𝜔 2 𝜏𝜏 2 0
From 𝜎𝜎(𝑡𝑡) = 𝛾𝛾0 [𝐺𝐺 ′ cos(𝜔𝜔𝜔𝜔) − 𝐺𝐺 ′′ sin(𝜔𝜔𝜔𝜔)] we deduce that:

𝜔𝜔2 𝜏𝜏 2
𝐺𝐺 ′ = 𝐺𝐺
1 + 𝜔𝜔 2 𝜏𝜏 2

166
 Chapter 7 Rheological behaviour of colloidal suspensions

𝜔𝜔𝜔𝜔
𝐺𝐺 ′′ = 𝐺𝐺
1 + 𝜔𝜔 2 𝜏𝜏 2

These functions are plotted here:

1
1 10

0.9
0
0.8 10

0.7

dimentionless moduli
G1/G
dimentionless moduli

-1
G2/G 10
0.6
apparent viscosity
0.5
-2
0.4 10

0.3
-3 G1/G
0.2 10
G2/G
0.1 apparent viscosity
-4
0 10
-2 -1 0 1 2
0 1 2 3 4 5 10 10 10 10 10
ωt ωt

Left: Linear plot of the dimensionless storage modulus 𝐺𝐺 ′ /𝐺𝐺 (blue line), loss modulus 𝐺𝐺 ′′ /𝐺𝐺
(blue dashed line) and apparent viscosity 𝜂𝜂′ /𝜂𝜂. Right: the same, but in log-log scale.

From the graphs it is clear that 𝜏𝜏 = 𝜂𝜂/𝐺𝐺 represents the time where one goes from a
viscous-dominated fluid (𝜂𝜂′ ~ 𝜂𝜂) to an elastic-dominated fluid (𝐺𝐺 ′ ~ 𝐺𝐺). The loss
modulus is highest at that transition.

Couette rheometer

A rheometer is a laboratory device used to measure how a liquid, suspension or

slurry flows in response to applied forces. One of the most used rheometer is the
concentric cylinder rheometer (also called Couette rheometer, after its inventor, the
physicist Maurice Couette (1858-1943)). Other types of geometry (plate-plate, which
was discussed at the beginning of the chapter or cone-plate,…) do also exist. In a
Couette rheometer a torque M (thus the shear stress) is imposed and the rotation
speed (thus the shear rate) is recorded. This type of rheometer is said to be stress-
controlled. Strain-controlled rheometers also exist.

We assume that the outer cylinder remains fixed and that the inner cylinder rotates
with an angular velocity 𝜔𝜔 = 𝑑𝑑𝑑𝑑/𝑑𝑑𝑑𝑑. The shear stress 𝜎𝜎(𝑟𝑟) represents the friction
between two layers located on either side of the distance 𝑟𝑟. The torque M on the
inner cylinder is equal to the force due to friction at the distance 𝑟𝑟 multiplied by 𝑟𝑟
as:

𝑴𝑴 = 𝑟𝑟𝒆𝒆𝒓𝒓 × 𝐹𝐹𝒆𝒆𝜽𝜽

𝑀𝑀 = 𝑟𝑟𝑟𝑟 sin(𝒆𝒆𝒓𝒓 , 𝒆𝒆𝜽𝜽 )

𝑀𝑀 = 𝑟𝑟𝑟𝑟(𝑟𝑟)

167
 Introduction to Colloid Science

The stress at the distance 𝑟𝑟 is therefore linked to the torque by:

𝐹𝐹(𝑟𝑟) 𝑀𝑀
𝜎𝜎(𝑟𝑟) = =
2𝜋𝜋𝜋𝜋ℎ 2𝜋𝜋𝑟𝑟 2 ℎ
If we assume that 𝑅𝑅𝑜𝑜𝑜𝑜𝑜𝑜 /𝑅𝑅𝑖𝑖𝑖𝑖 is close to one (which is the case in standard
rheometers), then both 𝜎𝜎 and 𝛾𝛾̇ are independent of the position in the gap between
the two cylinders. It is then easy to show that the shear rate is proportional to the
angular velocity.

𝑑𝑑𝑣𝑣𝜃𝜃 𝑣𝑣𝜃𝜃 𝑅𝑅𝑖𝑖𝑖𝑖 𝜔𝜔

𝛾𝛾̇ = = =
𝑑𝑑𝑑𝑑 (𝑅𝑅𝑜𝑜𝑜𝑜𝑜𝑜 − 𝑅𝑅𝑖𝑖𝑖𝑖 ) (𝑅𝑅𝑜𝑜𝑜𝑜𝑜𝑜 − 𝑅𝑅𝑖𝑖𝑖𝑖 )

Left: illustration of a rheometer; usually the ratio 𝑅𝑅𝑜𝑜𝑜𝑜𝑜𝑜 /𝑅𝑅𝑖𝑖𝑖𝑖 is very close to 1; Right:
concentric inner and outer cylinders with cylindrical coordinates

Rheology of suspensions
Until now, we have reviewed general principles about rheology. We did not yet
discuss the microscopic structure of these fluids, and considered the fluids as a
continuum, having bulk properties. Fluids displaying “real” continuum properties are
for instance water or oil, i.e. fluids composed of the same type of molecules. In the
remainder of the chapter we will discuss the rheological properties of suspensions
as function of the properties of the solvent and colloidal particles that form these
fluids. To start, we give an example 81 of the flow behaviour of a diluted clay
suspension in a 1-1 electrolyte:

81 See Handbook of Clay Science Edited by F. Bergaya, B.K.G. Theng and G. Lagaly

Developments in Clay Science, Vol. 1, Chapter 5

168
 Chapter 7 Rheological behaviour of colloidal suspensions

A: isolated particles; B: minimum of rheological properties (viscosity, yield stress) due to

electroviscous effect; C,D: aggregation in the form of networks; E,F: fragmentation of the
networks at high salt concentrations.

The effect of salt concentration can be understood as follows: both yield stress and
viscosity are low at no added salt, as the clay platelets are just suspended in water,
so the rheological properties measured should be close to (a bit larger than) the one
of water (as we consider a dilute suspension). A minimum is observed at low salt
concentration (in B) as a consequence of the secondary electroviscous effect: the
double layers around the particles are sufficiently compressed so as to ensure that
the particles do not “feel” each other over significant distances, so they can flow
better. When the amount of added salt becomes significant, aggregation occurs and
large aggregates are formed which hinder the flow, leading to an increase in viscosity
and yield stress. At even higher salt concentration, several scenarios are possible:
(D): the aggregates are very strong, and remain as they are, and the rheological
properties do not change upon addition of salt, (E,F): some delamination occurs, and
the size of the aggregates is reduced, leading to a reduction in rheological properties.

Dilute and semi-dilute suspensions

In his doctoral thesis Albert Einstein derived a relation between the viscosity of the
suspension and the volume fraction of the non-interacting spherical colloidal
particles inside the suspension. He found (after a colleague, Hopf, corrected a
mistake 82), that:

82 Einstein had found η = η (1 + ϕ) which would be the correct expression if the suspension
0
would be made of droplets or gas bubbles for which there is no friction at the particle’s surface

169
 Introduction to Colloid Science

𝜂𝜂 = 𝜂𝜂0 (1 + 2.5𝜙𝜙𝑠𝑠 )

where 𝜂𝜂0 is the viscosity of the fluid when there is no particles and 𝜙𝜙𝑠𝑠 the volume
fraction of particles. Even though the relation looks simple, it is quite difficult to
derive. The result is obtained by calculating the energy dissipation in a sphere of
radius R around the reference sphere and let R go to infinity. A rigorous
mathematical derivation has been provided by Batchelor and Green, who even
extended Einstein’s formula for interacting particles. An alternative (simpler)
derivation is outlined in the book of Theo van de Ven (Colloidal Hydrodynamics, cited
in the reference list), based on the work of Happel and Brenner. The extended
Einstein formula can be written:

𝜂𝜂 = 𝜂𝜂0 �1 + 2.5𝜙𝜙𝑠𝑠 + 𝑐𝑐2 𝜙𝜙𝑠𝑠 2 �

where 𝑐𝑐2 is a numerical value. The term 𝜙𝜙𝑠𝑠 2 behind reflects the fact that 𝑐𝑐2 should
account for particle-particle interactions. Depending on the assumptions made, 𝑐𝑐2
can therefore take different values.

Batchelor and the development of colloidal hydrodynamics

Batchelor and Green in an important article 83 have found the second order
correction to Einstein’s relation:

𝜂𝜂 = 𝜂𝜂0 �1 + 2.5𝜙𝜙𝑠𝑠 + 7.6𝜙𝜙𝑠𝑠 2 �

implying that 𝑐𝑐2 = 7.6. This was in the case where they considered an extensional
flow. For a shear flow of a semi-dilute suspensions of hard spheres with no Brownian
motion (i.e. large particles), assuming a random particle distribution they found 𝑐𝑐2 =
5.2.

The calculation of the 𝜙𝜙𝑠𝑠 2 term requires evaluating diverging integrals which can be
evaluated by using special mathematical techniques. Batchelor extended Einstein’s
original argument to show that Brownian motion in a suspension with two or more
particles can be represented as a statistical thermodynamic force.

(free slip). The other contribution to the viscosity comes from the distortion of the flow lines,
which exist in any case, due to the presence of the particles.
83 Batchelor, G. K., & Green, J. T. (1972). The determination of the bulk stress in a suspension

of spherical particles to order c2. Journal of Fluid Mechanics, 56(03), 401-427.

170
 Chapter 7 Rheological behaviour of colloidal suspensions

George Keith Batchelor FRS (1920 – 2000) was an

Australian applied mathematician and fluid dynamicist.
He was for many years the Professor of Applied
Mathematics in the University of Cambridge, and was
founding head of the Department of Applied
Mathematics and Theoretical Physics (DAMTP). In 1956
he founded the influential Journal of Fluid Mechanics
which he edited for some forty years. Prior to Cambridge
he studied in Melbourne High School.

As an applied mathematician (and for some years at

Cambridge a co-worker of Sir Geoffrey Taylor in the field of turbulent flow), he was
a keen advocate of the need for physical understanding and sound experimental
basis. His An Introduction to Fluid Dynamics (CUP, 1967) is still considered a classic
of the subject.

Batchelor also contributed greatly to make Kolmogorov ‘s theory of turbulence

understandable, by publishing 3 articles where he explains this theory in detail. In
the 1950/60s even students in the Soviet Union used Batchelor’s articles as an
introduction to the subject. He worked in Cambridge with G.I. Taylor, a British
physicist and mathematician and major figure in fluid dynamics and wave theory,
whose interest moved from turbulence to other fields, so the work of Batchelor in
turbulence was to a large extend independent. Realizing that he needed
experiments to improve his theoretical work, Batchelor wrote to an Australian
colleague and close friend A.A. Towsend “ You will come to Cambridge, study
turbulence and work with G.I. Taylor”. The answer came immediately: “ I agree, but
I have two questions: what is turbulence and who is G.I. Taylor?”. Townsend came
and soon revealed himself as one of the most remarkable experimentalists working
in turbulence.

Among the students of Batchelor, some of them made substantial contributions to

colloidal science, Edward Hinch and Richard Wyngham O’Brien in particular for their
work on electro-rheology. O’ Brien was at the origin of the numerical code that
enables to calculate the frequency-dependent electrophoretic mobility and
electrokinetic response of charged colloids (for all 𝜅𝜅𝜅𝜅 and 𝜁𝜁 potentials). He also
developed the theory of electroacoustics for colloids, and is at the origin of the
Electrokinetic Sonic Amplitude (ESA) measurement device, which enables to probe
the response of concentrated slurries.

When Brownian motion is taken into account 84, then an extra term accounting for
the stresses generated in the dispersion by the random movements of the particles

84 Mendoza, C. I., & Santamaria-Holek, I. (2009). The rheology of hard sphere suspensions at

arbitrary volume fractions: An improved differential viscosity model. The Journal of chemical
physics, 130(4), 044904.

171
 Introduction to Colloid Science

should be added. Numerical calculations have then showed that 𝑐𝑐2 = 6.17.
Depending on the assumptions made about the hydrodynamic flow and the particle-
particle interactions, one finds that in general that 5.00 ≤ 𝑐𝑐2 ≤ 6.17.

The extended Einstein formula is valid primarily at low volume fractions and low
shear frequencies (low ω). As many industrial applications concern non-diluted
suspensions at any shear rate, many semi-empirical formulae have been proposed
to fit experiments in the largest possible range of volume fractions.

Krieger – Dougherty model (for a large range of volume fraction)

This model was developed as an extension of Einstein’s model for higher volume
fractions. It is derived as follows. The addition of particles to a Newtonian medium
will increases the viscosity:

𝜂𝜂(𝜙𝜙𝑠𝑠 + 𝑑𝑑𝜙𝜙𝑠𝑠 ) = 𝜂𝜂(𝜙𝜙𝑠𝑠 ) + 𝑑𝑑𝑑𝑑

By analogy with Einstein’s formula, one would expect to have:

𝑑𝑑𝑑𝑑 = 2.5𝜂𝜂(𝜙𝜙𝑠𝑠 )𝑑𝑑𝜙𝜙𝑠𝑠

However, by adding particles, the space available is not the entire volume, but the
volume reduced by a factor proportional to the current volume fraction. We define
𝜙𝜙𝑐𝑐 as the filling fraction at maximum packing, and therefore obtain that:

2.5𝑑𝑑𝜙𝜙𝑠𝑠
𝑑𝑑𝑑𝑑 = 𝜂𝜂(𝜙𝜙𝑠𝑠 )
1 − 𝜙𝜙𝑠𝑠 /𝜙𝜙𝑐𝑐

One can verify that for low volume fraction one has indeed 𝑑𝑑𝑑𝑑 ~ 2.5𝜂𝜂(𝜙𝜙𝑠𝑠 )𝑑𝑑𝜙𝜙𝑠𝑠
whereas for 𝜙𝜙𝑠𝑠 → 𝜙𝜙𝑐𝑐 one gets 𝑑𝑑𝑑𝑑 → ∞ which is expected, as at maximum packing
the fluid cannot flow anymore.

We now have obtained:

𝜂𝜂(𝜙𝜙𝑠𝑠 + 𝑑𝑑𝜙𝜙𝑠𝑠 ) − 𝜂𝜂(𝜙𝜙𝑠𝑠 ) 𝑑𝑑𝑑𝑑 2.5𝑑𝑑𝜙𝜙𝑠𝑠

= = 𝑑𝑑�ln(𝜂𝜂)� = = −2.5𝜙𝜙𝑐𝑐 𝑑𝑑�ln(1 − 𝜙𝜙𝑠𝑠 /𝜙𝜙𝑐𝑐 )�
𝜂𝜂(𝜙𝜙𝑠𝑠 ) 𝜂𝜂 1 − 𝜙𝜙𝑠𝑠 /𝜙𝜙𝑐𝑐

which can be integrated:

ln�𝜂𝜂(𝜙𝜙𝑠𝑠 )� ln(1−𝜙𝜙𝑠𝑠 /𝜙𝜙𝑐𝑐 )

� 𝑑𝑑�ln(𝜂𝜂)� = −2.5𝜙𝜙𝑐𝑐 � 𝑑𝑑�ln(1 − 𝜙𝜙𝑠𝑠 /𝜙𝜙𝑐𝑐 )�
ln(𝜂𝜂0 ) ln(1)=0

This gives:

𝜂𝜂(𝜙𝜙𝑠𝑠 ) 1 − 𝜙𝜙𝑠𝑠 /𝜙𝜙𝑐𝑐

ln � � = −2.5𝜙𝜙𝑐𝑐 ln � �
𝜂𝜂0 1

172
 Chapter 7 Rheological behaviour of colloidal suspensions

from which we obtain the Krieger-Dougherty relation:

𝜙𝜙𝑠𝑠 −2.5𝜙𝜙𝑐𝑐
𝜂𝜂(𝜙𝜙𝑠𝑠 ) = 𝜂𝜂0 �1 − �
𝜙𝜙𝑐𝑐

This relation agrees reasonably well with the experimental data. Moreover, it
reduces to the Einstein’s equation for low 𝜙𝜙𝑠𝑠 as one can verify easily using the
relation (1 + 𝑥𝑥)𝑛𝑛 = 1 + 𝑛𝑛𝑛𝑛 which holds for small 𝑥𝑥.

Other models for a large range of volume fraction

Several authors have extended the Einstein relation by using an effective filling
fraction:

𝜙𝜙𝑠𝑠
𝜙𝜙eff =
1 − 𝑐𝑐𝜙𝜙𝑠𝑠

where the constant c takes into account the fact that the complete free volume
cannot be filled by the spheres. They obtain:

𝜂𝜂 = 𝜂𝜂0 (1 + 2.5𝜙𝜙eff )

using the same arguments as the ones presented in the derivation of the Krieger-
Dougherty relation, it is possible to find:

𝜂𝜂(𝜙𝜙𝑠𝑠 ) = 𝜂𝜂0 (1 − 𝜙𝜙eff )−2.5

This formula reduces to Einstein formula for low 𝜙𝜙𝑠𝑠 where 𝜙𝜙eff ~ 𝜙𝜙𝑠𝑠 and for the
critical packing 𝜙𝜙𝑐𝑐 one has 𝜙𝜙eff = 1. This last equation enables to find a relation for
the constant c:

1 − 𝜙𝜙𝑐𝑐
𝑐𝑐 =
𝜙𝜙𝑐𝑐

Here we give a plot of the theoretical behaviour of the functions we have described:

173
 Introduction to Colloid Science

3
10
Einstein
c2 = 6.17 Dimensionless viscosity 𝜂𝜂/𝜂𝜂0 as
K-D function of volume fraction following
φ eff (1)
the models:
dimentionless viscosity

2
10
φ eff (2)
2
𝜂𝜂 = 𝜂𝜂0 �1 + 2.5𝜙𝜙𝑠𝑠 + 𝑐𝑐2 𝜙𝜙𝑠𝑠 � where
𝑐𝑐2 = 0 (Einstein), 𝑐𝑐2 = 6.17, Krieger-
1
10 Dougherty relation (K-D),
𝜂𝜂 = 𝜂𝜂0 (1 + 2.5𝜙𝜙eff ) (called 𝜙𝜙eff (1))
and 𝜂𝜂(𝜙𝜙) = 𝜂𝜂0 (1 − 𝜙𝜙eff )−2.5 (called
0 𝜙𝜙eff (2)). We have taken 𝜙𝜙𝑐𝑐 = 0.74.
10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
φ

In practice, it appears that relations as K-D and 𝜙𝜙eff (2) are close to the
experimental data. Here we give an example with measurements on silica colloidal
particles of about 75 nm in radius dispersed in cyclohexane (an oil that enables to
nearly suppress all electrostatic and van der Waals interactions and ensures that the
suspension remains stable). The data is taken from an old article 85 and was plotted
with a linear-linear scale, which made it difficult to accurately estimate the points at
low volume fraction. This is the reason of the scatter of the data points at low volume
fraction.

3
10
φ c =0.63
φ c = 0.74
dimentionless viscosity

2
10

1
10

0
10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
φ

Dimensionless viscosity 𝜂𝜂/𝜂𝜂0 as function of volume fraction; symbols indicate measurements

and lines are plotted according to the model: 𝜂𝜂�𝜙𝜙𝑠𝑠 � = 𝜂𝜂0 (1 − 𝜙𝜙eff )−2.5 .

An important factor to account for, which makes a universal theoretical (non-

empirical) model for the whole volume fraction range quite impossible to realize is

85 de Kruif, C. D., Van Iersel, E. M. F., Vrij, A., & Russel, W. B. (1985). Hard sphere colloidal
dispersions: Viscosity as a function of shear rate and volume fraction. The Journal of chemical
physics, 83(9), 4717-4725.

174
 Chapter 7 Rheological behaviour of colloidal suspensions

that phase transitions can occur in the suspension (without any aggregation). This
has already been briefly discussed in Chapter 2, in the context of particle settling. In
Chapter 2 we have given the example of anisotropic particles, but phase transitions
also occur for spherical (isotropic) particles. This is discussed next.

Phase transition
The phase transition for (spherical) colloidal particles can be schematized as follows:

Hard-sphere phase diagram constructed from light diffraction measurements

Confocal micrographs of different phases in a colloidal suspension with 5%

polydispersity. Scale bar is 10 μm. From left to right: liquid, glass and crystal phase.

At low volume fraction the particles can diffuse freely and there is no long-range
ordering in particle position implying that the dispersion is in a fluid state. Increasing
concentration above 𝜙𝜙𝑠𝑠 = 0.5 liquid and crystalline phases coexist in equilibrium
and the fraction of crystalline phase increases until the sample is fully crystalline at
𝜙𝜙𝑠𝑠 = 0.55. With increasing volume fraction, the particle mobility is dramatically
reduced until the glassy state is reached. In a glassy state, the dispersion does not
flow anymore. One can make the analogy with a (liquid) glass window that solidifies
when the temperature is decreased 86, except that decreasing temperature is now

86 The fact that glass panels do flow after hundreds of years (see the bottom of stained glass

panels in old churches) cannot be explained from the simple statement that this is due to
gravity. If only gravity would play a role, glass panels would have significantly changed their

175
 Introduction to Colloid Science

equivalent with increasing volume fraction. Note that the existence of the glassy
state requires some polydispersity (at least some %) otherwise the system will
directly crystallise 87.

Some close packings happening in crystals and colloidal crystals

The upper limit for close packing (hexagonal close packing or face-centered cubic,
i.e. fcc) is given by 𝜙𝜙𝑐𝑐 = 𝜋𝜋/(3√2) = 0.74. Other close packings are the body-
centered cubic packing where 𝜙𝜙𝑐𝑐 = 𝜋𝜋√3/8 = 0.68, simple cubic packing where
𝜙𝜙𝑐𝑐 = 0.52.

We here briefly recall how these values are found with taking the
simple cubic as an example. In a simple cubic the particles are
arranged as in the figure. For one element of fluid (the black cube
– that we assume to have a length R), one can see that there is
1/8th of a colloidal particle (of radius a) in each corner that is
inside the cube. In total there is therefore 8 × (1/8) = one
particle’s volume inside the cube. In close packing, the particles are touching,
implying that 𝑅𝑅 = 2𝑎𝑎. The volume fraction can therefore be calculated by:

4𝜋𝜋𝑎𝑎3 /3 4𝜋𝜋𝑎𝑎3 /3 𝜋𝜋
𝜙𝜙𝑐𝑐 = = = = 0.52
𝑅𝑅3 (2𝑎𝑎)3 6

bottom thickness only after a period larger than the age of the universe.
(http://engineering.mit.edu/ask/how-does-glass-change-over-time)
87 Hunter, G. L., & Weeks, E. R. (2012). The physics of the colloidal glass transition. Reports on

progress in physics, 75(6), 066501 and https://arxiv.org/pdf/1106.3581.pdf

176
 Chapter 7 Rheological behaviour of colloidal suspensions

The other close packings are found in the same way. Close packings are usually used
to describe the geometry of crystals, made of atoms. As said in Chapter 2, this is why
colloidal suspensions are often used as model for crystals and are termed colloidal
crystals.

The transition from one packing order to another one can be observed
experimentally. The measurements presented here (right figure) are done by
torsional resonance spectroscopy, a method that enables to get the shear modulus
at very low frequencies using acoustic waves 88.

Conductivity (left) and shear modulus (right) as function of volume fraction, for spheres of a =
60 nm radius that are highly charged and dispersed in de-ionized (i.e. very pure) water. The
x-axis represents the number of spheres per μm3

Given the size of the particles the number of particles can easily be converted in
volume fraction. To give an idea, 1 particle/μm3 corresponds to 𝜙𝜙𝑠𝑠 = 10−3 . The fact
that these particles crystallise at such low volume fractions is linked to their charge.

Note that the conductivity measurements do not enable to give information about
the packing transition. However, the slope of the line can be used to determine the
charge of the spheres, according to the relation:

𝐾𝐾 = 𝑛𝑛(𝑍𝑍𝑍𝑍)2 �𝜇𝜇𝑝𝑝 + 𝜇𝜇𝐻𝐻 + � + 𝐾𝐾𝐵𝐵

where 𝑒𝑒 is the elementary charge, 𝜇𝜇𝑝𝑝 is the independently measured particle

mobility (from electrophoresis for example), 𝜇𝜇𝐻𝐻 + the electrophoretic mobility of H+
that can be found in handbooks from the limiting conductivity (see Chapter 3) and
Z is the valence of the particle. 𝐾𝐾𝐵𝐵 represents the background conductivity from the
self-dissociation of water and residual impurities. As the spheres are dispersed in

88Wette, P., Schöpe, H. J., & Palberg, T. (2003). Experimental determination of effective
charges in aqueous suspensions of colloidal spheres. Colloids and Surfaces A: Physicochemical
and Engineering Aspects, 222(1), 311-321.

177
 Introduction to Colloid Science

ultra-pure water, effectively only their counterions (H+) are present in the water, and
the total amount of counterion charges should be equal to the total amount of
charges on the colloidal particles, because of electroneutrality. In the present case,
one finds that 𝑍𝑍 = 685, which corresponds to a surface charge of 𝑍𝑍𝑒𝑒/(4𝜋𝜋𝑎𝑎2 ) = 0.24
C/m2.

The pair correlation function

In the next section we will be using a function called pair correlation function 𝑔𝑔(𝑟𝑟)
which is the radial distribution that describe positional correlations among particles
in the equilibrium fluid. One of the experimental way to assess the pair correlation
function 𝑔𝑔(𝑟𝑟) is to use (confocal) microscopic imaging. The positions of the particles
are then determined (usually with an accuracy of 0.05 μm) from the image analysis.
For small particles, neutron or x-ray scattering is used. The probability of finding
particles separated by r is then obtained, which is directly linked to 𝑔𝑔(𝑟𝑟). The
function 𝑔𝑔(𝑟𝑟) describe how density varies as a function of distance from a reference
particle. It is a normalized function, implying that 𝑔𝑔(𝑟𝑟) = 1 when the density is equal
to the bulk density. Let us consider a reference particle (black) and its neighbours (in
colour):

At position r/(2a) = 1 (the first peak), the probability to find neighbours is very high.
These are the nearest neighbours that are touching the reference particle (r = 2a).
Some minuscule distance away from these nearest neighbours there is statistically
no particles (if there would be, they would overlap with the nearest neighbours,
which is physically not possible), therefore one observes a dip between r/(2a) = 1
and r/(2a) = 2. At r/(2a) = 2 there is a next crowd of neighbours, and this continues
until we are so far from the particles that the correlation becomes very weak and
𝑔𝑔(𝑟𝑟) = 1. In an ideal gas, there is no relevant correlation between neighbouring
particles, which implies that 𝑔𝑔(𝑟𝑟) = 1 all the time.

A typical example of 𝑔𝑔(𝑟𝑟) is given here:

178
 Chapter 7 Rheological behaviour of colloidal suspensions

3.5
2 mM
3 12.5 mM
50 mM
2.5

g(r)
1.5

0.5

0
1 2 3 4 5
r/(2a)

g(r) for 100 nm particles with a surface potential of 25 mV and dispersed in a KCl solution
(concentrations are indicated in the legend). Volume fraction is 0.1.

One can note that increasing salt concentration decreases the separation between
nearest neighbours and the amplitude of the peaks: in the limit of 50 mM added salt,
the suspension behaves like a fluid (even nearly like a gas), and for 2 mM the
structure of g(r) is close to the one of a crystal, where the position of nearest
neighbours are extremely well defined.

It is also possible to perform computer simulation models to estimate g(r). One then
has to use an interaction potential between particles and determine the particles’
positions. One of the most classical function to describe the interaction between
particles is the Lennard –Jones potential:

𝑎𝑎 12 𝑎𝑎 6
Φ𝐿𝐿𝐿𝐿 (𝑟𝑟) = 4Φ0 �� � − � � �
𝑟𝑟 𝑟𝑟

which is used extensively to study molecules. Note the differences between this
expression and the DLVO expression introduced in Chapter 3. The main reason for
the differences lays in the fact that colloidal particles are much bigger than
molecules, and many more effects are therefore incorporated in the DLVO theory.

179
 Introduction to Colloid Science

Colloidal charge determination from high-frequency shear measurements

For concentrated suspensions of (charged) spheres, it is possible to derive

′
relationships between the high-frequency elastic modulus 𝐺𝐺∞ and the interaction
potential . The derivation of this relationship is beyond the scope of the present
89

book, but its dimensionless form is given by:

′
𝑎𝑎3 3𝜙𝜙𝑠𝑠 3𝜙𝜙𝑠𝑠 2 ∞ 𝑑𝑑 𝑑𝑑Φ(𝑟𝑟)/(𝑘𝑘𝐵𝐵 𝑇𝑇)
𝐺𝐺∞ = + � 𝑔𝑔(𝑟𝑟) �𝑟𝑟 4 � 𝑑𝑑𝑑𝑑
𝑘𝑘𝐵𝐵 𝑇𝑇 4𝜋𝜋 40𝜋𝜋 0 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑

where 𝑔𝑔(𝑟𝑟) is pair correlation function. From the presence of the electric interaction
potential Φ(𝑟𝑟) the formula can be linked to the surface charge of the particles and
the effects of double layer changes upon changes in salt concentration.

Typical frequency-dependent measurements are given here for latex nanospheres

with a = 38 nm, dispersed in 30 mM of KCl:

Left: normalized dynamic viscosity 𝜂𝜂′ (𝜔𝜔)/𝜂𝜂0 = 𝐺𝐺 ′′ (𝜔𝜔)/(𝜔𝜔𝜂𝜂0 ) and right: dynamic
storage modulus 𝐺𝐺 ′ (𝜔𝜔), both as function of volume fraction

For the theoretical analysis to be valid, it is required that the frequency of the
oscillation is high enough so that 𝜔𝜔/(2𝜋𝜋) ≫ (𝐷𝐷𝑆𝑆 /𝑑𝑑 2 ) where 𝐷𝐷𝑆𝑆 is the short-time
self-diffusion coefficient of a colloidal particle in the concentrated suspension and 𝑑𝑑
is the mean interparticle separation. If we assume, to get an order of magnitude,
that 𝐷𝐷𝑆𝑆 ~ 𝐷𝐷0 = 𝑘𝑘𝐵𝐵 𝑇𝑇/(6𝜋𝜋𝜋𝜋𝜋𝜋) (the theoretical diffusion coefficient of a colloidal
particle in a dilute suspension) and 𝑑𝑑 ~ 𝑎𝑎 we get 𝜔𝜔/(2𝜋𝜋) ≫ 3 kHz. Indeed, as can be
seen in the figures, above that frequency both the storage modulus and the dynamic
viscosity are fairly independent of 𝜔𝜔.

89 Bergenholtz, J., Willenbacher, N., Wagner, N. J., Morrison, B., Van den Ende, D., & Mellema,
J. (1998). Colloidal charge determination in concentrated liquid dispersions using torsional
resonance oscillation. Journal of colloid and interface science, 202(2), 430-440.

180
 Chapter 7 Rheological behaviour of colloidal suspensions

By performing a systematic study of 𝐺𝐺 ′ at high frequency, as function of volume

fraction and salt concentration, it is then possible to assess the surface charge of the
particle: 𝑔𝑔(𝑟𝑟) can be measured, or approximated, as can Φ(𝑟𝑟) (see Chapter 3 about
the theoretical evaluation of Φ(𝑟𝑟) and Chapter 6 for the AFM technique that enables
to measure it). In the theoretical expression of Φ(𝑟𝑟) the surface charge (or zeta
potential) is the only unknown quantity. The theoretical expression for Φ(𝑟𝑟) is
′
inserted in the 𝐺𝐺∞ formula and the experimental data is fitted, giving access to the
surface charge.

In the limiting case that the suspension is in a crystalline phase, it can be shown that
the formula can be approximated by 90:

′
𝑎𝑎3 𝜙𝜙𝑚𝑚𝑚𝑚𝑚𝑚 𝑁𝑁𝑛𝑛𝑛𝑛 𝜙𝜙𝑚𝑚𝑚𝑚𝑚𝑚 1/3 𝑎𝑎2 𝑑𝑑 2 Φ(𝑟𝑟)
𝐺𝐺∞ ~ � � � �
𝑘𝑘𝐵𝐵 𝑇𝑇 10𝜋𝜋 𝜙𝜙𝑠𝑠 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑑𝑑𝑟𝑟 2 𝑟𝑟=2𝑎𝑎
eff

In the above, 𝜙𝜙𝑚𝑚𝑚𝑚𝑚𝑚 = 0.74 if one assumes a face-centered cubic where each particle
has 𝑁𝑁𝑛𝑛𝑛𝑛 = 12 nearest neighbours. Moreover

𝜙𝜙𝑚𝑚𝑚𝑚𝑚𝑚 1/3
𝑎𝑎eff = 𝑎𝑎 � �
𝜙𝜙𝑠𝑠

is a measure for the minimal separation of the particles (which is not 2𝑎𝑎 but 2𝑎𝑎eff ).
In the limit of a fluid, one has the approximation:
4
′
𝑎𝑎3 3𝜙𝜙𝑠𝑠 3𝜙𝜙𝑠𝑠 2 𝜙𝜙𝑚𝑚𝑚𝑚𝑚𝑚 1/3 1 𝑑𝑑Φ(𝑟𝑟)
𝐺𝐺∞ ~ − 𝑔𝑔ℎ𝑠𝑠 �2 � � � � �
𝑘𝑘𝐵𝐵 𝑇𝑇 4𝜋𝜋 40𝜋𝜋 𝜙𝜙𝑠𝑠 𝑘𝑘𝐵𝐵 𝑇𝑇 𝑑𝑑𝑑𝑑 𝑟𝑟=2𝑎𝑎eff

where for 0 < 𝜙𝜙𝑠𝑠 < 0.5:

1 + 𝜙𝜙𝑆𝑆 + 𝜙𝜙𝑠𝑠2 − 𝜙𝜙𝑠𝑠3

𝑔𝑔ℎ𝑠𝑠 =
(1 − 𝜙𝜙𝑠𝑠 )3

(this formula is the Carnahan-Starling expression) and for 0.5 < 𝜙𝜙 < 0.64:

1 1.21 + 𝜙𝜙𝑆𝑆
𝑔𝑔ℎ𝑠𝑠 =
4𝜙𝜙𝑆𝑆 0.64 − 𝜙𝜙𝑆𝑆
′
Using this last expression for 𝐺𝐺∞ together with the DLVO expression developed in
Chapter 3, one can show that it is possible to fit the measured data, and obtain the
surface charge of particles:

90 Buscall, R., Goodwin, J. W., Hawkins, M. W., & Ottewill, R. H. (1982). Viscoelastic properties

of concentrated latices. Part 2.—Theoretical Analysis. J. Chem. Soc., Faraday Trans. 1,

1982,78, 2889-2899

181
 Introduction to Colloid Science

Left: Dimensionless high-frequency shear modulus as function of volume fraction for

PBMA/AA dispersions. Right: surface charge density for two types of particles, evaluated
following the method exposed above 91.

The PBMA/AA particles are latex spheres of size a = 43.5 nm, and the PBMA particles
have a size of 38 nm. From their estimated surface charge, one can estimate that the
PBMA/AA particles have a surface potential of -78 mV and the PBMA of -56 mV.

Shear thinning and shear thickening

A figure above showed the evolution of the rheological properties of a suspension in
terms of flocs of particles that form or break as function of salt. We will here discuss
a similar type of figure where the rheological property is given as function of shear
rate 92.

91 Bergenholtz, J., Willenbacher, N., Wagner, N. J., Morrison, B., Van den Ende, D., & Mellema,

J. (1998). Colloidal charge determination in concentrated liquid dispersions using torsional

resonance oscillation. Journal of colloid and interface science, 202(2), 430-440.

92 See the book: Mewis, J., & Wagner, N. J. (2012). Colloidal suspension rheology. Cambridge

University Press.

182
 Chapter 7 Rheological behaviour of colloidal suspensions

Sketch of the effect of shear rate on the viscosity of a stable (non-aggregating) concentrated
colloidal suspension

At low particle concentrations, the suspension viscosity is nearly independent of

shear. For higher particle concentrations, increasing the shear rate (and shear stress)
leads to a marked shear thinning behaviour. This is of great importance in many
industrial applications, as colloidal particles can be used made to flow, pour or
spread with less effort at these higher shear rates. At even higher shear rates, the
viscosity increases significantly. This shear thickening behaviour is often undesirable
as it may damage processing equipment.

Shear thinning : the application of a shear flow distorts the equilibrium structure
and leads to fewer particle interactions. The particles therefore re-arrange in the
flow and the adopt an organization that permits flow with fewer particle encounters
(the particles flow roughly in lanes, like cars on a highway). An example of shear
thinning is given on the next page : instead of starting from a suspension, one has a
gel made of clay and silt loosely aggregated in a fabric. By applying a pressure
gradient (the famous Rissa landslide 93 was triggered by a small excavation and
stockpiling along a lake-shore), the fabric is locally broken and the particles
rearrange in a (concentrated) suspension-like structure. The phenomenon
propagates rapidly as at the front of the broken fabric inter-particle interactions are
immediately changed.

Thixotropy : thixotropy is defined as the continuous decrease of viscosity with time

when flow is applied to a sample that has been previously at rest, and the subsequent
recovery of viscosity when the flow is discontinued. Thixotropy is therefore a time-

93 L’Heureux, J. S., et al. (2012). The 1978 quick clay landslide at Rissa, mid Norway:
subaqueous morphology and tsunami simulations. In Submarine mass movements and their
consequences (pp. 507-516). Springer Netherlands. For a movie of the landslide, see:
https://www.youtube.com/watch?v=3q-qfNlEP4A

183
 Introduction to Colloid Science

dependent shear thinning property. Many gels and colloids are thixotropic materials,
bentonite (montmorillonite) is a good example : when montmorillonite is dispersed
in water, the platelets bind together electrostatically to form a house-of-cards
structure and the liquid becomes viscous. When the structure develops further, the
montmorillonite-dispersed liquid becomes a gel. However, shearing the gel returns
it to a dispersed liquid. Landslides are generally thixotropic.

Breakdown of a 3D thixotropic structure 94

Shear thickening : when the shear forces become sufficient, particle motions
become hydrodynamically highly correlated. This induces the grouping of particles
in clusters called hydroclusters. Hydroclusters are not necessarily aggregates : the
coloured particles in the example are just in close contact, and if the shear rate
would be set to zero the hydroclusters would disappear. The colour was used just to
indicate the positions of the hydroclusters. The system would then return to its initial
state thanks to Brownian motion and repulsion forces between particles. Of course,
if the suspension would be unstable, particles could aggregate and the hydroclusters
would then become “real” aggregates. When the shear would be reduced, these
aggregates would remain and therefore there is no reversibility in a change of shear
rate for this situation. Increasing flocculation due to shear rate in fact can lead to
rheopecty.

94Barnes, H. A. (1997). Thixotropy—a review. Journal of Non-Newtonian fluid mechanics,

70(1), 1-33.

184
 Chapter 7 Rheological behaviour of colloidal suspensions

Rheopecty : rheopecty is defined as the

continuous increase of viscosity with
time when flow is applied to a sample
that has been previously at rest, and
the subsequent recovery of viscosity
when the flow is discontinued.
Rheopecty is therefore a time-
dependent shear thickening property.
Rheopectic fluids thicken or solidify
when shaken. Examples of rheopectic
fluids include gypsum pastes and
printer inks. The synovial fluid in the
joints that link our bones becomes
thick the moment shear is applied in
order to protect the joint and
subsequently thins back to normal
viscosity to resume its lubricating function.

Marine clay, quick clay and landslides

Marine clay is clay that was deposited in a salty environment. The clay particles can
self-assemble into different configurations (see Chapter 5) and the formed fabric can
have very different properties. Construction in marine clays therefore presents a
geotechnical engineering challenge. For example, swelling of marine clay can
destroy building foundations in only a few years. Quick clay is a clay, which originally
was a marine clay, with 'quick' properties. It is a fine-grained sediment where the
grain structure may collapse even if the sediment is initially quite firm. Quick clay
can be firm as long as it is undisturbed, but flows like liquid if it becomes overloaded
or stirred, causing the loose grain structure to
collapse. Quick clay landslides can developed rapidly
when the firm clay liquefies.

The clay and silt particles that constitute the quick

clay were left behind during the retreat of the glaciers
and deposited in a nearby sea. The particles are
loosely packed and form a card house structure.
When the sea retreats, the salt is washed out by fresh
water (from rainfalls). The fabric’s structure then
becomes unstable, as the double layers have
extended significantly, and repulsive forces started to act between the particles. Any
mechanical stress can initiate liquefaction. The failure then rapidly propagates,
leading to landslides. When a landslide encounters a river or a fjord, a tsunami can
also be created as the quick clay flow pushes water. Marine clay is most widespread

185
 Introduction to Colloid Science

in Norway, in the Trøndelag region and eastern Norway in particular. Quick clay are
also found in parts of Canada and Sweden.

Quick clay landslide 95 at Lyngseidet (Norway) in 2010 (220.000 m3). The landslide was likely
triggered by loading of fill along the shoreline. (See the white car in the upper part for scale)

Gels and hydrogels

We have so far implicitly spoken about the formation of a gel when the shear rate
is zero in the case of thixotropic and rheopectic fluids. There is an important
difference between a concentrated suspension and a gel. In general, gels are
apparently solid, jelly-like materials:

95 https://www.ngu.no/en/topic/quick-clay-and-quick-clay-landslides; Photo: Andrea

Taurisano, NVE (with permission)

186
 Chapter 7 Rheological behaviour of colloidal suspensions

A gel is usually defined in colloid science as a substantially dilute polymer

suspension, where the polymers are cross-linked. Cross-linking refers to the joining
of polymer chains with covalent bonds. Cross-linking can occur during polymer
synthesis or later with the addition of atoms or molecules which will share electrons
with a part of the polymer chain.

