SD R6e

Svensk Djupstabilisering
Swedish Deep Stabilization Research Centre
Report 6 Mixing Processes for

Ground Improvement
by Deep Mixing
Stefan Larsson
English
Translation
July 2002
The Swedish Deep Stabilization Research Centre coordinates research and development activities in
deep stabilization of soft soils with lime-cement columns. A joint research programme based on the
needs stated by the authorities and the industry is being conducted during the period 1996 – 2004.
Members of the Centre include authorities, lime and cement manufacturers, contractors, consultants,
research institutes and universities.
The work of the Swedish Deep Stabilization Research Centre is financed by its members and by
research grants.
The Swedish Deep Stabilization Research Centre is located at the Swedish Geotechnical Institute and
has a Steering Committee with representatives choosen from among its members.
Further information on the Swedish Deep Stabilization Research Centre can be obtained from the
Project Manager, Mr G Holm, tel: +46 13 20 18 61, +46 70 521 09 39, fax: +46 13 20 19 14 or
e-mail: [email protected], http://www.swedgeo.se/sd.htm.
Svensk Djupstabilisering
Report 6
Mixing Processes
for Ground Improvement
by Deep Mixing
Stefan Larsson
English Translation July 2002

Report Swedish Deep Stabilization Research Centre
c/o Swedish Geotechnical Institute
SE-581 93 Linkoping
Phone: +46 13 20 18 62
Fax: 013-20 19 13
E-mail: [email protected]
ISSN 1402-2036
ISRN SD-R--00/6--SE
2 Swedish Deep Stabilization Research Centre, Report 9

Foreword – English Translation
Translation of this document was made possible through collaborative effort of Swedish Deep Sta-
bilization Research Centre and the US National Deep Mixing (NDM) program. The mission of both
organizations is to facilitate advancement and implementation of deep mixing technology through
partnered research and dissemination of international experience.
The Swedish Deep Stabilization Research Center with its headquarter at the Swedish Geotechnical
Institute (SGI) coordinates research primarily on dry mix methods. An international conference on
“dry mix methods for deep soil stabilization” was held in Stockholm in 1999. The findings from
research and state-of-the practice in Europe, particularly on stabilization of organic soil, were pre-
sented and published in the proceedings. SGI is also a partner of EuroSoilStab, the R&D project
focusing on organic soils and infrastructure applications.
The National Deep Mixing program coordinates a program of deep mixing research in the US. The
international workshop on “Deep mixing technology for infrastructure development” was held in
Oakland, California in 2001. A forum of users and experts from industry, government and academia
examined the current practice and research needs; and identified the challenges ahead for imple-
mentation of deep mixing technology.
We hope dissemination of international experience serves as a step toward our better understanding
and promotion of this innovative technology in the construction industry.
July 2002
Göran Holm, Linkoping, Sweden Ali Porbaha, California, USA
Deep Mixing 3
Foreword
The present thesis is concerned with mixing mechanisms and mixing processes having application
to ground improvement by deep mixing. The main part of the work consists of a literature review
with particular emphasis on literature on the process industries. This review forms a basis for a de-
scription and discussion of the mixing process and factors affecting the process in connection with
deep mixing methods. The project also comprised field and laboratory trials.
The work was carried out at the Department of Civil and Environmental Engineering, Division of
Soil and Rock Mechanics, Royal Institute of Technology, Stockholm. The project is part of an in-
dustry-wide R&D programme run by the Swedish Deep Stabilization Research Centre.
The project was financed by the Development Fund of the Swedish Construction Industry (SBUF),
the Swedish Deep Stabilization Research Centre, Tyréns Infrakonsult AB, the Royal Institute of
Technology, Hercules Grundläggning, Stabilator, and LC-Markteknik.
My warmest thanks go to my supervisor Associate professor Sven-Erik Rehnman, to Professor

Bengt Broms, and to my colleague Mr. Morgan Axelsson for their active interest in the project.
Many thanks also to all the staff of the Division of Soil and Rock Mechanics.
Stockholm, May 2000

Stefan Larsson

Contents
Foreword – English Translation ................................................................................................ 3
Foreword ............................................................................................................................. 4
Summary ............................................................................................................................. 7
Definitions ............................................................................................................................. 9
1. Introduction ........................................................................................................................ 12
1.1 The purpose of mixing processes ................................................................................. 12
1.2 Deep mixing ................................................................................................................. 12
1.3 Problems and knowledge gaps connected with mixing mechanisms .......................... 14
1.4 Purpose and execution ................................................................................................. 14
1.5 Important literature ...................................................................................................... 15
2. Mixing mechanisms .......................................................................................................... 16