A hydrogel is a fabric made of polymeric chains that contains an enormous amount

of water. There are two kinds of hydrogels:

1 – chemical hydrogels: these are formed by cross-linked polymers : the polymers/

polyelectrolytes are chemically bound to each other and form a fabric. Because of
the presence of hydrophilic groups (-OH, -CONH, -CONH2, -SO3H) on the polymers
water is “bound” to this fabric (more than 90% in weight, since there is a huge
surface area available – the pores of the fabric can be as small as nanometers). The
links between the polymers are permanent, and therefore, when agitated or stirred
the hydrogel will break in pieces, and no water will be freed.

2 – physical hydrogels : the polymers are bound by reversible (non-permanent) links.

These links can be due to hydrogen bonds, van der Waals forces, entanglements
(when the polymer concentration is high the polymers are bound to touch each
other and form a fabric). In this case, changing the properties, the gelation is
reversible : one can think of adding water (in case of entanglement), changing the
salinity (increase electrostatic repulsion between polymers and hence counteract
the van der Waals attraction), changing pH, temperature… By agitating the gel, one
generally observe a solid/liquid phase transition. This is the case for montmorillonite
for example (see landslide example). Many of this type of hydrogels display
thixotropy : they become fluid when agitated, but re-solidify when resting.

Hydrogels are studied a lot in biology (pharmacy) and food science, but their study
is still limited in civil engineering : a flocculated clay for example can be seen as a
complex hydrogel. Water is then trapped inside the floc structure like in a
conventional hydrogel, but the flocs are forming a peculiar fabric of interacting
macroscopic particles. The consolidation of this fabric is discussed in Chapter 9 and
10, but is still an on-going topic of research.

187
 Introduction to Colloid Science

Sand/mud mixtures : fall velocity in shear flow

We now consider a colloidal mud gel in which sand particles are imbedded. (In a mud
suspension the sand would simply settle to the bottom of the settling column.) If no
shear is applied, the gel does not change over time and the sand particles stay
imbedded. If shear is applied, one will observe that the sand particles start to
settle 96. This property causes critical problems: if settling occurs, the material loses
its homogeneity, which can strongly affect its mechanical properties.

In slow flows, it is considered that the settling properties of suspended particles are
not significantly affected by the material flow, and the sedimentation velocity is
usually computed from the balance of gravity and drag forces. In order to avoid or
slow down sedimentation, the only practical solution consists in inducing a sufficient
agitation to the system which will induce some lift or dispersion forces to the
particles. This principle is typically used in fluidization processes, in which a vertical
flow of the interstitial fluid induces a drag force counterbalancing gravity force. For
horizontal flows in conduits one may also rely on turbulence effects.

For many materials, the situation is different: the denser particles do not settle at
rest because they are embedded in a yield stress fluid which is able to maintain the
particles in their position. This situation is typically encountered with mortars or
fresh concrete which are made of particles (sand or gravel) of density around 2.5
mixed with a cement-water paste of density around 1.5 (the density is here
expressed as 𝜌𝜌/𝜌𝜌𝑤𝑤 ). This is the same for toothpastes which contain silica particles of
density 2.5 suspended in a paste of density close to 1. In that case the gravity force
(weight of the particle) is counterbalanced by the elastic force from the fabric, as
long as the yield stress obeys 97 :

96 Talmon, A. M., & Huisman, M. (2005). Fall velocity of particles in shear flow of drilling fluids.
Tunnelling and underground space technology, 20(2), 193-201.
97 Ovarlez, G., Bertrand, F., Coussot, P., & Chateau, X. (2012). Shear-induced sedimentation in

yield stress fluids. Journal of Non-Newtonian Fluid Mechanics, 177, 19-28.

188
 Chapter 7 Rheological behaviour of colloidal suspensions

4
𝜋𝜋�𝜌𝜌 − 𝜌𝜌𝑝𝑝 �𝑔𝑔𝑎𝑎3
𝜎𝜎0 ≥ 3 ~ �𝜌𝜌 − 𝜌𝜌𝑝𝑝 �𝑔𝑔𝑔𝑔
𝜋𝜋𝑎𝑎2
When the fluid is at rest, the weight of the sand particle in the mud fabric (gel) is
compensated by the reaction of the gel underneath. When shear is applied, the
reaction (static force) of the gel is changed in a friction (dynamic force) which is not
enough to compensate for the weight anymore. The evaluation of the settling
velocity of the sand particle in the sheared mud is a complex task, due to the fact
that the variables become tensors.

Until now, we have simplified the definitions of the variables. The shear stress, for
example (see beginning of the chapter) should properly be defined as:

𝐹𝐹𝑥𝑥𝑥𝑥 𝑑𝑑𝑣𝑣𝑥𝑥
𝜎𝜎𝑥𝑥𝑥𝑥 = = 𝜂𝜂
𝑆𝑆 𝑑𝑑𝑑𝑑

where the subscript 𝑥𝑥𝑥𝑥 of the tensor 𝜎𝜎𝑥𝑥𝑥𝑥 indicates that we are looking at the stress
created by the the 𝑥𝑥 − component of the force 𝐹𝐹𝑖𝑖𝑖𝑖 (𝑖𝑖 = 𝑥𝑥, 𝑦𝑦, 𝑧𝑧) on the slice 𝑦𝑦. This
force is linked to the change in shear rate 𝑣𝑣𝑥𝑥 over a thickness 𝑑𝑑𝑑𝑑. Similarly, one can
define 𝜎𝜎𝑥𝑥𝑥𝑥 , 𝜎𝜎𝑥𝑥𝑥𝑥 , 𝜎𝜎𝑦𝑦𝑦𝑦 , etc…

Due the tensorial nature of the variables, the flow around the sand particle is
complex. We therefore refer to Ovarlez et al. (see footnote) for a detailed derivation
and explanations. In their experiments, two flows are present: the shear flow and
the settling flow. The settling flow is considered to be a secondary flow as compared
to the shear flow. This means that the shear flow is the major flow in the system.
This has consequences on the estimation of the settling velocity, as the Stokes
settling velocity has to be adapted. We assume a Herschel-Bulkley model for the
stress:

𝜎𝜎 = 𝜎𝜎0 + 𝜂𝜂𝐻𝐻𝐻𝐻 𝛾𝛾̇ 𝑛𝑛

where 𝜎𝜎 is the macroscopic shear in the cell (and not the shear around the sand
particle). The authors show that the settling of a sand particle is given by the
following modified Stokes velocity, when 𝜎𝜎0 ≪ 𝜂𝜂𝐻𝐻𝐻𝐻 𝛾𝛾̇ 𝑛𝑛

2 𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤
𝒗𝒗 = 𝛼𝛼 𝑎𝑎2 𝒈𝒈
9 𝜂𝜂𝐻𝐻𝐻𝐻 𝛾𝛾̇ 𝑛𝑛

where 𝛼𝛼 is an adjustable parameter that accounts for the complexity of the flow
around the sand particle.

Experiments, reported by Ovarlez et al. are given below. They are done on a
suspension of glass beads in an emulsion (to control the viscosity). Experiments are
performed in a Couette cell. There is a strong discrepancy between what is expected

189
 Introduction to Colloid Science

from the analysis and the experimental measurements. While the experimental
profiles show a rather narrow front with no broadening in time, the sedimentation
front of the theoretical profiles gets broader and broader in time. This suggests that
collective effects are at play, which tend to stabilize the front at a given speed. This
point is discussed in Chapter 8 where it is shown that sedimentation velocities are
correlated in the horizontal plane over very long distances, of order of 20𝑎𝑎𝜙𝜙𝑆𝑆 −1/3 .
For the particles in the study this lengthscale is of the order of 1 cm, i.e. of the order
of the gap size (which is 1.9 cm). This may thus explain the observations.

Vertical volume fraction profiles observed in the gap of a Couette geometry in a 5%

suspension of 275 μm glass beads in a concentrated emulsion of yield stress 𝜎𝜎0 = 8.5 Pa after
a 24 h rest (blue circles), after 15 min (orange circles), 30 min (grey circles) and 45 min
(yellow circles) of shear 𝛾𝛾̇ = 4 𝑠𝑠 −1 . The dotted lines are the theoretical profiles expected
from the modified Stokes equation, taking into account the heterogeneity of the apparent
viscosity in the sheared material, under the assumption that the sedimentation velocity of
the suspension at a given radial position is set by the local viscosity 𝜂𝜂(𝑟𝑟) = 𝜎𝜎(𝑟𝑟)/𝛾𝛾̇ (𝑟𝑟) of the
sheared yield stress fluid only, independently of the sedimentation velocity in its
surroundings.

190
 Chapter 7 Rheological behaviour of colloidal suspensions

A fifty cent rheometer for yield stress measurement 98

Even though samples can be brought to the lab and analysed there, it is often
convenient for an engineer to perform rapid and inexpensive tests in the field. The
slump test schematized underneath is used extensively by civil engineers to estimate
the ‘‘workability’’ of fresh concrete, but was shown to be useful for estimating the
yield stress of mud slurries as well. “Workability” means that the slurry should have
the proper yield stress and viscosity. For example, if fresh concrete is too stiff, the
mixture will not flow into tight corners of moulding and if the concrete is too runny
(it contains too much water), the concrete will flow better but the strength of the
final hardened concrete will be reduced. Similarly, the flow properties of tailings in
a waste disposal scheme need to be tailored for slope deposition: if the suspension
is too thin, the material will result in little if any slope, while if it is too thick a tailing
will result in the material being deposited around the discharge point and not
flowing over the disposal area.

Schematic diagram of the slump test : a bottomless column is filled with the material to be
tested. Lifting the column allows the material to collapse under its own weight. The height of
the final deformed (or “slumped”) material is measured. The difference between the initial
and final heights is called the slump height.

The estimation of the yield stress is made as follows:

At t = 0 and at a given height z (where z = 0 is at the bottom of the column), the

pressure is given by:

𝑃𝑃(𝑧𝑧) = 𝜌𝜌𝜌𝜌(𝐻𝐻 − 𝑧𝑧) + 𝑃𝑃0

98 Pashias, N., Boger, D. V., Summers, J., & Glenister, D. J. (1996). A fifty cent rheometer for
yield stress measurement. Journal of Rheology (1978-present), 40(6), 1179-1189.

191
 Introduction to Colloid Science

where 𝐻𝐻 = ℎ + ℎ0 and the atmospheric pressure is given by 𝑃𝑃0 . For simplicity we

assume that 𝑃𝑃0 ≪ 𝜌𝜌𝜌𝜌𝜌𝜌 and therefore 𝑃𝑃(𝑧𝑧) = 𝜌𝜌𝜌𝜌(𝐻𝐻 − 𝑧𝑧).

At t > 0 and at a certain height ℎ0 (see figure underneath) the material will
experience a stress that is higher than the yield stress and the part of the material
below that height will start to flow. The interface layer between the yielded and
unyielded material is assumed to be a flat surface that moves down as the material
beneath it flows. During the deformation stage it is assumed that all horizontal
sections remain horizontal, and slumping is only due to radial flow. At the end of the
slumping, the unyielded region represented by ℎ0 will have a stress distribution that
is identical to that of the undeformed material (before slumping), while the stress in
the remaining material is equal to the yield stress.

At t = 0, the sample is divided in slices with same thickness dz. Assuming

incompressibility, conservation of volume ensures that at the end of the slumping
the new thickness of each slice is given by

𝑟𝑟 2
𝑑𝑑𝑧𝑧1 = 𝑑𝑑𝑑𝑑
𝑟𝑟12

Schematic representation of the initial (left) and final (right) sediment distribution.
The sample is divided in N different layer with same thickness dz at t = 0. At the end
of the slumping, using conservation of volume, each layer thickness becomes dz1

The amount of material above any given plane z will be the same before and after
the slump (the flow only occurs in the cross-sectional area). This implies that the
force due to gravity 𝑚𝑚(𝑧𝑧)𝑔𝑔 where 𝑚𝑚(𝑧𝑧) is the mass of the material above 𝑧𝑧 and 𝑔𝑔 is
the gravity constant will be the same and lead to the following balance of forces at
equilibrium:

𝜎𝜎𝑟𝑟 2 = 𝜎𝜎0 𝑟𝑟12

where 𝜎𝜎0 is the yield stress. The shear stress that acts on a body when a pressure is
applied to its normal direction is proportional to this pressure:

192
 Chapter 7 Rheological behaviour of colloidal suspensions

𝜎𝜎(𝑧𝑧) = 𝛼𝛼𝛼𝛼(𝑧𝑧) = 𝛼𝛼 [𝜌𝜌𝜌𝜌(𝐻𝐻 − 𝑧𝑧)]

From the equations above we get:

𝜎𝜎0 𝜎𝜎0
𝑑𝑑𝑧𝑧1 = 𝑑𝑑𝑑𝑑 = 𝑑𝑑𝑑𝑑
𝜎𝜎 𝛼𝛼𝛼𝛼𝑔𝑔(𝐻𝐻 − 𝑧𝑧)

By integration:
ℎ1 ℎ
𝜎𝜎0
� 𝑑𝑑𝑧𝑧1 = � 𝑑𝑑𝑑𝑑
0 0 𝛼𝛼𝛼𝛼𝛼𝛼(𝐻𝐻 − 𝑧𝑧)
𝜎𝜎0
ℎ1 = [−ln(ℎ0 ) + ln(𝐻𝐻)]
𝛼𝛼𝛼𝛼𝛼𝛼

The height ℎ0 can be linked to the yield stress by:

𝜎𝜎(ℎ) = 𝜎𝜎0 = 𝛼𝛼𝛼𝛼𝛼𝛼ℎ0

As

𝑠𝑠 = 𝐻𝐻 − ℎ0 − ℎ1

We obtain:
𝜎𝜎0 𝜎𝜎0
𝑠𝑠 = 𝐻𝐻 − �1 + ln(𝐻𝐻) − ln � ��
𝛼𝛼𝛼𝛼𝛼𝛼 𝛼𝛼𝛼𝛼𝛼𝛼

𝜎𝜎0 𝛼𝛼𝛼𝛼𝛼𝛼𝛼𝛼
𝑠𝑠 = 𝐻𝐻 − �1 + ln � ��
𝛼𝛼𝛼𝛼𝛼𝛼 𝜎𝜎0

This expression is similar to the one found by Pashias and Boger cited above and the
same as eq.(7.12a) in the book of Coussot 99. Coussot uses a value of 𝛼𝛼 = 1/√3 which
he derives in the book based on the Von Mises criterion, whereas Pashias and Boger
uses 𝛼𝛼 = 1/2 which they say to hold for an ideal elastic solid. In fact, the rough
assumption that is behind the result of 𝛼𝛼 = 1/2 is that a flow starts or stop in a layer
when the maximum shear stress in the material reaches the yield stress value. One
can rewrite the previous equation as:

𝜎𝜎0∗ 𝜎𝜎0∗
𝑠𝑠 ∗ = 1 − �1 − ln � ��
𝛼𝛼 𝛼𝛼

where we define the dimensionless coefficients:

99 Coussot, P. (2005). Rheometry of pastes, suspensions, and granular materials: applications

in industry and environment. John Wiley & Sons.

193
 Introduction to Colloid Science

𝑠𝑠 𝜎𝜎0
𝑠𝑠 ∗ = ; 𝜎𝜎0∗ =
𝐻𝐻 𝜌𝜌𝜌𝜌𝜌𝜌

We here give an example, in which “theory” is the plot for which 𝛼𝛼 = 1/2 and
“theory2” for 𝛼𝛼 = 1/√3 (clearly a better fit) :

Dimensionless slump height as function of dimensionless yield stress for various slurries. The
yield strength was obtained from vane tests. The different “red mud” (bauxite residue which
was strongly flocculated with polyacrylamide) data corresponds to different structural states
of the mud (the mud was allowed to be stirred for more or less long periods, causing
structural decay, or process water was added which also lowered the yield stress).
Experiments with titania and zirconia were done at the isoelectric point to maximize the yield
stress (see next section).

Despite the fact that the “50 cent rheometer” test (also referred to as slump test)
seems to provide some good qualitative (and quantitative) estimations of the yield
stress for the given example, one should not forget that we did not address here the
experimental problems of slurry/wall effects, inhomogeneous spreading etc… that
can affect the results. The slump test is nonetheless an easy and fast test that can
be used in the field. Coussot showed that for large slumps (where there is hardly an
undeformed region) a spread test can be done to estimate the yield stress.

Maximum yield and zeta potential

In the legend of the previous figure, we have stated that at the isoelectric point the
yield stress is maximum. The isoelectric point is defined as the pH value for which
the zeta potential is minimum. At that point, van der Waals forces dominate, and
the particles aggregate. For two of the previous samples, it was found that:

194
 Chapter 7 Rheological behaviour of colloidal suspensions

Yield stress versus pH : at the isoelectric point (pH = 7.2 and 7.6) the yield stress is maximum.
At that point, the zeta potential is close to zero and the suspension is strongly flocculated.

A systematic study – theoretical and experimental - of the relation between yield

stress and zeta potential has be performed by other authors 100.

The model introduced by Scales and co-workers (see footnote) evaluates the shear
stress as a summation of all pair interactions calculated from the DLVO forces
between the particles:

𝜎𝜎0 = 𝐾𝐾𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (𝜙𝜙𝑠𝑠 )F𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 (𝑟𝑟)

where 𝐾𝐾𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 is a network structural term dependent upon the particle size, the
solids volume fraction 𝜙𝜙𝑠𝑠 and the mean coordination number. The term F𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷
represents the DLVO force between two particles and 𝑟𝑟 the distance between the
centres of these particles. We note that:

𝑑𝑑Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷
F𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 (𝑟𝑟) = −
𝑑𝑑𝑑𝑑
where Φ𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 is the DLVO interaction potential, see Chapter 3. The expression for
the DLVO used by Scales et al. is valid for identical spherical particles of radius 𝑎𝑎
where 𝑟𝑟 ≫ 2𝑎𝑎 and given by:

100 Scales,
P. J., Johnson, S. B., Healy, T. W., & Kapur, P. C. (1998). Shear yield stress of partially
flocculated colloidal suspensions. AIChE Journal, 44(3), 538-544. and Zhou, Z., Scales, P. J., &
Boger, D. V. (2001). Chemical and physical control of the rheology of concentrated metal oxide
suspensions. Chemical Engineering Science, 56(9), 2901-2920.

195
 Introduction to Colloid Science

𝐴𝐴𝐴𝐴 exp�−𝜅𝜅(𝑟𝑟 − 2𝑎𝑎)�

F𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 (𝑟𝑟) = − 2π𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜁𝜁 2 𝜅𝜅𝜅𝜅
12(𝑟𝑟 − 2𝑎𝑎)2 1 + exp�−𝜅𝜅(𝑟𝑟 − 2𝑎𝑎)�

where it was assumed that 𝜓𝜓(𝑎𝑎) = 𝜁𝜁 (the surface electric potential is the zeta
potential, i.e. there is no Stern layer).

The expression for 𝜎𝜎0 implies that 𝜎𝜎0 will be maximum when F𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 (𝑟𝑟) is maximum.
This will happen when the electrostatic repulsion is minimum, that is, when 𝜁𝜁 is
minimum (close to zero):

Yield stress (measured by the vane technique) and zeta potential (measured by
electroacoustics) as function of pH and volume fraction for alumina particles suspended in 10
mM background electrolyte: note the relation between maximum yield stress and the
isoelectric point (zero zeta potential).

For the same alumina suspension, one can plot the yield stress as function of the
zeta potential squared (see underneath). The yield stress is maximum at the
isoelectric point (where the zeta potential 𝜁𝜁 is zero) and decreases in a parabolic
manner as 𝜁𝜁 increases. This is consistent with an increase in the electrical double
layer repulsion, as the surface charge of the particles increases.

In addition to its dependence upon 𝜁𝜁 and 𝜙𝜙𝑆𝑆 , a number of workers have

demonstrated that the rheology of concentrated mineral suspensions is strongly
dependent upon the mean particle size. For a given 𝜙𝜙𝑆𝑆 and microstructural
arrangement, geometrical considerations dictate that the number of particles
resisting an applied stress along a given shear plane is inversely related to the square
of the particle size 101. Note that the DLVO forces are proportional to the particle size.
To complicate matters, the particle size can also impact upon microstructural factors

101Johnson, S. B., Franks, G. V., Scales, P. J., Boger, D. V., & Healy, T. W. (2000). Surface
chemistry–rheology relationships in concentrated mineral suspensions. International Journal
of Mineral Processing, 58(1), 267-304.

196
 Chapter 7 Rheological behaviour of colloidal suspensions

such as the mean particle coordination number 102 in a non-trivial manner. As a

result, the theoretical particle size dependence of the yield stress properties of
concentrated suspensions is not obvious although experimentally, systems
consisting of small particles produce higher yield stresses than those containing
larger colloids.

The shear yield stress of concentrated alumina as function of the zeta potential squared.

Illustrations
Batchelor (fair use)
http://www.damtp.cam.ac.uk/about/gkb/

Synovial joint (creative commons)

https://en.wikipedia.org/wiki/Synovial_joint

Gel (public domain)

https://fr.wikipedia.org/wiki/Dentifrice

102 the mean number of particles touching a given one. See Jouannot-Chesney, P., Jernot, J.
P., & Lantuéjoul, C. (2011). Practical determination of the coordination number in granular
media. Image Analysis & Stereology, 25(1), 55-61.

197
 Chapter 8
Settling of
(concentrated) suspensions
 Introduction to Colloid Science

In the previous chapters, we have discussed the properties of dilute suspensions in

the context of settling (Chapter 2) and colloid interactions (Chapter 3). In Chapter 7,
we investigated the rheological properties of concentrated suspensions. In this
chapter, and in the next ones we are going to review the settling and consolidation
properties of concentrated suspensions. Concentrated suspensions are very
complex systems, owing to their complicated particle-particle interactions, and their
influence on the hydrodynamics. Nevertheless concentrated suspensions are
encountered in natural environments, and it is therefore important to have some of
the keys to understand where lays the complexity of this type of systems and how
to model them.

Concentrated suspensions and fluid mud

By definition, in a suspension, particles should be suspended (i.e. not resting on
others). When all the particles are touching each other in some way, a large structure
is formed which gets the name of fabric or gel. In colloid science, a gel is usually
defined as a dilute cross-linked polymeric system, which exhibits no flow when in
steady-state (see also Chapter 7). The name gel originates the word gelatine, which
is an irreversible hydrolysed polymer (collagen). In civil engineering, when studying
mud, the word gel is used to design a mud with a substantial yield stress.

Examples of concentrated mud. Left: concentration between 30 and 500 g/L. Right:
concentration between 1 and 30 g/L

Fluid mud is a high concentration aqueous suspension of fine-grained sediment in

which settling is substantially hindered by the proximity of sediment grains and flocs,
but which has not formed an interconnected matrix of bonds strong enough to
eliminate the potential for mobility 103. Other terms used to denote fluid mud include
fluff (UK), vloeibare sliblaag (NL), and crème de vase (FR). Fluid mud typically
exhibits concentrations of tens to hundreds of grams per liter and bulk densities
between 1,080 and 1,200 kg m−3.

103McAnally, W.H. et al. 2007. Management of fluid mud in estuaries, bays and lakes. Part I:
Present state of understanding on character and behaviour. Journal of Hydraulic Engineering,
Vol. 133, No. 1, 9-22

200
 Chapter 8 Settling of (concentrated) suspensions

Fluid mud exists only because of its transient behaviour: if fluid mud is left at rest, it
will eventually consolidate and form a gel (fabric). If, on the other hand, fluid mud
can be picked-up by currents or waves, the local mud concentration will decrease as
mud particles are dispersed and the fluid mud will become an usual mud suspension.

Modelling concentrated suspensions

A lot of work has been performed to understand the settling of particles in non-
dilute conditions. Theoretical models for systems where more than two particles
interact are usually very complicated, and this is why their study remains an open
field of research.

Microscopic mechanics can be upscaled to give macroscopic (bulk) properties thanks

to statistical physics. The difficulties are not just linked to solving the hydrodynamic
equations with better, faster computers. The collective interactions between the
particles can give rise to quite unexpected qualitative behaviour, actually often much
simpler than the microscopic motions seem to suggest 104. A litre of gas for instance
could be studied by tracking the behaviour of its constituting molecules (about
𝑁𝑁~ 0.3 × 1023 particles for one litre for usual pressure) – a huge task – however the
macroscopic behaviour of a simple ideal gas is very well described by the simple
relation: 𝑃𝑃𝑃𝑃 = 𝑁𝑁𝑁𝑁𝑁𝑁 where 𝑃𝑃 is the pressure, 𝑉𝑉 the volume, 𝑘𝑘 the Boltzmann
constant, 𝑇𝑇 the temperature .

Statistical physics makes use of probability distribution functions. In Chapter 7, the

pair correlation function 𝑔𝑔(𝑟𝑟) was introduced. The pair correlation function is used
to establish the probability distribution function. For example, the probability
density function Ψ(𝑟𝑟) for finding a particle at 𝑟𝑟 given that there is a particle at 𝑟𝑟 = 0
is written:

𝑁𝑁
Ψ(𝑟𝑟) = 𝑔𝑔(𝑟𝑟)
𝑉𝑉
where 𝑁𝑁/𝑉𝑉 is the average number density of particles within the sample. As particles
do not interpenetrate, the probability to find a particle between 0 and 2a (where a
is the radius of one particle) is zero. This volume, that cannot be occupied by other
particles, is called the excluded volume. Moreover:
𝑉𝑉𝑡𝑡𝑡𝑡𝑡𝑡
� Ψ(𝑟𝑟)𝑑𝑑𝑑𝑑 = 𝑁𝑁 − 1
𝑉𝑉(𝑟𝑟>2𝑎𝑎)

as the number of particles in the total volume considered 𝑉𝑉𝑡𝑡𝑡𝑡𝑡𝑡 is equal to 𝑁𝑁 − 1 (the
one at 𝑟𝑟 = 0 is not counted). Because of the excluded volume, the reference particle

104Guazzelli, E., & Morris, J. F. (2011). A physical introduction to suspension dynamics (Vol.
45). Cambridge University Press.

201
 Introduction to Colloid Science

influences the particle positioning near itself: this leads to interesting and physically
significant microstructures in concentrated suspensions.

Settling of concentrated suspensions

Dilute suspensions

In Chapter 2, we have discussed the settling of individual particles according to

Stokes’ law. We recall that this settling velocity can be expressed as:

2 𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤
𝒗𝒗𝟎𝟎 = 𝑎𝑎2 𝒈𝒈
9 𝜂𝜂

We here added the subscript 0 to indicate that we refer to an individual particle

settling according to Stokes. In very dilute suspensions the particles can be assumed
to settle according to Stokes.

Doublet and triplet of spheres

Using reversibility and symmetry principles one can show that two identical spheres
close to each other fall at the same velocity and therefore do not change their
orientation and separation. The pair has a sideways motion as there is a horizontal
component of the fall motion (except when the angle between the centres of the
spheres and the vertical, 𝜃𝜃, is 0 or 𝜋𝜋/2) and they fall faster than when they are alone.
The flow of the two spheres can be obtained by the method of reflections. One can
show that:

𝒗𝒗𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝 3𝑎𝑎
=1+ for 𝜃𝜃 = 0
𝒗𝒗𝟎𝟎 2𝑟𝑟

and:

𝒗𝒗𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝𝐝 3𝑎𝑎 𝜋𝜋
=1+ for 𝜃𝜃 =
𝒗𝒗𝟎𝟎 4𝑟𝑟 2

From these equations we see that two spheres falling next to each other (in an
horizontal plane) will fall slower than two spheres falling one above the other. In the
limit of touching spheres, it is easy to verify that the particles will settle with a
velocity 1.75 𝑣𝑣0 for 𝜃𝜃 = 0 and 1.375 𝑣𝑣0 for 𝜃𝜃 = 𝜋𝜋/2.

202
 Chapter 8 Settling of (concentrated) suspensions

Schematic representation of the settling of two spheres separated by a distance r. Note that
for an any angle 𝜃𝜃 between 0 and 𝜋𝜋/2 there is an horizontal component to the velocity
(which is represented by the red arrows)

When a third sphere is introduced (the three-body problem) the behaviour is

different as the particles do not usually maintain a constant separation, and the
configuration is unstable. In a sedimenting triplet equally spaced on a vertical or an
horizontal line one can then show that the middle particle will settle the fastest.

Concentrated suspensions

For the sedimentation of a suspension of many spheres, things become more

complicated. One might try to compute the mean sedimentation velocity of particles
by summing the effects between pairs of particles. The velocity of a particle can then
be written:

𝒗𝒗 = 𝒗𝒗𝟎𝟎 + ∆𝒗𝒗

where ∆𝒗𝒗 is the incremental velocity due to a second particle located at distance r.
For accounting for more particles, one could average over all possible separations
which occur with a probability Ψ(𝑟𝑟)(the probability of finding a sphere with its
center at r given there is a particle at r = 0, see above). This would give:
𝑉𝑉𝑡𝑡𝑡𝑡𝑡𝑡
𝒗𝒗 = 𝒗𝒗𝟎𝟎 + � ∆𝒗𝒗 ∙ Ψ(𝑟𝑟)𝑑𝑑𝑑𝑑
𝑉𝑉(𝑟𝑟>2𝑎𝑎)

From a dimension analysis, one can infer that the flow field ∆𝒗𝒗 decreases at first
order as 1/r (see the doublet examples above), that Ψ(𝑟𝑟) scales at first order as a
constant independent of r since for a reasonably dilute suspension 𝑔𝑔(𝑟𝑟) = 1 and
that:

𝑁𝑁
Ψ(𝑟𝑟) ≈
𝑉𝑉
The element of volume 𝑑𝑑𝑑𝑑 scales as 𝑟𝑟 3 . This implies that

203
 Introduction to Colloid Science

𝑉𝑉𝑡𝑡𝑡𝑡𝑡𝑡
� ∆𝒗𝒗Ψ(𝑟𝑟)𝑑𝑑𝑑𝑑 ~ 𝐿𝐿2
𝑉𝑉(𝑟𝑟>2𝑎𝑎)

where L is a characteristic length of the settling column. This dependence would

imply a very strong divergence (imagine that the considered column would
represent the sea…), and is due to the fact we omitted the long-range hydrodynamic
interactions in our analysis. This divergence is not found experimentally, as, on the
contrary, the velocity of particles in a concentrated suspension is even lower than
the one of single particles. This is what is called hindered settling: the velocity of the
particles is decreasing, relative to the Stokes velocity, with increasing particle
concentration, as particles start to be in each other’s way. The mean settling velocity
of a particle in a concentrated suspension is defined as:

𝒗𝒗 = 𝑓𝑓(𝜙𝜙𝑆𝑆 )𝒗𝒗𝟎𝟎

where 𝑓𝑓(𝜙𝜙𝑆𝑆 ) is the hindered settling factor which is function of the volume fraction
𝜙𝜙𝑆𝑆 . For a dilute suspension, one has 𝑓𝑓(𝜙𝜙𝑆𝑆 ) = 1. A very widely used empirical
correlation function is the one attributed to Richardson and Zaki in 1954 105:

𝑓𝑓(𝜙𝜙𝑆𝑆 ) = (1 − 𝜙𝜙𝑆𝑆 )𝑛𝑛

where n is a coefficient generally found to be close to 5 for the settling of

monodisperse colloidal spheres (see figure below). The experimental data are from
Nicolai et al. 106

Hindered settling function: Richardson-Zaki (red line) with n = 5, and experimental data
(symbols) for colloidal spheres

105Richardson and Zaki, Sedimentation and fluidization: Part I. Trans. Inst. Chem. Engrs., 32,
35-53 (1954)
106 Nicolai, H., et al. "Particle velocity fluctuations and hydrodynamic self‐diffusion of

sedimenting non‐Brownian spheres." Physics of Fluids 7.1 (1995): 12-23.

204
 Chapter 8 Settling of (concentrated) suspensions

Experimental observations for the hindered settling velocity in the dilute regime
(𝜙𝜙𝑆𝑆 ≪ 1) have established that then 𝑓𝑓(𝜙𝜙𝑆𝑆 ) ≈ 1 − 5𝜙𝜙𝑆𝑆 , in accordance with
(1 − 𝜙𝜙𝑆𝑆 )𝑛𝑛 ≈ 1 − 𝑛𝑛𝜙𝜙𝑆𝑆 .

One cause for hindered settling is linked to the presence of the bottom of the settling
column which imposes a back-flow due to volume conservation. The condition to
express this constant volume flow is similar to the one that will be discussed in
Chapter 9 :

(1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 + 𝜙𝜙𝑆𝑆 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 0

The velocity of the solid phase (particles) in the rest frame of the laboratory is given
by 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 and the velocity of the water in the rest frame of the laboratory is given by
𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 . Newton’s equation, with the gravity acting as only external field gives:

4 𝜌𝜌
−6𝜋𝜋𝜋𝜋𝜋𝜋(𝒗𝒗𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 − 𝒗𝒗𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 ) + 𝜋𝜋𝑎𝑎3 𝜌𝜌𝑠𝑠 �1 − � 𝒈𝒈 = 0
3 𝜌𝜌𝑠𝑠

The first term represents the Stokes drag force and the second the force of gravity,
compensated for Archimedes. The fluid density 𝜌𝜌 is given by:

𝜌𝜌 = (1 − 𝜙𝜙𝑆𝑆 )𝜌𝜌𝑤𝑤 + 𝜙𝜙𝑆𝑆 𝜌𝜌𝑠𝑠

Using the fact that:

𝜌𝜌 𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤
1− = (1 − 𝜙𝜙𝑆𝑆 )
𝜌𝜌𝑠𝑠 𝜌𝜌𝑠𝑠

and taking into account the back-flow, one can rearrange the last equations to
obtain 107:

2 (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )
𝒗𝒗𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑎𝑎2 (1 − 𝜙𝜙𝑆𝑆 )2 𝒈𝒈
9 𝜂𝜂

𝒗𝒗𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝒗𝒗𝟎𝟎 (1 − 𝜙𝜙𝑆𝑆 )2

Therefore, if one would neglect particle-particle interactions and hydrodynamic

effects, we would find n = 2 in the Richardson and Zaki formulation. One of the
(1 − 𝜙𝜙𝑆𝑆 ) results from the return flow, and the other arises because a typical sphere

107 Oliver, D. R. "The sedimentation of suspensions of closely-sized spherical particles."

Chemical Engineering Science 15.3 (1961): 230-242. See also: Gourdin-Bertin, S., and C.
Chassagne. "Onsager’s reciprocal relations for electroacoustic and sedimentation: Application
to (concentrated) colloidal suspensions." The Journal of chemical physics 142.19 (2015):
194706. Note the mistake in eq.(29) of that last article as can be found by inserting eq.(26) in
eq.(28).

205
 Introduction to Colloid Science

was considered to be settling in a fluid with density equal to that of the suspension,
and not that of the water alone.

Other corrections can be done. One could in particular account for the change in
viscosity. We have seen in Chapter 7 that a possible expression for the viscosity is:
−2.5
𝜙𝜙𝑆𝑆
𝜂𝜂(𝜙𝜙𝑆𝑆 ) = 𝜂𝜂0 �1 − �
1 − 𝑐𝑐𝜙𝜙𝑆𝑆

where 𝜂𝜂0 is the viscosity of water. If we assume that 1 − 𝑐𝑐𝜙𝜙𝑆𝑆 ≈ 1 (which is valid for
not too concentrated suspensions), we get:

𝜂𝜂(𝜙𝜙𝑆𝑆 ) ≈ 𝜂𝜂0 (1 − 𝜙𝜙𝑆𝑆 )−2.5

Therefore if we define the settling of an individual particle 𝒗𝒗𝟎𝟎 by:

2 (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )
𝒗𝒗𝟎𝟎 = 𝑎𝑎2 𝐠𝐠
9 𝜂𝜂0
we now find:

𝒗𝒗𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝒗𝒗𝟎𝟎 (1 − 𝜙𝜙𝑆𝑆 )4.5

The exponent is now 4.5, quite close to the experimentally found one for spheres.
This approach is however questionable, as it is not obvious that the viscosity
experienced by a particle can correctly be approximated by the mean viscosity of the
suspension.

Non-spherical particles

Needless is to say that for particles that are not monodisperse spheres, as assumed
until here, the settling rate dependence on volume fraction is much more complex.
In numerical simulations, it has been shown that suspensions that are homogeneous
at t = 0, with random positions and orientations exhibit a peculiar settling behaviour.
A the sedimentation proceeds, a broad suspension front forms at the interface and
the clear fluid appears at the top. Careful observation of the bulk shows that the
suspension does not remain homogeneous: particles tend to group into clusters.

206
 Chapter 8 Settling of (concentrated) suspensions

Particles are also observed to rotate and to orient preferentially in the vertical
direction, with a strong correlation between centre-of-mass positions and
orientations; while the orientation remains random inside the dense clusters, the
alignment in the vertical direction is much clearer in their periphery, where a strong
vertical shear exists. For large 𝜙𝜙𝑆𝑆 the settling found experimentally could be
described with a Richardson-Zaki law with n = 9.

Fluctuations in settling 108

Large velocity fluctuations have been measured by tracking marked spheres in an

otherwise transparent sedimenting suspension. The measurements were not
collected immediately after the initial mixing of the suspension: the suspension was
allowed to settle for some time and reach a steady behaviour before tracking
started. The fluctuations did not appear to be affected by the size of the container
in the large containers used.

Tracking of two marked spheres in the

midst of a 30% volume fraction
sedimenting suspension of unmarked
spheres made optically transparent by
matching the index of refraction of the
suspending fluid to that of the glass
spheres. The particle trajectories are
tortuous and exhibit large as well as small
loops as the spheres sometimes moved
upward against gravity.

For many years, theoretical estimates and numerical simulations predicted the
steady-state fluctuations of the velocities of spheres to increase with the size of the
container, whereas experiments found no such variation. In fact, the correlation
length of the velocity fluctuations was found experimentally to be 20 interparticle
separations, 20𝑎𝑎𝜙𝜙𝑆𝑆 −1/3 . This value of the correlation length was observed for
volume fractions from 10-4 to 0.4. When the minimum dimension of the container L
was less than 20𝑎𝑎𝜙𝜙𝑆𝑆 −1/3 it was found that the velocity fluctuations Δ𝑣𝑣 would scale
as:

108Guazzelli, Élisabeth, and John Hinch. "Fluctuations and instability in sedimentation."

Annual review of fluid mechanics 43 (2011): 97-116.

207
 Introduction to Colloid Science

𝐿𝐿
Δ𝑣𝑣 ~ 𝑣𝑣0 �𝜙𝜙𝑆𝑆
𝑎𝑎

While the steady-state velocity fluctuations are independent of the size of the
container if 𝐿𝐿 > 20𝑎𝑎𝜙𝜙𝑆𝑆 −1/3 , the early fluctuations do depend on the container. The
initial magnitude of the velocities again scales as 𝑣𝑣0 �𝜙𝜙𝑆𝑆 𝐿𝐿/𝑎𝑎. Some experimental
results are given below. These are obtained for suspensions of glass spheres of
radius 150 microns, volume fraction 0.1% and density 4 g/cm3. They were suspended
in silicon oil to insure that the Reynolds number remain extremely small. The time t
= 0 corresponds to the time just after mixing the column.

Relaxation of the large-scale fluctuations. The velocity field is from particle-image

velocimetry sampling the whole-cell height and width within a laser sheet located in the
middle plane, and the concentration profile 𝜙𝜙𝑠𝑠 /𝜙𝜙0 is from light-attenuation measurements
through the suspension. The timescale is the Stokes time tS = a/v0, i.e. the time for an isolated
sphere to settle a distance equivalent to its radius.

The fluctuations magnitudes are strongly anisotropic with vertical velocities around
four times the horizontal velocities. The initial strong large-scale fluctuations decay
in time to weaker small-scale fluctuations. These small-scale fluctuations remain in
a steady state until the sedimentation front arrives. This reduction of the initially
large fluctuations to a smaller steady value independent of the size of the container,
seen initially at low volume fractions 𝜙𝜙𝑆𝑆 , has been also observed at larger 𝜙𝜙𝑆𝑆 .

This behaviour can be partially understood by modelling the system by heavy blobs
of particles falling to the bottom and light blobs rising to the top.

208
 Chapter 8 Settling of (concentrated) suspensions

Falling blobs

Let us consider a blob of size 𝑑𝑑. This blob would contain on average 𝑛𝑛𝑛𝑛 3 particles
(where 𝑛𝑛 is the number density of particles), but there would be statistical
fluctuations of √𝑛𝑛𝑛𝑛 3 in the number if the particles were positioned randomly and
independently. This fluctuation in the number gives a 𝑚𝑚√𝑛𝑛𝑛𝑛 3 fluctuation in the mass
of the blob, where 𝑚𝑚 is the mass of a particle compensated for Archimedes.
Balancing the fluctuation in weight with a Stokes drag on the blob yields a fluctuation
in velocity of:

𝑚𝑚𝑚𝑚√𝑛𝑛𝑛𝑛 3 𝑑𝑑
Δ𝑣𝑣 ~ = 𝑣𝑣0 �𝜙𝜙𝑆𝑆
6𝜋𝜋𝜋𝜋𝜋𝜋 𝑎𝑎

As larger blobs give larger fluctuations in velocity, one would expect to see those
corresponding to the largest spherical blob that can be fitted into the container. This
largest size corresponds to the largest of the height, width, and depth of the column,
which we have noted L above. We therefore find, in accordance with observations,
that the velocity fluctuations Δ𝑣𝑣 scale as:

𝐿𝐿
Δ𝑣𝑣 ~ 𝑣𝑣0 �𝜙𝜙𝑆𝑆
𝑎𝑎

Falling clouds

We have just introduced the concept of a “blob” of particles. One other example of
blob that is quite relevant for different studies is when the blob of particles is found
in water. This is the case for example when a concentrated suspension is pipetted
into a jar containing water. It is common to define as a cloud a blob containing a very
large number of particles surrounded by clear water.