2.1 Introduction .................................................................................................................. 16
2.2 Classification of mixing mechanisms .......................................................................... 16
2.3 Classification of flow types ......................................................................................... 17
2.4 Agglomeration and segregation ................................................................................... 18
2.5 Forces and mechanisms causing mixing ...................................................................... 19
2.6 Dispersion of fine particles in a liquid ......................................................................... 28
3. Mixing equipment .............................................................................................................. 32

3.1 Introduction .................................................................................................................. 32
3.2 Dispersion and dissolution in liquids ........................................................................... 32
3.3 Mixing of pastes and plastic materials ......................................................................... 40
4. Mixture quality .................................................................................................................... 47

4.1 Introduction .................................................................................................................. 47
4.2 Describing the quality of a mixture ............................................................................. 48
4.3 Scale of scrutiny ........................................................................................................... 49
4.4 Scale and intensity of segregation................................................................................ 49
4.5 Mixing indices.............................................................................................................. 53
4.6 Measurement of mixture quality .................................................................................. 56
5. Rheology ........................................................................................................................... 60
5.1 Introduction .................................................................................................................. 60
Deep Mixing 5
5.2 Rheological properties ................................................................................................. 60
5.3 Measurement of rheological properties ....................................................................... 67
5.4 Rheological properties in partially remoulded state .................................................... 69
5.5 Rheological properties on incorporation of lime and cement ...................................... 72
6. Deep mixing ........................................................................................................................ 75

6.1 Introduction .................................................................................................................. 75
6.2 Historical sketch ........................................................................................................... 75
6.3 The mixing process ...................................................................................................... 76
6.4 Factors affecting the mixing process ........................................................................... 87
6.5 Laboratory philosophy ............................................................................................... 102
6.6 Mixing tools ............................................................................................................... 106
6.7 Discussion of some characteristic defects .................................................................. 121
7. Assessment of mixture quality: Arboga field trial .................................................. 124

7.1 Introduction ................................................................................................................ 124
7.2 Methods and materials ............................................................................................... 125
7.3 Results ........................................................................................................................ 131
7.4 Discussion .................................................................................................................. 137
7.5 Conclusions ................................................................................................................ 140
8. Further research .............................................................................................................. 141
References ......................................................................................................................... 144
Appendix A ......................................................................................................................... 161

Appendix B ......................................................................................................................... 167
Appendix C ......................................................................................................................... 189

Summary
The lime-cement column method in deep mixing, where lime and cement are mixed into the soil to
form columns with a hardening effect on the soil, has been used extensively in the Scandinavian
countries to reduce settlements and stabilise soft soils. The method is under rapid development with
new applications such as the stabilisation of deep excavations and high embankments. These new
applications may require a better mixing quality and better control over the mixing process in the
stabilised soil. Both clients and contractors are in need of methods for evaluating the mixing quali-
ty. Different applications may require different properties and thereby different mixing qualities.
Although the method has been used for over 20 years and has become a large-scale industry,
knowledge about the fundamentals of the mixing process is limited. There is also a lack of methods
for investigating the different parts of the mixing process.
The present project is part of Svensk Djupstabilisering (the Swedish Deep Stabilization Research
Centre). The purpose of this thesis is to form a basis for development of the mixing process by in-
creasing knowledge of the fundamental theories and mechanisms when mixing binders and soft
soils.
The thesis begins with a presentation of an extensive literature survey in the process industry. Mix-
ing operations are used widely in industrial production processes where physical and chemical
changes occur. Present knowledge of the various mixing processes is mainly based on investiga-
tions by the chemical industry relating to the manufacture of pharmaceuticals, paper, plastics, ce-
ramics and rubber. In spite of extensive research over a long period, there is still a lack of funda-
mental understanding of complex mixing processes.
The first three chapters deal with mixing mechanisms, basic mixing operations and equipment, and
assessment of the mixing quality, respectively. The focus is on mixing of particle suspensions and
laminar mixing mechanisms. The principal conclusion from this part of the thesis is that the litera-
ture survey provides a key implement for the primary purpose, describing and discussing the mix-
ing process in the deep mixing method.
The process of mixing is highly dependent on the properties of the materials concerned. When dif-
ferent types of materials are mixed, it is important to have a thorough knowledge of the rheological
properties of the materials. When these are known, it is possible to forecast the mechanisms which
are essential for mixing and thereby to determine which mixing processes should be chosen. Nor-
mally, the behaviour of particle suspensions is very complex. With a dry binder, the soil mixture
becomes even more complex. The viscous elastic properties of a soft soil reduce the movements
around the mixing device. Relative motions between different elements in the mixture, which are a
prerequisite for mixing to occur, only take place in a narrow volume around the mixing device.
Another problem lies in the difficulties of measuring the rheological properties of a clay soil during
very rapid deformation. Basic knowledge is lacking and the technique becomes empirical.
Deep Mixing 7
The main part of the thesis deals with the application of deep mixing. The mixing process is very
complex, with many phases and several factors that influence the process and the results. Although
it is difficult to distinguish the different mechanisms during a mixing operation, it is important to
understand how they affect both each other and the outcome. The different parts of the mixing
process are described on the basis of the dry mixing method, used in the Scandinavian countries.
An important conclusion is that the first part of the process, the distribution of the binder from the
mixing device into the soft soil, is most critical for the process to perform as required.
The influence of various factors on the mixing process is described and discussed. Unfortunately,
no extensive studies have been published concerning mixing processes in deep mixing methods.
Most of the studies referred to in this thesis are fragmentary and of very limited scope.
A survey of different variants of deep mixing methods worldwide shows that the development of
the mixing process and mixing devices has been very slow and limited in the Scandinavian coun-
tries. In Japan, development of the method has resulted in a number of versions and thereby an
broad area of application.
During the project, a number of laboratory investigations have been performed. However, these are
very fragmentary compared with the literature survey, which deals with the whole mixing process.
In co-operation with a parallel project, a large number of samples were taken from a number of
lime-cement columns that had been extracted. The purpose was to study the distribution of binders
in the columns by measuring the binder content in the samples. The discussion deals with the possi-
bility of applying statistical analysis to measure the mixing quality when using the dry mixing
method. The study shows that mixing indices can be used as a quantitative measure of the mixing
quality. The results indicate that the sample size has considerable influence when the mixing quality
is poor and limited influence when the mixing quality is relatively good. It is also shown that it is
difficult to draw conclusions concerning the distribution of binder when using only a few samples,
since the distribution may vary. With more extensive knowledge, it might be possible to assess the
mixing quality without increasing the number of samples.