In Chapter 2 we have illustrated what happens when regions of high particle

concentration come into contact with regions of low particle concentration: the
particles will diffuse to regions of low particle concentration. Implicitly, we have
there assumed that the hydrodynamics could be neglected: the pipette is just put
into contact with the water and the particles will start to diffuse thanks to Brownian
motion (helped by gravity). This is also how the pipetting for particle size is
performed in settling studies (see Chapter 5).

If a blob (or cloud) of particles would be deposited as such (for instance pushed with
a syringe) in clear water, a different behaviour would be observed.

209
 Introduction to Colloid Science

A cloud of particles sedimenting in a viscous fluid

evolves into a torus that becomes unstable and breaks
up into secondary droplets, which deform into tori
themselves in a repeating cascade. This instability
occurs even in the complete absence of inertia and
without the need to perturb the initial shape. The
particles circulate in closed toroidal streamlines as
predicted by a continuum approach in which the cloud is
modelled as an effective medium of higher density.
Fluctuations arising from the multibody character of the
hydrodynamic interactions cause particles to depart
from these streamlines and to be carried into a
downstream tail. Because the lost particles are those
located in the circulation rim, this depletes the central
region and leads to torus formation. The mechanism
responsible for the further expansion of the torus
remains unclear, but the breakup can be described as a
change in the flow topology that occurs when the torus
reaches a critical aspect ratio. Simulations using a point-
particle approach containing the minimal physics of the
long-range interactions capture this dynamics. Faster
breakup is observed for clouds of fibers due to the self-
motion of the anisotropic particles.

A cloud of particles can primarily be seen as a heavy fluid drop which has a different
density and specific viscosity than the lighter water surrounding it. The problem was
originally solved by Hadamard and Rybczyński in 1911 109. The boundary conditions
at the particle’s surface are that the fluid velocities are continuous (contrary to the
no-slip condition when one considers a hard sphere), and that there is no surface
tension between the drop and the surrounding fluid. One finds that the terminal
velocity of the spherical cloud is then given by the balance of the Hadamard-
Rybczyński drag force and the weight (compensated for Archimedes):

4
𝑁𝑁 𝜋𝜋𝑎𝑎3 (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔 6𝑎𝑎
𝑈𝑈𝑆𝑆 = 3 = 𝑁𝑁 𝑣𝑣0
2 + 3𝜆𝜆 2 + 3𝜆𝜆
2𝜋𝜋𝜋𝜋 𝑅𝑅 2 𝑅𝑅
𝜆𝜆 + 1 𝜆𝜆 + 1
where 𝑁𝑁 is the number of particles in the cloud of radius 𝑅𝑅 and 𝜆𝜆 is the ratio between
the cloud viscosity and the viscosity of water. The factor 2(2 + 3𝜆𝜆)/(𝜆𝜆 + 1) ranges
from 5 for a cloud with a low volume fraction (where 𝜆𝜆 ≈ 1) to 6 for a concentrated
cloud (where 𝜆𝜆 ≫ 1).

109Hadamard, J. S. (1911). "Mouvement permanent lent d'une sphere liquide et visqueuse

dans un liquide visqueux". CR Acad. Sci. (in French). 152: 1735–1738 and Rybczynski, W.
(1911). "Über die fortschreitende Bewegung einer flüssigen Kugel in einem zähen Medium".
Bull. Acad. Sci. Cracovie, A. (in German): 40–46.

210
 Chapter 8 Settling of (concentrated) suspensions

The most remarkable feature observed during the cloud fall is the collective motion
followed by the particles. While settling, the particles circulate in a toroidal vortex
inside the cloud, similarly to the heavy fluid inside a drop sedimentating in a lighter
fluid. As a result the cloud remains a cohesive entity for long times, maintaining a
sharp boundary between its particle-filled interior and the clear fluid outside. It is
the chaotic fluctuations arising from the many-body character of the hydrodynamic
interactions that cause the particles to cross the boundary of the closed toroidal
circulation which, unlike the drop, is not a material surface. Some of the particles
may thus be carried by the outside flow into a vertical tail at the rear of the cloud
(see figure (a) above).

Sketch of the settling of a spherical cloud of particles

showing the toroidal circulation of the particles inside the
cloud and the particle leakage (in red) at the rear of the
cloud (in a frame relative to the moving particle)

Clouds made of a small number of particles are

found to keep their shape until they disintegrate
owing to the constant loss of particles. Clouds having
a large number of particles (N > 500) become
unstable (see figures (b,c,d) above).

The evolution of a cloud of particles is a good

example of how the long-range nature of many-body hydrodynamic interactions and
the coupling between hydrodynamics and the microscopic arrangement of particles
lead to a collective effect. While the suspension can be modelled as an effective
medium with excess mass, the discrete nature of the suspension is a fundamental
ingredient in understanding phenomena as leakage at the rear and destabilization
of the cloud.

Wall effects

We have already discussed above the influence of

the size of the column on the settling. Even though
the relative sedimentation velocity is independent
of the shape of the column (at steady-state) there
exists a global convection of the suspension called
intrinsic convection that was discovered by Prof.
Peter Mazur of Leiden University and coworkers in
1985 110. This convection effect exists even for dilute
suspensions. The principle is sketched on the figure
opposite: close to the column walls there is a small

110 Beenakker, C. W. J., and P. Mazur. "Is sedimentation container‐shape dependent?." Physics

of Fluids (1958-1988) 28.11 (1985): 3203-3206.

211
 Introduction to Colloid Science

depleted region (symbolized by the dashed lines) of size the radius of a particle.
Between the wall and each dashed line the centre of particles cannot penetrate
more than a distance corresponding to their radius. Conservation of volume and no-
slip condition at the wall (= the velocity of the fluid is zero at the walls) ensure that
the hydrodynamic profile is given by the blue line on the figure 111. This profile
corresponds to a Poiseuille flow with a slip velocity 𝑣𝑣𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = 9𝜙𝜙𝑆𝑆 𝑣𝑣0 /4 at a particle
radius distance from the walls, and a maximum downwards velocity in the centre of
the column equal to −𝑣𝑣𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 /2. This flow induces two vortices in the column with an
ascending velocity near the wall and a descending in the middle (red lines in the
figure). As this global convection is superimposed on the settling motion of the
particles relative to the suspension (black arrow), the particles should settle faster
in the centre of the cell than near the side walls. Experiments have shown the
existence of this intrinsic convection, however the effect was observed to be much
smaller than predicted. For a concentrated suspension, an ordering of particles near
the wall was observed, as particles piled up against the side walls. Existing models
do not take into account this effect.

Chemical potential, osmotic pressure and thermodynamics

Until now we have focussed on microscopic and mesoscopic descriptions of the
settling of (concentrated) suspensions. Even though progress has been made in
understanding the related observed features, it is clear that the complexity of
problem prevent the use of microscopic and mesoscopic theories for large-scale
engineering problems, such as predicting the settling velocity of a concentrated
suspension. We have however already shown that even though the systems might
be complex at a microscopic scale, it is possible to derive some simple relationships
for their mesoscopic and macroscopic behaviour. Examples are the empirical
Richardson-Zaki relationship and the ideal gas theory (𝑃𝑃𝑃𝑃 = 𝑁𝑁𝑘𝑘𝑇𝑇). The
mesoscopic/macroscopic description of a system can (in colloid science) be linked to
statistical physics : the numerical simulations of the falling fibers given above or of
interacting molecules in a gas are examples of it. There, the microscopic interactions
between particles are simulated (one example of such interactions is given in
Chapter 7, where we introduced the Lennard-Jones potential) and the macroscopic
behaviour of the suspension can be inferred from the results. Another approach to
macroscopic descriptions of a complex system is through thermodynamics.

In the previous chapters of this book, we have defined various concepts relevant for
colloid science without explicitly referring to thermodynamics. In Chapters 2 and 3,
we have discussed the osmotic pressure and the Boltzmann distribution. These
concepts are both linked to statistical physics and thermodynamics. Statistical
mechanics provides a framework for relating the microscopic properties of

111Bruneau, D., et al. "Intrinsic convection in a settling suspension." Physics of Fluids (1994-
present) 8.8 (1996): 2236-2238.

212
 Chapter 8 Settling of (concentrated) suspensions

individual atoms and molecules to the macroscopic or bulk properties. It enables to

explain thermodynamics as a natural result of statistics, classical mechanics, and
quantum mechanics at the microscopic level. Both statistical physics and
thermodynamics originate from the work of Gibbs (see underneath). It is not the
purpose of the present book to go too much into thermodynamic concepts, but it is
useful to get some very basic understanding of it. Indeed, it is for example through
thermodynamics that the osmotic pressure and the Boltzmann distribution can best
be understood. Theoretical work in thermodynamics and statistical physics is still
ongoing in order to understand the macroscopic behaviour of complex systems such
as concentrated suspensions, slurries and transport in porous media .

Chemical potential

An important thermodynamic parameter is the chemical potential of component i

(that can be an ion species, a colloid or the solvent), noted 𝜇𝜇𝑖𝑖 . The chemical potential
is a form of potential energy that can be absorbed or released during a chemical
reaction, and that can change during a phase transition. The chemical potential is
defined as the partial derivative of the free energy with respect to the amount of
the considered species (all other species’ concentrations in the mixture remaining
constant). At chemical equilibrium or in phase equilibrium the total sum of chemical
potentials is zero, as the free energy is at a minimum. Particles tend to go from a
higher chemical potential to a lower one : a simple example is the diffusion of
particles from high to low concentrations that we have introduced when we
discussed Fick’s law in Chapter 2. The microscopic explanation for this is based on
kinetic theory and the random motion of particles. However, it is simpler to describe
the process in terms of chemical potentials: for a given temperature, a particle has
a higher chemical potential in a higher-concentration area, and a lower chemical
potential in a low concentration area. Movement of particles from higher chemical
potential to lower chemical potential is accompanied by a release of free energy.

The total chemical potential can be split into an internal and an external chemical
potential, where the external chemical potential originates from external gradients
such as electric field and gravity.

For example let us consider an electrolyte solution in the presence of a charged wall
or colloid. The electrochemical potential 𝜇𝜇𝑖𝑖 of ion species i is a function of the electric
potential 𝜓𝜓(r) originating from the presence of the charged wall or colloid. It is given
by:

𝑛𝑛𝑖𝑖 (𝑟𝑟)
𝜇𝜇𝑖𝑖 (𝑟𝑟) = 𝑞𝑞𝑖𝑖 𝜓𝜓(𝑟𝑟) + 𝑘𝑘𝐵𝐵 𝑇𝑇ln � � + 𝜇𝜇𝑖𝑖0
𝑛𝑛0

where 𝜇𝜇𝑖𝑖0 is a reference value which only depends on temperature 𝑇𝑇, 𝑘𝑘𝐵𝐵 is the
Boltzmann constant and 𝑛𝑛0 a reference density. The other variables are defined in
Chapter 3.

213
 Introduction to Colloid Science

Josiah Willard Gibbs,

colloid science and thermodynamics

Josiah Willard Gibbs (1839 – 1903) was an American

scientist who made important theoretical contributions
to physics, chemistry, and mathematics. His work on
the applications of thermodynamics was instrumental
in transforming physical chemistry into a rigorous
deductive science. Together with James Clerk Maxwell
and Ludwig Boltzmann, he created statistical
mechanics (a term that he coined), explaining the laws
of thermodynamics as consequences of the statistical
properties of ensembles of the possible states of a physical system composed of
many particles. Gibbs also worked on the application of Maxwell's equations to
problems in physical optics. As a mathematician, he invented modern vector calculus
(independently of the British scientist Oliver Heaviside, who carried out similar work
during the same period).

In 1863, Gibbs received the first Doctorate of Philosophy (Ph.D.) from Yale in
engineering granted in the US, for a thesis entitled "On the Form of the Teeth of
Wheels in Spur Gearing", in which he used geometrical techniques to investigate the
optimum design for gears.

Gibbs traveled to Europe with his sisters where they spent the winter of 1866–67 in
Paris. Gibbs attended lectures at the Sorbonne and the Collège de France, given by
such distinguished mathematical scientists as Joseph Liouville and Michel Chasles.
Moving to Berlin, Gibbs attended the lectures taught by mathematicians Karl
Weierstrass and Leopold Kronecker, as well as by chemist Heinrich Gustav Magnus.
In Heidelberg, Gibbs was exposed to the work of physicists Gustav Kirchhoff and
Hermann von Helmholtz, and chemist Robert Bunsen. At the time, German
academics were the leading authorities in the natural sciences, especially chemistry
and thermodynamics.

After a three-year sojourn in Europe, Gibbs spent the rest of his career at Yale, where
he was professor of mathematical physics from 1871 until his death. Working in
relative isolation, he became the earliest theoretical scientist in the United States to
earn an international reputation and was praised by Albert Einstein as "the greatest
mind in American history".

Commentators and biographers have remarked on the contrast between Gibbs's

quiet, solitary life in turn of the century New England and the great international
impact of his ideas. Though his work was almost entirely theoretical, the practical
value of Gibbs's contributions became evident with the development of industrial
chemistry during the first half of the 20th century. According to Robert A. Millikan,
in pure science Gibbs "did for statistical mechanics and for thermodynamics what

214
 Chapter 8 Settling of (concentrated) suspensions

Laplace did for celestial mechanics and Maxwell did for electrodynamics, namely,
made his field a well-nigh finished theoretical structure."

When Dutch physicist J. D. van der Waals received the 1910 Nobel Prize "for his
work on the equation of state for gases and liquids" he acknowledged the great
influence of Gibbs's work on that subject.

When no external electric field is applied there are no ionic fluxes, the system is in
equilibrium and the electrochemical potential is constant in the whole system. This
implies that:

𝜇𝜇𝑖𝑖 (𝑟𝑟) = 𝜇𝜇𝑖𝑖 (𝑟𝑟 → ∞)

𝑛𝑛𝑖𝑖 (𝑟𝑟) 𝑛𝑛𝑖𝑖 (∞)

𝑞𝑞𝑖𝑖 𝜓𝜓(𝑟𝑟) + 𝑘𝑘𝐵𝐵 𝑇𝑇ln � � + 𝜇𝜇𝑖𝑖0 = 𝑘𝑘𝐵𝐵 𝑇𝑇ln � � + 𝜇𝜇𝑖𝑖0
𝑛𝑛0 𝑛𝑛0

This leads to:

−𝑞𝑞𝑖𝑖 𝜓𝜓(𝑟𝑟)
𝑛𝑛𝑖𝑖 (𝑟𝑟) = n𝑖𝑖 (∞)exp � �
𝑘𝑘𝐵𝐵 𝑇𝑇

which is the Boltzmann distribution given in Chapter 3.

Osmotic pressure

The derivation of the osmotic pressure from the chemical

potential is due to the Dutch chemist Jacobus Henricus van’t
Hoff (1852 – 1911). Van’t Hoff got the first Nobel prize in
Chemistry in 1901.

His derivation is as follows:

Let us consider a suspension that is brought into contact with

a semi-permeable membrane, as sketched in Chapter 2. The
system is at equilibrium when the chemical potential of the
solvent is equal on both sides of the membrane (there is then also no flux). The
chemical potentials of the solvent for each side of the membrane are then equal:
0 (𝑃𝑃)
𝜇𝜇𝑤𝑤 = 𝜇𝜇𝑤𝑤 (𝑥𝑥𝑤𝑤 , 𝑃𝑃 + Π)
0
where 𝜇𝜇𝑤𝑤 is the chemical potential of the pure solvent, 𝑃𝑃 is the pressure at one side
of the membrane and 𝑃𝑃 + Π is the pressure at the other side.

The variable 𝑥𝑥𝑤𝑤 stands for the mole fraction of the solvent:

215
 Introduction to Colloid Science

𝑛𝑛𝑤𝑤
𝑥𝑥𝑤𝑤 =
𝑛𝑛𝑡𝑡𝑡𝑡𝑡𝑡

where 𝑛𝑛𝑤𝑤 is the number of moles of solvent and 𝑛𝑛𝑡𝑡𝑡𝑡𝑡𝑡 the total number of moles of
the system. If 𝑥𝑥𝑠𝑠 is the mole fraction of the solute, we have:

𝑥𝑥𝑤𝑤 + 𝑥𝑥𝑠𝑠 = 1

For an ideal mixture, the chemical potential can be written:

0 (𝑃𝑃
𝜇𝜇𝑤𝑤 (𝑥𝑥𝑤𝑤 , 𝑃𝑃 + Π) = 𝜇𝜇𝑤𝑤 + Π) + 𝑅𝑅𝑅𝑅ln(𝑥𝑥𝑤𝑤 )

where 𝑅𝑅 = 𝑘𝑘𝐵𝐵 𝑁𝑁𝐴𝐴 is the gas constant, with 𝑁𝑁𝐴𝐴 being Avogadro’s number. Moreover,
the addition to the pressure can be seen as an energy of expansion:
𝑃𝑃+Π
0 (𝑃𝑃 0 (𝑃𝑃)
𝜇𝜇𝑤𝑤 + Π) = 𝜇𝜇𝑤𝑤 +� 𝑉𝑉(𝑃𝑃)𝑑𝑑𝑑𝑑
𝑃𝑃

where 𝑉𝑉 is the molar volume (m3/mol). Combining the above equations, we get:
𝑃𝑃+Π
−𝑅𝑅𝑅𝑅ln(𝑥𝑥𝑤𝑤 ) = � 𝑉𝑉(𝑃𝑃)𝑑𝑑𝑑𝑑
𝑃𝑃

If the liquid is incompressible the molar volume is constant, and the integral is then
equal to ΠV. Thus, we get:

−𝑅𝑅𝑅𝑅 −𝑅𝑅𝑅𝑅 𝑅𝑅𝑅𝑅

Π= ln(𝑥𝑥𝑤𝑤 ) = ln(1 − 𝑥𝑥𝑠𝑠 ) ≈ 𝑥𝑥 = 𝑛𝑛𝑘𝑘𝐵𝐵 𝑇𝑇
𝑉𝑉 𝑉𝑉 𝑉𝑉 𝑠𝑠
We have here used the fact that 𝑥𝑥𝑠𝑠 ≪ 1.

Kinetics of sedimentation in colloidal suspensions

We have seen in the previous sections the importance of the existence of a backflow
and convection in the settling column, which is due to the hydrodynamic influence
of the walls of the container. In the laboratory coordinate frame, the volume flux of
colloidal material through a cross sectional surface area perpendicular to the
sedimentation velocity is always compensated by fluid flowing in opposite direction.
This is expressed as

(1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 + 𝜙𝜙𝑆𝑆 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 0

Clearly, backflow tends to decrease sedimentation velocities, especially at larger

volume fractions. Although the fluid backflow may be considered constant at a local
scale, allowing statistical mechanical analysis for a uniform backflow, it certainly
varies significantly from point to point over distances comparable to the size of the

216
 Chapter 8 Settling of (concentrated) suspensions

container as we have discussed earlier. The backflow may be considered

homogeneous over distances small compared to the size of the sample container
and at the same time large compared to the average distance between Brownian
particles. This assumption (called the “chemical approximation” by some
authors 112) allows the chemical potential and the drag on a particle to be seen as
pre-averaged functions of the local mean volume fraction of particles in a region of
size 𝐿𝐿 around the particle, where

particle radius ≪ 𝐿𝐿 ≪ length over which 𝑛𝑛(𝑧𝑧) varies

In volume elements of size 𝐿𝐿, the chemical potential and the drag are thus equal to
the values they would have in a macroscopically uniform suspension of equivalent
concentration 𝑛𝑛0 = 𝑛𝑛(𝑧𝑧). This approximation is only valid if the rate of diffusion of
the particle is greater than its rate of sedimentation. This is represented by the
Péclet number Pe which is a measure of the ratio of the gravitational force on a
particle to its Brownian motion:

4𝜋𝜋(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝑎𝑎3

𝑃𝑃𝑃𝑃 = 𝐿𝐿
3𝑘𝑘𝐵𝐵 𝑇𝑇

The chemical approximation is thus valid for slowly sedimenting dispersions.

Settled particles

Before we turn to the settling phase, let us first come back to the case where the
particles have settled. In Chapter 2, we have found that for non-interacting particles
a density profile could be established:

4 𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤
𝑛𝑛(𝑧𝑧) = 𝑛𝑛0 exp �− 𝜋𝜋𝑎𝑎3 � � 𝑔𝑔𝑔𝑔�
3 𝑘𝑘𝐵𝐵 𝑇𝑇

This relation can be written in the more general form:

𝑭𝑭𝑒𝑒𝑒𝑒𝑒𝑒 ∙ 𝒓𝒓 −𝐹𝐹𝑒𝑒𝑒𝑒𝑒𝑒 𝑧𝑧
𝑛𝑛(𝑧𝑧) = 𝑛𝑛0 exp � � = 𝑛𝑛0 exp � �
𝑘𝑘𝐵𝐵 𝑇𝑇 𝑘𝑘𝐵𝐵 𝑇𝑇

where 𝑭𝑭𝑒𝑒𝑒𝑒𝑒𝑒 is the sum of the external forces applied to a colloidal (Brownian)
particle. In Chapter 2, we have also already said that in the context of colloids, the
pressure associated to the barometric profile is the osmotic pressure:

Π = 𝑛𝑛𝑘𝑘𝐵𝐵 𝑇𝑇

112Buscall, Richard. "The sedimentation of concentrated colloidal suspensions." Colloids and

surfaces 43.1 (1990): 33-53.

217
 Introduction to Colloid Science

Note from the two last equations we find that:

𝑑𝑑Π 1 𝜕𝜕ln(𝑛𝑛) 𝑑𝑑Π

𝐹𝐹𝑒𝑒𝑒𝑒𝑒𝑒 = − =−
𝑑𝑑𝑑𝑑 𝑛𝑛 𝜕𝜕𝜕𝜕 𝑑𝑑𝑑𝑑
which demonstrates that the external force is associated to the osmotic pressure
gradient.

Osmotic pressure and external force

The link between the external force and osmotic pressure can be done using a
thermodynamic approach. Again, we here only briefly sketch the derivation, and
refer to the footnote – and complementary books / courses about thermodynamics
- for more details 113.

In equilibrium the chemical potential is a constant, independent of position. The

driving force for diffusion is equal to gradients in the chemical potential:

𝑭𝑭𝑤𝑤 = −𝛁𝛁𝜇𝜇𝑤𝑤

𝑭𝑭𝑠𝑠 = −𝛁𝛁𝜇𝜇𝑠𝑠

where the minus sign is introduced to indicate that the diffusion current is directed
towards regions of lower chemical potential, so as to minimize the free energy. The
two chemical potentials are not independent quantities: they are related by the
Gibbs-Duhem relation (at constant mechanical pressure and temperature):

𝑛𝑛(𝑧𝑧)𝛁𝛁𝜇𝜇𝑠𝑠 + 𝑛𝑛𝑤𝑤 (𝑧𝑧)𝛁𝛁𝜇𝜇𝑤𝑤 = 0

where 𝑛𝑛𝑤𝑤 is the local number density of solvent molecules. As 𝜙𝜙𝑠𝑠 + 𝜙𝜙𝑤𝑤 = 1 we get:

𝑉𝑉𝑤𝑤 𝑛𝑛𝑤𝑤 + 𝑉𝑉𝑠𝑠 𝑛𝑛 = 1

where 𝑉𝑉𝑤𝑤 and 𝑉𝑉𝑠𝑠 are the volume of a solvent molecule and a colloidal particle.

We note that any force per unit of volume, acting on the solvent and the particles
alike do not produce a relative velocity. The force per unit volume on the fluid is
equal to 𝑭𝑭𝑤𝑤 /𝑉𝑉𝑤𝑤 . The force per unit volume of particle that generate a relative motion
is therefore given by:

𝑭𝑭𝒆𝒆𝒆𝒆𝒆𝒆 𝑭𝑭𝑠𝑠 𝑭𝑭𝑤𝑤

= −
𝑉𝑉𝑠𝑠 𝑉𝑉𝑠𝑠 𝑉𝑉𝑤𝑤

113 See especially Dhont, J. KG. (1996). An introduction to dynamics of colloids. Vol. 2. Elsevier.,
Chapter 7

218
 Chapter 8 Settling of (concentrated) suspensions

Using the equations given above:

𝑭𝑭𝒆𝒆𝒆𝒆𝒆𝒆 𝑛𝑛𝑤𝑤 𝛁𝛁𝜇𝜇𝑤𝑤 −𝛁𝛁𝜇𝜇𝑤𝑤 1

= − = 𝛁𝛁𝜇𝜇
𝑉𝑉𝑠𝑠 1 − 𝑉𝑉𝑤𝑤 𝑛𝑛𝑤𝑤 𝑉𝑉𝑤𝑤 𝑉𝑉𝑠𝑠 𝑛𝑛𝑉𝑉𝑤𝑤 𝑤𝑤

which gives:

1
𝑭𝑭𝒆𝒆𝒆𝒆𝒆𝒆 = 𝛁𝛁𝜇𝜇
𝑛𝑛𝑉𝑉𝑤𝑤 𝑤𝑤

Since the local osmotic pressure is by definition (see the definition above) equal to:
0
𝜇𝜇𝑤𝑤 − 𝜇𝜇𝑤𝑤
Π=−
𝑉𝑉𝑤𝑤
0
where 𝜇𝜇𝑤𝑤 is the chemical potential of pure solvent, we find indeed that:

1
𝑭𝑭𝑒𝑒𝑒𝑒𝑒𝑒 = − 𝛁𝛁Π
𝑛𝑛
1 𝑑𝑑Π
𝐹𝐹𝑒𝑒𝑒𝑒𝑒𝑒 = −
𝑛𝑛 𝑑𝑑𝑑𝑑
The force is therefore directly linked to the osmotic pressure. This force symbolizes
the fact that there is an interaction between the colloidal particles, associated to the
pair-correlation function 𝑔𝑔. The local osmotic pressure can be written (the derivation
is beyond the scope of this book):
∞
2𝜋𝜋 2 𝑑𝑑𝑑𝑑(𝑅𝑅)
Π(𝒓𝒓) = 𝑛𝑛(𝒓𝒓)𝑘𝑘𝐵𝐵 𝑇𝑇 − 𝑛𝑛 (𝒓𝒓) � 𝑅𝑅3 𝑔𝑔(𝑅𝑅)𝑑𝑑𝑑𝑑
3 0 𝑑𝑑𝑑𝑑

where 𝜑𝜑 is the pair potential of interaction between two particles. The first term on
the right-hand side is related to Brownian motion, the second to the interaction
between particles. When 𝜑𝜑 = 0 (no interaction between particles), one finds:

Π = 𝑛𝑛𝑘𝑘𝐵𝐵 𝑇𝑇

which corresponds to the barometric height distribution as given in Chapter 2.

Settling profiles
Let us now discuss the settling phase. The chemical approximation explained above
implies that the thermodynamic and drag forces on a particle are only dependent on
the local average concentration 𝜙𝜙𝑆𝑆 . This, in particular, is at the origin of the
continuity equation:

219
 Introduction to Colloid Science

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕�𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 𝜙𝜙𝑆𝑆 �

+ =0
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
that will be used in Chapter 9. Note that the continuity equation for one particle
reads:
𝑁𝑁
𝜕𝜕Ψ𝑚𝑚 𝜕𝜕�𝑢𝑢𝑛𝑛/𝑙𝑙𝑙𝑙𝑙𝑙 Ψ𝑛𝑛 �
+� =0
𝜕𝜕𝜕𝜕 𝜕𝜕𝑥𝑥𝑛𝑛
𝑛𝑛=1

where Ψ𝑚𝑚 is the probability of finding a particle at a point 𝑥𝑥𝑚𝑚 (and is linked to a pair-
correlation function), 𝑢𝑢𝑛𝑛/𝑙𝑙𝑙𝑙𝑙𝑙 is the velocity of the nth particle.

The forces acting upon a settling particle in a concentrated suspension are: the
gravitational force (compensated for Archimedes) and a force due to the interaction
with other particles. This last one is expressed through the osmotic pressure. The
viscous drag force exerted on the colloidal sphere is balanced by the sum of these
forces:

6𝜋𝜋𝜋𝜋𝜋𝜋 𝑉𝑉𝑝𝑝 𝜕𝜕Π

(𝑣𝑣𝑠𝑠 − 𝑣𝑣𝑤𝑤 ) = −(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝑉𝑉𝑝𝑝 −
𝜒𝜒(𝜙𝜙𝑆𝑆 ) 𝜙𝜙𝑆𝑆 𝜕𝜕𝜕𝜕

The ratio 𝑉𝑉𝑝𝑝 /𝜙𝜙𝑆𝑆 = 1/𝑛𝑛 represents the mean volume element per particle (𝑛𝑛 is the
mean number of particles per unit of volume). The factor 𝜒𝜒(𝜙𝜙𝑆𝑆 ) accounts for the
hydrodynamic interaction effects on the particle mobility 114 and 𝑉𝑉𝑝𝑝 = 4𝜋𝜋𝑎𝑎3 /3 is the
volume of one particle. We have:

𝜒𝜒(𝜙𝜙𝑆𝑆 ) = 1 for 𝜙𝜙𝑆𝑆 → 0

𝜒𝜒(𝜙𝜙𝑆𝑆 ) = 0 for 𝜙𝜙𝑆𝑆 → 1

The last equation implies that for 𝜙𝜙𝑆𝑆 → 1 one has 𝑣𝑣𝑠𝑠 = 𝑣𝑣𝑤𝑤 (= 0). Using 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 =
−(1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑠𝑠 = (1 − 𝜙𝜙𝑆𝑆 )(𝑣𝑣𝑠𝑠 − 𝑣𝑣𝑤𝑤 ), the expression can be rewritten:

6𝜋𝜋𝜋𝜋𝜋𝜋 𝑉𝑉𝑝𝑝 𝜕𝜕Π

𝑣𝑣 = −(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝑉𝑉𝑝𝑝 −
𝑓𝑓(𝜙𝜙𝑆𝑆 ) 𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 𝜙𝜙𝑆𝑆 𝜕𝜕𝜕𝜕

where

𝑓𝑓(𝜙𝜙𝑆𝑆 ) = (1 − 𝜙𝜙𝑆𝑆 )𝜒𝜒(𝜙𝜙𝑆𝑆 )

One relation for 𝑓𝑓(𝜙𝜙𝑆𝑆 ) is for example the Richardson-Zaki expression given above.

114 This implies that 𝜂𝜂 is the viscosity of the medium.

220
 Chapter 8 Settling of (concentrated) suspensions

The equation we have just proposed is not obtained from following a single particle
in time. This would be too complex to do, as it would require to know, at each time,
the interactions of all the surroundings particles with the particle we track, and
moreover we would need the particle’s initial position and velocity which is
impossible to determine. In fact, even though the equation has been written as if it
is for one particle, it is obtained from the pre-averaged functions of the local mean
volume fraction of particles in a region of size L around the particle (see discussion
about the chemical approximation above).

Note

At equilibrium,𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 0 and we find:

𝜕𝜕Π
= (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆
𝜕𝜕𝜕𝜕
from which it can be inferred that it is the gradient in osmotic pressure (i.e. the
interactions between particles) that keeps the colloidal particles suspended. For
large particles (non-colloidal), the osmotic pressure term is negligible in comparison
with the gravity force : this was discussed in Chapter 2 in terms of volume and
surface forces, see discussion for Case (C) and Case (D). We find in that case:

6𝜋𝜋𝜋𝜋𝜋𝜋
𝑣𝑣 = −(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝑉𝑉𝑝𝑝
𝑓𝑓(𝜙𝜙𝑆𝑆 ) 𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙

This implies that the only way to get equilibrium (𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 0) is to have 𝜌𝜌𝑠𝑠 = 𝜌𝜌𝑤𝑤 .

The particle flux along the vertical direction is given by:

𝐽𝐽𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝜙𝜙𝑆𝑆 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = −𝜙𝜙𝑆𝑆 (1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑠𝑠 = −𝜙𝜙𝑆𝑆 𝐽𝐽𝑤𝑤/𝑠𝑠

𝑓𝑓(𝜙𝜙𝑆𝑆 ) 𝑉𝑉𝑝𝑝 𝜕𝜕Π

𝐽𝐽𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = −𝜙𝜙𝑆𝑆 �(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝑉𝑉𝑝𝑝 + �
6𝜋𝜋𝜋𝜋𝜋𝜋 𝜙𝜙𝑆𝑆 𝜕𝜕𝜕𝜕

The osmotic pressure can be written:

𝑘𝑘𝐵𝐵 𝑇𝑇
Π= 𝜙𝜙 𝑍𝑍(𝜙𝜙𝑆𝑆 ) = 𝑍𝑍(𝜙𝜙𝑆𝑆 )𝑛𝑛𝑘𝑘𝐵𝐵 𝑇𝑇
𝑉𝑉𝑝𝑝 𝑆𝑆

where Z(ϕS ) is the compressibility factor that accounts for interparticle interactions.
A widely used expression for a homogeneous hard-sphere suspension is given by the
Carnahan-Starling equation of state 115:

115 Dhont, J. KG. (1996). An introduction to dynamics of colloids. Vol. 2. Elsevier., Chapter 7

221
 Introduction to Colloid Science

1 + 𝜙𝜙𝑆𝑆 + 𝜙𝜙𝑠𝑠2 − 𝜙𝜙𝑠𝑠3

𝑍𝑍(𝜙𝜙𝑆𝑆 ) =
(1 − 𝜙𝜙𝑆𝑆 )3

Incidentally, one can note that in Chapter 7 we introduced the Carnahan-Starling

expression without referring explicitly to the osmotic pressure. We define the
diffusion coefficient and velocity:

𝑘𝑘𝐵𝐵 𝑇𝑇 (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝑉𝑉𝑝𝑝

𝐷𝐷0 = ; 𝑣𝑣0 =
6𝜋𝜋𝜋𝜋𝜋𝜋 6𝜋𝜋𝜋𝜋𝜋𝜋

[𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑣𝑣0 at infinite dilution], and we get:

𝜕𝜕[𝜙𝜙𝑆𝑆 𝑍𝑍(𝜙𝜙𝑆𝑆 )]
𝐽𝐽𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = −𝜙𝜙𝑆𝑆 𝑣𝑣0 𝑓𝑓(𝜙𝜙𝑆𝑆 ) − 𝑓𝑓(𝜙𝜙𝑆𝑆 )𝐷𝐷0
𝜕𝜕𝜕𝜕
A characteristic length can be obtained from 𝐷𝐷0 and 𝑣𝑣0 , namely:

𝑘𝑘𝐵𝐵 𝑇𝑇
𝑙𝑙𝑔𝑔 = 𝐷𝐷0 /𝑣𝑣0 =
(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝑉𝑉𝑝𝑝

which may be seen as the length a particle has to settle before the sedimentation
drift becomes equal to the root mean square displacement due to Brownian motion.

The continuity equation imposes:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕𝐽𝐽𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙
+ =0
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
leading to the general equation that has to be solved in order to find the full
hindered settling profile 𝜙𝜙𝑆𝑆 (𝑧𝑧, 𝑡𝑡):

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 𝜕𝜕 𝜕𝜕[𝜙𝜙𝑆𝑆 𝑍𝑍(𝜙𝜙𝑆𝑆 )]

− 𝑣𝑣0 [𝜙𝜙𝑆𝑆 𝑓𝑓(𝜙𝜙𝑆𝑆 )] = 𝐷𝐷0 �𝑓𝑓(𝜙𝜙𝑆𝑆 ) �
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕

In general, the full solution of this equation can only be obtained numerically 116. The
equation has a Burger-like structure. Burgers’ equation is from the family of
conservation equations that can develop discontinuities and used to characterize
shock waves. The ”shock” in our context is the propagation of the suspension/bed
interface through the fluid.

116 Auzerais, F. M., Jackson, R., & Russel, W. B. (1988). The resolution of shocks and the effects

of compressible sediments in transient settling. J. Fluid Mech, 195(1), 437-462.

222
 Chapter 8 Settling of (concentrated) suspensions

Johannes Martinus Burgers (1895-1981) was a Dutch

physicist, who studied in Leiden under Paul Ehrenfest.
Three months before his PhD graduation he already got
a position as professor at the Technische Hogeschool
Delft (nowadays the Technical University Delft), where
he worked in the laboratory for aero- and
hydrodynamics. In particular, he designed the ventilation
system for the Maastunnel in Rotterdam. In 1955 he
emigrated to the United States for the facilities he was
offered there. The Dutch Burgerscentrum 117 is named
after him.

We will here only discuss some particular cases of the equation for which general
behaviours can be defined. A (numerical) full solution is discussed at the end of the
chapter.

Dilute regime

In the dilute limit 𝜙𝜙𝑆𝑆 ≪ 1 and 𝑓𝑓(𝜙𝜙𝑆𝑆 ) = 𝑍𝑍(𝜙𝜙𝑆𝑆 ) = 1. At equilibrium, 𝜕𝜕𝜙𝜙𝑆𝑆 /𝜕𝜕𝜕𝜕 = 0 and
the equation reduces to:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 2 𝜙𝜙𝑆𝑆
−𝑣𝑣0 = 𝐷𝐷0
𝜕𝜕𝜕𝜕 𝜕𝜕𝑧𝑧 2
This equation can be solved and yields:

𝜙𝜙𝑆𝑆 (𝑧𝑧) = 𝜙𝜙0 𝑒𝑒𝑒𝑒𝑒𝑒�−𝑧𝑧/𝑙𝑙𝑔𝑔 �

where the integration constant 𝜙𝜙0 is obtained by knowing the initial conditions. In
Chapter 2, we have found the same result when we analysed the settling of non-
interacting particles, but expressed it as

4 𝜌𝜌𝑝𝑝 − 𝜌𝜌𝑤𝑤
𝑛𝑛(𝑧𝑧) = 𝑛𝑛0 𝑒𝑒𝑒𝑒𝑒𝑒 �− 𝜋𝜋𝑎𝑎3 � � 𝑔𝑔𝑔𝑔�
3 𝑘𝑘𝐵𝐵 𝑇𝑇

One can verify that 𝑙𝑙𝑔𝑔 = 𝐷𝐷0 /𝑣𝑣0 ≪ 1 for clay particles larger than 100 nm. In that
case 𝜙𝜙𝑆𝑆 (𝑧𝑧 ≠ 0) = 0 and all the particles have settled at the layer 𝑧𝑧 = 0.

Initial settling rate for an initial homogeneously mixed suspension

The initial settling rate of the water/suspension interface is now discussed for the
special case of a suspension consisting of spherical colloidal (Brownian) particles that

117 http://www.jmburgerscentrum.nl/

223
 Introduction to Colloid Science

experience hindered settling. Let us assume that the suspension is well mixed at t =
0:

𝜙𝜙𝑆𝑆 (𝑧𝑧, 𝑡𝑡 = 0) = 𝜙𝜙0

This implies that there are no gradients in concentration, and consequently:

𝜕𝜕𝜕𝜕𝑆𝑆
� � =0
𝜕𝜕𝜕𝜕 𝑧𝑧,𝑡𝑡=0

As the osmotic pressure depends on 𝜙𝜙𝑆𝑆 and:

𝜕𝜕Π 𝑑𝑑Π 𝜕𝜕𝜙𝜙𝑆𝑆

=
𝜕𝜕𝜕𝜕 𝑑𝑑𝜙𝜙𝑆𝑆 𝜕𝜕𝜕𝜕

we find that:

𝜕𝜕Π
� � =0
𝜕𝜕𝜕𝜕 𝑧𝑧,𝑡𝑡=0

We recall that:

𝑓𝑓(𝜙𝜙𝑆𝑆 ) 𝑉𝑉𝑝𝑝 𝜕𝜕Π

𝐽𝐽𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝜙𝜙𝑆𝑆 𝑣𝑣𝑠𝑠 = −𝜙𝜙𝑆𝑆 �(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝑉𝑉𝑝𝑝 + �
6𝜋𝜋𝜋𝜋𝜋𝜋 𝜙𝜙𝑆𝑆 𝜕𝜕𝜕𝜕

From which we deduce that:

𝑓𝑓(𝜙𝜙0 )
(𝒗𝒗𝒔𝒔 )𝑧𝑧,𝑡𝑡=0 = (𝜌𝜌 − 𝜌𝜌𝑤𝑤 )𝑉𝑉𝑝𝑝 𝒈𝒈
6𝜋𝜋𝜋𝜋𝜋𝜋 𝑠𝑠

This is the familiar Stokes’ s law for sedimentation rate of an isolated sphere in a
medium of viscosity 𝜂𝜂 modified through 𝑓𝑓(𝜙𝜙0 ) to account for the increase in fluid
drag due to the hydrodynamic interactions. Note that as soon as 𝑡𝑡 > 0 we will have
𝜕𝜕Π⁄𝜕𝜕𝜕𝜕 ≠ 0 at the bottom of the settling column (as particles cannot penetrate the
bottom). In time, the condition 𝜕𝜕Π⁄𝜕𝜕𝜕𝜕 ≠ 0 will propagate to the top of the column.

Boycott effect

We have (and will) assume that the suspensions we consider are settling vertically.
In 1920 Boycott discovered that the sedimentation of blood components under
gravity force occurred faster in inclined tubes than in vertical tubes. An analytical

224
 Chapter 8 Settling of (concentrated) suspensions

model was later developped, called the PNK theory, after Ponder (1925), Nakamura
and Kuroda (1937) who independently proposed it 118.

Figure opposite: (A) region of particle-free fluid above the

suspension, (B) interface between the particle-free fluid and
the suspension, (C) suspension, (D) thin particle-free fluid
layer ·beneath the downward-facing surface, (E)
concentrated sediment.