Definitions
Agglomerate A collection of primary particles and aggregates which are

connected at their ends and corners is called an agglomerate.
Its surface area is not much less than the total surface area of
its constituents.
Aggregate Aggregates are groups of primary particles connected by their

surfaces. An aggregate has a surface area that is less than the total
surface area of its constituent particles.
Bingham plastic fluid A fluid which differs from a Newtonian fluid only in that the linear
plot of shear stress against shear rate does not pass through the
origin.
Dough A non-Newtonian material having elastic and viscous flow

properties.
Diffusive mixing Natural or spontaneous mixing. The components blend by

themselves.
Dilatant fluid With this type of fluid, the apparent viscosity increases as shear
rate is increased.
Disperse To distribute (become distributed) through a liquid in the form

of fine particles.
Dispersion A chemical mixture of fine particles dispersed in a liquid.
Dispersive mixing Mixing in which the size of the ultimate particles is reduced.
Distributive mixing Mixing by redistribution through slicing and replacing.
Dynamic viscosity The viscosity µ (Pa·s) is now called the dynamic viscosity.
Flocculation Suspended particles bound together by weak structures in a liquid
Fluid (adj. and noun) Liquid.
Deep Mixing 9
Flow In this report: material in motion
Granular viscous material A suspension having both the properties of a solid and viscous flow
properties
Intensity of segregation The intensity of segregation (IS) is a measure of the difference in

concentration between neighbouring lumps in a mixture. It
describes the effect of molecular diffusion in the mixing process.
Laminar mixing Reduction of the scale of segregation or layer thickness by laminar

deformation (shearing, stretching, slicing and folding).
Layer thickness The average distance between adjacent boundary layers of

materials mixed by laminar mixing mechanisms.
Liquid (n.) A substance in the melted state. (adj.) In the melted state.
Macromixing Mixing by turbulent or laminar motion on a scale larger than the

smallest layer thickness or eddy size
Micromixing Mixing by molecular diffusion on a scale smaller than the smallest

layer thickness or eddy size
Mixing time The time it takes to mix from a given initial state to a prescribed
final mixing quality, i.e. the time required for the mixture of the
component substances to reach a specified degree of uniformity.
Molecular diffusion Reduction in the intensity of segregation. The transport or

movement of individual molecules through a fluid by the random
motion of individual molecules. E.g. calcium ions which react with
the silica and aluminium in clay minerals over a long period.
Newtonian fluid A fluid in which the shear stress is proportional to the shear strain
is classed as a Newtonian fluid.
Paste A non-Newtonian dispersion with a high concentration of solid

particles in a liquid. The shear stress in the material is not
proportional to the shear rate.
Plastic fluid Plastic fluids are characterised by a yield stress which must be
overcome in order for movement to take place.
Pseudoplastic fluid With this type of fluid, the apparent viscosity varies inversely as
the shear rate.

Rheology The study of the deformation and flow properties of materials.
Scale of segregation The scale of segregation (LS) is a measure of the size of unmixed
lumps of pure consistency.
Shear rate The velocity gradient of a fluid under shear or elongation.
• du dv m −1 
γ = =  ms = s 
dxdt dx
Suspension Fine particles of an insoluble substance uniformly dispersed in a

liquid or gas.
Thixotropic fluid These fluids display a reversible decrease in shear stress over time
at constant shear rate.
Turbulent mixing Reduction of the scale of segregation by random turbulent

movement of the material.
Residence time The time which an element spends in a mixer etc. between entry
and exit.
Viscoplastic fluid These fluids behave in part as a viscous liquid and in part as an
elastic solid, i.e. they can exhibit elastic recovery.
Viscosity Internal friction, resistance to flow (from Latin viscosus ‘sticky’).
Viscous fluid A fluid having high viscosity (sticky, thick-flowing)
Viscoelastic material A material having both elastic and viscous properties. The material
shows elastic recovery from the deformation that takes place on
movement.
Wetting Replacement of gas (or vapour) with a liquid from the surface of
a solid.
Yield stress The minimum stress required to break up structure of a material

sufficiently to produce movement. The yield stress of a suspension
is such that its properties resemble those of a solid as long as the
stresses in the material do not exceed the yield stress.
Deep Mixing 11
1. Introduction
1.1 THE PURPOSE OF MIXING PROCESSES

Mixing and homogenization is an important process in numerous industries, e.g. the food, chemical
engineering, paper, rubber and concrete industries. The principal purpose of mixing processes is to
bring about changes in physical, chemical and biological properties.
The term mixing is applied to operations which tend to reduce nonuniformities or gradients in com-
position, properties, or temperature of material in bulk. (Uhl & Gray 1966).
The objective of mixing is homogenization, manifesting itself in a reduction of concentration or

temperature gradients, or of both simultaneously, within the agitated system... Rapid mixing neces-
sitates motion of the ingredients. (Štìrbáèek & Tausk 1965).
Movement produced by e.g. mechanical mixing may be necessary e.g. in order to break up aggre-
gates so that solid particles can disperse in a liquid. The different forms of movement constitute the
mechanisms of mixing (Nienow et al. 1992).
A uniform distribution of ingredients may be required e.g. to produce a final product with a certain
texture or to bring about a desired chemical reaction. Requirements for uniformity in a mixing proc-
ess vary depending on the requirements on the final product or function.
Mixing takes place between two materials when they come in contact with each other and are de-
formed so that the contact area between them increases. High-viscosity liquids have high resistance
to deformation, and molecular movement and diffusion are low. Relatively large forces are neces-
sary to overcome this resistance. Mixing also takes place when materials are cut into smaller ele-
ments and redistributed. A satisfactory mixing process produces a mixture of the required uniformi-
ty in the shortest possible time and at the lowest possible cost in terms of machinery, power and
labour.
This thesis describes the basic theories and mechanisms of mixing, primarily of liquids and solids.
The literature review concentrates mainly on literature published in the process industries, where
research has been done in this field since the 1940s.
1.2 DEEP MIXING

Ground improvement by deep mixing is very widely used, especially in infrastructure construction,
for increasing stability and reducing settlement in loose soils. Experience has been positive and the
method has great development potential. The method is undergoing rapid development, particularly
with regard to its applicability, cost effectiveness and export potential. Increasingly stringent de-