When the sediment particles fall onto the side wall of

the column, they form a thin sediment layer that slides
rapidly towards the bottom of the column, hereby
increasing the settling speed. Because of conservation
of volume, clear fluid will appear at the opposite wall
and propagate to the top of the column. The PNK
model states that:

𝑑𝑑𝑑𝑑 𝐻𝐻
= −𝑣𝑣𝑠𝑠 �1 + sin(𝜃𝜃)�
𝑑𝑑𝑑𝑑 𝑏𝑏

When a settling tube is inclined rather than vertical, the

velocity at which colloidal particles sedimentate is
enhanced by several orders of magnitude: this
phenomenon is called the Boycott effect, named after
Arthur Edwin Boycott (1877-1938), an eminent
pathologist and naturalist. In childhood, Boycott
displayed a taste for natural history and was particularly
interested in snails. When he was 15 years of age, he
made a list of those of Herefordshire which was published
in Science Gossip in 1892. His interest in the topic never
faded and the last scientific papers he published were
two memoirs on the habitats of the land and freshwater mollusca in Britain. In 1912
he was appointed professor at the University of Manchester. His colleagues
expected the professor to display a practical interest in the application of pathology
to the everyday problems of clinical work. Boycott would not accept this view. He
mainained that a professor must devote all his energies to the advancement of
science, and that the application of laboratory methods to clinical medicine was not

118Davis, R. H., & Acrivos, A. (1985). Sedimentation of noncolloidal particles at low Reynolds
numbers. Annual Review of Fluid Mechanics, 17(1), 91-118 and Xu, Z. J., & Michaelides, E. E.
(2005). A numerical simulation of the Boycott effect. Chem. Eng. Comm., 192(4), 532-549.

225
 Introduction to Colloid Science

part of his duty. His uncompromised attitude was resented by his medical colleagues
and rendered his tenure of the chair less happy than it should have been. As he grew
older, Boycott arrived at the conclusion that it would be better for the progress of
science if its workers spent less time piling up fresh facts and more in the
contemplation of the implication of those already discovered.

When gravity forces are larger than thermodynamic forces

For hindered settling, assuming smooth initial conditions 119 i.e. 𝜙𝜙𝑆𝑆 (𝑧𝑧, 0) = 𝜙𝜙0 and
𝜙𝜙𝑆𝑆 (ℎ, 𝑡𝑡) = 0 (where the characteristic length ℎ represents the height of the column)
and negligible thermodynamic forces, one can assume that the front’s speed is
constant. When the thermodynamic force is negligible compared to the gravitational
force, one has 𝑃𝑃𝑃𝑃 = 𝑣𝑣0 ℎ/𝐷𝐷0 = ℎ/𝑙𝑙𝑔𝑔 ≫ 1. Noting that the hindered settling relation
can be written:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 𝐷𝐷0 𝜕𝜕 𝜕𝜕[𝜙𝜙𝑆𝑆 𝑍𝑍(𝜙𝜙𝑆𝑆 )]

− [𝜙𝜙 𝑓𝑓(𝜙𝜙𝑆𝑆 )] = �𝑓𝑓(𝜙𝜙𝑆𝑆 ) �
𝜕𝜕(𝑡𝑡𝑣𝑣0 /ℎ) 𝜕𝜕(𝑧𝑧/ℎ) 𝑆𝑆 𝑣𝑣0 ℎ 𝜕𝜕(𝑧𝑧/ℎ) 𝜕𝜕(𝑧𝑧/ℎ)

we introduce the dimensionless variables :

𝑧𝑧 ∗ = 𝑧𝑧/ℎ ; 𝑡𝑡 ∗ = 𝑡𝑡 ∙ 𝑣𝑣0 /ℎ

and therefore the equation becomes:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 1 𝜕𝜕 𝜕𝜕[𝜙𝜙𝑆𝑆 𝑍𝑍(𝜙𝜙𝑆𝑆 )]

− [𝜙𝜙 𝑓𝑓(𝜙𝜙𝑆𝑆 )] = �𝑓𝑓(𝜙𝜙𝑆𝑆 ) �
𝜕𝜕𝑡𝑡 ∗ 𝜕𝜕𝑧𝑧 ∗ 𝑆𝑆 𝑃𝑃𝑃𝑃 𝜕𝜕𝑧𝑧 ∗ 𝜕𝜕𝑧𝑧 ∗

For large Pe the hindered settling equation reduces to:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕
− 𝑣𝑣0 [𝜙𝜙𝑆𝑆 𝑓𝑓(𝜙𝜙𝑆𝑆 )] = 0
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
𝜕𝜕𝜙𝜙𝑆𝑆 𝑑𝑑 𝜕𝜕𝜙𝜙𝑆𝑆
− 𝑣𝑣0 [𝜙𝜙𝑆𝑆 𝑓𝑓(𝜙𝜙𝑆𝑆 )] =0
𝜕𝜕𝜕𝜕 𝑑𝑑𝜙𝜙𝑆𝑆 𝜕𝜕𝜕𝜕

This equation can be solved with the method of characteristics (see Chapter 10 for
details about this method) and leads to:

𝜙𝜙𝑆𝑆 (𝑧𝑧, 𝑡𝑡) = 𝜙𝜙𝑆𝑆 (𝑧𝑧 + 𝑣𝑣𝑣𝑣)

119 Buzzaccaro, S., Tripodi, A., Rusconi, R., Vigolo, D., & Piazza, R. (2008). Kinetics of
sedimentation in colloidal suspensions. Journal of Physics: Condensed Matter, 20(49),
494219. See also Russel W B, Saville D A and Schowalter W R 1992 Colloidal Dispersions
(Cambridge: Cambridge University Press), chapter 12.

226
 Chapter 8 Settling of (concentrated) suspensions

with

𝑑𝑑
𝑣𝑣 = 𝑣𝑣0 [𝜙𝜙 𝑓𝑓(𝜙𝜙𝑆𝑆 )]
𝑑𝑑𝜙𝜙𝑆𝑆 𝑆𝑆

Consequently, in the hindered settling region, 𝜙𝜙𝑆𝑆 is constant along the curves with
slopes 𝑑𝑑𝑑𝑑/𝑑𝑑𝑑𝑑 = 𝑣𝑣.

Substituting 𝜙𝜙𝑆𝑆 (𝑧𝑧, 𝑡𝑡) = 𝜙𝜙𝑆𝑆 (𝑧𝑧 + 𝑣𝑣𝑣𝑣) in the full hindered settling equation and
integrating from z to infinity yields:
∞ ∞ ∞
𝜕𝜕𝜙𝜙𝑆𝑆 (𝑧𝑧, 𝑡𝑡) 𝜕𝜕 𝜕𝜕 𝜕𝜕[𝜙𝜙𝑆𝑆 𝑍𝑍(𝜙𝜙𝑆𝑆 )]
� 𝑑𝑑𝑑𝑑 − 𝑣𝑣0 � [𝜙𝜙𝑆𝑆 𝑓𝑓(𝜙𝜙𝑆𝑆 )]𝑑𝑑𝑑𝑑 = 𝐷𝐷0 � �𝑓𝑓(𝜙𝜙𝑆𝑆 ) � 𝑑𝑑𝑑𝑑
𝑧𝑧 𝜕𝜕𝜕𝜕 𝑧𝑧 𝜕𝜕𝜕𝜕 𝑧𝑧 𝜕𝜕𝑧𝑧 𝜕𝜕𝜕𝜕
∞
𝜕𝜕𝜙𝜙𝑆𝑆 (𝑧𝑧 + 𝑣𝑣𝑣𝑣) 𝜕𝜕[𝜙𝜙𝑆𝑆 𝑍𝑍(𝜙𝜙𝑆𝑆 )]
� 𝑑𝑑𝑑𝑑 − 𝑣𝑣0 𝜙𝜙𝑆𝑆 𝑓𝑓(𝜙𝜙𝑆𝑆 ) = 𝐷𝐷0 𝑓𝑓(𝜙𝜙𝑆𝑆 )
𝑧𝑧 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕

We find that:
∞ ∞
𝜕𝜕𝜙𝜙𝑆𝑆 (𝑧𝑧 + 𝑣𝑣𝑣𝑣) 𝜕𝜕𝜙𝜙𝑆𝑆 (𝑧𝑧 + 𝑣𝑣𝑣𝑣) 𝜕𝜕(𝑧𝑧 + 𝑣𝑣𝑣𝑣)
� 𝑑𝑑𝑑𝑑 = � 𝑑𝑑(𝑧𝑧 + 𝑣𝑣𝑣𝑣) = 𝑣𝑣𝜙𝜙𝑆𝑆 (𝑧𝑧 + 𝑣𝑣𝑣𝑣)
𝑧𝑧 𝜕𝜕𝜕𝜕 (𝑧𝑧+𝑣𝑣𝑣𝑣) 𝜕𝜕(𝑧𝑧 + 𝑣𝑣𝑣𝑣) 𝜕𝜕𝜕𝜕

Therefore:

𝜕𝜕[𝜙𝜙𝑆𝑆 𝑍𝑍(𝜙𝜙𝑆𝑆 )]
[𝑣𝑣 − 𝑣𝑣0 𝑓𝑓(𝜙𝜙𝑆𝑆 )]𝜙𝜙𝑆𝑆 = 𝐷𝐷0 𝑓𝑓(𝜙𝜙𝑆𝑆 )
𝜕𝜕𝜕𝜕
At the water/suspension interface, 𝜙𝜙𝑆𝑆 = 𝜙𝜙0 and the right-hand side of the equation
is equal to zero. This implies that we then have:

𝑣𝑣 = 𝑣𝑣0 𝑓𝑓(𝜙𝜙0 )

This steady-state, time-independent velocity is a direct consequence of the

dependence of the settling velocity on particle concentration. As 𝑣𝑣 is a
monotonically decreasing function of 𝜙𝜙𝑆𝑆 the particles “left behind” by diffusion will
sediment faster (with a velocity 𝑣𝑣0 ) and catch up with the front region of the settling
profile which settles with a velocity 𝑣𝑣 < 𝑣𝑣0 . This front therefore self-sharpens.
Eventually, this leads to a time-invariant shape, dictated by the balance between
settling and diffusion and settling at the uniform speed 𝑣𝑣. In practice, depending on
the size of the column, the time-invariant state might or not be reached as the gelling
point might or not be reached before the steady-state is established. The initial
concentration is also a variable: for a very dilute system for instance, the stationary
profile is uniform in the whole column since then all particles settle with the same
velocity 𝑣𝑣0 .

227
 Introduction to Colloid Science

A similar reasoning enables to find the settling velocity of the suspension/bed

interface: this time the integral should be done from 0 to z. We then also assume
that the settled colloidal particles that form the bed ensure that the bed has a
constant volume fraction 𝜙𝜙𝑚𝑚 . We get:
𝑧𝑧
𝜕𝜕𝜙𝜙𝑆𝑆 (𝑧𝑧 + 𝑣𝑣𝑣𝑣) 𝜕𝜕[𝜙𝜙𝑆𝑆 𝑍𝑍(𝜙𝜙𝑆𝑆 )]
� 𝑑𝑑𝑑𝑑 − 𝑣𝑣0 [𝜙𝜙𝑆𝑆 𝑓𝑓(𝜙𝜙𝑆𝑆 ) − 𝜙𝜙𝑚𝑚 𝑓𝑓(𝜙𝜙𝑚𝑚 )] = 𝐷𝐷0 𝑓𝑓(𝜙𝜙𝑆𝑆 )
0 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕

with
𝑧𝑧 (𝑧𝑧+𝑣𝑣𝑣𝑣)
𝜕𝜕𝜙𝜙𝑆𝑆 (𝑧𝑧 + 𝑣𝑣𝑣𝑣) 𝜕𝜕𝜙𝜙𝑆𝑆 (𝑧𝑧 + 𝑣𝑣𝑣𝑣) 𝜕𝜕(𝑧𝑧 + 𝑣𝑣𝑡𝑡)
� 𝑑𝑑𝑑𝑑 = � 𝑑𝑑(𝑧𝑧 + 𝑣𝑣𝑣𝑣)
0 𝜕𝜕𝜕𝜕 𝑣𝑣𝑣𝑣 𝜕𝜕(𝑧𝑧 + 𝑣𝑣𝑣𝑣) 𝜕𝜕𝜕𝜕
= 𝑣𝑣[𝜙𝜙𝑆𝑆 (𝑧𝑧 + 𝑣𝑣𝑣𝑣) − 𝜙𝜙𝑚𝑚 ]

Noting that 𝑓𝑓(𝜙𝜙𝑚𝑚 ) = 0 as the colloidal particles in the bed do not settle (they are
hard spheres in contact) and that 𝜙𝜙𝑆𝑆 = 𝜙𝜙0 at the suspension/bed interface as in the
hindered settling phase all the particles move with the same average speed, and
therefore the volume fraction is equal to the initial volume fraction, we get the
velocity of the suspension/bed interface:

𝜙𝜙0 𝑓𝑓(𝜙𝜙0 )
𝑣𝑣𝑏𝑏𝑏𝑏𝑏𝑏 = 𝑣𝑣0
𝜙𝜙0 − 𝜙𝜙𝑚𝑚

We here expressed the absolute values of the velocities. As we have taken z = 0 at

the bottom of the column, 𝑣𝑣𝑏𝑏𝑏𝑏𝑏𝑏 is oriented along the z-axis unit vector, whereas 𝑣𝑣 is
antiparallel to it.

Finding the hindered settling function 𝒇𝒇(𝝓𝝓𝑺𝑺 ) from the equilibrium profile

The equation we found for the water/suspension interface velocity

𝜕𝜕[𝜙𝜙𝑆𝑆 𝑍𝑍(𝜙𝜙𝑆𝑆 )]
[𝑣𝑣 − 𝑣𝑣0 𝑓𝑓(𝜙𝜙𝑆𝑆 )]𝜙𝜙𝑆𝑆 = 𝐷𝐷0 𝑓𝑓(𝜙𝜙𝑆𝑆 )
𝜕𝜕𝜕𝜕
can be rearranged into:
−1
𝑓𝑓(𝜙𝜙𝑆𝑆 ) 𝜕𝜕[𝜙𝜙𝑆𝑆 𝑍𝑍(𝜙𝜙𝑆𝑆 )] 1
= �1 + 𝑙𝑙𝑔𝑔 �
𝑓𝑓(𝜙𝜙0 ) 𝜕𝜕𝜕𝜕 𝜙𝜙𝑆𝑆

where we used the fact that 𝑣𝑣 = 𝑣𝑣0 𝑓𝑓(𝜙𝜙0 ) for the interface. Particularly interesting
from an experimental point of view is that this equation enables to extract extensive
dynamic information, i.e. the settling behaviour function 𝑓𝑓(𝜙𝜙𝑆𝑆 ) from a single static
measurement of the equilibrium settling profile for a concentrated suspension. The
alternative would be to measure the sedimentation velocity as function of 𝜙𝜙𝑆𝑆 which
are measurements that require careful temperature control to avoid fluctuations in

228
 Chapter 8 Settling of (concentrated) suspensions

the solvent viscosity. The right-hand-side of the equation contains variables that are
directly measureable from the settling profile and the equilibrium measurements as
we are going to discuss now.

First, we show the measurement results on the settling of a suspension (0.23 volume
fraction) of spherical monodisperse (non-aggregated) colloidal particles of size 77
nm made of MFA (a copolymer). Their density is high, which enables a better
accuracy in particle volume fraction determination by density measurements.
Electrostatic interactions were minimized by the addition of salt. Strong electrostatic
interactions would minimize the settling, as discussed in Chapter 2, case (B).

Left: stationary settling profile close to the water/suspension interface for a suspension of
colloidal hard spheres with 𝜙𝜙0 = 0.23. The measurement is done after the time-invariant
shape is reached. (Note that the bulk volume fraction is still 0.23 close to the interface).
Right: semilog plot of the low volume fraction region. The profile is barometric close to 𝜙𝜙𝑆𝑆 =
𝑑𝑑𝑑𝑑𝑑𝑑
0. The slope (full line) of the data points give 𝑙𝑙𝑔𝑔 (defined underneath) which is sensibly
larger than 𝑙𝑙𝑔𝑔 (dashed line)

229
 Introduction to Colloid Science

Equilibrium sedimentation profile for the same suspension of colloidal hard spheres. The
barometric profile just at the sediment/water interface (between 5 and 6 mm) is similar to
the previous figure.

Dynamic gravitational length

In the region around the water/suspension interface, see the stationary settling
profile, one has 𝜙𝜙𝑆𝑆 ≪ 𝜙𝜙0 and 𝑍𝑍(𝜙𝜙𝑆𝑆 ) = 𝑓𝑓(𝜙𝜙𝑆𝑆 ) = 1. The hindered settling relation
becomes:

𝜕𝜕𝜙𝜙𝑆𝑆 𝑣𝑣0 −𝜙𝜙𝑆𝑆

= [𝑓𝑓(𝜙𝜙0 ) − 1]𝜙𝜙𝑆𝑆 = 𝑑𝑑𝑑𝑑𝑑𝑑
𝜕𝜕𝜕𝜕 𝐷𝐷0 𝑙𝑙𝑔𝑔

where we have defined a dynamic gravitational length:

𝑑𝑑𝑑𝑑𝑑𝑑 𝑙𝑙𝑔𝑔
𝑙𝑙𝑔𝑔 =
1 − 𝑓𝑓(𝜙𝜙0 )

Therefore the top part of the settling profile still has an exponential (barometric)
𝑑𝑑𝑑𝑑𝑑𝑑
shape but with a length 𝑙𝑙𝑔𝑔 > 𝑙𝑙𝑔𝑔 since 𝑓𝑓(𝜙𝜙0 ) < 1. The smaller 𝜙𝜙0 the closer 𝑓𝑓(𝜙𝜙0 )
𝑑𝑑𝑑𝑑𝑑𝑑
comes to 1. In the limit of very dilute suspensions, 𝑙𝑙𝑔𝑔 therefore goes to infinity,
implying that all particles settle according to Stokes, as expected.

The osmotic pressure can be obtained from the equilibrium sedimentation profile.
We have shown earlier that at equilibrium:

𝜕𝜕Π
= (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆
𝜕𝜕𝜕𝜕
Integrating this equation gives:
∞
Π(z) = (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔 � 𝜙𝜙𝑆𝑆 (𝑧𝑧)𝑑𝑑𝑑𝑑
𝑧𝑧

From this integration, the compressibility factor 𝑍𝑍�𝜙𝜙𝑆𝑆 (𝑧𝑧)� can be evaluated using:

Π(𝑧𝑧) = 𝑛𝑛(𝑧𝑧)𝑘𝑘𝐵𝐵 𝑇𝑇 ∙ 𝑍𝑍�𝜙𝜙𝑆𝑆 (𝑧𝑧)�

We recall that 𝑛𝑛(𝑧𝑧) = 𝜙𝜙𝑆𝑆 (𝑧𝑧)/𝑉𝑉𝑝𝑝 . As we have seen, 𝑓𝑓(𝜙𝜙𝑆𝑆 ) is a function of 𝑓𝑓(𝜙𝜙0 ), 𝑙𝑙𝑔𝑔 ,
𝜙𝜙𝑆𝑆 (𝑧𝑧) and 𝑍𝑍(𝜙𝜙𝑆𝑆 ). For each 𝜙𝜙𝑆𝑆 (𝑧𝑧) one can estimate these required variables and
therefore obtain 𝑓𝑓(𝜙𝜙𝑆𝑆 ) for 𝜙𝜙𝑆𝑆 ≤ 𝜙𝜙0 .

230
 Chapter 8 Settling of (concentrated) suspensions

Left: hydrodynamic factor obtained from the stationary settling profile given above. Right:
the osmotic pressure is obtained by integrating the equilibrium sedimentation profile.

Numerical solution for sedimentation of an initially uniform suspension

We here give a simple example of the numerical solution of

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 𝜕𝜕 𝜕𝜕[𝜙𝜙𝑆𝑆 𝑍𝑍(𝜙𝜙𝑆𝑆 )]

− 𝑣𝑣0 [𝜙𝜙𝑆𝑆 𝑓𝑓(𝜙𝜙𝑆𝑆 )] = 𝐷𝐷0 �𝑓𝑓(𝜙𝜙𝑆𝑆 ) �
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕

where we use the Carnahan-Starling equation of state:

1 + 𝜙𝜙𝑆𝑆 + 𝜙𝜙𝑠𝑠2 − 𝜙𝜙𝑠𝑠3

𝑍𝑍(𝜙𝜙𝑆𝑆 ) =
(1 − 𝜙𝜙𝑆𝑆 )3

and the Richardson and Zaki relation:

𝑓𝑓(𝜙𝜙𝑆𝑆 ) = (1 − 𝜙𝜙𝑆𝑆 )6

with the boundary conditions 𝜙𝜙𝑠𝑠 = 𝜙𝜙0 at t = 0. Note that the exponent is now taken
to be 6 for a change (earlier it was shown that 5 was a good value for the data
presented). The exponent depends on the type of particles in suspension. There is
no flux at the bottom and top of the column of height h. The following dimensionless
variables are used:

𝐷𝐷0
𝜏𝜏 = 𝑡𝑡
ℎ2
𝑍𝑍 = 𝑧𝑧/ℎ

From numerical integration one obtains:

231
 Introduction to Colloid Science

The volume fraction versus the distance from the bottom of the column (Z = 0)

As can be seen from that figure, after a short time a clear fluid layer develops at the
top of the container, while particles accumulate at the bottom. An interface at the
top of the container is then seen to move downwards with a constant velocity until
it meets the sediment that is formed at the bottom. From then on the sediment
compacts relatively slowly until the sedimentation-diffusion equilibrium is reached.

Illustrations
Josiah Willard Gibbs
https://en.wikipedia.org/wiki/Josiah_Willard_Gibbs

Jacobus H. van 't Hoff

https://www.nobelprize.org/prizes/chemistry/1901/hoff/biographical/
https://en.wikipedia.org/wiki/Jacobus_Henricus_van_%27t_Hoff

Burgers
http://www.burgers.umd.edu/burgers.html

Arthur Boycott
https://www.npg.org.uk/collections/search/portrait/mw97885/Arthur-Edwin-Boycott

232
 Chapter 9
Permeability of slurries
 Introduction to Colloid Science

One important parameter in soil science is permeability, which quantifies the ability
of a soil to have water (or any fluid) permeate (diffuse through) it. This property is
not only important for agriculture but also for construction: the bottom of ponds for
example should be impermeable to water and so are dikes which should be built
with a soil that will ensure a good water retention. Soils are also generally made up
of different sediment layers and soil quality varies greatly from one layer to another.
An important variable for permeability is grain size: the coarser the grain, the more
permeable a sediment layer is. This led to the following classification:

Some estimations 120 of the permeability K (m/s) and related properties

The measurement of permeability in soils is studied in geoengineering, and then

concerns soils where particles have formed a fabric, also called a skeleton, implying
that the particles are touching each other in a kind of network. We will here discuss
a part of consolidation that lays at the frontier between hydraulic and geo-
engineering: the very early part of consolidation, which happens when a bed is
formed by the settling of (flocculated) particles. We will show that from the time
evolution of the interfaces (blue for the suspension/bed interface and red for the
water/suspension interface in the figure below) important information can be
obtained about the structure of the fabric and the permeability in particular.

120Sobolewski, M. (2005). Various methods of the measurement of the permeability

coefficient in soils-possibilities and application. Electronic Journal of Polish Agricultural
Universities. Series Civil Engineering, 8(2).

234
 Chapter 9 Permeability of slurries

For illustration, we will, in this chapter, study the behaviour of a mud suspension in
an undrained settling column. Undrained means that there is conservation of the
total mass (water + sediment), as the fluid cannot escape the column. This condition
can be relaxed for describing real systems by adapting the boundary conditions.

To describe this initial stage of consolidation of soils, we will distinguish three

different regimes:

t < t1 : the settling regime. The particles are non-touching, and due to gravity, move
to the bottom of the settling column. Both a suspension phase and a bed are
observed.

t1 < t < t2 : the primary consolidation regime. The suspension phase has disappeared.
The particles are in very close vicinity of each other and as they adjust their position
with respect to each other, a significant amount of water is expelled from the fabric.

t > t2: the secondary compression regime. The particles are trapped in their
positions, and significant internal stresses develop at the points where the particles
are touching. Over time, some links between particles can fail or change, leading to
a new reordering of particles and a slow reduction in bed height.

The terms primary consolidation and secondary compression come from soil
mechanics, where traditionally large loads are applied to soils, and their
compression index is studied by measuring the deformation resulting from applying
a load. Here the “load” applied to our settling mud is gravity. The dynamics of the
system are solely governed by gravity and the (complex) interactions between the
moving water and the particles. A full discussion about the settling regime is given
in Chapter 8.

At the bed, the stresses developed between particles are extremely weak compared
to those found in soil mechanics. The “load” felt by a particle in the bed is due to the
weight of other particles and the water above it. To give an order of magnitude, a 1
m high mud suspension with a density of 1200 kg m−3 generates a pressure of
approximately 2 kPa which is 4 or 5 orders of magnitude lower than the pressures
applied in a standard oedometer test (this test is briefly discussed at the end of
Chapter 10). Over time, the load felt by the particles in the bed increases as the
density of the bed increases, but it is evident that the pressures will never come in
range with those applied in soil mechanic tests.

Within the forming bed, the distribution of stresses will be depth-dependent, the
particles closer to the bottom of the column experiencing higher stresses than
particles higher up in the column. One can also argue that as soon as particles are
touching and forming a fabric the problem becomes a 3 dimensional one as the
stresses will be distributed in all directions. This in particular raises the question of
dependence of the final bed height on the column diameter.

235
 Introduction to Colloid Science

Observation of a settling suspension

The careful observation of a settling suspension shows that many factors should be
accounted for in order to get a complete description of the system. A highly cited
article 121 of Coe and Clevenger of 1916 gives an idea of the complexity of the
consolidation process. We here give the description given by the authors in their
article, in which they observe the settling of a pulp suspension.

Settling experiment of Coe and Clevenger, done with a settling pulp

“The first particles which reach the bottom of the cylinder are the coarser granular
sand which may be present in the pulp. Immediately following this and somewhat
contemporaneously with the settling of the sand, the slime flocs nearest the bottom
settle, filling the interstitial spaces between the sand particles, and build up, one
upon another, in a zone of increasing depth. This we term zone D, which may be
defined as that portion of pulp wherein the flocs, considered as integral bodies, have
settled to a point where they rest directly one upon another. After pulp enters zone
D, further separation of liquid must come through liquid pressed out of the flocs and
out of the interstitial spaces between the flocs. Immediately above zone D is a
transition zone C. The pulp in zone C decreases in percentage of solids from the
bottom, where the flocs enter zone D, to the top, where the consistency of the
flocculated pulp is the same as the original pulp. In speaking of flocculated pulp, it is
intended to eliminate from consideration the coarser portion of the contained sand
which falls directly through the overlying zones into zone D. Above zone C is zone B,
of constant consistency of flocculated pulp and of the same consistency as the
flocculated pulp in the feed pulp. Zone A, overlying zone B, is clear water or solution.
In the case of a very rapidly settling time, particularly with material which has been
roasted, zone A in the earlier stages may be turbid, due to the finely divided matter
remaining in suspension. Later, this very fine material settles and the liquid becomes

121 Coe, H. S. "Methods for determining the capacities of slime settling tanks." Transactions

American Institute of Mining Engineerring 55 (1916).

236
 Chapter 9 Permeability of slurries

clear, although there are cases, especially when the liquid contains very little
electrolyte, where it remains turbid for a long time. […]

Schematic representation of a column filled with a suspension with high concentration (high
volume fraction).

We designate as free settling pulp all of the pulp in zones B and C, wherein the sand
and flocs are falling freely through the liquid without pressing on the layers of flocs
beneath, although it is evident, from the peculiar interlocking structure of flocculated
pulp, that there are points of contact between the flocs even in these zones. We
designate as Critical Settling Point the top of zone D just as zone C disappears. At
this point the flocs at the surface just rest upon each other, but compression has not
yet commenced in the surface layer. It is therefore obvious that any elimination of
liquid from zone D cannot be accomplished by free settling but must be effected by
compression of flocs. The water liberated by compression finds its way out of zone D
through tubes or channels which form drainage systems upward through the zone.
The trunk channel for any system has its outlet at the top of zone D. Since zone B is
made of flocculated pulp of constant consistency, the flocs in this zone will settle at
a constant rate so long as the zone exists. If zone C remains shallow and of constant
depth, the liquid being expelled from zone D may be ejected through zone C with
little admixture with this zone. This upward current of liquid which has diffused
through zone B during the first stages of settling may also be ejected through zone
B when B becomes very shallow, and thereby considerably increase the apparent
settling rate of the pulp at this stage. This does not indicate that during this period
more pulp passes into zone D. It is merely a surface phenomenon, indicating that
zone C is very shallow. A curve plotted from the results of a settling test such as is
illustrated in the figure will show a constant settling rate throughout the stages
represented in E, F and G, an increased rate when the stage shown in H is reached,
followed by a rapid retardation in the settling rate after zone B disappears, until zone
C has also disappeared. Following this is a very slow rate of settling which gradually
becomes slower as the water is compressed from zone D. The final stage of settling
is reached when no further liquid is expelled from zone D by compression.”

237
 Introduction to Colloid Science

We are now going to set-up the set of equations required to find the evolution of
the water/suspension and the suspension/bed interfaces as function of time. We
will consider the case where there is no sand inside the mud suspension and
therefore zone D contains only flocs that are so compressed that water can barely
escape from the fabric. We make no distinction between zone C and D.

Link between settling and consolidation

First, we have to understand the role of pressure and stresses in the observed
behaviour. We will start by deriving a Darcy-like equation for settling particles. In the
previous chapter, we have already found a set of equations to be solved for the
settling of colloidal particles. In this chapter, we will make the link between the
formulation found in Chapter 8 and the formulation for settling and consolidation
we are going to derive.

First, we will recall the traditional Darcy equation for porous media, and then we
will show how this equation can be adapted for settling particles.

The Darcy equation

Water, under a pressure gradient ∇P, is forced into a porous medium (for instance a
sandstone, which is used often as a reference stone in petroleum industry). The flux of water
𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 (m/s) coming out of the stone is measured.

The Darcy equation links the flux of water that passes through a porous stone to the
pressure gradient ∇P that is applied to push the water. It is given by:

−𝑘𝑘
𝑱𝑱𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝛁𝛁P
𝜂𝜂

where k (m2) is the permeability and 𝜂𝜂 the viscosity. We have given here the general
Darcy equation in terms of vectors (bold notation). The scalar 𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 expresses the
fact that the flux of the water (subscript w) is defined in the frame of reference of
the laboratory (lab).

The gradient in pressure is a vector that is opposite to the vector of the water flow,
as the flow of water goes from regions of high pressure to regions of low pressure.

238
 Chapter 9 Permeability of slurries

To ensure that we have k > 0 we have a minus sign in front of k. In the following we
will consider only unidirectional flows. For that reason, we will drop the vector
notation for convenience.

The flux 𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 represents a macroscopic flux and is defined as the volume of water
exiting the porous media per unit of surface. Its units are therefore (m3/s)/m2= m/s.
It can be linked to the microscopic flux of water 𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 that flows inside the stone
pores by realising that:

𝑑𝑑𝑑𝑑𝑤𝑤
𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑣𝑣 = 𝜙𝜙𝑤𝑤 𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 = (1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙
𝑑𝑑𝑑𝑑 𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙
where 𝑑𝑑𝑑𝑑𝑤𝑤 is the small element of volume of water at a position z (we use z as we
are going to discuss the case where the flow is vertical in the next section). The total
small element of volume at position z is given by: 𝑑𝑑𝑑𝑑 = 𝑑𝑑𝑑𝑑𝑤𝑤 + 𝑑𝑑𝑑𝑑𝑠𝑠 where 𝑑𝑑𝑑𝑑𝑠𝑠 is the
small element of volume of soil (or solids) at z.

Definitions : volume fraction, porosity and void ratio

In this chapter we have chosen to use

the volume fraction of soil as the
relevant variable of the system. The
reason is that this is the common
variable in colloid science. In other
fields of research, and engineering in
particular, different variables are
preferred which we give here. All the
equations given in this chapter can
therefore be re-written as function of any of these variables.

We define 𝑉𝑉𝑠𝑠 as the total volume of the soil particles, 𝑉𝑉𝑝𝑝 as the total volume of the
pores (the voids filled with water) and 𝑉𝑉 as the total volume of the matrix (soil +
pores): 𝑉𝑉 = 𝑉𝑉𝑠𝑠 + 𝑉𝑉𝑝𝑝

The volume fraction 𝜙𝜙𝑆𝑆 of the soil is defined by:

𝑉𝑉𝑠𝑠
𝜙𝜙𝑆𝑆 =
𝑉𝑉
The porosity 𝜙𝜙 (also called void fraction) is defined by:

𝑉𝑉𝑝𝑝
𝜙𝜙 =
𝑉𝑉
The void ratio 𝑒𝑒 is defined by:

239
 Introduction to Colloid Science

𝑉𝑉𝑝𝑝
𝑒𝑒 =
𝑉𝑉𝑠𝑠

The relation between these variables is given by:

𝜙𝜙𝑆𝑆 + 𝜙𝜙 = 1 ; 𝜙𝜙 = 𝑒𝑒/(1 + 𝑒𝑒) ; 𝜙𝜙𝑆𝑆 = 1/(1 + 𝑒𝑒) ; 𝑒𝑒 = 𝜙𝜙/(1 − 𝜙𝜙)

Theoretically the porosity and volume fractions can vary between 0 and 1. The void
ratio, on the other hand, can vary between zero (for 𝜙𝜙 = 0) and infinity (for 𝜙𝜙 = 1).

The volume fraction is usually determined from the mass fraction (amount of dry
mass of soil present in the water). The dry mass of soil of a sample can be estimated
from drying the sample in an oven at 105°C for several hours. The density of a clay
gives the relation between mass and volume.

The size of the voids in a soil can be estimated by mercury intrusion and nitrogen
desorption techniques.

Mercury is a non-wetting fluid and when it is injected at a measured pressure into a

material the pore size can be estimated based on the relation (Washburn’s
equation) between pressure and pore size: the pressure difference is inversely
proportional to pore size.

Nitrogen is a non-corrosive gas that can be used to determine the specific surface
area of the pores, where the BET (Brunauer – Emmett – Teller) theory is used to
quantify the adsorption of the gas molecules on the solid surfaces. The BET theory
is an extension for polylayers of the monolayer Langmuir theory for the adsorption
of molecules.

In a similar way, we can define the volume fraction of water by:

𝑉𝑉𝑤𝑤
𝜙𝜙𝑤𝑤 =
𝑉𝑉
We recall that in the relation above we have defined:

𝑑𝑑𝑑𝑑𝑤𝑤
𝜙𝜙𝑤𝑤 =
𝑑𝑑𝑑𝑑
There is no contradiction in these definitions. The first one is a general one: the
volume fraction of water in a given system is the ratio between the volume of water
present and the total volume. The second definition is more specific: the volume
ratio at a given position z is the ratio between the small volume of water at z and the
small total volume at z. To emphasize that the second relation is at a given z, we
could write:

240
 Chapter 9 Permeability of slurries

𝑑𝑑𝑑𝑑𝑤𝑤
𝜙𝜙𝑤𝑤 (𝑧𝑧) =
𝑑𝑑𝑑𝑑
The flux of water can also be expressed in the rest frame of the particles. If the
particles are not moving, they are at rest in the laboratory frame their velocity is zero
and therefore:

𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 0

From which we deduce:

𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑣𝑣𝑤𝑤/𝑠𝑠 + 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑣𝑣𝑤𝑤/𝑠𝑠

where 𝑣𝑣𝑤𝑤/𝑠𝑠 is the velocity of water inside the pores in the frame of reference of the
soil (s). The relation between the macroscopic water flux coming out of the porous
media (measured in the reference frame of the laboratory) can now be expressed as
function of the water flux in the reference frame of the particles as:

−𝑘𝑘
𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 = (1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑠𝑠 = ∇P
𝜂𝜂

where 𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 is the macroscopic velocity of the fluid (m/s), 𝑘𝑘 the permeability (m2).

A Darcy equation for settling particles

We want to write a similar relation, but now in the case that the particles are moving
(settling) in a column. We do this by setting-up a thought experiment. This is
illustrated by the picture opposite. The left figure depicts, in the frame of the
laboratory, what is happening during settling: both water and particles are moving
with velocities that we define as 𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙
and 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 respectively. On the right figure
(our thought experiment), we have applied
a (fictive) macroscopic counter flow
−𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 generated by a (not yet defined)
pressure gradient ∇P ∗ . In that case, the
new velocities of water and particles
∗ ∗
𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 and 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 become:
∗
𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 − 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑣𝑣𝑤𝑤/𝑠𝑠
∗
𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 − 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 0

241
 Introduction to Colloid Science

We are now satisfying Darcy conditions, as only water is moving and particles are
immobile. We can therefore write:

∗
−𝑘𝑘 ∗
𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 = −𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = ∇P
𝜂𝜂

In a next step, we would like to express 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 (which is both the microscopic velocity
∗
of the settling particles and the minus the macroscopic flow 𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 applied in the
mind experiment) as function of 𝑣𝑣𝑤𝑤/𝑠𝑠 , i.e. the microscopic velocity of the water in
the frame of the particles. We make the hypothesis that there is not net volume flux
in the settling column (this implies that no water or soil is entering or leaving the
column during the experiment 122). This condition imposes that the volume flux of
water should be minus the volume flux of soil at any height z:

𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 = −𝐽𝐽𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙

Using the same lines of derivation as above, we have:

𝐽𝐽𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝜙𝜙𝑆𝑆 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙

which leads to:

(1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 + 𝜙𝜙𝑆𝑆 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 0

This equation can further be developed into:

(1 − 𝜙𝜙𝑆𝑆 )(𝑣𝑣𝑤𝑤/𝑠𝑠 + 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 ) + 𝜙𝜙𝑆𝑆 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 0

(1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑠𝑠 = −𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙

The relation between the macroscopic water flux inside the column in the mind
experiment and the macroscopic one in the real experiment is given by:
∗
𝐽𝐽𝑤𝑤/𝑠𝑠 = (1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑠𝑠 = 𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙

Note the difference between the fluxes: one is in the frame of reference of the
laboratory, the other in the frame of reference of the soil. Using the Darcy equation
∗
obtained for 𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 we now obtain a modified Darcy equation for settling particles
in an undrained column:

122Usually one speaks of an undrained system: in most experimental cases (also in-situ) the
question is whereas water is leaving or not the system, either by seepage at the bottom of
the column, or by evaporation at the top. The soil concentration is not varying in typical
experiments.

242
 Chapter 9 Permeability of slurries

−𝑘𝑘 ∗
𝐽𝐽𝑤𝑤/𝑠𝑠 = ∇P
𝜂𝜂

The question remains so as to properly define ∇P ∗ . We have said that it is the

pressure gradient necessary to generate the (fictive) counter flow −𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 in the case
of our mind experiment. Conversely, it can be seen as minus the pressure gradient
created by the settling of the particles in the real experiment. Note that ∇P ∗ is a
macroscopic pressure difference, generating the macroscopic fluid velocity 𝐽𝐽𝑤𝑤/𝑠𝑠 .
Other relations could be set-up to express the microscopic fluid flow 𝑣𝑣𝑤𝑤/𝑠𝑠 and
microscopic pressure gradients could then be defined, acting within the pores of the
fabric. These relations are beyond our present scope. We are now going to discuss
more into details the form of ∇P ∗ .

Pressure gradient

In Darcy experiments, a pressure difference is applied across the sample and the
water flux is measured. The pressure gradient can then be defined as:

∆P
∇P =
𝐿𝐿
where ∆P is the pressure difference between one end and the other of the sample,
in the direction of the water flow, and 𝐿𝐿 is the length of that section of sample. In
the case of settling particles in an undrained column, no external pressure is applied.
There is the hydrostatic pressure, due to gravity:

Pℎ𝑦𝑦𝑦𝑦 (𝑧𝑧) = 𝜌𝜌𝑤𝑤 𝑔𝑔(ℎ − 𝑧𝑧) + 𝑃𝑃0

where 𝜌𝜌𝑤𝑤 is the density of water, 𝑔𝑔 is the gravitational acceleration, ℎ is the height
of the liquid in the column, 𝑧𝑧 = 0 is the position at the bottom of the column and 𝑃𝑃0
is the atmospheric pressure. This hydrostatic pressure does not contribute to 𝐽𝐽𝑤𝑤/𝑠𝑠 .
It is easy to convince oneself of this fact: let’s assume that the column is filled by a
porous media such as a rock or a pile of glass beads. The fabric is not moving, and
neither is the water, even though a hydrostatic pressure is present.

There is however another pressure borne by the water and caused by the settling of
the particles. We define P𝑤𝑤 as the total pressure borne by the water. In order to
express P𝑤𝑤 we will use the concept of stress. We recall that:

Pressure is a force applied perpendicular to a surface per unit area.

Stress is a force per unit of surface, not necessarily perpendicular to that surface.
Stress can arise from the internal forces that neighbouring particles exert on each
other: stress might therefore exist in the absence of external forces.

243
 Introduction to Colloid Science

Strain is the measure of the deformation of a material related to a stress. It has no

dimension. For instance it can be the difference between the length of a bar before
and after applying a stress, divided by the length of the bar (before or after applying
the stress). Stress can exist in a system without noticeable strain.

In the case of a column of soil, fully saturated with water, the stress in the vertical
direction is defined by:

σ𝑧𝑧𝑧𝑧 = 𝑃𝑃𝑤𝑤 + σ𝑠𝑠𝑠𝑠

where σ𝑠𝑠𝑠𝑠 is the stress associated to the skeleton. In order to get a better
understanding of σ𝑧𝑧𝑧𝑧 let us first consider the case where the particles are not
settling (a rock saturated with water for instance) with everywhere the same volume
fraction 𝜙𝜙𝑆𝑆 . We then have hydrostatic conditions (the water is not moving) and
hence:

P𝑤𝑤 (𝑧𝑧) = Pℎ𝑦𝑦𝑦𝑦 (𝑧𝑧)

We then also have:

σ𝑧𝑧𝑧𝑧 (𝑧𝑧) = 𝜌𝜌𝜌𝜌(ℎ − 𝑧𝑧) + 𝑃𝑃0

with

𝜌𝜌 = (1 − 𝜙𝜙𝑆𝑆 )𝜌𝜌𝑤𝑤 + 𝜙𝜙𝑆𝑆 𝜌𝜌𝑠𝑠

where 𝜌𝜌 is the density of the sample (water + soil) and 𝜌𝜌𝑠𝑠 is the density of the soil
itself. Recombining the equations, we find:

σ𝑠𝑠𝑠𝑠 (𝑧𝑧) = (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔(ℎ − 𝑧𝑧)𝜙𝜙𝑆𝑆

If the soil and the water have the same density (𝜌𝜌𝑠𝑠 = 𝜌𝜌𝑤𝑤 ), we get σ𝑠𝑠𝑠𝑠 = 0 and the
total vertical stress is the same as the one given by a column full of water:

σ𝑧𝑧𝑧𝑧 (𝑧𝑧) = 𝜌𝜌𝑤𝑤 𝑔𝑔(ℎ − 𝑧𝑧) + 𝑃𝑃0

which is expected. In the other limit, when the column is only filled with soil (no
water), we have 𝜙𝜙𝑆𝑆 = 1 and:

σ𝑧𝑧𝑧𝑧 (𝑧𝑧) = 𝜌𝜌𝑠𝑠 𝑔𝑔(ℎ − 𝑧𝑧) + 𝑃𝑃0

In that case σ𝑧𝑧𝑧𝑧 = σ𝑠𝑠𝑠𝑠 as one can consider that 𝑃𝑃𝑤𝑤 = 0 since there is no water. In
the case of settling particles in a column (the water is moving), we now have
P𝑤𝑤 (𝑧𝑧) ≠ Pℎ𝑦𝑦𝑦𝑦 (𝑧𝑧) and the pressure P ∗ we are looking for in order to define the
pressure gradient can be defined as:

244
 Chapter 9 Permeability of slurries

P ∗ (𝑧𝑧) = P𝑒𝑒 (z) = P𝑤𝑤 (𝑧𝑧) − Pℎ𝑦𝑦𝑦𝑦 (𝑧𝑧)

This pressure is called the excess pore pressure and is usually noted P𝑒𝑒 . In
hydrostatic conditions, we have already discussed that P𝑤𝑤 = Pℎ𝑦𝑦𝑦𝑦 and therefore
P𝑒𝑒 = 0.

It is important to realize that the excess pore pressure is solely generated by the
settling velocity of the particles. Due to their slow motion, stresses caused by vertical
accelerations are not considered. The excess pore pressure is the pressure leading
to the water flow 𝑣𝑣𝑤𝑤/𝑠𝑠 .

We have:

∇P𝑒𝑒 = ∇P𝑤𝑤 − ∇Pℎ𝑦𝑦𝑦𝑦

Using the definition σ𝑧𝑧𝑧𝑧 = 𝑃𝑃𝑤𝑤 + σ𝑠𝑠𝑠𝑠 we get:

∇P𝑒𝑒 = ∇(σ𝑧𝑧𝑧𝑧 − σ𝑠𝑠𝑠𝑠 ) − ∇Pℎ𝑦𝑦𝑦𝑦

In order to estimate σ𝑧𝑧𝑧𝑧 we make the hypothesis that the consolidation is slow and
that we have nearly hydrostatic conditions (fluid accelerations are neglected):

𝑑𝑑σ𝑧𝑧𝑧𝑧 (𝑧𝑧) = −��1 − 𝜙𝜙𝑆𝑆 (𝑧𝑧)�𝜌𝜌𝑤𝑤 + 𝜙𝜙𝑆𝑆 (𝑧𝑧)𝜌𝜌𝑠𝑠 �𝑔𝑔𝑔𝑔𝑔𝑔

We find:

𝜕𝜕𝑃𝑃𝑒𝑒 𝜕𝜕σ𝑧𝑧𝑧𝑧 𝜕𝜕σ𝑠𝑠𝑠𝑠

= 𝜌𝜌𝑤𝑤 𝑔𝑔 + −
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
Leading to:

𝜕𝜕𝑃𝑃𝑒𝑒 𝜕𝜕σ𝑠𝑠𝑠𝑠
− = + (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕

The Gibson equation

The Gibson equation gives the changes of volume fraction 𝜙𝜙𝑆𝑆 as function of time and
space. The solution of this equation is used to fit the consolidation data and obtain
relevant parameters for the system, one of which being the permeability. We will
now set-up this equation by combining the Darcy equation for settling particles and
the continuity equation.

The continuity equation imposes that mass is conserved. This can be written like:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕�𝑣𝑣𝑠𝑠/𝑙𝑙𝑎𝑎𝑏𝑏 𝜙𝜙𝑆𝑆 �

+ =0
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕

245
 Introduction to Colloid Science

General derivation of the continuity equation

In order to derive the continuity equation, one has to put in equation the following
observation: conservation of mass (a very fundamental principle, that only breaks
down when one studies nuclear reactions) requires that no mass can be lost or
gained from a small volume element. Let us first define mass as:

𝑚𝑚 = 𝜌𝜌𝜌𝜌

where 𝜌𝜌 is the density of the medium and 𝑉𝑉 symbolizes the (small) volume
considered which has a (small) mass 𝑚𝑚. We now consider a small volume element 𝑉𝑉
(symbolized by the red cylinder underneath), through which a fluid is moving in the
𝑥𝑥 -direction and we define the change of mass inside this volume as function of time
as:

𝑑𝑑𝑑𝑑 = 𝑚𝑚𝑖𝑖𝑖𝑖 − 𝑚𝑚𝑜𝑜𝑜𝑜𝑜𝑜

Both the entering (“in”) and exiting (“out”) mass are defined as 𝑚𝑚𝑘𝑘 = 𝜌𝜌 ∙ 𝑉𝑉𝑘𝑘 where
the volumes 𝑉𝑉𝑘𝑘 are defined in grey in the figure and 𝑘𝑘 stands for “in” or “out”.

The velocity of the fluid is given by 𝑣𝑣. We get:

𝑑𝑑𝑑𝑑 = 𝜌𝜌(𝑥𝑥) ∙ 𝑣𝑣(𝑥𝑥) ∙ 𝑑𝑑𝑑𝑑 ∙ 𝑑𝑑𝑑𝑑 − 𝜌𝜌(𝑥𝑥 + 𝑑𝑑𝑑𝑑) ∙ 𝑣𝑣(𝑥𝑥 + 𝑑𝑑𝑑𝑑) ∙ 𝑑𝑑𝑑𝑑 ∙ 𝑑𝑑𝑑𝑑

𝑑𝑑[𝜌𝜌(𝑥𝑥) ∙ 𝑣𝑣(𝑥𝑥)]
𝑑𝑑𝑑𝑑 = − ∙ 𝑑𝑑𝑑𝑑 ∙ 𝑑𝑑𝑑𝑑 ∙ 𝑑𝑑𝑑𝑑
𝑑𝑑𝑑𝑑
𝑑𝑑𝑑𝑑 𝑑𝑑[𝜌𝜌(𝑥𝑥) ∙ 𝑣𝑣(𝑥𝑥)]
=− ∙ 𝑑𝑑𝑑𝑑
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑
The same reasoning can be made if the fluid would move in the other directions
(𝑦𝑦, 𝑧𝑧) as well. If we add all the mass differences from all directions, we get:

𝑑𝑑(𝑚𝑚𝑥𝑥 + 𝑚𝑚𝑦𝑦 + 𝑚𝑚𝑧𝑧 ) 𝑑𝑑(𝜌𝜌 ∙ 𝑣𝑣𝑥𝑥 ) 𝑑𝑑�𝜌𝜌 ∙ 𝑣𝑣𝑦𝑦 � 𝑑𝑑(𝜌𝜌 ∙ 𝑣𝑣𝑧𝑧 )

= −� + + � ∙ 𝑑𝑑𝑑𝑑
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑

In the illustration we have taken 𝑑𝑑𝑑𝑑 to be circular. This is not a requirement. The
only assumption that is needed is that the velocity vector should always be taken

246
 Chapter 9 Permeability of slurries

perpendicular to the surface (if it is not, only the component perpendicular to the
surface should be used). Mathematically, this means that one can write the previous
equation in the general form:

𝑑𝑑𝑑𝑑
= −𝛁𝛁 ∙ (𝜌𝜌 ∙ 𝒗𝒗) ∙ 𝑑𝑑𝑑𝑑
𝑑𝑑𝑑𝑑
where 𝑑𝑑𝑑𝑑 is the total mass change inside the volume element considered (which
can be of any shape). This equation can be expressed in any coordinate system. In
Cartesian coordinates, one will find the equation we derived above.

As we are looking at a very general case, we may assume that inside the volume
element considered there exist sources and sinks, where matter (and mass) can be
appearing or disappearing: for instance roots that can take up water, or a leaking
pipe. This leads to a change in mass that can be quantified by fluxes:

𝑑𝑑𝑚𝑚𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑜𝑜𝑜𝑜 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = 𝜌𝜌 ∙ (𝑞𝑞𝑠𝑠𝑠𝑠𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 − 𝑞𝑞𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 ) ∙ 𝑑𝑑𝑑𝑑 ∙ 𝑑𝑑𝑑𝑑

where 𝑞𝑞 represent the flux in (s-1). We therefore get:

𝑑𝑑𝑚𝑚𝑡𝑡𝑡𝑡𝑡𝑡
= −𝛁𝛁 ∙ (𝜌𝜌 ∙ 𝒗𝒗) ∙ 𝑑𝑑𝑑𝑑 + 𝜌𝜌 ∙ (𝑞𝑞𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 − 𝑞𝑞𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 ) ∙ 𝑑𝑑𝑑𝑑
𝑑𝑑𝑑𝑑
Realizing that 𝑑𝑑𝑚𝑚𝑡𝑡𝑡𝑡𝑡𝑡 = 𝜌𝜌𝜌𝜌𝜌𝜌 and

𝜕𝜕𝑚𝑚𝑡𝑡𝑡𝑡𝑡𝑡 𝜕𝜕𝜕𝜕
𝑑𝑑𝑚𝑚𝑡𝑡𝑡𝑡𝑡𝑡 = � � 𝑑𝑑𝑑𝑑 = ∙ 𝑑𝑑𝑑𝑑 ∙ 𝑑𝑑𝑑𝑑
𝜕𝜕𝜕𝜕 at V fixed 𝜕𝜕𝜕𝜕

we get:

∂ρ
+ 𝛁𝛁 ∙ (ρ ∙ 𝐯𝐯) = ρ ∙ (qsource − qsink )
∂t
This is the general form of the continuity equation. In the specific case that the
particle are incompressible (ρ𝑠𝑠 is constant), and that there are no sink and source
terms, and that the movement is along the z-axis, we may use ρ = ρ𝑠𝑠 ϕ𝑠𝑠 and we find

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕�𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 𝜙𝜙𝑆𝑆 �

+ =0
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕

One can show that the continuity equation can also be expressed as:

𝜕𝜕𝜕𝜕 𝜕𝜕 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙
− + (1 + 𝑒𝑒)2 � �=0
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 1 + 𝑒𝑒

247
 Introduction to Colloid Science

This is the form most used in geoengineering. In hydraulic engineering, where the
free settling phase of the particles is also considered, it is preferable to work with
volume fraction. The reasons are expressed by Eric Toorman in 1996 123:

“In this text the solids volume fraction 𝜙𝜙𝑆𝑆 is used as the dependent variable to
describe the solids content. There are several reasons for this. The solids
concentration is preferred to the void ratio because in the sedimentation phase e
becomes infinite when the concentration decreases to zero. Furthermore, if one has
to consider the behaviour of a polydisperse sediment, the particle concentration must
be taken as the dependent variable, since the mass balance must be solved now for
each fraction separately. The volumetric solids concentration is also the best variable
to relate other suspension properties (e.g. rheological properties) with.”

We recall the modified Darcy equation obtained above:

−𝑘𝑘 𝜕𝜕𝑃𝑃𝑒𝑒
−𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 =
𝜂𝜂 𝜕𝜕𝜕𝜕

Combining these equations, one obtains:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 𝑘𝑘 𝜕𝜕𝑃𝑃𝑒𝑒
=− � 𝜙𝜙 �
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜂𝜂 𝜕𝜕𝜕𝜕 𝑆𝑆

which can further be developed into:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 𝑘𝑘 𝑘𝑘 𝜕𝜕σ𝑠𝑠𝑠𝑠
= � (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑠𝑠2 + 𝜙𝜙𝑆𝑆 �
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜂𝜂 𝜂𝜂 𝜕𝜕𝜕𝜕

This equation is called the Gibson equation. Usually a new permeability is defined
as:
𝑔𝑔𝜌𝜌𝑤𝑤
𝐾𝐾 (m⁄s) = 𝑘𝑘(m2 )
𝜂𝜂

where 𝐾𝐾 (m⁄s) is usually called hydraulic conductivity and 𝑘𝑘(m2 ) permeability. At

the beginning of the chapter we have seen that 𝐾𝐾 (m⁄s) is also called coefficient of
permeability by some authors.

The Gibson equation 124 can be written:

123 Toorman, E.A. (1996) “Sedimentation and self-weight consolidation: general unifying
theory”, Géotechnique 46, No 1, 103-113.
124 Gibson, R. E., G. L. England, and M. J. L. Hussey. "The Theory of one-dimensional

consolidation of saturated clays: 1. finite non-Linear consildation of thin homogeneous

layers." Geotechnique 17.3 (1967): 261-273.

248
 Chapter 9 Permeability of slurries

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 2 𝐾𝐾 𝜕𝜕σ𝑠𝑠𝑠𝑠

= �𝐾𝐾 𝜙𝜙𝑠𝑠 + 𝜙𝜙𝑆𝑆 �
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜌𝜌𝑤𝑤 𝑔𝑔𝜌𝜌𝑤𝑤 𝜕𝜕𝜕𝜕

Consolidation when gas is present in the soil

We now consider the consolidation of a soil in which gas bubbles are trapped 125. We
will only study the case where t > t1 (the interface water/suspension has
disappeared). If the gas bubbles are fixed to the soil skeleton one simply gets:

𝐽𝐽𝑤𝑤/𝑠𝑠 = 𝑆𝑆𝑟𝑟 (1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑠𝑠

where 𝑆𝑆𝑟𝑟 is the degree of saturation : 𝑆𝑆𝑟𝑟 = 1 is for the case there is no gas inside the
soil.

We will now consider the general case, where gas bubbles are distributed within the
soil in an undetermined way. One defines the porosity as:

𝜙𝜙 = 𝜙𝜙𝑔𝑔 + 𝜙𝜙𝑤𝑤

where:

𝜙𝜙𝑔𝑔 = (1 − 𝑆𝑆𝑟𝑟 )𝜙𝜙 ; 𝜙𝜙𝑤𝑤 = 𝑆𝑆𝑟𝑟 𝜙𝜙 = 𝑆𝑆𝑟𝑟 (1 − 𝜙𝜙𝑆𝑆 )

The total vertical stress can be estimated, for t > t1, by:
ℎ𝑏𝑏 𝐿𝐿
σ𝑧𝑧𝑧𝑧 (𝑧𝑧) = � 𝜌𝜌(𝑧𝑧)𝑔𝑔𝑔𝑔𝑔𝑔 + � 𝜌𝜌(𝑧𝑧)𝑔𝑔𝑔𝑔𝑔𝑔 + 𝑃𝑃0
𝑧𝑧 ℎ𝑏𝑏

Between 𝑧𝑧 and ℎ𝑏𝑏 we have :

𝜌𝜌(𝑧𝑧) = 𝜙𝜙𝑤𝑤 𝜌𝜌𝑤𝑤 + 𝜙𝜙𝑆𝑆 𝜌𝜌𝑠𝑠 + 𝜙𝜙𝑔𝑔 𝜌𝜌𝑔𝑔

If we assume that we can neglect the weight of the gas (𝜌𝜌𝑔𝑔 ≪ 𝜌𝜌𝑠𝑠 , 𝜌𝜌𝑤𝑤 ), we get:

𝜌𝜌(𝑧𝑧) = 𝜙𝜙𝑤𝑤 𝜌𝜌𝑤𝑤 + 𝜙𝜙𝑆𝑆 𝜌𝜌𝑠𝑠

Therefore:
ℎ𝑏𝑏 (𝑧𝑧)
σ𝑧𝑧𝑧𝑧 (𝑧𝑧) = � [𝜙𝜙𝑤𝑤 𝜌𝜌𝑤𝑤 + 𝜙𝜙𝑆𝑆 𝜌𝜌𝑠𝑠 ]𝑔𝑔𝑔𝑔𝑔𝑔 + 𝜌𝜌𝑤𝑤 𝑔𝑔(𝐿𝐿 − ℎ𝑏𝑏 ) + 𝑃𝑃0
𝑧𝑧

125 For more details, see “ Consolidation behaviour of gassy mud: theory and experimental
validation”, B. Wichman, PhD thesis (1999), TU Delft

249
 Introduction to Colloid Science

For numerical reasons, one likes to avoid having ℎ𝑏𝑏 , which is depending on z, as an
upper bound of an integral. To get a fixed upper bound, the following change in
variable is introduced:

𝑑𝑑𝑑𝑑𝑠𝑠 = 𝜙𝜙𝑠𝑠 𝑑𝑑𝑑𝑑

The variable 𝑧𝑧𝑠𝑠 is called the material coordinate (for the soil).

Material coordinates

A small volume 𝑑𝑑𝑑𝑑 = 𝑆𝑆 𝑑𝑑𝑑𝑑 where 𝑆𝑆 is the cross section area of the column. The volume can
be expressed as : 𝑑𝑑𝑑𝑑 = 𝑑𝑑𝑑𝑑𝑤𝑤 + 𝑑𝑑𝑑𝑑𝑠𝑠 where 𝑑𝑑𝑑𝑑𝑤𝑤 represents the volume of water and 𝑑𝑑𝑑𝑑𝑠𝑠 the
volume of soil Left figure: soil particles in suspension in 𝑑𝑑𝑑𝑑. Right figure: the height 𝑑𝑑𝑑𝑑𝑤𝑤 and
𝑑𝑑𝑧𝑧𝑠𝑠 are defined by: 𝑑𝑑𝑑𝑑𝑤𝑤 = 𝑆𝑆 𝑑𝑑𝑑𝑑𝑤𝑤 and 𝑑𝑑𝑑𝑑𝑠𝑠 = 𝑆𝑆 𝑑𝑑𝑑𝑑𝑠𝑠 .

Let us consider a slice 𝑑𝑑𝑑𝑑 of the column of cross area 𝑆𝑆. This slice is composed of
water and soil, therefore we can define two (fictive) heights such that 𝑆𝑆𝑆𝑆𝑆𝑆𝑤𝑤
corresponds to the volume of water inside the slice and 𝑆𝑆𝑆𝑆𝑆𝑆𝑠𝑠 to the volume of soil.
We have:

𝑆𝑆𝑆𝑆𝑆𝑆 = 𝑆𝑆𝑆𝑆𝑆𝑆𝑤𝑤 + 𝑆𝑆𝑆𝑆𝑆𝑆𝑠𝑠

Dividing both sides of this equation by 𝑆𝑆𝑆𝑆𝑆𝑆 we get:

𝑆𝑆𝑑𝑑𝑑𝑑𝑤𝑤 𝑆𝑆𝑆𝑆𝑆𝑆𝑠𝑠
1= +
𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆𝑆𝑆
realising that:

𝑆𝑆𝑆𝑆𝑆𝑆𝑠𝑠 𝑉𝑉𝑠𝑠 𝑑𝑑𝑑𝑑𝑤𝑤

= = 𝜙𝜙𝑠𝑠 and = 𝜙𝜙𝑤𝑤
𝑆𝑆𝑆𝑆𝑆𝑆 𝑉𝑉 𝑑𝑑𝑑𝑑
we get:

𝜙𝜙𝑠𝑠 + 𝜙𝜙𝑤𝑤 = 1

The material coordinate is defined, is given by:

𝑧𝑧 𝑧𝑧𝑠𝑠
𝑧𝑧𝑠𝑠 (𝑧𝑧) = � 𝜙𝜙𝑆𝑆 𝑑𝑑𝑑𝑑 = � 𝑑𝑑𝑧𝑧𝑠𝑠
0 0

250
 Chapter 9 Permeability of slurries

It is the height of solid there would be theoretically found in the column if all the
solids between 0 and z would be squeezed at the bottom. Note that:

𝑧𝑧𝑠𝑠 (ℎ ) = 𝐿𝐿𝐺𝐺

where ℎ is the height of the suspended particles + bed. (we assume to be in the
special case where ℎ = ℎ𝑏𝑏 ). Following a similar reasoning one can define the
material coordinate for water:
𝑧𝑧 𝑧𝑧𝑤𝑤
𝑧𝑧𝑤𝑤 (𝑧𝑧) = � 𝜙𝜙𝑤𝑤 𝑑𝑑𝑑𝑑 = � 𝑑𝑑𝑧𝑧𝑤𝑤
0 0

The material coordinate 𝑧𝑧𝑠𝑠 is linked to the Gibson height (also called material
height) that we will now define.

If all the particles of the left could be

squeezed at the bottom of the column, a
height 𝐿𝐿𝐺𝐺 would be obtained. Because of
conservation of mass, this is a constant
whereas ℎ𝑏𝑏 varies in time as the particles
settle.

The Gibson height 𝐿𝐿𝐺𝐺 is a (constant) length

obtained from the conservation of mass
within the undrained column. It can be seen
as the average volume fraction of soil per
unit of area.

The Gibson height can easily be obtained using the fact that at t = 0:
ℎ
𝐿𝐿𝐺𝐺 = � 𝜙𝜙0 𝑑𝑑𝑑𝑑 = 𝐿𝐿𝜙𝜙0
0

where 𝐿𝐿 is the height of the fluid in the column.

We get:
𝐿𝐿𝐺𝐺
𝜙𝜙𝑤𝑤 𝜌𝜌𝑤𝑤 + 𝜙𝜙𝑆𝑆 𝜌𝜌𝑠𝑠
σ𝑧𝑧𝑧𝑧 (𝑧𝑧𝑠𝑠 ) = � � � 𝑔𝑔𝑔𝑔𝑧𝑧𝑠𝑠 + 𝜌𝜌𝑤𝑤 𝑔𝑔(𝐿𝐿 − ℎ𝑏𝑏 ) + 𝑃𝑃0
𝑧𝑧𝑠𝑠 𝜙𝜙𝑠𝑠

As we have 𝑧𝑧𝑠𝑠 (ℎ𝑏𝑏 ) = 𝐿𝐿𝐺𝐺 (the Gibson height) which is a constant (it represents the
soil height that would be obtained in the column if no voids were present) the upper
bound now does not anymore depend on z.

251
 Introduction to Colloid Science

The material coordinates can conveniently be used to find again some of the results
we have derived earlier, when setting-up the Darcy equation for settling particles.
For a water/soil/gas system, one gets:

𝑑𝑑𝑑𝑑 = 𝑑𝑑𝑑𝑑𝑤𝑤 + 𝑑𝑑𝑑𝑑𝑠𝑠 + 𝑑𝑑𝑑𝑑𝑔𝑔

Using the fact that the volume is conserved (implying that each volume element is
conserved as function of time i.e. 𝑑𝑑𝑑𝑑/𝑑𝑑𝑑𝑑 = 0) we obtain

`𝑑𝑑𝑧𝑧𝑤𝑤 𝑑𝑑𝑧𝑧𝑔𝑔 𝑑𝑑𝑧𝑧𝑠𝑠

� � +� � = −� �
𝑑𝑑𝑑𝑑 𝑧𝑧 𝑑𝑑𝑑𝑑 𝑧𝑧 𝑑𝑑𝑑𝑑 𝑧𝑧

which should be read as: at a given position z in the column, the flux of water and
gas is minus the flux of soil. In the notations adopted above, in the case there is no
gas, we find as previously:

𝐽𝐽𝑤𝑤/𝑙𝑙𝑙𝑙𝑙𝑙 = −𝐽𝐽𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙

The change in time of 𝑧𝑧𝑤𝑤 at a constant 𝑧𝑧𝑠𝑠 is the velocity of water for a fixed soil
skeleton. This is what we can define as 𝐽𝐽𝑤𝑤/𝑠𝑠 (𝑧𝑧):

𝜕𝜕𝑧𝑧𝑤𝑤
𝐽𝐽𝑤𝑤/𝑠𝑠 (𝑧𝑧) = � �
𝜕𝜕𝜕𝜕 𝑧𝑧𝑠𝑠

Note that this equation is valid independent of the fact that gas is present or not in
the soil. Using

𝜙𝜙𝑤𝑤
𝑑𝑑𝑑𝑑𝑤𝑤 = 𝑑𝑑(𝜙𝜙𝑤𝑤 𝑧𝑧) = 𝑑𝑑 � 𝑧𝑧𝑠𝑠 � = 𝑑𝑑(𝑒𝑒𝑤𝑤 𝑧𝑧𝑠𝑠 )
𝜙𝜙𝑠𝑠

we get:

𝜕𝜕𝑒𝑒𝑤𝑤
𝐽𝐽𝑤𝑤/𝑠𝑠 (𝑧𝑧) = 𝑧𝑧𝑠𝑠 � �
𝜕𝜕𝜕𝜕 𝑧𝑧𝑠𝑠

Leading to:

𝜕𝜕𝜕𝜕𝑤𝑤/𝑠𝑠 𝜕𝜕𝑒𝑒𝑤𝑤
� � =� �
𝜕𝜕𝑧𝑧𝑠𝑠 𝑡𝑡 𝜕𝜕𝜕𝜕 𝑧𝑧𝑠𝑠

Using the Darcy equation for settling particles we get:

𝜕𝜕𝑒𝑒𝑤𝑤 𝜕𝜕 𝑘𝑘 𝜕𝜕𝑃𝑃𝑒𝑒 𝜕𝜕 𝑘𝑘 𝜕𝜕𝑃𝑃𝑒𝑒

� � = � �= �𝜙𝜙 �
𝜕𝜕𝜕𝜕 𝑧𝑧𝑠𝑠 𝜕𝜕𝑧𝑧𝑠𝑠 𝜂𝜂 𝜕𝜕𝜕𝜕 𝜕𝜕𝑧𝑧𝑠𝑠 𝑆𝑆 𝜂𝜂 𝜕𝜕𝑧𝑧𝑠𝑠

252
 Chapter 9 Permeability of slurries

We recall that the derivative of the excess pore pressure is given by:

𝜕𝜕𝑃𝑃𝑒𝑒 𝜕𝜕σ𝑧𝑧𝑧𝑧 𝜕𝜕σ𝑠𝑠𝑠𝑠

= 𝜌𝜌𝑤𝑤 𝑔𝑔 + −
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
From which we get:

𝜕𝜕𝑃𝑃𝑒𝑒 𝜕𝜕σ𝑠𝑠𝑠𝑠
= ��𝜙𝜙𝑆𝑆 (𝑧𝑧) + 𝜙𝜙𝑔𝑔 (𝑧𝑧)� 𝜌𝜌𝑤𝑤 − 𝜙𝜙𝑆𝑆 (𝑧𝑧)𝜌𝜌𝑠𝑠 � 𝑔𝑔 −
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
𝜕𝜕𝑃𝑃𝑒𝑒 𝜕𝜕σ𝑠𝑠𝑠𝑠
= −𝜙𝜙𝑆𝑆 (𝑧𝑧) �𝜌𝜌𝑠𝑠 − �1 + 𝑒𝑒𝑔𝑔 (𝑧𝑧)� 𝜌𝜌𝑤𝑤 � 𝑔𝑔 −
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
𝜕𝜕𝑃𝑃𝑒𝑒 𝜕𝜕σ𝑠𝑠𝑠𝑠
= −�𝜌𝜌𝑠𝑠 − �1 + 𝑒𝑒𝑔𝑔 �𝜌𝜌𝑤𝑤 �𝑔𝑔 −
𝜕𝜕𝑧𝑧𝑠𝑠 𝜕𝜕𝑧𝑧𝑠𝑠

This leads to:

𝜕𝜕𝑒𝑒𝑤𝑤 𝜕𝜕 𝑘𝑘 𝜕𝜕σ𝑠𝑠𝑠𝑠
� � =− �𝜙𝜙𝑠𝑠 � + �𝜌𝜌𝑠𝑠 − �1 + 𝑒𝑒𝑔𝑔 �𝜌𝜌𝑤𝑤 �𝑔𝑔𝜙𝜙𝑆𝑆 ��
𝜕𝜕𝜕𝜕 𝑧𝑧𝑠𝑠 𝜕𝜕𝑧𝑧𝑠𝑠 𝜂𝜂 𝜕𝜕𝑧𝑧𝑠𝑠

This is the general formulation of the Gibson equation, in the case that gas bubbles
are distributed within the soil matrix. We also deduce that:

𝑘𝑘 𝜕𝜕σ𝑠𝑠𝑠𝑠
𝐽𝐽𝑤𝑤/𝑠𝑠 = −𝜙𝜙𝑠𝑠 � + �𝜌𝜌𝑠𝑠 − �1 + 𝑒𝑒𝑔𝑔 �𝜌𝜌𝑤𝑤 �𝑔𝑔𝜙𝜙𝑆𝑆 �
𝜂𝜂 𝜕𝜕𝑧𝑧𝑠𝑠

If 𝑒𝑒𝑔𝑔 = 0 (there is no gas in the soil), then:

𝜕𝜕𝑒𝑒𝑤𝑤 𝜕𝜕 𝑘𝑘 𝜕𝜕σ𝑠𝑠𝑠𝑠
� � =− �𝜙𝜙𝑠𝑠 � + (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆 ��
𝜕𝜕𝜕𝜕 𝑧𝑧𝑠𝑠 𝜕𝜕𝑧𝑧𝑠𝑠 𝜂𝜂 𝜕𝜕𝑧𝑧𝑠𝑠

𝜕𝜕𝜙𝜙𝑠𝑠 𝜕𝜕 𝑘𝑘 𝜕𝜕σ𝑠𝑠𝑠𝑠
� � = 𝜙𝜙𝑠𝑠2 �𝜙𝜙𝑠𝑠 � + (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆 ��
𝜕𝜕𝜕𝜕 𝑧𝑧𝑠𝑠 𝜕𝜕𝑧𝑧𝑠𝑠 𝜂𝜂 𝜕𝜕𝑧𝑧𝑠𝑠

Using once more 𝑑𝑑𝑑𝑑𝑠𝑠 = 𝜙𝜙𝑠𝑠 𝑑𝑑𝑑𝑑 and realizing that :

𝜕𝜕𝜙𝜙𝑠𝑠 𝑑𝑑𝜙𝜙𝑠𝑠
� � =
𝜕𝜕𝜕𝜕 𝑧𝑧𝑠𝑠 𝑑𝑑𝑑𝑑

we get:

𝑑𝑑𝜙𝜙𝑠𝑠 𝜕𝜕 𝑘𝑘 𝜕𝜕σ𝑠𝑠𝑠𝑠
= 𝜙𝜙𝑠𝑠 � � + (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆 ��
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝜂𝜂 𝜕𝜕𝜕𝜕

which is the original Gibson equation (when no gas is present).

253
 Introduction to Colloid Science

We found earlier for the Gibson equation:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 𝑘𝑘 𝜕𝜕σ𝑠𝑠𝑠𝑠
= � �𝜙𝜙 + (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑠𝑠2 ��
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜂𝜂 𝑆𝑆 𝜕𝜕𝜕𝜕

Where does the difference come from? To find the answer, we have to look back to
one of the equation we have used to set-up the Gibson equation when we
introduced it the first time: the continuity equation. We have expressed this
equation as:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕�𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 𝜙𝜙𝑆𝑆 �

+ =0
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
This equation is the Eulerian form of the continuity equation. This implies that we
follow the changes in 𝜙𝜙𝑆𝑆 at a given position 𝑧𝑧 as function of time. Doing so, we say
that the changes in 𝜙𝜙𝑆𝑆 in time at a position 𝑧𝑧 are caused by the inflow and outflow
of matter at position 𝑧𝑧. The rate of inflow/outflow is in our case the velocity 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 .
The Lagrangian form of the same equation gives the changes in 𝜙𝜙𝑆𝑆 for a given
volume element within the laboratory frame of reference that we follow in space
and time. The link between the Eulerian and Lagrangian descriptions is given by:

𝑑𝑑𝜙𝜙𝑆𝑆 𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕𝜙𝜙𝑆𝑆

= + 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑏𝑏
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
where 𝑑𝑑𝜙𝜙𝑆𝑆 /𝑑𝑑𝑑𝑑 gives the Lagrangian derivative and 𝜕𝜕𝜙𝜙𝑆𝑆 /𝜕𝜕𝜕𝜕 is the Eulerian one. This
equation tells us that the changes in 𝜙𝜙𝑆𝑆 of the volume element we follow and
currently is located at (𝑧𝑧, 𝑡𝑡) is given by the change of 𝜙𝜙𝑆𝑆 (term 𝜕𝜕𝜙𝜙𝑆𝑆 /𝜕𝜕𝜕𝜕) at position 𝑧𝑧
plus the change associated to the fact that the volume element we follow is leaving
position 𝑧𝑧 at the instant 𝑡𝑡 (term 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 (𝜕𝜕𝜙𝜙𝑆𝑆 ⁄𝜕𝜕𝜕𝜕)). In Lagrangian form, the continuity
equation is therefore:

𝑑𝑑𝜙𝜙𝑆𝑆 𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕�𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 𝜙𝜙𝑆𝑆 �

− 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 + =0
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
𝑑𝑑𝜙𝜙𝑆𝑆 𝜕𝜕𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙
+ 𝜙𝜙𝑆𝑆 =0
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕
We can now combine the Lagrangian form of the continuity equation with the Darcy
equation

𝑘𝑘 𝜕𝜕𝑃𝑃𝑒𝑒
𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 =
𝜂𝜂 𝜕𝜕𝜕𝜕

and we get:

254
 Chapter 9 Permeability of slurries

𝑑𝑑𝜙𝜙𝑆𝑆 𝜕𝜕 𝑘𝑘 𝜕𝜕𝑃𝑃𝑒𝑒
= −𝜙𝜙𝑆𝑆 � �
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝜂𝜂 𝜕𝜕𝜕𝜕

which can further be developed into:

𝑑𝑑𝜙𝜙𝑆𝑆 𝜕𝜕 𝑘𝑘 𝑘𝑘 𝜕𝜕σ𝑠𝑠𝑠𝑠
= 𝜙𝜙𝑆𝑆 � (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆 + �
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝜂𝜂 𝜂𝜂 𝜕𝜕𝜕𝜕

This is the original form of the Gibson equation – and the one we have found above.

There exists numerical schemes that enable to solve the Gibson equation, in the
presence or not of gas. In order to solve these equations, hypothesis have to be
made for the dependence of 𝑘𝑘 and σ𝑠𝑠𝑠𝑠 on z (and for the dependence of 𝑒𝑒𝑔𝑔 on z is
gas is considered).

In the next chapter (Chapter 10) we are going to present analytical and numerical
solutions for the Gibson equation, in the absence of gas.

Lagrange and Euler derivatives

To illustrate the difference between the Lagrangian and Eulerian derivatives, let us
consider a variable that occupies the space we study. For instance, in a fluid, we
could measure the temperature 𝑇𝑇 at any position in time and space.

For simplicity, we will consider a stripe of fluid in direction 𝑥𝑥. We are measuring the
temperature with a thermometer A that we place at a fixed position 𝑥𝑥 and a
thermometer B that we are moving along 𝑥𝑥. With thermometer B we get the
temperature as a function of 𝑥𝑥(𝑡𝑡), i.e. the trajectory of the thermometer. This
implies that for a given time 𝑡𝑡 we are at a specific location 𝑥𝑥, i.e. 𝑥𝑥 and 𝑡𝑡 are
correlated. In compact notation, we say that we get 𝑇𝑇(𝑥𝑥(𝑡𝑡)), that is the temperature
along the trajectory 𝑥𝑥(𝑡𝑡). With thermometer A, we get the temperature at a fixed
position 𝑥𝑥, where this 𝑥𝑥 is therefore not dependent on time. In that case, in compact
notation, we say that we get 𝑇𝑇(𝑥𝑥, 𝑡𝑡), that is the temperature for a fixed position 𝑥𝑥
as a function of time. Of course, when thermometer B is at the same position as
thermometer A, the temperatures are the same:

255
 Introduction to Colloid Science

𝑇𝑇�𝑥𝑥(𝑡𝑡)� = 𝑇𝑇(𝑥𝑥, 𝑡𝑡)

However, the variations in temperature measured with thermometers A and B can

be different. From mathematics, we get:

𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝑑𝑑𝑑𝑑

= +
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝑑𝑑𝑑𝑑
where 𝑑𝑑𝑑𝑑 ⁄𝑑𝑑𝑑𝑑 is the change in temperature measured with thermometer B and
𝜕𝜕𝜕𝜕 ⁄𝜕𝜕𝜕𝜕 the change in temperature measured with thermometer A. 𝜕𝜕𝜕𝜕 ⁄𝜕𝜕𝜕𝜕
represents the change in temperature with displacement 𝑑𝑑𝑑𝑑 around time 𝑡𝑡 of the
measurement and 𝑑𝑑𝑑𝑑 ⁄𝑑𝑑𝑑𝑑 = 𝑣𝑣𝐵𝐵 is the velocity of thermometer B. From the relation
above, we see that if the temperature does not dependent on time (but is not the
same at every position 𝑥𝑥) we get

𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕
= 𝑣𝑣
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝐵𝐵
which means that even though for a given 𝑥𝑥 the temperature does not change
(𝜕𝜕𝜕𝜕 ⁄𝜕𝜕𝜕𝜕 = 0), the thermometer B will indicate a change of temperature in time as it
is moving. Only if the thermometer B is not moving or if the temperature is the same
at any 𝑥𝑥 do we get 𝑑𝑑𝑑𝑑 ⁄𝑑𝑑𝑑𝑑 = 0.

Finding the permeability using colloid science

So far we have discussed the consolidation of soft soils so as to determine the
permeability from the physical compaction (and in particular the water/mud
interface as function of time) of sediment beds. Quite some research has been
performed over the years to devise other methods for finding information about the
permeability and porosity of consolidated soils, also as function of environmental
parameters such as salinity, temperature and clay surface properties.

A simple and extremely popular relation is provided by the empirical equation found
by Gustavus E. Archie (1907-1978) in 1942. Archie’s law states that the electric
conductivity 𝜎𝜎 of a porous media 126 saturated by an electrolyte is proportional to its
porosity such that:

𝜎𝜎 = 𝜙𝜙 𝑚𝑚 𝜎𝜎𝑒𝑒

where 𝜎𝜎𝑒𝑒 is the conductivity of the electrolyte, 𝜙𝜙 = (1 − 𝜙𝜙𝑆𝑆 ) is the porosity and m
is an exponent to be fitted which varies usually between 1 and 4.

126
In this chapter we will use the symbol 𝜎𝜎 for the conductivity to avoid any confusion with 𝑘𝑘
and 𝐾𝐾 that are used for the hydraulic conductivity. In Chapter 3 we used 𝐾𝐾 for expressing the
conductivity and 𝜎𝜎 for designing a surface charge.

256
 Chapter 9 Permeability of slurries

Quite some work has also been devoted to relate the electric conductivity 𝜎𝜎 to the
hydraulic permeability k. In a first approach, one makes use of the relation:

𝑘𝑘 ≈ 𝐴𝐴 𝜙𝜙𝑟𝑟 2

where 𝐴𝐴 is an unknown shape factor (that might or not depend on 𝜙𝜙) and 𝑟𝑟 a
representative length for a pore radius (that might or not depend on 𝜙𝜙). The relation
originates from the Hagen-Poiseuille relation we have already mentioned above.
Combining the two equations, one finds the link between conductivity and
permeability to be:

𝜎𝜎 1/𝑚𝑚 2
𝑘𝑘 ≈ 𝐴𝐴 � � 𝑟𝑟
𝜎𝜎𝑒𝑒

More elaborate models have been created, for instance by assuming that the porous
media is formed by a collection of colloidal particles. The parameter m can then be
calculated 127.

If a surface is charged, as we have seen in Chapter 3, applying an electric field will

provide information about this charge. This information can then be correlated to
the permeability as we will see now for the simple example of a bundle of pores.

Bundle of charged pores in an electric field

If one applies and electric field 𝑬𝑬 = 𝐸𝐸 𝒆𝒆𝒚𝒚 in the direction parallel to a charged pore
filled with an electrolyte, the ions inside the pore will start to move. In the stationary
state, the ions will have a constant velocity, implying that the total sum of forces on
each ion must be zero. The electric force is compensated by the drag force exerted
by the liquid.

If we now consider the forces exerted on the liquid, one finds that:

- in the stationary state, each layer of fluid dx moves with a uniform velocity parallel
to the walls because the total force on such layer is zero.

- In virtue of Newton’s third law (action = reaction) the force exerted by the ions on
the liquid is equal in magnitude to the (electric) force exerted directly on the ions
and which produce the movement.