mands are made on the uniformity of properties along the length and across the section of the col-
umns produced. There is also a particular requirement for uniformity of properties between the col-
umns in a structure. In order to improve the production process, we need a better knowledge of the
basic theories and mechanisms involved in the incorporation of binders, such as lime and cement,
into soil.
In the Nordic countries compressed air is used almost exclusively as the medium for the transport
of lime and cement from the tank to the soil. This is known as “dry deep mixing”. The alternative
today is to premix the binder, usually cement, with water to form a grout, thus distributing the bind-
er to the soil in liquid form. This is known as “wet deep mixing” and is the dominant technology in
Japan and elsewhere.
The use of compressed air as the medium for transporting the binder has the advantage that it takes
a relatively small amount of binding agent to achieve the requisite strength gain. Given that loose
soils already contain a lot of water, it appears logical not to add still more water to the soil, as is
done when the wet method is employed. However, the addition of air adds to the difficulty of the
mixing process in a material, loose soil, whose rheological properties are already very complex. In
the mixing process an air-borne binder complicates the dispersion process with regard to the wet-
ting of lime and cement particles and the breaking up of agglomerates.
The objective of the mixing process in deep mixing is to transport and distribute the binder in such
a way as to produce uniform columns having the properties required for the function they are to
perform.
A series of operations and mixing mechanisms occurs in deep mixing. Compared with e.g. the mix-
ing of two miscible liquids in a vessel, the mixing process that occurs in deep mixing is very com-
plex and it is difficult to monitor the entire production process continuously. The wide variations in
local geology, the difficulty of predicting the rheological properties of the mixture, field conditions
affecting equipment etc. all make it difficult to monitor and study the mixing process and mixing
mechanisms in the field.
The mixing process affects the physical and chemical properties of the columns, including:
1 Uniformity. The properties of columns can vary both along their length and across their section.
Heterogeneities in various properties have been observed across column sections, e.g. grainy
consistency, accumulations of binder along the column periphery, and tapered columns.
2 Strength gain and permeability. The main purpose of the mixing process is to disperse a binder
in the soil. The potential strength gain and permeability of the column vary depending on how
efficiently the binder is dispersed.
Deep Mixing 13
1.3 PROBLEMS AND KNOWLEDGE GAPS CONNECTED
WITH MIXING MECHANISMS
As mixing processes become more complex, with the mixing of multiphase systems such as suspen-
sions, the demands for quality and cost effectiveness grow. Although many studies have been car-
ried out on mixing processes, the mixing of different materials is something of an art. There are still
gaps in our theoretical knowledge of mixing processes. In practice, in almost all mixing processes
the agitator is called upon to perform many tasks, any one of which may be the most critical for the
product.
The design of mixing processes in the process industries is not based on well-established science.
The choice and design of mixers is still done by a trial calculation procedure giving due weight to
previous experience (Short & Etchells 1982). In design, the mixing plant is selected that is expected
to give the best results. Laboratory scale tests are then often carried out with the selected equip-
ment. Some specific factor, depending on which one is expected to be the most critical, may be kept
constant during scale-up. This approach works relatively well for the design of mixing processes
involving liquids, but it works poorly for the mixing of concentrated suspensions, as is the case
when incorporating lime and cement into loose soils such as clays.
The deep mixing industry is changing in order to be able to compete by producing a good product
with maximum efficiency. The efficiency of a specific mixing process varies with the properties of
the soil and of the binding agent. Thus more knowledge and understanding are needed in order to
develop the method.
Some problems which need to be solved in the improvement or design of a mixing process:
1 Design of mixing tools
2 Control of incorporation and spreading procedures
3 Rheological properties of the materials being mixed
4 Standards for comparison of different mixing tools
5 Classification and measurement of mixing quality
6 Description of mixing processes
7 Description of mixing mechanisms
1.4 PURPOSE AND EXECUTION

The purpose of the present thesis is to provide a foundation to assist contractors and machinery
manufacturers to develop the mixing process in deep mixing, by making a contribution to knowl-
edge of the basic theories and mechanisms in the incorporation of binding agents into soil. The
main part of the work consists of a literature review.
Chapter 2, Mixing mechanisms, deals with basic theories and mechanisms of mixing, particularly of
liquids and solids.

Chapter 3, Mixing equipment, is a review of the basic designs of mixing equipment used in the
process industries for mixing liquids and solids.
Chapter 4, Mixing quality, is a compilation of simple statistical analyses and mixing indices for the
assessment of the mixing quality of different types of mixtures.
Chapter 5, Rheology, discusses the rheological properties of loose soils and their influence on a
mixing process.
Chapter 6, Deep mixing, describes and discusses the mixing process, mixing mechanisms and fac-
tors influencing the process of incorporating binding agents into loose soils. It further discusses the
mixing equipment used in deep mixing, primarily in the Nordic countries and in Japan.
Chapter 7, Assessment of mixing quality – field test at Arboga, presents a statistical analysis of mix-
ing quality based on chemical assays of a large number of samples taken from four lime–cement
columns.
Chapter 8, Further research, discusses further research needs in the field.
1.5 IMPORTANT LITERATURE