127 P.N.Sen et al. “A self-similar model for sedimentary rocks with application to the dielectric
constant of fused glass beads”, Geophysics, 46,5,781-795 (1981); S. Kostek et al. “Fluid
permeability in porous media: Comparison of electrical estimates with hydrodynamical
calculation”, Physical Review B, 45, 1 (1992); Kirichek, A., C. Chassagne, and R. Ghose.
"Dielectric spectroscopy of granular material in an electrolyte solution of any ionic strength."
Colloids and Surfaces A: Physicochemical and Engineering Aspects 533 (2017): 356-370.

257
 Introduction to Colloid Science

- A frictional force is exerted on a layer by the neighbouring layers which move with
different velocities.

Schematic representation of a section of a capillary. The capillary walls are negatively

charged. Double layers are represented by black dashed lines. Inside the double layers, one
finds a majority of couterions (+), even though co-ions (-) are also present (but not
represented for simplification). In the bulk of the capillary, far away of the double layers, the
ionic concentration is independent of the surface charge of the pores.

This enables to write:

𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 𝑑𝑑 2 𝑢𝑢
𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 = 𝜂𝜂 � � − 𝜂𝜂 � � = 𝜂𝜂 � 2 � 𝑑𝑑𝑑𝑑
𝑑𝑑𝑑𝑑 𝑥𝑥+𝑑𝑑𝑑𝑑 𝑑𝑑𝑥𝑥 𝑥𝑥 𝑑𝑑𝑥𝑥

where 𝜌𝜌 is the ionic concentration.

Using Poisson’s equation introduced in Chapter 3, here in Cartesian coordinates,

one finds:

𝑑𝑑 2 𝜓𝜓 𝑑𝑑 2 𝑢𝑢
−𝜀𝜀0 𝜀𝜀𝑟𝑟 𝐸𝐸 � � = 𝜂𝜂 � �
𝑑𝑑𝑥𝑥 2 𝑑𝑑𝑥𝑥 2

This equation can be integrated a first time:

𝑥𝑥
𝑑𝑑 2 𝜓𝜓 𝑥𝑥
𝑑𝑑 2 𝑢𝑢
−𝜀𝜀0 𝜀𝜀𝑟𝑟 𝐸𝐸 � � 2
� 𝑑𝑑𝑑𝑑 = 𝜂𝜂 � � 2 � 𝑑𝑑𝑑𝑑
𝑟𝑟/2 𝑑𝑑𝑥𝑥 𝑟𝑟/2 𝑑𝑑𝑥𝑥

and 𝑟𝑟/2 is the middle of the capillary. Using the fact that, by symmetry, both the
derivatives of 𝜓𝜓 and 𝑢𝑢 are zero at 𝑟𝑟/2 we get:

𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑
−𝜀𝜀0 𝜀𝜀𝑟𝑟 𝐸𝐸 � � = 𝜂𝜂 � �
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑

By integrating a second time:

𝑥𝑥 𝑥𝑥
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑
−𝜀𝜀0 𝜀𝜀𝑟𝑟 𝐸𝐸 � � � 𝑑𝑑𝑥𝑥 = 𝜂𝜂 � � � 𝑑𝑑𝑑𝑑
𝑥𝑥𝑠𝑠 𝑑𝑑𝑑𝑑 𝑥𝑥𝑠𝑠 𝑑𝑑𝑑𝑑

258
 Chapter 9 Permeability of slurries

Using the boundary conditions:

𝜓𝜓(𝑥𝑥𝑠𝑠 ) = 𝜁𝜁 ; 𝑢𝑢(𝑥𝑥𝑠𝑠 ) = 0

where 𝑥𝑥𝑠𝑠 is the position of the slip plane (which is very close to the capillary wall)
and 𝜁𝜁 the zeta potential of the capillary tube, we get:

−𝜀𝜀0 𝜀𝜀𝑟𝑟 𝐸𝐸 𝜁𝜁 = 𝜂𝜂𝜂𝜂(𝑥𝑥)

Rearranging this equation, we can write:

𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜁𝜁
𝑢𝑢(𝑥𝑥) = 𝐸𝐸 for 𝑥𝑥 > 𝑥𝑥𝑠𝑠
𝜂𝜂

This formula is similar to the one we have found for the Smoluchowski
electrophoretic mobility of a charged sphere in an electric field. Similarly to what we
have seen for spheres, a double layer exists close to the pore walls. In our example,
we have assumed that the double layers do not overlap, which is most certainly true
for not too small pores and reasonable ionic strength. In case of overlapping double
layers, numerical solutions exist.

One should also realize that the derivation presented here is implicitly done for very
thin double layers 𝜅𝜅 −1 compared to 𝑟𝑟 as we have used Cartesian coordinates. For
arbitrary 𝜅𝜅𝜅𝜅 one should work in cylindrical coordinates, assuming of course that the
pores are cylindrical, which is generally a valid assumption.

The flux of water exiting the capillary can be estimated by:

𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜁𝜁
𝐽𝐽 = 𝑆𝑆𝑆𝑆 = 𝑆𝑆 𝐸𝐸
𝜂𝜂

where S is the cross section of the capillary. For a bundle of capillaries (all parallel to
each other), one simply gets:

𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜁𝜁
𝐽𝐽𝑤𝑤/𝑠𝑠 = 𝜙𝜙𝑆𝑆 𝐸𝐸
𝜂𝜂

Contrary to the Darcy equation, the relation we found is a function of the applied
electric field, since no pressure gradient was applied in the experiment. We have
thus found that it is also possible to create a flow of water by applying an electric
field. This is due to the fact that the ions are set in movement by the electric field
and, by friction, set the water in motion. If the capillaries are uncharged (𝜁𝜁 = 0)
there is no net water flux (𝐽𝐽𝑤𝑤/𝑠𝑠 = 0). This does not mean that the ions are not moving
or that the water is not moving, but that the water flux created by the negative ions
compensate the flux created by the positive ions such that no net flux is created.
When the pores are charged (here negatively), there is a dissymmetry between the

259
 Introduction to Colloid Science

fluxes: there are more positive ions that are mobile, compared to negative ones (of
which some are fixed on the pores’ walls) and hence a net flux is created.

Onsager relations
Is it possible to link the water flux we have found to the permeability of the porous
medium? In order to find out, we will have to discuss more into details about what
happens when one apply either a pressure gradient or an electric field to the same
porous media. As before, we will here only consider a bundle of capillaries, fully
saturated with electrolyte.

When an electric field 𝐸𝐸 is applied to this porous medium, we have just found that
we could create a water flux. By applying a pressure gradient ∇𝑃𝑃, we also will create
a water flux 𝐽𝐽𝑤𝑤/𝑠𝑠 (m/s) and the relation between water flux and pressure gradient is
then simply the Darcy equation. In general we can therefore write:

𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜁𝜁 𝑘𝑘
𝐽𝐽𝑤𝑤/𝑠𝑠 = 𝜙𝜙𝑆𝑆 𝐸𝐸 − ∇𝑃𝑃
𝜂𝜂 𝜂𝜂

This equation could be expanded by considering that a water flux can also be created
by applying a temperature gradient or a ionic concentration gradient, etc… We will
limit ourselves to electric and pressure gradients.

Another flux is created by the application of an electric field: a macroscopic electric

flux (electric current 𝐽𝐽𝑒𝑒 in A/m2) which is proportional to the electric field following
Ohm’s law:

𝐽𝐽𝑒𝑒 = 𝜎𝜎𝜎𝜎

where 𝜎𝜎 is the electric conductivity (of the porous medium saturated with
electrolyte). One could then make the following reasoning: by applying a pressure
gradient, water will be set in motion. As this water contains ions, these will also be
set into motion. The movement of charges create an electric current, and therefore
an electric current can be created by applying a pressure gradient. When both
electric and pressure gradients are considered, one can therefore write:

𝐽𝐽𝑒𝑒 = 𝜎𝜎𝜎𝜎 − 𝐿𝐿𝑒𝑒 ∇𝑃𝑃

where 𝐿𝐿𝑒𝑒 is a nameless parameter. For commodity, we will rewrite the water flux in
a more general form as:

𝑘𝑘
𝐽𝐽𝑤𝑤/𝑠𝑠 = 𝐿𝐿𝑤𝑤 𝐸𝐸 − ∇𝑃𝑃
𝜂𝜂

260
 Chapter 9 Permeability of slurries

This formulation is in fact valid for any porous media, as we have not specified the
values of 𝐿𝐿𝑒𝑒 , 𝐿𝐿𝑤𝑤 , 𝑘𝑘 or 𝜎𝜎.

In 1968 Lars Onsager got the Nobel prize in Chemistry for having theoretically found
that:

𝐿𝐿𝑒𝑒 = 𝐿𝐿𝑤𝑤

The formulation of Onsager was of course more general than the one given here,
which is just one of the many examples of the Onsager reciprocal relations.

Conductivity of a porous medium consisting of a bundle of charged pores

Let us assume that a porous medium is submitted to a pressure gradient ∇𝑃𝑃. By the
movement of ions, an electric field 𝐸𝐸 is then created. An example is given right
underneath.

Streaming potential coupling coefficient in sandstones as function of NaCl concentration.

Data from Jaafar et al. 128 The pH models are from Glover at al. 129 .The red line represents the
classical dependence, explained in the text.

At steady state, assuming that no electrochemical reactions occur, and that no

charges are created or destroyed, a bulk electric current will be created that opposes

128 Jaafar, M. Z., J. Vinogradov, and M. D. Jackson, 2009, Measurement of streaming potential
coupling coefficient in sandstones saturated with high salinity NaCl brine: Geophysical
Research Letters, 36, L21306.
129 P.W.J. Glover et al. “Streaming-potential coefficient of reservoir rock: A theoretical model”,

Geophysics, 77, 2 (2012).

261
 Introduction to Colloid Science

the electric current along the walls (in the double layers). This leads to the fact that
the (macroscopic) electric current is zero: 𝐽𝐽𝑒𝑒 = 0. It follows that:

𝐿𝐿𝑒𝑒
(𝐸𝐸)𝐽𝐽𝑒𝑒=0 = (∇𝑃𝑃)𝐽𝐽𝑒𝑒=0
𝜎𝜎
This relation defines the streaming potential: an electric field is created by the
application of a pressure gradient in the absence of electric current.

The streaming potential coupling coefficient 𝑆𝑆𝑆𝑆 as given on the y-axis of the figure
is defined by:

−∆𝑉𝑉 𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜁𝜁

𝑆𝑆𝑆𝑆 = = −𝜙𝜙𝑆𝑆
∆𝑃𝑃 𝜂𝜂𝜎𝜎𝑒𝑒

where 𝜎𝜎𝑒𝑒 is the conductivity of the electrolyte inside the pores. This expression can
be found using the relation given above and realising that 𝐸𝐸 = −∇𝑉𝑉 and

𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜁𝜁
𝐿𝐿𝑒𝑒 = 𝐿𝐿𝑤𝑤 = 𝜙𝜙𝑆𝑆
𝜂𝜂

Furthermore ∇𝑉𝑉/∇𝑃𝑃 = ∆𝑉𝑉/∆𝑃𝑃 for a linear system (∇𝑃𝑃 = ∆𝑃𝑃 /L where L is the length
of the bundle of pores). For a bundle of pores, the electric current at no pressure
difference is given by:

𝐽𝐽𝑒𝑒 = 𝜙𝜙𝑆𝑆 𝜎𝜎𝑒𝑒 𝐸𝐸 = 𝜎𝜎𝜎𝜎

One therefore finds:

−∆𝑉𝑉 𝜀𝜀0 𝜀𝜀𝑟𝑟 𝜁𝜁

𝑆𝑆𝑆𝑆 = =−
∆𝑃𝑃 𝜂𝜂𝜎𝜎𝑒𝑒

Recalling (see Chapter 3) that 𝜎𝜎𝑒𝑒 ~ 𝐶𝐶 where 𝐶𝐶 is the salt concentration, one finds:

𝑙𝑙𝑙𝑙𝑙𝑙(−𝑆𝑆𝑆𝑆) = −𝑙𝑙𝑙𝑙𝑙𝑙(𝐶𝐶) + constant

This implies that for the simple model of a bundle of pores, the logarithm of the
streaming coupling coefficient should scale as minus the logarithm of the salt
concentration. In the log-log plot above this corresponds to the red line of slope -1.
The data presented on the figure has been measured on a sandstone saturated with
high salinity NaCl. We can therefore conclude that the simple “bundle of pores”
model is quite successful to explain the obtained results. From the data, we can also
estimate the zeta potential: we find (in absolute value) 19 mV, which is a realistic
value for such a system. Note that the models proposed by Glover et al., even though
more elaborated, are less successful. Many models include so many adjustable
parameters that they cannot provide an estimate for the zeta potential as they have

262
 Chapter 9 Permeability of slurries

made the zeta potential dependent on parameters that are quite difficult to
estimate.

Inserting the general streaming potential relation into the expression for 𝐽𝐽𝑤𝑤/𝑠𝑠 and
using the fact that 𝐿𝐿𝑒𝑒 = 𝐿𝐿𝑤𝑤 leads to:

𝐿𝐿2𝑒𝑒 𝑘𝑘
�𝐽𝐽𝑤𝑤/𝑠𝑠 � =� − � (∇𝑃𝑃)𝐽𝐽𝑒𝑒=0
𝐽𝐽𝑒𝑒 =0 𝜎𝜎 𝜂𝜂

It is usually the case that

𝐿𝐿2𝑒𝑒 𝑘𝑘
≪
𝜎𝜎 𝜂𝜂

which is why, in first approximation, Darcy’s equation is generally used to express a

flow of water. We note however, that the water flux exiting the porous medium
could be smaller than the one expected on the basis of Darcy alone. The permeability
can be correctly estimated by a series of experiments that accounts for both electric
and hydraulic effects. One can easily show that:

𝑘𝑘 ∇𝑃𝑃
𝐿𝐿𝑤𝑤 = � �
𝜂𝜂 𝐸𝐸 𝐽𝐽𝑤𝑤/𝑠𝑠=0

𝐸𝐸
𝐿𝐿𝑒𝑒 = 𝜎𝜎 � �
∇𝑃𝑃 𝐽𝐽𝑒𝑒=0

Combining these two equations using 𝐿𝐿𝑒𝑒 = 𝐿𝐿𝑤𝑤 one finds:

(𝐸𝐸/∇𝑃𝑃)𝐽𝐽𝑒𝑒=0
𝑘𝑘 = 𝜂𝜂𝜂𝜂
(∇𝑃𝑃/𝐸𝐸)𝐽𝐽𝑤𝑤/𝑠𝑠=0

This means that two series of experiments (one at zero water flux and one at zero
electric flux) should be performed, for which the conductivity, electric and hydraulic
gradients should be measured. One then can find the hydraulic permeability 𝑘𝑘.

263
 Introduction to Colloid Science

Lars Onsager and the reciprocal relations

Lars Onsager (1903 –1976) was a Norwegian-born

American physical chemist and theoretical physicist. He
held the Gibbs Professorship of Theoretical Chemistry
at Yale University. He was awarded the Nobel Prize in
Chemistry in 1968 for his work on the reciprocal
relations he derived. After completing secondary
school in Oslo, he attended the Norwegian Institute of
Technology (NTH) in Trondheim (nowadays the NTNU –
Norges Teknisk-Naturvitenskapelige Universitet, i.e.
Norwegian University of Science and Technology),
graduating as a chemical engineer in 1925.

In 1925 he arrived at a correction to the Debye-Hückel theory of electrolytic

solutions, to specify Brownian movement of ions in solution, and during 1926
published it. He travelled to Zürich, where Peter Debye was teaching, and told him
his theory was wrong. He impressed Debye so much that he was invited to become
Debye's assistant at the Eidgenössische Technische Hochschule (ETH), where he
remained until 1928.

In 1928 he went to the United States and held a position at the Johns Hopkins
University (JHU) in Baltimore and then at the Brown University in Providence. While
clearly Onsager was extremely good at developing
theories in physical chemistry, he was a very poor
teacher. This is why he was dismissed at JHU after one
semester and in 1933 at Brown. After a trip to Europe he
was hired by Yale University, where he remained for most
of the rest of his life, retiring in 1972. The only graduate
student who could really understand his lectures on
electrolyte systems, Raymond Fuoss, worked under him
and eventually joined him on the Yale chemistry faculty.
The 1932 paper that they wrote together on irreversible
processes in electrolytes took up eighty-nine pages in the
Journal of Physical Chemistry and remained the definitive
treatment of the topic until it was taken up again by the
two of them in the 1950s.

a classic book about irreversible thermodynamics 130

The reciprocal relations of Onsager are derived in the frame of irreversible

thermodynamics. First, relations between fluxes and forces are expressed by

130
De Groot, Sybren Ruurds, and Peter Mazur. Non-equilibrium thermodynamics. Courier
Corporation, 2013

264
 Chapter 9 Permeability of slurries

considering the entropy production. Then the reciprocal relations can be given.
These relations are a consequence of the time reversibility of microscopic
dynamics. The idea of microreversibility was introduced when the kinetics of gases
was studied. In 1872, Ludwig Boltzmann represented kinetics of gases as statistical
ensemble of elementary collisions. Equations of mechanics are reversible in time,
hence, the reverse collisions obey the same laws. According to Boltzmann, this
microreversibility implies the principle of detailed balance for collisions: at the
equilibrium ensemble each collision is equilibrated by its reverse collision. Another
macroscopic consequence of microscopic reversibility is the symmetry of kinetic
coefficients, the so-called reciprocal relations. The reciprocal relations were
discovered in the 19th century by Thomson and Helmholtz for some phenomena but
the general theory was proposed by Lars Onsager in 1931. He found also the
connection between the reciprocal relations and detailed balance.

Illustrations

Lars Onsager
https://www.nobelprize.org/prizes/chemistry/1968/onsager/biographical/

265
 Chapter 10
Modelling
the consolidation
of slurries
 Introduction to Colloid Science

In this chapter, we are going to discuss some solutions of the Gibson equation for
slurries that was set-up in Chapter 9. The Gibson equation reads:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 𝑘𝑘 𝜕𝜕σ𝑠𝑠𝑠𝑠
= � �𝜙𝜙 + (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑠𝑠2 ��
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜂𝜂 𝑆𝑆 𝜕𝜕𝜕𝜕

We are first going to give analytical solutions, and the solutions will be critically
analysed at the end of the chapter. We will then also make the link between Chapter
8 (settling) and the consolidation processes discussed in Chapter 9 and in the present
chapter.

Consolidation of slurries : the fractal approach

The permeability 𝑘𝑘 (and 𝐾𝐾) and the skeleton stress σ𝑠𝑠𝑠𝑠 depend both on 𝜙𝜙𝑆𝑆 . In order
to solve the Gibson equation it is therefore primordial to first find a constitutive
relation for both 𝐾𝐾 and σ𝑠𝑠𝑠𝑠 as function of 𝜙𝜙𝑆𝑆 . An innovative approach has been
introduced by Merckelbach and Kranenburg in 2000 131 to model the consolidation
of a column filled with (flocculated) clay. They make use of the self-similarity
properties of fractals, and hence assume that the column is filled with fractal flocs.

In their approach, at the initial time t = 0, the column is filled with fractal flocs that
are touching one another and all the flocs have the same diameter. A schematic
representation of a column filled by actual (“real”) fractal flocs is given here for times
larger than zero:

Schematic representation of the column filled with fractal flocs; the flocs are “real” flocs that
are physically present in the column, with a defined (measurable) size and fractal dimension

“Consolidation and strength evolution of soft mud layers”, Lucas Merckelbach, PhD thesis
131

TU Delft, 2000

268
 Chapter 10 Modelling the consolidation of slurries

On purpose, we have represented in the column a suspended phase: in the region

the flocs are suspended, they are, by definition, not touching and hence will
experience settling. Their settling velocity is in general different from the velocity of
the flocs that are in the bed : a settling floc “has no clue” there is a bed underneath
it, in other words: the settling velocity of a floc is independent of the settling velocity
of the bed composed of the same flocs. Settling and consolidation should therefore
in general be treated as two separate processes.

There is a special case, however, where settling and consolidation can be treated
with a single model, which is the case we present here.

If we assume that the sediment concentration in the column is high, and

homogeneous at t = 0, it is possible to mathematically define a fractal floc size, even
when the sediment is not flocculated or the flocs (if any) are not fractal. In the case
the sediment is flocculated, the mathematical floc size will always be larger or equal
to the actual floc size and the fractal dimension of the mathematical floc size will not
necessarily be the fractal dimension of the actual floc (if the actual floc is fractal).
The reason for this is linked to an important requirement of the model: the
mathematical flocs should be space-filling – also in the settling region. The proper
illustration for the model is therefore the following:

Schematic representation of the column filled by mathematical fractal flocs; contrary to the
previous illustration, the flocs are here only mathematical objects. The real particles inside
the column are not (necessarily) flocs or fractal and their size is usually smaller than the size
of the mathematical fractal flocs, as the real particles do not have to be space-filling.

The advantage of this approach is that there is no mathematical discontinuity

between the settling and the consolidation phases (the same parameters, i.e. the
same permeability and effective stress are used for both phases). We will see with
examples at the end of the chapter the limitations of this approach.

269
 Introduction to Colloid Science

As we assume that at a position z in the column (also for times larger than zero) the
whole slice is filled by flocs, we can say that the volume fraction of clay at height z is
in good approximation equal to the fraction of clay within a floc:

volume clay in a floc

𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 =
volume of a floc
In Chapter 5 we have introduced fractals. We have said that the number of primary
particles inside the floc is given by:

𝑅𝑅𝑁𝑁 𝐷𝐷
𝑁𝑁 = � �
𝑎𝑎
where D is the fractal dimension, a is the size of a primary particle and 𝑅𝑅𝑁𝑁 the size
of the fractal floc. In the case that the primary particles are not spheres and have a
radius of gyration Rp, the exact size one has to take for a primary particle is often
unknown:

Depending on the way clay platelets aggregate, the characteristic size R of a primary
particle (that can be approximated by the centre-to-centre distance between two
platelets) can be quite different from the radius of gyration Rp of the platelet. An
unknown factor λ1 that ranges between a very small number and one can be
introduced to reflect the fact that the number of primary particles in a floc is
unknown:
𝐷𝐷
𝑅𝑅𝑁𝑁
𝑁𝑁 = � �
𝜆𝜆1 𝑅𝑅𝑝𝑝

Similarly, the amount of clay contained in the radius of gyration Rp is depending on

the shape of the clay particle, which leads to the introduction of a parameter 𝜆𝜆2 :

270
 Chapter 10 Modelling the consolidation of slurries

4
volume clay in a floc = 𝜆𝜆2 𝜋𝜋𝑅𝑅𝑝𝑝3
3
One therefore obtains:

𝜆𝜆2 𝑁𝑁𝑅𝑅𝑝𝑝3 𝜆𝜆2 𝑅𝑅𝑝𝑝 3−𝐷𝐷

𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = = 𝐷𝐷 � �
𝑅𝑅𝑁𝑁3 𝜆𝜆1 𝑅𝑅𝑁𝑁

From this last equation, we find a relation between the volume fraction of clay and
the size of a floc 𝑅𝑅𝑁𝑁 . As on the right-hand side, only 𝑅𝑅𝑁𝑁 is a function of z, we get:
𝐷𝐷−3
𝑅𝑅𝑁𝑁 (𝑧𝑧, 𝑡𝑡)
𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (𝑧𝑧, 𝑡𝑡) ~ � �
𝑅𝑅𝑝𝑝

This small digression about non-spherical primary particles is only relevant for
estimating the clay volume fraction of “real” flocs. For the “mathematical” flocs, it is
simpler to assume the primary particles are spherical. The shape and size of flocs is
irrelevant for the model.

A small discussion about the fractal dimension of “mathematical” flocs

Let us recapitulate briefly the way we connected the volume fraction 𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 to the
fractal dimension D (for simplicity we assume that there are no other particles
present and hence 𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = 𝜙𝜙𝑠𝑠 ):

The clay volume fraction 𝜙𝜙clay in a floc of a floc made of clay primary particles is equal
to the number of primary particles times the volume of a primary particle divided by
the volume of a floc:

𝑁𝑁 ∙ 𝑅𝑅𝑝𝑝3
𝜙𝜙clay in a floc ~
𝑅𝑅𝑁𝑁3

As we have made the assumption that the flocs are space-filling, 𝜙𝜙clay in a floc is also
the volume fraction of a slice of the column at a given height z:

𝜙𝜙clay in a floc (𝑧𝑧) = 𝜙𝜙𝑠𝑠 (𝑧𝑧)

As, by definition of a fractal,

𝐷𝐷
𝑅𝑅𝑁𝑁
𝑁𝑁 ~ � �
𝑅𝑅𝑝𝑝

one gets

271
 Introduction to Colloid Science

𝑅𝑅𝑝𝑝 3−𝐷𝐷
𝜙𝜙𝑠𝑠 ~ � �
𝑅𝑅𝑁𝑁

which is the relation we have been using so far.

There is, however, a little subtlety: we have attributed the value 𝜙𝜙𝑠𝑠 (𝑧𝑧) to a slice of
thickness 𝑑𝑑𝑑𝑑 of the column. This slice 𝑑𝑑𝑑𝑑 should of course be small for mathematical
reasons, but as we have explicitly taken the fractal floc to be spherical, the slice
should be larger than the radius 𝑅𝑅𝑁𝑁 for all positions and all times, otherwise 𝜙𝜙𝑠𝑠 (𝑧𝑧)
would vary unrealistically:

Improper thickness 𝑑𝑑𝑑𝑑 of a slice: as 𝑑𝑑𝑑𝑑 < 𝑅𝑅𝑁𝑁 the volume fraction 𝜙𝜙𝑠𝑠 (𝑧𝑧) would vary
from a low, high and again low number for three consecutive 𝑑𝑑𝑑𝑑

In fact, as one assumes the volume fraction 𝜙𝜙𝑠𝑠 (𝑧𝑧) to be constant for a small slice 𝑑𝑑𝑑𝑑,
one should preferably want to assume that the considered flocs are isotropic in that
direction, i.e. implying that they should not be spherical but cylindrical, with
translational symmetry along z for each thickness 𝑑𝑑𝑑𝑑. This would put no restriction
on how small 𝑑𝑑𝑑𝑑 could be. If we take such cylindrical flocs, with their base being a
2D fractal, one would get for the clay volume fraction:

𝑁𝑁 ∙ 𝑅𝑅𝑝𝑝2 ∙ 𝑑𝑑𝑑𝑑
𝜙𝜙clay in a floc ~
𝑅𝑅𝑁𝑁2 ∙ 𝑑𝑑𝑑𝑑

Along the same derivation lines as above, this would give:

𝑅𝑅𝑝𝑝 2−𝐷𝐷
𝜙𝜙𝑠𝑠 ~ � �
𝑅𝑅𝑁𝑁

For primary particles that are space-filling, the fractal dimension would therefore be
2 in this case and not 3. A fractal dimension of 1 corresponds also to the case where
the primary particles are space-filling, i.e. occupying the whole given line, curve or
volume considered. In fact all fractal dimensions that are integers (1, 2 or 3)
correspond to “primary particles that are space-filling”:

Let us imagine that we have primary particles such that

1
𝑅𝑅𝑁𝑁
𝑁𝑁 ~ � �
𝑅𝑅𝑝𝑝

272
 Chapter 10 Modelling the consolidation of slurries

This also implies that

𝐷𝐷
𝑅𝑅𝑁𝑁
𝑁𝑁 𝐷𝐷 ~ � �
𝑅𝑅𝑝𝑝

If D = 2 or 3, this means that we have completely filled the area or volume (𝑅𝑅𝑁𝑁 )𝐷𝐷
with primary particles (which are therefore “space-filling”). If the dimension D is not
an integer, then (𝑅𝑅𝑁𝑁 )𝐷𝐷 does not represent a curve, an area or a volume anymore,
but simply relates the amount of primary particles to the size 𝑅𝑅𝑁𝑁 :

𝑅𝑅𝑁𝑁 = 𝑁𝑁 1/𝐷𝐷 ∙ 𝑅𝑅𝑝𝑝

The important conclusion here is that the fractal dimension is simply a convenient
tool to model the system: one may chose the mathematical flocs to be spherical or
cylindrical. In the end, after fitting the data, one should find the same values for the
fractal dimension with a 1 difference: if one finds, for example D = 2,63 with the
spherical approach, one will find D = 1,63 with the cylindrical approach. A value of D
= 2 (cylindrical approach) or D = 3 (spherical approach) implies that the primary
particles are space-filling.

In order to be consistent with the work of Merckelbach, we will continue here to use
the following definition:
𝐷𝐷−3
𝑅𝑅𝑁𝑁 (𝑧𝑧, 𝑡𝑡)
𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (𝑧𝑧, 𝑡𝑡) ~ � �
𝑅𝑅𝑝𝑝

At time t = 0, we have 𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = 𝜙𝜙0 and therefore

𝐷𝐷−3
𝑅𝑅𝑁𝑁 (𝑧𝑧, 𝑡𝑡 = 0)
𝜙𝜙0 ~ � �
𝑅𝑅𝑝𝑝

In the limiting case that D = 3, one gets

𝜙𝜙0 = 1

and no fractal approach can be used, as there is no proper way to define the volume
of a floc. In the case 𝜙𝜙0 = 1, there is no settling and no consolidation, as all the
primary particles (of radius 𝑅𝑅𝑝𝑝 ) are touching, and by hypothesis, cannot deform. The
case 𝜙𝜙0 = 1 is corresponding to a fully gelled state of primary particles. Depending
on the shape of particles 𝜙𝜙0 will be smaller than 1 in most cases (𝜙𝜙0 = 0.74 for
compacted spheres in their most compacted form).

273
 Introduction to Colloid Science

If the characteristic size 𝑅𝑅𝑝𝑝 of the primary particle can be evaluated (by static light
scattering for instance), it is then possible to have an estimation of the mathematical
fractal floc size at t = 0:
1
𝑅𝑅𝑁𝑁 ~ 𝑅𝑅𝑝𝑝 (𝜙𝜙0 )𝐷𝐷−3

If we assume that the fractal dimension is given by D = 2.63 (which is a reasonable

value, close to what is generally found in experiments), we get:

This implies that for an initial concentration of 150 g/L (𝜙𝜙0 = 150/2650 = 0.0566),
having an homogeneous suspension of particles of size 0.1 microns would give an
effective fractal floc radius of 0.23 mm whereas 10 microns particles would give 2.35
cm. For lower volume fractions, the floc radius becomes clearly unrealistic. For 50
g/L for example and 10 microns particles, one would get a radius of 45 cm, much
larger than the settling column radius!

From above, we deduce that the fractal approach will only be appropriate for high
concentrations of clay, as one of the assumption made in the model (see
underneath) is that the size of a pore scales as the size of a floc.

Note that if the density of the suspension is known and equal to 𝜌𝜌 and that we
assume that the flocs are filling all the space, one finds, from the definition 𝜌𝜌 =
𝜌𝜌𝑠𝑠 𝜙𝜙0 + 𝜌𝜌𝑤𝑤 (1 − 𝜙𝜙0 ) that:
𝜌𝜌 − 𝜌𝜌𝑤𝑤
𝜙𝜙0 =
𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤

which is a convenient way to estimate the initial volume fraction.

The hypotheses required for the model are summarized here:

274
 Chapter 10 Modelling the consolidation of slurries

(a) the sample should be homogeneous at t = 0. This implies that if the sample is
polydisperse and contains both silt and clay (for simplicity we assume here there is
no sand fraction – which anyhow would settle quite quickly at the bottom of the
column), that at any height in the column the ratio 𝜙𝜙𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 /𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 does not depend on
z.

(b) The portion of the column under consideration is filled by mathematical fractal
flocs (so, no real flocs, see discussion above) that are touching one another. These
flocs are made of clay particles. If silt particles are present, they act as filling material
for the flocs:

Schematic representation of flocs at two different heights in the columns. The flocs are made
of clay particles (yellow) and some silt particles are imbedded inside the flocs (brown)

(c) at a same height, all flocs settle with the same velocity. As we also consider that
all flocs have the same fractal dimension and are made of the same primary particles,
this implies that all the flocs at a given height have the same size.

The volume fraction of clay is linked to the total solid volume fraction by:

𝜙𝜙𝑠𝑠 (𝑧𝑧) = 𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (𝑧𝑧) + 𝜙𝜙𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (𝑧𝑧)

By dividing this relation by 𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 one gets:

𝜙𝜙𝑠𝑠 (𝑧𝑧) 𝜙𝜙𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (𝑧𝑧)

=1+
𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (𝑧𝑧) 𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (𝑧𝑧)

As we have made the hypothesis that 𝜙𝜙𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 /𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 does not depend on z, we deduce
that the ratio 𝜙𝜙𝑠𝑠 /𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 does neither. This implies that at any given height the solid
volume fraction is only proportional to the clay fraction:

𝜙𝜙𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠
𝜙𝜙𝑠𝑠 (𝑧𝑧, 𝑡𝑡) = �1 + � 𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (𝑧𝑧, 𝑡𝑡)
𝜙𝜙𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐

275
 Introduction to Colloid Science

Relation for 𝑲𝑲(𝝓𝝓𝑺𝑺 )

In order to find the dependence of the

permeability on z we will use a dimension
analysis. If one calculates the permeability of
a system of pores as sketched opposite, one
can show that the permeability is
proportional to 132:

𝐾𝐾 ~ (1 − 𝜙𝜙𝑆𝑆 )𝑟𝑟 2

where r is the size of a pore.

Using this equivalence, we obtain:

2
𝐾𝐾(𝑧𝑧, 𝑡𝑡) ~ �𝑅𝑅𝑁𝑁 (𝑧𝑧, 𝑡𝑡)�

where we have used the assumptions that:

(a) 𝜙𝜙𝑆𝑆 ≪ 1. This hypothesis is certainly true at the beginning of consolidation, as we

have seen that, to give an order of magnitude, an initial concentration of 150 g/L
gives 𝜙𝜙0 = 150/2650 = 0.0566 ≪ 1. the hypothesis should however be checked
at the end of consolidation, for the deepest layers in the column.

(b) The size of the largest connecting pores scales as 𝑅𝑅𝑁𝑁 :

the blue circle is the middle (connecting pore) has the same size as the flocs (dashed circles)

This is a consequence of the scale invariance (a property of fractals). The fractal

approach is therefore a convenient and mathematically elegant way to describe the
change of volume fraction with position 𝑧𝑧 and time 𝑡𝑡. As both 𝑅𝑅𝑝𝑝 and D are assumed
be constant, mathematically if 𝜙𝜙𝑠𝑠 is increasing, 𝑅𝑅𝑁𝑁 is decreasing. This does not mean
that the “real” flocs (if any) in the system are becoming smaller: physically, what is

2
𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 ~ 𝑟𝑟 for one pore is the Poiseuille equation. For a bundle
132 The equation that gives 𝐾𝐾

of pores as sketched here, one has: 𝐾𝐾 = (1 − 𝜙𝜙𝑆𝑆 )𝐾𝐾𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 from simple geometrical
consideration.

276
 Chapter 10 Modelling the consolidation of slurries

expected to happen is that real flocs becomes squeezed and their shape (and
density) is changing. If there are no flocs but only suspended particles in the column
what happens is that the distance between particles is decreasing until they are
touching one another.

The fractal approach: the “flocs” are assumed to be space-filling, and the clay volume
fraction is increasing when the size of the flocs is decreasing. Note that the size of the pores
(voids between primary particles) indeed scales as 𝑅𝑅𝑁𝑁 .

Linking 𝑅𝑅𝑁𝑁 to 𝜙𝜙𝑠𝑠 , the permeability is defined by:

2
𝐾𝐾(𝑧𝑧, 𝑡𝑡) = 𝐾𝐾𝑘𝑘 [𝜙𝜙𝑠𝑠 (𝑧𝑧, 𝑡𝑡)]−𝑛𝑛 with 𝑛𝑛 =
3 − 𝐷𝐷
where 𝐾𝐾𝑘𝑘 is a factor not depending on z or t.

Relation for 𝛔𝛔𝒔𝒔𝒔𝒔 (𝝓𝝓𝑺𝑺 )

We now would like to find a similar type of relation for σ𝑠𝑠𝑠𝑠 . First we will define the
isotropic stress in the skeleton as:

σ0𝑠𝑠𝑠𝑠 = 𝐹𝐹(𝑛𝑛𝑡𝑡𝑡𝑡𝑡𝑡 − 𝑛𝑛𝑖𝑖 )

where 𝐹𝐹 is the force of one bond, 𝑛𝑛𝑡𝑡𝑡𝑡𝑡𝑡 is the total number of bonds per unit area:

𝑛𝑛𝑡𝑡𝑡𝑡𝑡𝑡 = 𝑛𝑛𝑎𝑎 + 𝑛𝑛𝑖𝑖

where 𝑛𝑛𝑎𝑎 is the number of bonds per unit area on which a stress is exerted (“active
bond”) and 𝑛𝑛𝑖𝑖 the number of bonds per unit area on which no stress is exerted
(“inactive bond”).

277
 Introduction to Colloid Science

In the fractal approach we take here, with touching fractal flocs, the total number of
bonds per unit area is given by:
𝑛𝑛0
𝑛𝑛𝑡𝑡𝑡𝑡𝑡𝑡 (𝑧𝑧, 𝑡𝑡) = 2
�𝑅𝑅𝑁𝑁 (𝑧𝑧, 𝑡𝑡)�

where 𝑛𝑛0 is the number of bonds per floc: this is invariant for any floc size (scale
invariance is a property of fractals):

the number of bonds (in this case 4) does not change with the size of flocs

We deduce that:
𝑛𝑛0
σ0𝑠𝑠𝑠𝑠 (𝑧𝑧, 𝑡𝑡) = 𝐹𝐹 � − 𝑛𝑛𝑖𝑖 �
(𝑅𝑅𝑁𝑁 (𝑧𝑧, 𝑡𝑡))2

Note that 𝐹𝐹, 𝑛𝑛0 and 𝑛𝑛𝑖𝑖 do not depend on z. From the dependence of 𝑅𝑅𝑁𝑁 on 𝜙𝜙𝑠𝑠 we
get:

2
σ0𝑠𝑠𝑠𝑠 (𝑧𝑧, 𝑡𝑡) ≈ 𝐴𝐴[𝜙𝜙𝑠𝑠 (𝑧𝑧, 𝑡𝑡)]𝑛𝑛 − 𝐹𝐹𝑛𝑛𝑖𝑖 with 𝑛𝑛 =
3 − 𝐷𝐷
where A is a factor that does not depend on z. The total effective stress σ0𝑠𝑠𝑠𝑠 is related
to the vertical effective (skeleton) stress σ𝑠𝑠𝑠𝑠 and the horizontal one σℎ𝑠𝑠𝑠𝑠 by:

1 2
σ0𝑠𝑠𝑠𝑠 = σ𝑠𝑠𝑠𝑠 + σℎ𝑠𝑠𝑠𝑠
3 3
This relation can be re-written:

1 + 2σℎ𝑠𝑠𝑠𝑠 /σ𝑠𝑠𝑠𝑠
σ0𝑠𝑠𝑠𝑠 = σ𝑠𝑠𝑠𝑠
3

Assuming that σℎ𝑠𝑠𝑠𝑠 /σ𝑠𝑠𝑠𝑠 does not depend on 𝜙𝜙𝑠𝑠 the skeleton stress is defined as:

σ𝑠𝑠𝑠𝑠 (𝑧𝑧, 𝑡𝑡) = 𝐾𝐾𝜎𝜎 [𝜙𝜙𝑠𝑠 (𝑧𝑧, 𝑡𝑡)]𝑛𝑛 − 𝐾𝐾𝜎𝜎0

where 𝐾𝐾𝜎𝜎 and 𝐾𝐾𝜎𝜎0 are factors that do not depend on z or t.

278
 Chapter 10 Modelling the consolidation of slurries

We have now found two constitutive equations 𝐾𝐾�𝜙𝜙𝑠𝑠 (𝑧𝑧, 𝑡𝑡)� and σ𝑠𝑠𝑠𝑠 �𝜙𝜙𝑠𝑠 (𝑧𝑧, 𝑡𝑡)� which
enables us to solve (analytically) the Gibson equation. We will do this for three
different cases. The first and second solutions given underneath correspond to the
primary consolidation regime defined in Chapter 9. The first solution is given for the
case there is both a settling and a consolidation phase whereas the second solution
corresponds to the case the settling phase has disappeared and there is only
consolidation. The transition between the first and second solution occurs at the
gelling time t1 that will be defined below. The last solution corresponds to the end
of consolidation, at equilibrium. For this solution creep effects are neglected.
Mathematically, this implies that 𝐾𝐾𝜎𝜎 and 𝐾𝐾𝜎𝜎0 are not time-dependent, i.e. that over
time the fabric (the clay skeleton) does not deform. The secondary compression
regime, as defined in Chapter 9 cannot be solved analytically and will be discussed
at the end of the chapter, where numerical solutions will be presented.

The setting regime (below the gelling time, t < t1)

At the initial stage of consolidation, we make the hypothesis that the self-weight is
almost entirely borne by the pore water, thus:

𝜕𝜕σ𝑠𝑠𝑠𝑠
≈0
𝜕𝜕𝜕𝜕
for all heights. Intuitively, one can already predict that this hypothesis is
questionable as time increases for the deepest layers close to the bottom of the
column, which we define by the position z = 0. The validity of the assumption and its
consequences will be discussed at the end of the chapter. From the previous
subsection, we have:

𝐾𝐾(𝑧𝑧, 𝑡𝑡) = 𝐾𝐾𝑘𝑘 [𝜙𝜙𝑠𝑠 (𝑧𝑧, 𝑡𝑡)]−𝑛𝑛

The Gibson equation therefore becomes:

𝜕𝜕𝜙𝜙𝑆𝑆 𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 𝜕𝜕𝜙𝜙𝑠𝑠

= 𝐾𝐾𝑘𝑘 (2 − 𝑛𝑛)𝜙𝜙𝑠𝑠1−𝑛𝑛
𝜕𝜕𝜕𝜕 𝜌𝜌𝑤𝑤 𝜕𝜕𝜕𝜕

This equation is a nonlinear convection equation, similar to Kynch’s sedimentation

equation, and can be solved by the method of characteristics.