Although no well-organized, definitive research has been done on mixing processes, there is a good
deal of review literature on the field. Some of the literature that may be recommended and on which
this report is based is listed below.
Harnby, N., Edwards, M.F. and Nienow, A.W., 1992. Mixing in the Process Industries. Second
Edition. Provides a very good overview of different mixing mechanisms and applications.
Štìrbáèek, Z. & Tausk, P., 1965. Mixing in the Chemical Industry. Although written over thirty
years ago, this book remains highly relevant as a good overview of different mixing mechanisms
and their applications.
Barnes, H.A., Hutton, J.F. and Walters, K., 1989. An Introduction to Rheology. Rheology
Series, 3. A good introduction to the science of rheology.
Yonekura, R., Terashi, M. and Shibazaki, M. 1996. Grouting and Deep Mixing. Proceedings of
IS–Tokyo´96. The 2nd International Conference on Ground Improvement Geosystems, Tokyo,
14–17 May 1996. Vol. 1–2. A very extensive compilation of papers on deep mixing with appli-
cations, properties and design.
Bredenberg, H., Holm, G. and Broms, B.B. 1999. Dry Mix Methods for Deep Soil Stabilization.
Proceedings of the International Conference on Dry Mix Methods for Deep Soil Stabilization,
Stockholm, Sweden, 13–15 Oct. 1999. A good complement to the Japanese conference,
IS–Tokyo ’96, focussing mainly on dry deep mixing.
Deep Mixing 15
2. Mixing mechanisms
2.1 INTRODUCTION
As mentioned in the introductory chapter, mixing processes reduce nonuniformities or gradients in
composition, properties, or temperature of material in bulk (Uhl & Gray 1966). An understanding
of the movements which achieve these reductions in liquids and solids is an essential prerequisite
for the design or improvement of a mixing process. The different forms of movement constitute the
mechanisms of mixing (Nienow et al. 1992).
The present chapter sets out to review the basic mechanisms involved in mixing, particularly that of
liquids and suspensions. Since loose soils are suspensions, and generally highly concentrated sus-
pensions, the emphasis is placed on laminar mixing mechanisms. The process of dispersion of solid
particles in liquids is also touched upon.
The literature cited in this chapter comes primarily from the process industries, where the available
research on mixing mechanisms has been done. Mixing mechanisms in connection with deep mix-
ing will be discussed in Chapter 6.
2.2 CLASSIFICATION OF MIXING MECHANISMS

A general classification of mixing processes is independent of the product or industry. Mixing proc-
esses can be classified as follows (Nienow et al. 1992):
1 Single-phase liquid mixing
2 Liquid-liquid
3 Solid-liquid mixing
4 Gas-liquid mixing
5 Three-phase contacting
6 Solids mixing
Dry deep mixing by means of columns of lime, lime-cement, and cement involves solids (lime and
cement particles), a concentrated clay suspension, and a gas (air). The phases differ greatly in their
chemical and physical properties, and as a result these systems have extremely complex properties
which can vary under different conditions.
While most types of mixing equipment are used to mix materials having a range of flow properties,
the materials' properties are important to the design of the mixing plant. It may be that each type of
equipment provides optimal performance with only a narrow range of mixture properties.

Rheology is the study of the deformation and flow properties of materials. From a rheological view-
point all materials will flow if one waits long enough. Liquids (fluids) to be mixed may be classi-
fied as (Lindley 1991a):
1 Newtonian
2 non-Newtonian
3 Viscoelastic
Some simple models of flow are discussed in Appendix A.
Additional materials classifications might be (Lindley 1991a):

1 Cohesive solids
2 Powders
3 Sticky solids
4 Dry solids
2.3 CLASSIFICATION OF FLOW TYPES

Suspensions, such as clay suspensions and suspensions of lime and cement particles in clay, show
properties that differ from those of Newtonian fluids, i.e. most relatively concentrated suspensions
have non-Newtonian or viscoelastic characteristics.
In an incompressible material the characteristics of plane flows (relative movement between the
elements of the material) can be described by two parameters: the rate of deformation ε and the rate
of rotation ω. A number R which characterizes the flow type is written as (Giesekus 1983):
(ε 2 − ω 2 ) (1 − r02 )
R= 2 = (2.1)
(ε + ω 2 ) (1 + r02 )
where r0 = ω / ε
Two classes of flow types are defined as:

hyperbolic flows: 0<ω<ε 0 < r0 < 1 0<R<1
elliptic flows: ε<ω<∞ 1 < r0 < ∞ -1 < R < 0
Deep Mixing 17
The limiting cases of these are:
pure shear flow: ω=0 r0 = ∞ R=1
simple shear flow: ω=ε r0 = 1 R=0
pure rotational flow: ε = 0 r0 = ∞ R = –1
Simple shear flow can be conceived as a relative sliding between rigid layers of infinitesimal thick-
ness, see Fig. 2.1. The distances between particles remain constant for all particles situated in the
same layer and change only if they are situated in different layers. The rate of rotation is equal to
the rate of deformation.
In a mixing process in a non-Newtonian liquid there is usually a combination of all types of flow.
(a) Simple shear (b) Pure rotation (c) Elliptic (d) Pure shear
Fig. 2.1 Types of plane flow (after Giesekus 1983).
The rheological behaviour of the material depends on the type of flow (Giesekus 1983). Thus in a
mixing process, different parts of the mixture may show different rheological behaviour.
2.4 AGGLOMERATION AND SEGREGATION