The method of characteristics

The method can be applied to solve equations of the type:

𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
𝑎𝑎(𝑥𝑥, 𝑡𝑡, 𝑢𝑢) + 𝑏𝑏(𝑥𝑥, 𝑡𝑡, 𝑢𝑢) = 𝑐𝑐(𝑥𝑥, 𝑡𝑡, 𝑢𝑢)
𝜕𝜕𝑡𝑡 𝜕𝜕𝜕𝜕

279
 Introduction to Colloid Science

This method relies on the property that the function 𝑢𝑢(𝑥𝑥, 𝑡𝑡) can be differentiated
with respect to a variable 𝑠𝑠 (not yet defined) as:

𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕

= +
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
If we now define the variable s as the curvilinear coordinate along a curve in the (𝑥𝑥, 𝑡𝑡)
plane by:

𝜕𝜕𝜕𝜕
= 𝑎𝑎(𝑥𝑥, 𝑡𝑡, 𝑢𝑢)
𝜕𝜕𝜕𝜕
𝜕𝜕𝜕𝜕
= 𝑏𝑏(𝑥𝑥, 𝑡𝑡, 𝑢𝑢)
𝜕𝜕𝜕𝜕
Then we simply have to solve, along with the 2 previous equations:

𝑑𝑑𝑑𝑑
= 𝑐𝑐(𝑥𝑥, 𝑡𝑡, 𝑢𝑢)
𝑑𝑑𝑑𝑑
In other words: we have transformed an equation with partial differentials in a series
of 3 ordinary differential equations that are easier to solve.

In our case, we can apply the method of characteristics using the function 𝜙𝜙𝑆𝑆 instead
of 𝑢𝑢 and the variables 𝑧𝑧, 𝑡𝑡 instead of 𝑥𝑥, 𝑡𝑡. We then find that we have to solve:

𝜕𝜕𝜕𝜕
=1
𝜕𝜕𝜕𝜕
𝜕𝜕𝜕𝜕 𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤
= −𝐾𝐾𝑘𝑘 (2 − 𝑛𝑛)𝜙𝜙𝑠𝑠1−𝑛𝑛
𝜕𝜕𝜕𝜕 𝜌𝜌𝑤𝑤

𝑑𝑑𝜙𝜙𝑆𝑆
=0
𝑑𝑑𝑑𝑑
The last equation can be solved into:

𝜙𝜙𝑆𝑆 = constant

This implies that the lines of constant 𝜙𝜙𝑆𝑆 are representing the curves associated with
the curvilinear coordinate s. These lines are represented by the black lines on the
figure underneath.

280
 Chapter 10 Modelling the consolidation of slurries

Variation of the interface with time for a suspension that at start is below the gelling
concentration. 𝜙𝜙𝑆𝑆 represents the volume fraction of the particles, 𝜙𝜙0 is their initial volume
fraction and 𝜙𝜙𝑔𝑔𝑔𝑔𝑔𝑔 = 𝜙𝜙0 is the volume fraction at the suspension/bed interface, below which
𝜙𝜙𝑆𝑆 > 𝜙𝜙𝑔𝑔𝑔𝑔𝑔𝑔 . Until the gelling time t1 there is a suspension/bed interface. Above t1 only the
water/bed interface remains.

Above the suspension/bed line the volume fraction is by definition constant and
equal to 𝜙𝜙0 so that the lines of constant 𝜙𝜙𝑆𝑆 occupy the whole area below the
water/suspension interface.

The variable s can be eliminated by combining the two first equations from which
𝑧𝑧(𝑡𝑡) is obtained:
𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤
𝑧𝑧(𝑡𝑡) = 𝐾𝐾𝑘𝑘 (𝑛𝑛 − 2)𝜙𝜙𝑠𝑠1−𝑛𝑛 𝑡𝑡 + 𝑧𝑧(0)
𝜌𝜌𝑤𝑤

Solving this equation for 𝜙𝜙𝑆𝑆 with 𝑧𝑧(0) = 0, one finds for 0 < 𝑧𝑧 ≤ ℎ𝑏𝑏 :
1
𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 𝑡𝑡 𝑛𝑛−1
𝜙𝜙𝑆𝑆 (𝑧𝑧, 𝑡𝑡) = �𝐾𝐾𝑘𝑘 (𝑛𝑛 − 2) �
𝜌𝜌𝑤𝑤 𝑧𝑧

An interesting case if for 𝑧𝑧 = ℎ𝑏𝑏 for which 𝜙𝜙𝑆𝑆 = 𝜙𝜙0 . It enables to find the height of
the bed with time:
𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤
ℎ𝑏𝑏 (𝑡𝑡) = 𝐾𝐾𝑘𝑘 (𝑛𝑛 − 2)𝜙𝜙01−𝑛𝑛 𝑡𝑡
𝜌𝜌𝑤𝑤

from which we deduce that:

1
ℎ𝑏𝑏 (𝑡𝑡) 𝑛𝑛−1
𝜙𝜙𝑆𝑆 (𝑧𝑧, 𝑡𝑡) = 𝜙𝜙0 � �
𝑧𝑧

281
 Introduction to Colloid Science

As drawn at the beginning of the chapter, ℎ𝑏𝑏 (𝑡𝑡) is a linearly increasing function of
time and hence 𝑣𝑣𝑏𝑏𝑏𝑏𝑏𝑏 is constant as function of time. Using conservation of mass (by
evaluating the Gibson height 𝐿𝐿𝐺𝐺 defined in Chapter 9) allows to determine ℎ(𝑡𝑡), i.e.
the water/suspension interface height:
ℎ ℎ𝑏𝑏 ℎ
𝐿𝐿𝐺𝐺 = � 𝜙𝜙𝑆𝑆 𝑑𝑑𝑑𝑑 = � 𝜙𝜙𝑆𝑆 (𝑧𝑧, 𝑡𝑡)𝑑𝑑𝑑𝑑 + � 𝜙𝜙0 𝑑𝑑𝑑𝑑 = 𝜙𝜙0 𝐻𝐻
0 0 ℎ𝑏𝑏

where 𝐻𝐻 is the height of the suspension in the column at 𝑡𝑡 = 0. After integration

one finds:
1
𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 𝑛𝑛−1 1 − 𝑛𝑛 𝑛𝑛−2
𝐿𝐿𝐺𝐺 = �𝐾𝐾𝑘𝑘 (𝑛𝑛 − 2)𝑡𝑡� (ℎ𝑏𝑏 )𝑛𝑛−1 + 𝜙𝜙0 (ℎ − ℎ𝑏𝑏 )
𝜌𝜌𝑤𝑤 2 − 𝑛𝑛

Combining the expressions for 𝐿𝐿𝐺𝐺 and ℎ𝑏𝑏 one finds:

1
𝐿𝐿𝐺𝐺 = � � 𝜙𝜙 ℎ + 𝜙𝜙0 ℎ
𝑛𝑛 − 2 0 𝑏𝑏
𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 1−𝑛𝑛
ℎ(𝑡𝑡) = 𝐻𝐻 − 𝐾𝐾𝑘𝑘 𝜙𝜙0 𝑡𝑡
𝜌𝜌𝑤𝑤

The gelling time

The time t1 is called the “gelling time” as the settling phase has now disappeared. At
𝑡𝑡 = 𝑡𝑡1 one has ℎ = ℎ𝑏𝑏 (the interface water/suspension has disappeared) and from
the relation found above we get:

𝜌𝜌𝑤𝑤 𝐿𝐿𝐺𝐺 𝜙𝜙0𝑛𝑛−2

𝑡𝑡1 =
𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 (𝑛𝑛 − 1)𝐾𝐾𝑘𝑘

The height at the gelling time is given by

𝑛𝑛 − 2
ℎ𝑏𝑏 (𝑡𝑡1 ) = ℎ(𝑡𝑡1 ) = 𝐻𝐻
𝑛𝑛 − 1
Even though this height is independent on 𝐾𝐾𝑘𝑘 and could be used to determine 𝑛𝑛, in
practice, it is very difficult to estimate. A better way to estimate 𝑛𝑛 is given in the next
section.

The primary compression regime (above the gelling time, t > t1)

When t > t1 the interface water/suspension has disappeared. The density profile is
then independent on the initial condition. In that case, we have:

282
 Chapter 10 Modelling the consolidation of slurries

ℎ
𝐿𝐿𝐺𝐺 = � 𝜙𝜙𝑆𝑆 (𝑧𝑧, 𝑡𝑡)𝑑𝑑𝑑𝑑
0

One then finds:

1
𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 𝑛𝑛−1 1 − 𝑛𝑛 𝑛𝑛−2
𝐿𝐿𝐺𝐺 = �𝐾𝐾𝑘𝑘 (2 − 𝑛𝑛)𝑡𝑡� (ℎ)𝑛𝑛−1
𝜌𝜌𝑤𝑤 2 − 𝑛𝑛

This leads to:

1−𝑛𝑛 1
2 − 𝑛𝑛 2−𝑛𝑛 𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 2−𝑛𝑛
ℎ(𝑡𝑡) = �𝐿𝐿𝐺𝐺 � �𝐾𝐾𝑘𝑘 (𝑛𝑛 − 2)𝑡𝑡�
1 − 𝑛𝑛 𝜌𝜌𝑤𝑤

We have now found that

1 3−𝐷𝐷
ℎ(𝑡𝑡) ~ 𝑡𝑡 2−𝑛𝑛 = 𝑡𝑡 4−2𝐷𝐷

The fractal dimension D can therefore be determined from the water/gel interface
as function of time. When plotting the data on a double logarithmic scale, one gets
(3 − 𝐷𝐷)/(4 − 2𝐷𝐷) from the slope of the obtained line.

The unknown parameter 𝐾𝐾𝑘𝑘 can subsequently be found by fitting the data for any
time. In practice it is often found that the values for 𝐾𝐾𝑘𝑘 differ when the data is
plotted for t < t1 or for t > t1 as settling and consolidation are unrelated. From 𝐾𝐾𝑘𝑘
the permeability 𝐾𝐾(𝑧𝑧) can be estimated.

Examples

The analytical solution for t < t2

A typical interface profile is given underneath where we used the found analytical
expressions for ℎ(𝑡𝑡) and ℎ𝑏𝑏 (𝑡𝑡) and the following representative parameters:

𝜌𝜌𝑤𝑤 = 103 kg/m3 𝜌𝜌𝑠𝑠 = 2600 kg/m3 H=1m g = 9.81 m2/s

D = 2.63 𝜙𝜙0 = 2.5 % 𝐾𝐾𝑘𝑘 = 1.5 × 10−9 m/s

From the figure, one can see that the analytical solution gives unrealistic values for
the interface at long times. In fact, at infinite time the size of the interface will be
zero, as can be easily verified from the expression for ℎ(𝑡𝑡) given above. At long
times, it is not correct to assume that the effects of the skeleton stress can be
omitted. This is discussed further in the next example, where we compare the
analytical solution to the numerical solution, found by solving the full Gibson
equation.

283
 Introduction to Colloid Science

The final size of the bed (when the consolidation has stopped and 𝜙𝜙𝑠𝑠 (𝑧𝑧) is
independent of time) can also be evaluated analytically, as shown in the next
section.

Water / sediment interface with time. In the settling phase the volume fraction at any height
is given by the initial volume fraction. Darker region: build-up of the bed-sol interface. Note
that we used a semilog scale. The two lines join at t = t1 (gel point). In this example t1 = 8.2 s.

Comparison between the analytical and numerical solution of the Gibson equation
for t < t1

The analytical solutions given in the sections above are here compared with the
numerical solution of the full Gibson equation. We recall that the analytical solution
has been derived under the assumption that 𝜕𝜕σ𝑠𝑠𝑠𝑠 ⁄𝜕𝜕𝜕𝜕 ≈ 0.

284
 Chapter 10 Modelling the consolidation of slurries

Volume fraction as function of height in the column after 290 days. Dashed red: initial
volume fraction; Dotted black: volume fraction end profile; dashed green: analytical solution;
blue line: numerical solution. 𝐾𝐾𝑘𝑘 = 1.74 × 10−13 m/s, 𝐾𝐾𝜎𝜎 = 1.26 × 107 Pa, D = 2.7, 𝜙𝜙0 =
0.2; other parameters as in the table given above.

The volume fraction profile is for the situation we are below the gel(ling) point. This
is confirmed on the figure by the fact that between 0.4 and 0.8 m the volume fraction
is still equal to the initial volume fraction. The blue line, which corresponds to the
numerical solution of Gibson equation, using both the permeability and effective
stress terms, would reduce to the green dashed line, which corresponds to the
analytical profile if the effective stress term would be set equal to zero.

Because of the effective stress term, which mathematically corresponds to a

diffusion term (the permeability term is comparable to an advection term), the
interfaces between water / suspension and suspension / bed are not clearly defined
for the numerical solution. One can therefore already conclude that if one wants to
use the analytical method detailed above in combination with measurements of
interfaces against time this will only be possible for systems in which the effective
stress terms are quite small – these kind of systems are for instance suspensions of
concentrated hard (unflocculated) particles. Experimentally, as we have seen
already in Chapter 8 (see section “Finding the hindered settling function from the
equilibrium profile”) the change in volume fraction between 0 and 𝜙𝜙0 occurs over a
distance of the order of 1 or 2 mm (in the numerical example we showed above, we
used a very large 𝐾𝐾𝜎𝜎 to exaggerate the effect).

The discrepancy between analytical and numerical solution at the bottom of the
column clearly demonstrate that even below the gelling point (in the primary
consolidation regime) the assumption

285
 Introduction to Colloid Science

𝜕𝜕σ𝑠𝑠𝑠𝑠
≈0
𝜕𝜕𝜕𝜕
is not valid. In the absence of diffusion, there is no constraint to prevent all the
particles reaching the bottom of the column as we have implicitly treated the
particles as having no volume (point-like objects). Mathematically, this is symbolised
by the fact that the analytical solution (in the absence of effective stress – i.e. a
diffusion term), scales as
1
1 𝑛𝑛−1
𝜙𝜙𝑆𝑆 (𝑧𝑧, 𝑡𝑡) ~ � �
𝑧𝑧
At the bottom of the column one gets 𝜙𝜙𝑆𝑆 (𝑧𝑧 = 0, 𝑡𝑡) → ∞. This unphysical result
implies that the water/suspension interface, which is calculated from mass
conservation through the evaluation of the Gibson height, will become lower than
the actual interface at longer times. In the case of the settling of hard particles,
where only advection (settling) play a role, it is numerically easy to limit the filling of
the settling column by imposing that each slice of the column should not exceed a
given volume fraction, i.e. the maximum packing volume fraction of particles.

In conclusion, the fractal approach method explained in this chapter can be applied
for t < t2 in the case of concentrated hard particles that are settling without
significant diffusion. From the procedures explained in the previous sections for
below and just above the gelling time, 𝐾𝐾𝑘𝑘 and the fractal parameter 𝑛𝑛 dimension can
be obtained from which the permeability can be deduced.

As discussed in the introduction, these fractal parameter 𝑛𝑛 and 𝐾𝐾𝑘𝑘 parameter are
however not expected to be the same above and below t1 as in the model we have
made the hypothesis that the flocs are space-filling both in the settling and in the
consolidation phase. In fact, the space-filling criterion is in most cases expected to
be valid above t1 and not below.

In a later section, we will show how this cross-over between settling and
consolidation is found in experiments, and how we can model it.

Determination of the effective stress : end of consolidation (t > t2)

From the results of this subsection, we will get an estimated of the final bed height
when consolidation is finished.

In the final stage of consolidation, the structure is bearing and we can estimate that

𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 ≈ 0

From the modified Darcy equation (see Chapter 9), we had:

286
 Chapter 10 Modelling the consolidation of slurries

𝑘𝑘 𝜕𝜕𝑃𝑃𝑒𝑒 𝐾𝐾 𝐾𝐾 𝜕𝜕σ𝑠𝑠𝑠𝑠
𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = = (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆 +
𝜂𝜂 𝜕𝜕𝜕𝜕 𝑔𝑔𝜌𝜌𝑤𝑤 𝑔𝑔𝜌𝜌𝑤𝑤 𝜕𝜕𝜕𝜕

Which now gives:

1 𝜕𝜕σ𝑠𝑠𝑠𝑠
𝜙𝜙𝑆𝑆 + ≈0
(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔 𝜕𝜕𝜕𝜕

Using the equation

σ𝑠𝑠𝑠𝑠 = 𝐾𝐾𝜎𝜎 [𝜙𝜙𝑠𝑠 (𝑧𝑧)]𝑛𝑛 − 𝐾𝐾𝜎𝜎0

We substitute 𝜙𝜙𝑠𝑠 and find:

𝑛𝑛 𝜕𝜕𝜙𝜙𝑠𝑠
𝜙𝜙𝑆𝑆 + 𝐾𝐾 [𝜙𝜙 ]𝑛𝑛−1 =0
(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔 𝜎𝜎 𝑠𝑠 𝜕𝜕𝜕𝜕

This yields

𝜕𝜕[𝜙𝜙𝑠𝑠 ]𝑛𝑛−1 −(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔

= (𝑛𝑛 − 1)
𝜕𝜕𝜕𝜕 𝑛𝑛𝐾𝐾𝜎𝜎

At long times, the bed/water interface has reached a constant value (𝑧𝑧 = ℎ∞ ) and
the final volume fraction profile can be estimated to be, using σ𝑠𝑠𝑠𝑠 (ℎ∞ ) = 0:
1
𝐾𝐾𝜎𝜎0 𝑛𝑛
𝜙𝜙𝑠𝑠 (ℎ∞ ) = � �
𝐾𝐾𝜎𝜎

We will assume that creep effects are neglected, hence 𝐾𝐾𝜎𝜎0 = 0 and

𝜙𝜙𝑠𝑠 (ℎ∞ ) = 0

We can now evaluate:

ℎ∞
𝜕𝜕[𝜙𝜙𝑠𝑠 ]𝑛𝑛−1 ℎ∞
−(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔
� 𝑑𝑑𝑑𝑑 = � (𝑛𝑛 − 1)𝑑𝑑𝑑𝑑
𝑧𝑧 𝜕𝜕𝜕𝜕 𝑧𝑧 𝑛𝑛𝐾𝐾𝜎𝜎

which gives:
1
(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔 𝑛𝑛−1
𝜙𝜙𝑠𝑠 (𝑡𝑡 → ∞, 𝑧𝑧) = � (𝑛𝑛 − 1)(ℎ∞ − 𝑧𝑧)�
𝑛𝑛𝐾𝐾𝜎𝜎

1 (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔

log[𝜙𝜙𝑠𝑠 (𝑡𝑡 → ∞, 𝑧𝑧)] = �log[(ℎ∞ − 𝑧𝑧)] + log � (𝑛𝑛 − 1)��
𝑛𝑛 − 1 𝑛𝑛𝐾𝐾𝜎𝜎

287
 Introduction to Colloid Science

By plotting the particle volume fraction, obtained from a measured density profile
in the final stage of consolidation, versus the distance below the interface (ℎ∞ − 𝑧𝑧)
on a double logarithmic plot the fractal dimension (via 𝑛𝑛) can be derived from the
slope. Subsequently, the parameter 𝐾𝐾𝜎𝜎 can be obtained from any value of 𝑧𝑧. This
method can lead to very large estimation errors however.

The final bed height can be obtained by using once more the Gibson height:
ℎ∞
𝐿𝐿𝐺𝐺 = 𝜙𝜙0 𝐻𝐻 = � 𝜙𝜙𝑆𝑆 (𝑡𝑡 → ∞, 𝑧𝑧)𝑑𝑑𝑑𝑑
0

which gives:
1
𝑛𝑛 𝐾𝐾𝜎𝜎 𝑛𝑛
ℎ∞ = 𝜙𝜙 𝐻𝐻 � �
(𝑛𝑛 − 1) 0 (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙0 𝐻𝐻

The volume fraction at the bottom of the settling column is given by

𝑛𝑛 𝐻𝐻
𝜙𝜙𝑠𝑠 (𝑡𝑡 → ∞, 𝑧𝑧 = 0) = 𝜙𝜙0
(𝑛𝑛 − 1) ℎ∞

which also enables to find the fractal dimension in a simple way. The analytical
expression is found to correctly predict the evolution of density as function of height,
as shown in the following example:

Final density profile of settled beds. Symbols: density measurements. Only the data for the
bed of initial concentration 200 g/L (𝜙𝜙0 = 0.077) has been fitted which gave the values 𝐾𝐾𝜎𝜎 =
6.5 × 106 Pa and D = 2.65. All the other lines are plotted using these parameters.

288
 Chapter 10 Modelling the consolidation of slurries

The sediment used is from lake Markermeer 133. The density measurements were
done by Ultra-sonic High Concentration Meter (UHCM). We recall that the density is
related to the volume fraction 𝜙𝜙𝑠𝑠 by: 𝜌𝜌 = 𝜌𝜌𝑠𝑠 𝜙𝜙𝑠𝑠 + (1 − 𝜙𝜙𝑠𝑠 )𝜌𝜌𝑤𝑤 . By fitting one dataset,
the one corresponding to the initial concentration of 200 g/L, it was possible to
predict the bed height for all other initial concentrations.

It remains to be seen if the fractal dimension found by fitting the “end of

consolidation” data is the same as the one found by fitting the data above t1 . We
will show that this is the case with the data and protocol we present.

Settling and consolidation of natural mud

In the “end of consolidation” section given above, we have shown the fit of a density
profile for a mud suspension of concentration 200 g/L, from which it was found that
𝐾𝐾𝜎𝜎 = 6.5 × 106 Pa and D = 2.65. The parameter 𝐾𝐾𝑘𝑘 = 2.0 ∙ 10−12 m/s is found by
fitting the time evolution of the interface for the 200 g/L mud sample using the full
Gibson equation, keeping 𝐾𝐾𝜎𝜎 and 𝐷𝐷 constant and equal to the values found from
fitting the density profile. We give here the theoretical evolution of density and
excess pore water pressure for this sample:

van den Bosch, B.A.P. The effect of initial concentration on the consolidation behaviour of
133

mud: A study on lake Markermeer sediment, Master thesis, TU Delft, 2016

289
 Introduction to Colloid Science

The full numerical solution is represented in blue. The analytical solution of

Merkelbach and Kranenburg (MK) is in pink. In contrast to the example given earlier
in this chapter, the analytical and numerical solutions agree with each other below
the gelling point (in the period before 70 hours), as the 𝐾𝐾𝜎𝜎 is ten times smaller than
in the previous example. At the end of consolidation the numerical solution is not to
be distinguished from the exact analytical solution (dashed black line) and the excess
pore water pressure is zero as 𝑃𝑃𝑤𝑤 = 𝑃𝑃ℎ𝑦𝑦𝑦𝑦 . The initial pore water pressure (red
dashed line) is obtained by realizing that 𝜕𝜕𝜕𝜕𝑠𝑠𝑠𝑠 ⁄𝜕𝜕𝜕𝜕 ~ 0, which gives:

𝑃𝑃𝑒𝑒 (𝑡𝑡 = 0) = (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑠𝑠 (𝐻𝐻 − 𝑧𝑧)

where 𝐻𝐻 is the height of fluid in the column. The theoretical time evolution of the
water/suspension and suspension/bed for 40 g/L and 200 g/L is given here:

290
 Chapter 10 Modelling the consolidation of slurries

There is a good agreement between the numerical and analytical profiles until 104 s,
but, as expected, the analytical solution becomes incorrect when diffusion (effective
stress) plays a significant role. This happens much before the end of consolidation
for highly concentrated mud suspensions.

In contrast to these theoretical profiles, when recording the time evolution of the
interfaces of real suspensions, an observable transition is expected between the
settling and the consolidation phases, in particular for the most diluted suspension
(40 g/L). This is indeed what has been found, see figure below. For concentrations
less than 200 g/L, there is a discontinuity in the profile at the gel point. The
theoretical predictions, using the same parameters 𝐾𝐾𝑘𝑘 , 𝐾𝐾𝜎𝜎 and 𝐷𝐷 are given in full
lines. For samples of higher concentrations (200 g/L, 300 g/L and 400 g/L) one can
observe that data and predictions are in very good agreement.

We checked that the procedure described at the beginning of this chapter, i.e. using
the analytical approximations just above t1, works on these samples and that the
numerical solution of the full Gibson equation reduces to the analytical solution in
its range of validity (above t1 but well below t2).

Time evolution of water/sediment interface for different sediment concentrations. Symbols

are measurements. Lines are fits (see text). Full lines: 𝐾𝐾𝜎𝜎 = 6.5 ∙ 106 Pa, 𝐾𝐾𝑘𝑘 = 2.0 ∙ 10−12
m/s and 𝐷𝐷 = 2.65. The sediment is from lake Markermeer.

For lower concentrations, the flocs are not interconnected at start (reflected in the
change in slope around 103 s for the 40 g/L sample). The settling phase can then be
fitted (dashed lines) using
𝑚𝑚
ℎ = ℎ0 − 𝑡𝑡 ∙ �1 − 𝜙𝜙𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 � 𝑣𝑣𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆

291
 Introduction to Colloid Science

where ℎ(𝑚𝑚) is the position of the water / suspension, ℎ0 (𝑚𝑚) the initial height. We
used 𝑣𝑣𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 = 1.5 mm/s which is the Stokes settling velocity for 40 µm particles, in
agreement with the fact that the clay was sieved through a 63 µm sieve prior
dispersion. From the fit we found ϕgel = 0.075 (195 g/L), which agrees with the fact
that the 200 g/L sample is in a gel state.

Using the same ϕgel and 𝑣𝑣𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 , the settling phase of all samples below gelling
concentrations (40 g/L and 100 g/L) could reasonably be predicted. For the lowest
clay concentration (40 g/L) there is a good transition between the settling and the
consolidation model. This is less the case for the 100 g/L sample. In order to improve
the fit, it was necessary to double the value of 𝐾𝐾𝑘𝑘 (red dashed line). The reason of
the mismatch could be linked to the limitation of the model in the suspension/gel
transition range or due to the fact that the 100 g/L sample was not well-mixed at the
onset of the experiment and that a sediment layer was already present at the
bottom of the column. To check this last hypothesis, one should then know the
density profile at start (not known for the present set of experiments).

Link with Chapter 8 : how to couple settling and consolidation?

In Chapter 8, we found for settling that the velocity of the suspension/bed interface
is given by

𝜙𝜙0 𝑓𝑓(𝜙𝜙0 )
𝑣𝑣𝑏𝑏𝑏𝑏𝑏𝑏 = 𝑣𝑣0
𝜙𝜙𝑚𝑚 − 𝜙𝜙0

where 𝜙𝜙𝑚𝑚 is the solid volume fraction in the bed. Using the fractal approach we find
that for the early stage of consolidation

𝑑𝑑ℎ𝑏𝑏 𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤

𝑣𝑣𝑏𝑏𝑏𝑏𝑏𝑏 = = 𝐾𝐾𝑘𝑘 (𝑛𝑛 − 2)𝜙𝜙01−𝑛𝑛
𝑑𝑑𝑑𝑑 𝜌𝜌𝑤𝑤

for the settling of flocs. Depending on the sample and the interest of the researcher,
one or the other of these formulations are used. Both formulations depend on a-
priori unknown parameters: 𝑓𝑓(𝜙𝜙0 ) or 𝐾𝐾𝑘𝑘 and 𝑛𝑛, and both are decreasing functions
of 𝜙𝜙0 . These parameters can be found by fitting the function 𝑣𝑣𝑏𝑏𝑏𝑏𝑏𝑏 (𝜙𝜙0 ).

In Chapter 8, we found for the settling phase that the velocity of the
water/suspension interface is given by

𝑣𝑣𝑠𝑠 = 𝑓𝑓(𝜙𝜙0 )𝑣𝑣0

We here find that (in absolute values):

𝑑𝑑ℎ 𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 1−𝑛𝑛

𝑣𝑣𝑠𝑠 = = 𝐾𝐾𝑘𝑘 𝜙𝜙0
𝑑𝑑𝑑𝑑 𝜌𝜌𝑤𝑤

292
 Chapter 10 Modelling the consolidation of slurries

By comparing the expressions for 𝑣𝑣𝑠𝑠 and 𝑣𝑣𝑏𝑏𝑏𝑏𝑏𝑏 found in Chapter 8 and the fractal
approach used here, we find the following equivalence:

2 𝜙𝜙0
𝑛𝑛 = =2+
3 − 𝐷𝐷 𝜙𝜙𝑚𝑚 − 𝜙𝜙0

In Chapter 8 we only considered a settling column filled with hard spheres. This
means that in the bed, these hard spheres are in contact (they do not consolidate)
and hence the volume fraction 𝜙𝜙𝑚𝑚 is a constant. In the fractal approach, there is, on
the contrary, a smooth variation in volume fraction 𝜙𝜙𝑠𝑠 in the bed as function of z
and t. One can estimate the average volume fraction in the bed from:

1 ℎ𝑏𝑏
𝜙𝜙𝑚𝑚 = � 𝜙𝜙 𝑑𝑑𝑑𝑑
ℎ𝑏𝑏 0 𝑠𝑠

which yields:

𝑛𝑛 − 1
𝜙𝜙𝑚𝑚 = 𝜙𝜙0
𝑛𝑛 − 2
(note that 𝜙𝜙𝑚𝑚 does not depend on time). Inserting this equation in the relation
above gives the consistent relation 𝑛𝑛 = 𝑛𝑛.

In the early stage of consolidation, in the case the volume fractions at the bottom of
the column remain realistic, one can make the assumption that

𝜕𝜕σ𝑠𝑠𝑠𝑠 𝜕𝜕𝑃𝑃𝑒𝑒
≪
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
and this leads to

𝑔𝑔𝜌𝜌𝑤𝑤 (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )

(1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑠𝑠 = 𝑘𝑘 𝜙𝜙𝑆𝑆
𝜂𝜂 𝜌𝜌𝑤𝑤

We recall that 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 is the velocity of the settling particles measured in the frame of
the laboratory (see Chapter 9) and that

(1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑠𝑠 = −𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙

𝑔𝑔𝜌𝜌𝑤𝑤
𝐾𝐾(𝑚𝑚⁄𝑠𝑠) = 𝑘𝑘(𝑚𝑚2 )
𝜂𝜂

The settling velocity 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 is pointing downwards in the z direction, and therefore a
minus sign appears while doing the substitution. We here gave the absolute value of
𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 :

293
 Introduction to Colloid Science

(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )
𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 = 𝐾𝐾 𝜙𝜙𝑆𝑆
𝜌𝜌𝑤𝑤

This equation is independent of any formulation for the permeability. In the present
chapter, so far, we have analysed the settling and consolidation phases with the
same model. One can however wonder if this is the most suited approach.
Intuitively, one can see that in the settling phase (where 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 can be measured by
recording the water/suspension interface) there will be other forces acting on the
particles than in the consolidation phase, and hence 𝑣𝑣𝑠𝑠/𝑙𝑙𝑙𝑙𝑙𝑙 will be different in the
settling and in the consolidation regime. The (hindered) settling regime was analysed
in Chapter 8 whereas the consolidation regime was defined in Chapter 9. We will
now discuss about how to link these two regimes.

In Chapter 8 we have discussed the settling of hard spheres that were experiencing
repulsion (which lead to the introduction of an osmotic pressure term Π). We found
the following equation to describe the settling behaviour:

6𝜋𝜋𝜋𝜋𝜋𝜋 𝑉𝑉𝑝𝑝 𝜕𝜕Π

(𝑣𝑣𝑠𝑠 − 𝑣𝑣𝑤𝑤 ) = −(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝑉𝑉𝑝𝑝 −
𝜒𝜒(𝜙𝜙𝑆𝑆 ) 𝜙𝜙𝑆𝑆 𝜕𝜕𝜕𝜕

This equation can be rewritten:

𝜕𝜕Π 𝜙𝜙𝑆𝑆 6𝜋𝜋𝜋𝜋𝜋𝜋

= 𝑣𝑣 − (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆
𝜕𝜕𝜕𝜕 𝑉𝑉𝑝𝑝 𝜒𝜒(𝜙𝜙𝑆𝑆 ) 𝑤𝑤/𝑠𝑠

We recall the definition of the excess pore water pressure we have defined in
Chapter 9, and the associated Darcy-like equation:

𝜕𝜕𝑃𝑃𝑒𝑒 𝜕𝜕σ𝑠𝑠𝑠𝑠
− = + (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
−𝑘𝑘 𝜕𝜕𝑃𝑃𝑒𝑒
(1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑠𝑠 =
𝜂𝜂 𝜕𝜕𝜕𝜕

Combining these two last equations leads to:

𝜕𝜕σ𝑠𝑠𝑠𝑠 𝜂𝜂
= (1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑠𝑠 − (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆
𝜕𝜕𝜕𝜕 𝑘𝑘
Comparing the equation found in Chapter 8 given above and the last equation, it is
tempting to identify:

2 2 1 − 𝜙𝜙𝑆𝑆 2 𝑓𝑓(𝜙𝜙𝑆𝑆 )
𝑘𝑘(𝜙𝜙𝑆𝑆 ) = 𝑎𝑎 � � 𝜒𝜒(𝜙𝜙𝑆𝑆 ) = 𝑎𝑎2
9 𝜙𝜙𝑆𝑆 9 𝜙𝜙𝑆𝑆

294
 Chapter 10 Modelling the consolidation of slurries

where we used 𝑉𝑉𝑝𝑝 = 4𝜋𝜋𝑎𝑎3 /3 and 𝑓𝑓(𝜙𝜙𝑆𝑆 ) = (1 − 𝜙𝜙𝑆𝑆 )𝜒𝜒(𝜙𝜙𝑆𝑆 ). We then can set-up a
general force balance for a water-sediment mixture:

𝜕𝜕σ𝑠𝑠𝑠𝑠 𝜕𝜕Π 𝜂𝜂
+ = (1 − 𝜙𝜙𝑆𝑆 )𝑣𝑣𝑤𝑤/𝑠𝑠 − (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑔𝑔𝜙𝜙𝑆𝑆
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝑘𝑘
In the soil phase, one has Π = 0 and in the suspension σ𝑠𝑠𝑠𝑠 = 0. This last equation
was proposed by Toorman in 1996 134.

This equation simply tells that the initial stage of consolidation of soils, as defined in
Chapter 9 (with a settling regime, a primary consolidation regime and a secondary
compression regime) can be modelled by an advection-diffusion equation, but that
the advective and diffuse terms are different in each regime.

Depending on the composition of the suspension (hard or soft particles) and the
initial volume fraction (under or above the gelling state), different advective and
diffusive terms will be chosen, and the transition between the settling regime and
the primary consolidation regime has to be defined carefully. In the extreme case of
the settling of hard heavy spheres for example, the transition between the settling
phase and consolidated bed is extremely abrupt, as the velocity of a hard sphere
goes from a finite value to zero over an extremely small distance when it is hitting
the bed. This can be challenging to program numerically.

Hindered settling of flocculated clay

The derivations in Chapter 8 are done assuming that the settling particles are hard
spheres. If Brownian motion would be neglected, this would imply that at infinite
times the profile would be 𝜙𝜙𝑆𝑆 = 𝜙𝜙𝑚𝑚𝑚𝑚𝑚𝑚 (the maximum packing volume fraction) until
the interface where 𝜙𝜙𝑆𝑆 would jump to zero. Setting-up the Richardson-Zaki hindered
settling function, we have, for simplicity, assumed that 𝜙𝜙𝑚𝑚𝑚𝑚𝑚𝑚 = 1 which leads to
𝑓𝑓(𝜙𝜙𝑆𝑆 ) = 0 for 𝜙𝜙𝑆𝑆 = 1. A more realistic Richarson-Zaki profile 135 for hard spheres is

𝑓𝑓(𝜙𝜙𝑆𝑆 ) = (1 − 𝜙𝜙𝑆𝑆 /𝜙𝜙𝑚𝑚𝑚𝑚𝑚𝑚 )𝑚𝑚

where 𝜙𝜙𝑚𝑚𝑚𝑚𝑚𝑚 is close to 0.74 for packed spheres. For flocculated clay or mud
suspensions, at 𝜙𝜙𝑆𝑆 = 𝜙𝜙𝑚𝑚𝑚𝑚𝑚𝑚 or more precisely at 𝜙𝜙𝑆𝑆 = 𝜙𝜙𝑔𝑔𝑔𝑔𝑔𝑔 the settling is certainly
not zero as consolidation occurs: the particles (flocs) are squeezed at the bottom of
the column, which makes the water/bed interface move down. In this case, one
usually speaks of settlement (a term coming from soil science) as all the particles are

134 Toorman, E.A. (1996) “Sedimentation and self-weight consolidation: general unifying
theory”, Géotechnique 46, No 1, 103-113.
135 In Chapter 8, the exponent, following standard notations, was defined as n. To avoid

confusions with the variable n = 2/(3 − D) defined in the present chapter, we have chosen
to rename the exponent m.

295
 Introduction to Colloid Science

touching and form a fabric moving downwards under the action of gravity instead of
settling which usually refers to individual particles.

The first illustration at the beginning of the chapter represents a column filled with
real flocs that are settling and depositing at the bottom of the column, where they
are squeezed. In the illustration we have assumed that the flocs are fractal, and to
represent the squeezing at the bottom, we have drawn flocs of smaller dimension
as this amounts to increase the local volume fraction (see the fractal approach
detailed above). Instead of assuming that the flocs are mathematical space-filling
objects, we will in this subsection assume that the particles are real fractal flocs of
radius 𝑅𝑅𝑁𝑁 and that the clay concentration (volume fraction) 𝜙𝜙𝑠𝑠 is below the gelling
concentration, i.e. that the flocs are not touching. (There will therefore be a jump in
velocity at the suspension/bed interface.)

In order to evaluate the settling velocity of a fractal floc, we first need to define some
specific volume fractions:

Definitions regarding suspensions of fractal flocs

The suspended particles, made of primary particles of radius 𝑅𝑅𝑝𝑝 , are fractal flocs of
radius 𝑅𝑅𝑁𝑁 with a fractal dimension D. The flocs are not necessarily assumed to be
space-filling as was assumed in the first part of the chapter. Their fractal dimension
is therefore in general different from the space-filling case.

We define:

volume of flocs volume of clay

𝜙𝜙𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 = ×
volume of clay total volume

If we assume that all the clay is contained in the flocs:

volume of a floc
𝜙𝜙𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 = × 𝜙𝜙𝑠𝑠
volume of clay in a floc

We recall that

296
 Chapter 10 Modelling the consolidation of slurries

𝑅𝑅𝑝𝑝 3−𝐷𝐷
𝜙𝜙clay in a floc = � �
𝑅𝑅𝑁𝑁

which leads to
3−𝐷𝐷
𝑅𝑅𝑁𝑁
𝜙𝜙𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 =� � × 𝜙𝜙𝑠𝑠
𝑅𝑅𝑝𝑝

If the flocs are space-filling 𝜙𝜙𝑠𝑠 = 𝜙𝜙clay in a floc and one gets 𝜙𝜙𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 = 1. This implies
that we can define a gelling concentration (= concentration of clay) for which the
flocs are space-filling, which is given by

𝑅𝑅𝑝𝑝 3−𝐷𝐷
𝜙𝜙𝑔𝑔𝑔𝑔𝑔𝑔 = 𝜙𝜙clay in a floc = � �
𝑅𝑅𝑁𝑁

When the flocs are space-filling, it is clear that the fractal dimension is the same as
the one discussed in the first part of this chapter.

We can now evaluate the Stokes settling velocity of a fractal sphere:

2�𝜌𝜌𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 − 𝜌𝜌𝑤𝑤 �𝑅𝑅𝑁𝑁 2 𝑔𝑔

𝑣𝑣𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆(𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓) =
9𝜂𝜂

The density of a fractal floc can be calculated as follows:

mass clay and mass water in a floc

𝜌𝜌𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 =
volume of a floc
volume clay in a floc volume clay in a floc
𝜌𝜌𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 = 𝜌𝜌𝑠𝑠 + 𝜌𝜌𝑤𝑤 �1 − �
volume of a floc volume of a floc
Using the definition of 𝜙𝜙𝑔𝑔𝑔𝑔𝑙𝑙 introduced above, we find that

𝜌𝜌𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 − 𝜌𝜌𝑤𝑤 = (𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝜙𝜙𝑔𝑔𝑔𝑔𝑔𝑔

We get

2(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑅𝑅𝑁𝑁 2 𝑔𝑔

𝑣𝑣𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆(𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓) = 𝜙𝜙𝑔𝑔𝑔𝑔𝑔𝑔
9𝜂𝜂

This relation for the Stokes settling of a floc has been found by Kranenburg 136.

Kranenburg, C. (1994). The fractal structure of cohesive sediment aggregates. Estuarine,

136

Coastal and Shelf Science, 39(5), 451-460.

297
 Introduction to Colloid Science

In order to find a hindered settling velocity of a floc, 𝑣𝑣ℎ𝑖𝑖𝑖𝑖𝑖𝑖 , we can adapt the
Richardson-Zaki formula:
𝑚𝑚
𝑣𝑣ℎ𝑖𝑖𝑖𝑖𝑖𝑖 = �1 − 𝜙𝜙𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 � 𝑣𝑣𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆(𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓)

where 𝜙𝜙𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 is the volume fraction of flocs in the hindered settling zone. In that zone,
we have everywhere
3−𝐷𝐷
𝑅𝑅𝑁𝑁
𝜙𝜙𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 = 𝜙𝜙𝑠𝑠 /𝜙𝜙𝑔𝑔𝑔𝑔𝑔𝑔 = 𝜙𝜙𝑠𝑠 � �
𝑅𝑅𝑝𝑝

One can verify that for D = 3 (the particle is a solid particle) one gets 𝜙𝜙𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 = 𝜙𝜙𝑠𝑠 . The
adapted Richardson-Zaki expression then reduces to the standard Richardson-Zaki
expression. In the limit of space-filling flocs, 𝜙𝜙𝑠𝑠 = 𝜙𝜙𝑔𝑔𝑔𝑔𝑔𝑔 and the adapted Richardson-
Zaki formula gives the result 𝑣𝑣ℎ𝑖𝑖𝑖𝑖𝑖𝑖 = 0. There is no settling velocity as the
Richardson-Zaki formulation does not account for compression.