Since mixing processes involve different materials having different rheological properties, it is im-
portant to be aware of factors that influence the properties of the mixture. Particles of materials
which tend to agglomerate have poor flow characteristics. Particles agglomerate or adhere to one
another because of interacting forces. The following mechanisms can lead to agglomeration (Coul-
son et al. 1979):
a) Mechanical interlocking: masses of long, thin particles may become totally interlocked.
b) Surface attraction: these include Van der Waal's forces; substantial bonds may be formed (high
specific surface to volume ratio).
c) Plastic welding: results from high pressures which may develop in small contact area between
irregular particles.
d) Electrostatic attraction: movement of fine particles can result in charge build-up.

e) Effect of moisture. Moisture collects near points of contact between particles and results in sur-
face tension effects. It may also partially the particles, which then act as a bonding agent on sub-
sequent evaporation.
f) Temperature fluctuation may result in changes in particle structure and greater cohesiveness.
Perfect mixing is the state in which all the parts of the mixture have the same composition.
Segregation occurs because of differences in particle size, density, shape and/or resilience. Size
difference is the most important; density sometimes is important; other effects are generally unim-
portant. The following mechanisms lead to segregation (Williams 1983):
a) Particle trajectories where different size particles have different velocities.
b) Percolation of fine particles where fine particles move between gaps in the particle mass.
c) Rise of coarse particles on vibration.
2.5 FORCES AND MECHANISMS CAUSING MIXING

The fundamental processes for the transport of materials in a mixing process are: momentum trans-
fer in mobile media, heat transfer, and mass transfer (Geankoplis 1993).
Molecular transport process produce transfer or movement of a given property by molecular move-
ment through a system or medium (Geankoplis 1993). The properties transferred may be mass, heat
or momentum. Each molecule in a system possesses a given quantity of mass, heat or momentum
associated with it. If there is a concentration gradient of the between one region and an adjacent
region, transport of this property will take place. The flux will be proportional to the concentration
gradient and inversely proportional to the resistance.
The pattern of movement around an impeller is generally very complex. However, according to
Metzner & Otto (1957), the resistance to the impeller can be examined by comparison with simple
flow situations. In laminar flow around a sphere there is no separation of the flow and viscous dissi-
pation of energy is the controllable factor. The total force on the sphere is due to the shearing forces
on the surface and the differences in pressure between the front and the rear of the sphere. The pres-
sure differences are due to the viscous dissipation of energy in the fluid stream (Metzner & Otto
1957).
The mixing mechanisms can be summarized as follows (Lindley 1991a):

a) Turbulence and diffusion for low viscosity liquids
b) Breaking and recombining, deformation, laminar flow, shear flow and streamline flow for very
viscous liquids
c) Convection, diffusion, and shear for dry, non-cohesive solids
Deep Mixing 19
Turbulent mixing
For mixing to take place there must be relative movement between particles or flow elements.
When the flow rate increases the flow becomes turbulent. Mixing of materials takes place through
either spontaneous motion (diffusion or free convection) or forced motion (Šterbácek & Tausk
1965). External forces must be applied to overcome the forces of resistance in the mixture. These
forces are: inertial forces which resist changes in direction or velocity, and viscous resistance or
shear forces. For liquids of low viscosity in which turbulence has developed, inertial forces offer
the greatest resistance to mixing.
Turbulent flow is a hydrodynamic process in which eddies are produced. The elements move not
only in parallel layers but also on erratic paths. This promotes mixing in that not only molecules but
also fluid elements pass between the individual layers. The analysis of the turbulent flow in e.g. a
mixing vessel is very difficult because of the rapid three-dimensional motions. To determine wheth-
er the flow is turbulent or laminar one can use the Reynolds number criterion, but determining the
Reynolds number is very difficult. For mixing processes in vessels the maximum value of the Rey-
nolds number for laminar flow is Re = 10–20.
Turbulent motion can be considered as a superposition of a spectrum of velocity fluctuations and

eddy sizes in the mixture. The relation between shear stress and shear rate in turbulent flow can be
written as (Šterbácek & Tausk 1965)
dv
τ = (µ + ε ) (2.2)
dy
where µ is the viscosity (laminar internal friction), ε is the eddy viscosity, and dv/dy is the shear
rate or shear velocity gradient. The eddy viscosity e is not constant for a fluid at a given tempera-
ture. It is not constant in all parts of the fluid, but varies from small values at the periphery of the
flow to high values in the centre. For fully developed turbulence the influence of laminar friction m
may be neglected and the above equation can be written as
dv
τ =ε (2.3)
dy
The mixing of suspensions and solid-liquid systems are dependent on large eddies which contain a
large amount of energy, so-called anisotropic turbulence (Nienow et al. 1992). Large eddies have
large velocity fluctuations of low frequency and are of a size related to the geometry of the mixing
tool. Eddies are of great importance for both distributive and dispersive mixing.
Like the internal friction, the diffusion rate also increases as the flow becomes more turbulent.
Laminar mixing
The mechanisms that bring about mixing are laminar and turbulent flow, eddy diffusion and molec-
ular diffusion. Internal forces and shear forces are responsible for reduction in particle size and
hence an increase in interface area between the components. If aggregates, (strongly bonded lumps)
are present in the material, laminar mixing acts as a dispersing process, i.e. high shear velocities
break up the aggregates, bringing about homogenization.