At volume fractions such that 𝜙𝜙𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 < 1, one can make the approximation
𝐷𝐷−3
𝑅𝑅𝑁𝑁 2(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑅𝑅𝑁𝑁 2 𝑔𝑔
𝑣𝑣ℎ𝑖𝑖𝑖𝑖𝑖𝑖 ~ �� � − 𝑚𝑚𝜙𝜙0 �
𝑅𝑅𝑝𝑝 9𝜂𝜂

This implies that the function 𝑣𝑣ℎ𝑖𝑖𝑖𝑖𝑖𝑖 as function of 𝜙𝜙0 is a straight line of slope

2(𝜌𝜌𝑠𝑠 − 𝜌𝜌𝑤𝑤 )𝑅𝑅𝑁𝑁 2 𝑔𝑔

𝑚𝑚
9𝜂𝜂

from which 𝑚𝑚 can theoretically be deduced if the average size of flocs are known.

298
 Chapter 10 Modelling the consolidation of slurries

Beyond self-weight consolidation

The Gibson equation, which is the topic of Chapter 9 and the present chapter was
originally created to study the large strains occurring during the self-weight
consolidation of slurries. In Chapter 9, we have already stated that the stresses felt
by particles during self-weight consolidation are very small compared to the
pressures applied in an oedometer test.

The oedometer test is used to investigate the 1D consolidation of fine-grained soils.

From the oedometer test one gets the relation between vertical effective stress (the
load applied) and vertical strain (the deformation), from which the void ratio of the
sample can be determined.

The sample is placed between two porous disks at the top and bottom. The disks are
in contact with the same bath of water. This implies that at the end of the test, the
water pressure will be hydrostatic through the sample. A compressive stress is
applied from the top, by a vertical load, which is assumed to act uniformly over the
area of the soil sample:

Initially, all the vertical load is taken by pore water, because, due to the low
permeability of the soil, the pore water is unable to flow out of the voids quickly.
Therefore, there is very little compression of the soil sample immediately after
placing the load. After a few seconds, the pore water begins to flow out. This results
in a decrease in pore water pressure. At the same time, the effective stress increases.
As a result, the sample settles. Several increments of vertical stress are applied in an
oedometer test, usually by doubling the previous increment. For each load, the final
settlement of the soil sample and the time taken to reach this final settlement are
recorded.

When analysing these types of experiments, several hypothesis are made:

1 – the sample does not undergo self-weight consolidation; the settlement is solely
due to the applied stress.

2 – the volume fraction is constant over the height of the sample.

299
 Introduction to Colloid Science

3 – The particles are incompressible

4 – the water is incompressible

Using hypothesis 1 (and hypothesis 3 and 4 that have been used implicitly until now,
also in Chapters 9 and 10), the full Gibson equation reduces to

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 𝐾𝐾 𝜕𝜕σ𝑠𝑠𝑠𝑠
= � 𝜙𝜙𝑆𝑆 �
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝑔𝑔𝜌𝜌𝑤𝑤 𝜕𝜕𝜕𝜕

where, to remove the effect of self-weight consolidation we have taken 𝜌𝜌𝑠𝑠 = 𝜌𝜌𝑤𝑤
(which mathematically amounts to the same). This equation can further be
developed into

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 𝐾𝐾 𝜕𝜕σ𝑠𝑠𝑠𝑠 𝜕𝜕𝜙𝜙𝑆𝑆

= � 𝜙𝜙𝑆𝑆 �
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝑔𝑔𝜌𝜌𝑤𝑤 𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕𝜕𝜕

The so-called consolidation coefficient is defined by

𝐾𝐾 𝜕𝜕σ𝑠𝑠𝑠𝑠
𝐶𝐶𝑣𝑣 = 𝜙𝜙
𝑔𝑔𝜌𝜌𝑤𝑤 𝑆𝑆 𝜕𝜕𝜙𝜙𝑆𝑆

(the curl d’s are here equivalent to the straight d’ as σ𝑠𝑠𝑠𝑠 only depends on 𝜙𝜙𝑆𝑆 ). Using
the fractal approach given in the present chapter, one finds that

𝐾𝐾𝑘𝑘 𝐾𝐾𝜎𝜎
𝐶𝐶𝑣𝑣 = 𝑛𝑛
𝑔𝑔𝜌𝜌𝑤𝑤

which is a constant. The fact it is a constant depends on the peculiar expressions that
were chosen for 𝐾𝐾(~ 𝜙𝜙𝑆𝑆 −𝑛𝑛 ) and σ𝑠𝑠𝑠𝑠 (~ 𝜙𝜙𝑆𝑆 𝑛𝑛 ) . In soil science, it is generally assumed
that the consolidation coefficient is constant.

With 𝐶𝐶𝑣𝑣 constant, the equation reduces to

𝜕𝜕𝜙𝜙𝑆𝑆 𝜕𝜕 2 𝜙𝜙𝑆𝑆
= 𝐶𝐶𝑣𝑣
𝜕𝜕𝜕𝜕 𝜕𝜕𝑧𝑧 2
This equation is mathematically equivalent to the one called Fick’s second law in
Chapter 2 and can be solved analytically.

The hypothesis used to derive this equation were (1,3,4). We did not assume
hypothesis 2, which implies that this equation is valid for samples exhibiting a change
in volume fraction over height. This equation is useless in the case we assume
hypothesis 2, as it then reduces to

300
 Chapter 10 Modelling the consolidation of slurries

𝜕𝜕𝜙𝜙𝑆𝑆
=0
𝜕𝜕𝜕𝜕
This equation tells us that there is no local (at a position z) variation of 𝜙𝜙𝑆𝑆 in time,
which is compatible with hypothesis 2 : there is only a variation in time of the total
volume fraction of the sample between before and after the loading (but no
variation within the sample).

Usually, in an oedometer test, the vertical strain is recorded as function of the

vertical effective (skeleton) stress. The strain is defined by

ℎ − ℎ0
𝑑𝑑𝑑𝑑 =
ℎ0

where ℎ is the height of the sample after the test and ℎ0 the height before. We used
the symbol 𝑑𝑑𝑑𝑑 to indicate that the change is height is assumed to be quite small. As
the sample does not deform laterally, we also have

𝑉𝑉 − 𝑉𝑉0 𝑑𝑑𝑑𝑑
𝑑𝑑𝑑𝑑 = =
𝑉𝑉0 𝑉𝑉0

where the volumes are related to the heights by 𝑉𝑉 = 𝐴𝐴ℎ and 𝑉𝑉0 = 𝐴𝐴ℎ0 with 𝐴𝐴 being
the cross-sectional area of the sample. By definition

𝑉𝑉 = 𝑉𝑉𝑤𝑤 + 𝑉𝑉𝑠𝑠

where 𝑉𝑉𝑤𝑤 is the volume of water and 𝑉𝑉𝑠𝑠 the volume of solids.

The volume 𝑉𝑉 is changing because water is flowing out of the sample. Therefore

𝑑𝑑𝑉𝑉𝑤𝑤
𝑑𝑑𝑑𝑑 =
𝑉𝑉0

and the volume 𝑉𝑉𝑠𝑠 of solids is not changing. From the definition of volume fraction,
we have

𝑉𝑉𝑠𝑠 𝑉𝑉𝑠𝑠 𝑉𝑉𝑠𝑠 𝑉𝑉 − 𝑉𝑉0

𝑑𝑑𝜙𝜙𝑆𝑆 = − =− = −𝜙𝜙𝑆𝑆,end 𝑑𝑑𝑑𝑑
𝑉𝑉 𝑉𝑉0 𝑉𝑉 𝑉𝑉0

where 𝜙𝜙𝑆𝑆,end is the volume fraction after loading. As the incremental change 𝑑𝑑𝑑𝑑 is
very small, we have

ℎ = ℎ0 (1 + 𝑑𝑑𝑑𝑑) ~ ℎ0

and therefore

𝑉𝑉 ~ 𝑉𝑉0

301
 Introduction to Colloid Science

𝜙𝜙𝑆𝑆,end ~ 𝜙𝜙𝑆𝑆,begin ~ 𝜙𝜙𝑆𝑆

where 𝜙𝜙𝑆𝑆,begin is the volume fraction before loading. One has to realize that by
defining the strain, we have implicitly defined it as a macroscopic quantity, a
property affecting the whole sample. Therefore the volume fraction 𝜙𝜙𝑆𝑆 specified
here is representing the volume fraction of the sample (not a slice of the sample).
This implies that we fulfil hypothesis 2 (we assume that 𝜙𝜙𝑆𝑆 is constant over the
height). We get

𝑑𝑑𝜙𝜙𝑆𝑆 = −𝜙𝜙𝑆𝑆 𝑑𝑑𝑑𝑑

𝑑𝑑𝜙𝜙𝑆𝑆 𝑑𝑑𝑑𝑑
= −𝜙𝜙𝑆𝑆
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑
This last equation indicates that there is a (small) change in the total volume fraction
of the sample between before and after applying the loading. Note that we have
here used the Lagrangian derivatives, as 𝜙𝜙𝑆𝑆 and 𝜀𝜀 only depend on time only. Using
the Lagrangian form of Gibson equation, which we simplify using 𝜌𝜌𝑠𝑠 = 𝜌𝜌𝑤𝑤 for the
reason given above, we get

𝑑𝑑𝜙𝜙𝑆𝑆 𝑑𝑑𝑑𝑑 𝜕𝜕 𝐾𝐾 𝜕𝜕σ𝑠𝑠𝑠𝑠

= −𝜙𝜙𝑆𝑆 = 𝜙𝜙𝑆𝑆 � �
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝑔𝑔𝜌𝜌𝑤𝑤 𝜕𝜕𝜕𝜕

𝑑𝑑𝑑𝑑 𝜕𝜕 𝐾𝐾 𝜕𝜕σ𝑠𝑠𝑠𝑠
= � �
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝑔𝑔𝜌𝜌𝑤𝑤 𝜕𝜕𝜕𝜕

This equation links the change in strain to the change in stress. We also get

𝑑𝑑𝜙𝜙𝑆𝑆 𝑑𝑑σ𝑠𝑠𝑠𝑠 𝑑𝑑𝜙𝜙𝑆𝑆 𝜕𝜕 𝐾𝐾 𝜕𝜕σ𝑠𝑠𝑠𝑠

= = 𝜙𝜙𝑆𝑆 � �
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 𝑑𝑑σ𝑠𝑠𝑠𝑠 𝜕𝜕𝜕𝜕 𝑔𝑔𝜌𝜌𝑤𝑤 𝜕𝜕𝜕𝜕

𝑑𝑑σ𝑠𝑠𝑠𝑠 𝑑𝑑σ𝑠𝑠𝑠𝑠 𝜕𝜕 𝐾𝐾 𝜕𝜕σ𝑠𝑠𝑠𝑠

= 𝜙𝜙𝑆𝑆 � �
𝑑𝑑𝑑𝑑 𝑑𝑑𝜙𝜙𝑆𝑆 𝜕𝜕𝜕𝜕 𝑔𝑔𝜌𝜌𝑤𝑤 𝜕𝜕𝜕𝜕

In order to define the consolidation coefficient, it is now necessary to assume that

the permeability 𝐾𝐾 is constant, so it can get out of the bracket term. This condition
is fulfilled if we assume that 𝜙𝜙𝑆𝑆 is constant (our hypothesis 2). We then get

𝑑𝑑σ𝑠𝑠𝑠𝑠 𝜕𝜕 2 σ𝑠𝑠𝑠𝑠
= 𝐶𝐶𝑣𝑣
𝑑𝑑𝑑𝑑 𝜕𝜕𝑧𝑧 2
This equation enables to find the change in stress, from which the change in strain
can be obtained. The curl d’s indicate that the derivative in 𝑧𝑧 should be taken at 𝑡𝑡
constant. We see that even if 𝜙𝜙𝑆𝑆 is constant over the whole sample there is a change
in σ𝑠𝑠𝑠𝑠 over height, which originates from the fact that we apply a load at the top.

302
 Chapter 10 Modelling the consolidation of slurries

This implies that the stress between particles is higher at the bottom of the sample,
but that this does not influence their local (= at given 𝑧𝑧) compaction.

We also have

𝑑𝑑σ𝑠𝑠𝑠𝑠 𝜕𝜕σ𝑠𝑠𝑠𝑠 𝑑𝑑z 𝜕𝜕σ𝑠𝑠𝑠𝑠

= +
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕 𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕
where 𝑑𝑑z/dt is the velocity of the settlement. If that settlement is slow, we have in
good approximation

𝑑𝑑σ𝑠𝑠𝑠𝑠 𝜕𝜕σ𝑠𝑠𝑠𝑠
=
𝑑𝑑𝑑𝑑 𝜕𝜕𝜕𝜕
The change in effective stress over time is then the same at any 𝑧𝑧.

The load applied to the sample in a test is constant. From the definitions of stresses
(see Chapter 9), we get

σ𝑧𝑧𝑧𝑧 = σ𝑠𝑠𝑠𝑠 + 𝑃𝑃𝑒𝑒 + 𝑃𝑃ℎ𝑦𝑦𝑦𝑦

As inertia is neglected, one can say that the total stress σ𝑧𝑧𝑧𝑧 of the sample is equal to
the stress from the load. As neither σ𝑧𝑧𝑧𝑧 (which is constant as the load is constant)
nor 𝑃𝑃ℎ𝑦𝑦𝑦𝑦 depend on time, and that 𝑃𝑃ℎ𝑦𝑦𝑦𝑦 is linear in 𝑧𝑧 we get

𝑑𝑑σ𝑠𝑠𝑠𝑠 𝑑𝑑𝑃𝑃𝑒𝑒
=−
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑
𝜕𝜕 2 σ𝑠𝑠𝑠𝑠 𝜕𝜕 2 𝑃𝑃𝑒𝑒
2
=− 2
𝜕𝜕𝑧𝑧 𝜕𝜕𝑧𝑧
which gives

𝑑𝑑𝑃𝑃𝑒𝑒 𝜕𝜕 2 𝑃𝑃𝑒𝑒
= 𝐶𝐶𝑣𝑣
𝑑𝑑𝑑𝑑 𝜕𝜕𝑧𝑧 2
This equation is only fulfilled when all hypothesis (1-4) are obeyed.

Numerical solution

The differential equation

𝑑𝑑𝑃𝑃𝑒𝑒 𝜕𝜕 2 𝑃𝑃𝑒𝑒
= 𝐶𝐶𝑣𝑣
𝑑𝑑𝑑𝑑 𝜕𝜕𝑧𝑧 2

303
 Introduction to Colloid Science

can be solved numerically using the so-called finite differences method 137. Originally
it was solved analytically by Terzaghi 138 in 1923. For numerical stability it is always
better to convert the equation to a dimensionless one:

𝑑𝑑𝑃𝑃𝑒𝑒 ∗ 2 ∗
∗ 𝜕𝜕 𝑃𝑃𝑒𝑒
= 𝐶𝐶𝑣𝑣
𝑑𝑑𝑡𝑡 ∗ 𝜕𝜕𝑧𝑧 ∗ 2
with

𝑡𝑡 ∗ = 𝑡𝑡/𝜏𝜏

𝑧𝑧 ∗ = 𝑧𝑧/𝐻𝐻

𝐶𝐶𝑣𝑣
𝐶𝐶𝑣𝑣 ∗ =
𝐻𝐻2 /𝜏𝜏

𝑃𝑃𝑒𝑒 ∗ = 𝑃𝑃𝑒𝑒 /𝑃𝑃0

where 𝜏𝜏 and 𝐻𝐻 are characteristic time and height and 𝑃𝑃0 a reference pressure.
Writing the equation in dimensionless form also enables to find that the final
deformation has been reached when

𝑑𝑑𝑃𝑃𝑒𝑒 ∗
~0
𝑑𝑑𝑡𝑡 ∗
which occurs when

𝐶𝐶𝑣𝑣
≪ 1
𝐻𝐻 2 /𝜏𝜏

In other words, the consolidation time 𝜏𝜏 will take 4 times as long if the layer (𝐻𝐻) is
twice as thick.

For ease of notation, we drop the *, but in the following we only refer to the
dimensionless equation.

The equation is discretized as follows:

𝑑𝑑𝑃𝑃𝑒𝑒 𝑃𝑃𝑒𝑒 (𝑡𝑡 + 𝑑𝑑𝑑𝑑) − 𝑃𝑃𝑒𝑒 (𝑡𝑡)

=
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑

137 Verruijt, Arnold. An Introduction to Soil Mechanics. Springer, Theory and Applications of
Transport in Porous Media, Vol.30 (2018)
138 Karl von Terzaghi (1883 - 1963) was an Austrian geotechnical engineer known as “the father

of soil mechanics”

304
 Chapter 10 Modelling the consolidation of slurries

𝜕𝜕 2 𝑃𝑃𝑒𝑒 𝑃𝑃𝑒𝑒 (𝑧𝑧 + 𝑑𝑑𝑑𝑑, 𝑡𝑡) − 2𝑃𝑃𝑒𝑒 (𝑧𝑧, 𝑡𝑡) + 𝑃𝑃𝑒𝑒 (𝑧𝑧 − 𝑑𝑑𝑑𝑑, 𝑡𝑡)
=
𝜕𝜕𝑧𝑧 2 (𝑑𝑑𝑑𝑑)2

This leads to

𝑑𝑑𝑑𝑑
𝑃𝑃𝑒𝑒 (𝑧𝑧, 𝑡𝑡 + 𝑑𝑑𝑑𝑑) = 𝑃𝑃𝑒𝑒 (𝑧𝑧, 𝑡𝑡) + 𝐶𝐶𝑣𝑣 [𝑃𝑃 (𝑧𝑧 + 𝑑𝑑𝑑𝑑, 𝑡𝑡) − 2𝑃𝑃𝑒𝑒 (𝑧𝑧, 𝑡𝑡) + 𝑃𝑃𝑒𝑒 (𝑧𝑧 − 𝑑𝑑𝑑𝑑, 𝑡𝑡)]
(𝑑𝑑𝑑𝑑)2 𝑒𝑒

In an oedometer test the sample is usually drained at the top, using a thin sheet of
filter paper and a porous stone. In the container in which the sample and its
surrounding ring are placed, the water level is kept constant. This implies that at the
top of the sample the excess pore pressure is zero, which results in the boundary
conditions:

𝑃𝑃𝑒𝑒 (𝑧𝑧 = 1, 𝑡𝑡) = 0

where 𝑧𝑧 = 1 represents the top of the sample.

The sample is usually undrained at the bottom where an impermeable plate

prevents the water flow. The boundary condition at the bottom of the sample is
therefore

𝜕𝜕𝑃𝑃𝑒𝑒 (𝑧𝑧 = 0, 𝑡𝑡)

=0
𝜕𝜕𝜕𝜕
where the curl d represents the fact that the derivative is taken at a given time: for
all times, the boundary condition is respected.

Numerically, this last boundary condition is usually implemented by defining a

(fictive) pressure 𝑃𝑃𝑒𝑒 (−𝑑𝑑𝑑𝑑, 𝑡𝑡) located at a position 𝑑𝑑𝑑𝑑 below the sample and imposing
the condition

𝑃𝑃𝑒𝑒 (−𝑑𝑑𝑑𝑑, 𝑡𝑡) = 𝑃𝑃𝑒𝑒 (+𝑑𝑑𝑑𝑑, 𝑡𝑡)

This condition ensures that whatever the value of 𝑃𝑃𝑒𝑒 (0, 𝑡𝑡) the boundary condition
𝜕𝜕𝑃𝑃𝑒𝑒 (𝑧𝑧 = 0, 𝑡𝑡)/𝜕𝜕𝜕𝜕 = 0 is satisfied. Numerically, this means that the value of 𝑃𝑃𝑒𝑒 (0, 𝑡𝑡)
will be determined by

𝑑𝑑𝑑𝑑
𝑃𝑃𝑒𝑒 (0, 𝑡𝑡 + 𝑑𝑑𝑑𝑑) = 𝑃𝑃𝑒𝑒 (0, 𝑡𝑡) + 𝐶𝐶𝑣𝑣 [𝑃𝑃 (𝑑𝑑𝑑𝑑, 𝑡𝑡) − 2𝑃𝑃𝑒𝑒 (0, 𝑡𝑡) + 𝑃𝑃𝑒𝑒 (𝑑𝑑𝑑𝑑, 𝑡𝑡)]
(𝑑𝑑𝑑𝑑)2 𝑒𝑒

At 𝑡𝑡 = 0 we also assume that the pressure within the sample is known:

𝑃𝑃𝑒𝑒 (𝑧𝑧 < 1, 𝑡𝑡 = 0) = 1

In order to get a stable numerical solution it is important to ensure that

305
 Introduction to Colloid Science

(𝑑𝑑𝑑𝑑)2
𝑑𝑑𝑑𝑑 ≪
𝐶𝐶𝑣𝑣

Using 𝐶𝐶𝑣𝑣 = 0.1 and 𝑑𝑑𝑑𝑑 = 0.1 , this implies that 𝑑𝑑𝑑𝑑 ≪ 0.1. In the Matlab code given
underneath we used dt = 0.01.

The dissipation of pore water pressure over time (in dimensionless units) is given by:

From the definition of the consolidation coefficient and the strain, we get

−𝐾𝐾 𝑑𝑑σ𝑠𝑠𝑠𝑠
𝐶𝐶𝑣𝑣 =
𝑔𝑔𝜌𝜌𝑤𝑤 𝑑𝑑𝑑𝑑

The compressibility 𝑚𝑚𝑣𝑣 is defined as

𝐾𝐾
𝑚𝑚𝑣𝑣 =
𝑔𝑔𝜌𝜌𝑤𝑤 𝐶𝐶𝑣𝑣

and this gives

𝑑𝑑𝑑𝑑 = −𝑚𝑚𝑣𝑣 𝑑𝑑σ𝑠𝑠𝑠𝑠

The minus sign indicate the fact that an increase in effective stress is coupled with a
decrease in strain (the sample is compressed). As we fulfil hypothesis 1- 4, 𝑚𝑚𝑣𝑣 is a
constant.

As discussed above, the strain (and the volume fraction) is defined as a property
applied to the whole sample. We have found that the effective stress (as the excess
pore water pressure) is changing over the height. The equation we have derived for
the strain should therefore be written

306
 Chapter 10 Modelling the consolidation of slurries

−𝑚𝑚𝑣𝑣 ℎ
𝑑𝑑𝑑𝑑 = � dσ𝑠𝑠𝑠𝑠 𝑑𝑑𝑑𝑑
ℎ 0

which indicates that the strain is a function of the average skeleton stress. We get
from the definition of the strain:

ℎ(𝑡𝑡 + 𝑑𝑑𝑑𝑑) − ℎ(𝑡𝑡) −𝑚𝑚𝑣𝑣 ℎ(𝑡𝑡)

= � σ𝑠𝑠𝑠𝑠 𝑑𝑑𝑑𝑑
ℎ(𝑡𝑡) ℎ(𝑡𝑡) 0
ℎ(𝑡𝑡)
𝑑𝑑ℎ = ℎ(𝑡𝑡 + 𝑑𝑑𝑑𝑑) − ℎ(𝑡𝑡) = −𝑚𝑚𝑣𝑣 � σ𝑠𝑠𝑠𝑠 (𝑧𝑧, 𝑡𝑡)𝑑𝑑𝑑𝑑
0

We define

𝑚𝑚𝑣𝑣 ∗ = 𝑚𝑚𝑣𝑣 𝑃𝑃0

σ𝑠𝑠𝑠𝑠 ∗ = σ𝑠𝑠𝑠𝑠 /𝑃𝑃0

where 𝑃𝑃0 is, as above, a reference pressure. We get

ℎ(𝑡𝑡) ℎ(𝑡𝑡)
𝑑𝑑ℎ = −𝑚𝑚𝑣𝑣 ∗ �� �σ𝑧𝑧𝑧𝑧 ∗ − 𝑃𝑃ℎ𝑦𝑦𝑦𝑦 ∗ �𝑑𝑑𝑑𝑑 − � 𝑃𝑃𝑒𝑒 ∗ 𝑑𝑑𝑑𝑑�
0 0

The first integral on the right-hand side is a constant as the load and average
hydrostatic pressure are both constant in time. At the end of consolidation the
excess pore water pressure is zero, as demonstrated above and therefore we define
ℎ(𝑡𝑡=∞)
𝑑𝑑ℎ∞ = −𝑚𝑚𝑣𝑣 ∗ � �σ𝑧𝑧𝑧𝑧 ∗ − 𝑃𝑃ℎ𝑦𝑦𝑦𝑦 ∗ �𝑑𝑑𝑑𝑑
0

We get
ℎ(𝑡𝑡)
𝑑𝑑ℎ − 𝑑𝑑ℎ∞ = −𝑚𝑚𝑣𝑣 ∗ � 𝑃𝑃𝑒𝑒 ∗ 𝑑𝑑𝑑𝑑
0

We also have, with 𝑑𝑑ℎ0 = ℎ(𝑑𝑑𝑑𝑑) − ℎ(0) = ℎ(𝑑𝑑𝑑𝑑) − ℎ0

𝑑𝑑ℎ0 − 𝑑𝑑ℎ∞ = −𝑚𝑚𝑣𝑣 ∗ ℎ0

as by definition 𝑃𝑃𝑒𝑒 ∗ (𝑡𝑡 = 0) = 1, which implies that

𝑑𝑑ℎ(𝑡𝑡) − 𝑑𝑑ℎ∞ 1 ℎ(𝑡𝑡) ∗ ℎ(𝑡𝑡)/ℎ0

= � 𝑃𝑃𝑒𝑒 𝑑𝑑𝑑𝑑 = � 𝑃𝑃𝑒𝑒 ∗ 𝑑𝑑𝑑𝑑
𝑑𝑑ℎ0 − 𝑑𝑑ℎ∞ ℎ0 0 0

307
 Introduction to Colloid Science

The degree of consolidation 𝑈𝑈 is defined as

𝑑𝑑ℎ0 − 𝑑𝑑ℎ(𝑡𝑡) 𝑑𝑑ℎ(𝑡𝑡) − 𝑑𝑑ℎ∞

𝑈𝑈 = =1−
𝑑𝑑ℎ0 − 𝑑𝑑ℎ∞ 𝑑𝑑ℎ0 − 𝑑𝑑ℎ∞

yielding
ℎ(𝑡𝑡)⁄ℎ0
𝑈𝑈 = 1 − � 𝑃𝑃𝑒𝑒 ∗ 𝑑𝑑𝑑𝑑
0

The degree of consolidation U as function of time is given by:

308
 Chapter 10 Modelling the consolidation of slurries

The Matlab code is given by:

clear all
close all

nh=10; % number of steps in z direction

zbin(1) = 0.0d0; % bottom of the sample
dz = 1/(nh-1); % dimensionless increment in z direction
%zbin(1) = 0 ; zbin(2) = dz ; … ; zbin(nh) = (nh-1)*dz = 1

for m = 2:nh
zbin(m) = zbin(m-1)+dz;
Pe(m) = 1; % dimensionless excess pore water pressure
end

Pe(1)=1;
Pe(nh)=0;

Cv = 0.1; % dimensionless consolidation coefficient

tnum(1) = 0.01;
Pestart = Pe; % Pestart is Pe(t=0)

for kt = 2:5000
dt = 0.01;
tnum(kt) = tnum(kt-1)+dt;
tnumplot(kt) = tnum(kt);
a = Cv*dt/dz.^2;
for m = 2:nh-1
Penew(m) = Pe(m)+a*(Pe(m+1)-2*Pe(m)+Pe(m-1));
% Penew(m) is Pe(t+dt)
% Pe(m) is Pe(t)
end
Penew(nh) = Pe(nh); % Pe(H) = 0
Penew(1) = Pe(1)+a*(Pe(2)-2*Pe(1)+Pe(2));
Pe = Penew;

height(kt) = trapz(zbin,Pe);

if kt == 10 % t=0.1
Pe1 = Pe;
end

if kt == 100 % t=1
Pe2 = Pe;
end

if kt == 500 % t=5
Pe3 = Pe;
end

if kt == 1000 % t=10
Pe4 = Pe;
end

if kt == 5000 % t= 50
Pe5 = Pe;
end

309
 Introduction to Colloid Science

end

plot(Pestart,zbin,'-ob','LineWidth',2)
hold on
plot(Pe1,zbin,'-ok','LineWidth',2)
hold on
plot(Pe2,zbin,'-oc','LineWidth',2)
hold on
plot(Pe3,zbin,'-og','LineWidth',2)
hold on
plot(Pe4,zbin,'-om','LineWidth',2)
hold on
plot(Pe5,zbin,'-or','LineWidth',2)
hold on

legend('t/\tau = 0','t/\tau = 0.1','t/\tau = 1','t/\tau = 5','t/\tau =

10','t/\tau = 50');
ylim([0 1])
xlabel('P_e/P_0','FontSize',12);
ylabel('z/H','FontSize',12);
set(gca,'LineWidth',2,'FontSize',12);
set(gcf,'Color',[1 1 1])

figure

semilogx(tnumplot,1-height,'-or','LineWidth',2)
hold on

xlabel('t/\tau','FontSize',12);
ylabel('degree of consolidation','FontSize',12);
set(gca,'LineWidth',2,'FontSize',12);
set(gcf,'Color',[1 1 1])

310
 Conclusion
from colloid science
to large-scale applications
 Introduction to Colloid Science

As stated in the introduction, mud and clayey systems are studied in different fields
of research. We mentioned soil science and sediment dynamics, but these generic
names are associated to specific expertise such as agronomy, chemistry, geology,
physical geography, biology, civil engineering… In this book, we focussed primarily
on colloid science.

Length scales and associated particle sizes and field of science in the context of sediment
characterization. The term “bulk” refers to the fact that above a given length scale no
distinction is made between the carrying fluid and the sediment: water and sediment is then
seen as a continuum, with bulk properties (density, viscosity, …)

Despite the promise of the subtitle of the book (Applications to sediment

characterization), which hints to the study of natural muddy suspensions, the reader
will have noticed that many findings described in the book apply to well-
characterized systems only, i.e. particles with a defined size, shape and surface
charge. Even though colloid scientists are working on more complex systems, for
instance mixtures or anisotropically shaped particles (see articles cited and books in
the reference list), it is clear that fundamental research is at present not able to
bridge completely the knowledge between well-characterized and natural systems.
Even if some models, like the flocculation models presented in Chapter 6 can – in
some cases – be applied quite successfully to the flocculation of natural systems,
these models suffer from a severe restriction: their validity is limited to a scale that
is far below the scale usually desired from an engineering point of view.

The primary reason elaborated flocculation models cannot be applied in large-scale

transport models is due to the fact that, when coupled to a large-scale hydrodynamic
model, the resulting sediment transport model would be too CPU intensive to run.
It would be impossible to make long-term predictions with it. It is therefore
necessary to use less elaborated models 139,140,141, that will be able to reproduce the

139 Verney, Romaric, et al. "Behaviour of a floc population during a tidal cycle: laboratory
experiments and numerical modelling." Continental Shelf Research 31.10 (2011): S64-S83.
140 Lee, Byung Joon, et al. "A two-class population balance equation yielding bimodal

flocculation of marine or estuarine sediments." Water research 45.5 (2011): 2131-2145.

141 Manning, Andrew J., et al. "Flocculation settling characteristics of mud: sand

mixtures." Ocean dynamics 60.2 (2010): 237-253.

312
 Conclusion

in-situ measured data (mean floc size, settling velocities) to a reasonable accuracy,
with a limited number of adjustable parameters.

Doing so, one is forced to leave the bottom-up approach usually adopted in
fundamental (colloid) science and presented in this book in order to use a top-down
approach. The bottom-up approach consists in starting with simplified models and
add layers of complexity, whereas the top-down approach consists in taking a
system as a whole and breaking it down in sub-systems in a reverse engineering
fashion. Both approaches are not mutually exclusive, and there is often much to gain
to combine them. A lot of work is currently devoted to bridge the gap between
colloid science and engineering.

Illustration of the impact of a microscopic process (flocculation) on large-scale (km) sediment

transport. The area represents the mouth of the Western Scheldt estuary and the colour
indicates a three-month average suspended sediment concentration (SSC) with and without
taking flocculation into account. The differences in SSC are particularly obvious close to the
mouth. The flocculation model is taken from Manning et al. 142

142
Manning, A. J., and K. R. Dyer. "Mass settling flux of fine sediments in Northern European
estuaries: measurements and predictions." Marine Geology 245.1-4 (2007): 107-122.

313
References for colloid science books
Kruyt, H. R. editor. (1952)
Colloid Science. Vol. 1. Irreversible Systems. Elsevier.
A classic reference, with large contributions of J. Th. Overbeek

Russel, W. B., Saville, D. A., & Schowalter, W. R. (1989)

Colloidal dispersions. Cambridge university press.
A complete reference, especially suited for the theoretical modelling of suspensions.

Van de Ven, T. G. (1989)

Colloidal hydrodynamics. Academic Press.
A great reference to understand the principles of colloidal hydrodynamics from first
principles.

Hunter, R. J. (1993)
Introduction to modern colloid science. Oxford University Press.
A classic and easy to read reference to understand the fundamentals of colloid
science, with limited derivations.

Hunter, R. J. (2000)
Foundations of Colloid Science. Oxford University Press.
A completely revised and updated edition of a classic textbook on colloid science

Shaw, D. J., & Costello, B. (1993)

Introduction to colloid and surface chemistry. Butterworth-Heinemann, Oxford.
A concise overall coverage of colloid and surface chemistry, well suited for getting a
broad background on the topic.

Dhont, J. KG. (1996)

An introduction to dynamics of colloids. Vol. 2. Elsevier.
A great reference for detailed mathematical derivations and interpretation thereof

Hiemenz, P. C., & Rajagopalan, R. (Eds.). (1997)

Principles of Colloid and Surface Chemistry, revised and expanded (Vol. 14). CRC
press.
A very complete book that help to familiarize with the fundamentals of colloid and
surface science, characterization and measurements techniques.

Ohshima, H
Theory of colloid and interfacial electric phenomena. Vol. 12. Academic Press, 2006.
A classic book to understand the fundamentals and current developments in colloidal
electrokinetic phenomena
 Introduction to Colloid Science

G.J.M. Koper (2011)

An introduction to Interfacial Engineering. VSSD.
A concise but broad overview of various interfacial properties, including colloidal
aspects

Mewis, J., & Wagner, N. J. (2012)

Colloidal suspension rheology. Cambridge University Press.
A recent book focussing on the link between microstructure and rheology, with
limited derivations but multiple references.

Hunter, R. J. (2013)
Zeta potential in colloid science: principles and applications (Vol. 2). Academic press.
A recent reprint of a classic book focusing on electrokinetics, both theoretically and
experimentally

316
 Conclusion

Index
A
activity of clays 34
aggregate 62
aggregation (band-type) 118
aggregation (bridging) 95
aggregation (examples) 119
aggregation (orthokinetic) 142
aggregation (patching) 95
aggregation (perikinetic) 142
aggregation (with clay) 96
amphiphilic molecules 98
Archie's law 256
atomic force microscopy 151
Atterberg limits 30, 32
B
barometric profile 57, 217, 223
bilayer sheet 101
Bingham fluid 161
Bingham number 163
Bjerrum length 147
blob and cloud 209-211
Boltzmann distribution 67, 215
Boycott effect 224
Brownian motion 42
Burger's equation 222
C
Carnahan-Starling relation 181, 211
Casson equation 162
chemical potential 213
chlorite 26
clay (definition) 20, 22, 23
clay mineralogy 25
coalescence 101
colloid (definition) 17

317
 Introduction to Colloid Science

colloidal crystal 176

compressibility 306
conductivity (electric) 86, 177
conductivity (hydraulic) 248
conductiviy (limiting ionic) 87
consolidation coefficient 300
contact angle 102
continuity equation 245, 254
Couette rheometer 167
Coulombic forces 67
Coulombic repulsion 71
cyanobacteria 107
D
Darcy equation 238, 241
Debye forces 65
Debye length 68, 77
degree of consolidation 308
delamination 117
depletion effect 96
dialysis 51
diatoms 105
dielectric permittivity 68
diffusion 50
diffusion limited cluster aggregation (DLCA) 114, 144
dinoflagellate 107
dissolved inorganic carbon (DIC) 109
dissolved organic carbon (DOC) 109
DLVO theory 71, 147
drag force 45
dynamic gravitational length 230
E
electric dipole 63
electric double layer 67
electrophoretic mobility 81, 83
electroviscous effect 169
Euler derivative 255
excess pore pressure 245

318
 Conclusion

extensional flow 170

F
Fick's first law 47
Fick's second law 48
floc 62, 114
floc size and polyelectrolyte 126
floc size and salt 124
floc size and shear rate 122
fluid mud 200
fractal (pseudo-) dimension of flocs 131
fractal and self-similarity 128, 271
fractal approach theory 268
fractal dimension 130
G
gas in soil 249
Gauss' law 82
gelling time 282
gels and hydrogels 186, 187
Gibson equation 245, 248, 268
Gibson height 251
H
Hamaker constant 65
Herschel-Bulkley model 161, 189
Hooke's law 164
hydrogen bond 94, 120
hydrometer 39, 40
hypertonicity 59
hypotonicity 60
I
illite 26
illite (Hamaker constant) 66
isotonocity 59
K
kaolinite 26, 27, 29, 116
kaolinite (Hamaker constant) 66
kaolinite (interaction energy) 73
Keesom forces 64

319
 Introduction to Colloid Science

Kolmogorov microscale 144

Kynch equation 279
L
Lagrange derivative 255
laminar regime 159
Lennard-Jones potential 179
light scattering (dynamic) 46
light scattering (static) 46, 123, 154
lipids 99
liposome 101
liquid crystal 53
liquid state 29, 30
London forces 65
loss angle 164
loss modulus 164
L-system 129
M
marine snow 110
material coordinate 250
Maxwell model 165
method of characteristics 279
mica 26
micelles 99, 100
Mie theory 19, 47
modulus (elastic) 180
modulus (shear) 164, 166
modulus (storage) 164
moisture content 31
mole (definition) 20
montmorillonite 26, 116
montmorillonite (Hamaker constant) 66
montmorillonite (interlayer) 122
mud (definition) 22,23
mudflat 22
O
Oden formula 41
oedometer test 235, 299

320
 Conclusion

oil sands 95
Onsager relations 260
osmotic pressure 58, 70, 120, 215
P
pair correlation function 178, 201
particulate inorganic carbon (PIC) 109
particulate organic carbon (POC) 109
pelagic zone 104
permeability 234, 276
phase transition 175
phospholipids 100
phyllosilicates 28
plankton 104-106, 108
plastic limit 31
plastic state 29, 30
plasticity chart 33
plasticity index 32
Poisson equation 67
polycyclic aromatic carbons (PAH) 103
polyelectrolyte 90
polymer 90-94
population balance equation (PBE) 144
porosity 239
pressure (hydrostatic) 243
pressure (water) 244
primary consolidation regime 235, 282
primary minimum 74
Q
quick clay 185
R
Rayleigh theory 19, 46, 47
reaction limited cluster aggregation (RLCA) 114, 144
Reynolds number 38
rheopecty 184, 185
Richardson-Zaki relation 204, 220
Rouse profile 56

321
 Introduction to Colloid Science

S
sand/mud mixtures 188
Schulze-Hardy rule 76
secondary consolidation regime 235
secondary minimum 74
sedigraph 39, 40
sediment balance 41
settling (fluctuation) 207-208
settling (hindered) 204, 228, 295
settling (non-spherical particles) 206
settling (Stokes) 39, 56, 202, 224
settling regime 235
settling velocity 152
shear flow 170
shear rate 160
shear strain 160
shear stress 160
shear stress-shear rate 161
shear thickening 182, 184
shear thinning 182, 183
shock wave 222
shrinkage limit 31
silicon, silica 110
silt (definition) 23
skeleton stress 244, 277
slipping plane 70
Smoluchowski formula (electrophoresis) 80
Smoluchowski model (aggregation) 136-141
Sodium Dodecyl Sulfate 100
solid state 29, 30
solution (definition) 17
stability ratio 144-148
stabilization 97
Stern layer 70, 71, 79, 121
Stokes-Einstein relation 45
strain 244
streaming potential 262

322
 Conclusion

stress 243
stress-strain 32
surface charge 78
surface of shear 70
surface potential 67-69
surface tension 101, 102
surfactant 98-101
suspension (concentrated) 203
suspension (definition) 19
swelling (crystalline) 121
swelling (osmotic) 121
swelling of clays 120
T
thixotropy 183
tidal flat 22, 24
turbulent regime 159
Tyndall effect 18
V
Van der Waals forces 63
viscoelastic fluid 163
viscosity 158
viscosity 160
viscosity (Einstein's formula) 170
viscosity (extended Einstein formula) 170
viscosity (Krieger-Dougherty) 172
viscosity (other models) 173
void ratio 239
volume fraction 239
W
wall effects 211
water purification 60
Wilhelmy plate 102
Y
yield stress 162
yield stress (slump test) 191
yield stress maximum 194

323
 Introduction to Colloid Science

Z
zeta potential 70, 71
zeta potential (measurement) 80

324