If rapid mixing is necessary, the ability of the mixing method to deform the materials becomes im-
portant (Gray 1963). Some fluid movements are more effective than others for deforming and mix-
ing viscous materials.
A flow pattern is desirable that produces movement throughout the entire volume to be mixed. In
viscous liquids the kinetic energy of the moving fluid is consumed by the viscous drag. Viscous
liquids need to be pushed into and out of all parts of the volume by a moving surface or moving
fluid. This displacement should be done in a way which rapidly deforms the shape of and extends
the area between the materials being mixed (Gray 1963).
The mechanisms which can be active in laminar flows are laminar shear, extensional flow, distribu-
tive mixing, molecular diffusion and shear stresses (Edwards 1992). Laminar flow is normally asso-
ciated with high-viscosity liquids. At typical rates of energy input, viscosities greater than about
10 Pa·s are required if the flow is to be truly laminar (Nienow et al. 1992). High-viscosity liquids
are as a rule rheologically complex. Under laminar flow conditions, intermediary forces quickly die
out under the action of the high viscosity. The mixing divice must therefore occupy a large part of
the volume to be mixed. Close to the tool high shear rates exist where fluid elements can be de-
formed and stretched by laminar flow. The mixing quality is determined by the total shear, which is
the product of shear velocity and time. Mixing is improved by a combination of high shear forces
and long residence time.
Laminar shear flow

In laminar shear there is elongation and increase in the area of a plane which is perpendicular to the
direction of shear, see Fig. 2.2.
In order to set up a relation between deformation in laminar motions and the mixing quality of a
two-component mixture, we may treat the layer thickness as a measure of the state of the mixture.
If S represents the area between two components and t the thickness of the layer, the relation shown
in Fig. 2.3 is obtained.
Velocity gradient, ∆v/∆y Decreasing thickness t1>t2>t3>t4

•
= shear rate γ Increasing area
t1 t2 t3
t4
∆y
∆v
Fig. 2.2 The thinning of fluid elements due to laminar shear flow (after Nienow et al. 1992).
Deep Mixing 21
Fig. 2.3 Laminar shear flow (after Hold 1982)
A general analysis of mixing in laminar shear flow was presented by Spencer and Wiley (1951).
The process is illustrated in Fig. 2.4, and it can be shown that the ratio of the final to the initial in-
terfacial area is given by (Hold 1982):
S
= 1 − 2γ cos α cos β + γ 2 cos 2 α (2.4)
S0
where: S = interfacial area per unit of mixture (m2/m3)

S0 = interfacial area before shearing, area per unit of mixture (m2/m3)
α = angle between the normal to the original plane and the x axis (rad)
β = angle between the normal and the y axis (rad)
γ = net strain (-)
y y
∆x •
=γ t =γ
v ∆y
n
S0 β S
α
x x
z z
Fig. 2.4 Increase in interfacial area between parallel plates in laminar shear flow
(after Hold 1982).

The influence of orientation is brought out by the equation above (Edwards 1992):
a) for surfaces initially parallel to the x,y plane
S S0 = 1 (2.5)
that is, no mixing occurs;
b) for surfaces initially parallel to the x,z plane
S S0 = 1 (2.6)
and again no mixing occurs;
c) for surfaces initially parallel to the y,z plane, i.e. perpendicular to the plane of shear
•
S S 0 = (1 + (γ t ) 2 )1 2 = (1 + γ 2 )1/ 2 (2.7)
giving an increase in interfacial area between the components and hence an improvement in mix-
ing.
For very large deformations g Hold (1982) shows that:
S S 0 = (1 + γ 2 cos 2 α )1 2 ≅ γ cos α (2.8)
The interfacial area is increased if the direction of shearing changes during the shearing process
(Hold 1982). It is therefore more advantageous to deliberately change the direction of shear than to
apply a unidirectional shear field continuously. Simple shearing can thus be optimized by changing
the direction of shear on each cycle, see Fig. 2.5. The increase in interfacial area can be optimized
by changing the direction of shear so that it always maintains angles α = β = 45°, see Fig. 2.4.
Deep Mixing 23
(a) → (b) (b) → (c) (b) → (c)´
y vy y
vx x vx
(c)
(a) (b) (c)´
x
Fig. 2.5 Effect of simple shearing with change in direction. Two consecutive simple
shearing steps at right angles (after Edwards 1992 and Hold 1982).
Extensional flow
In most practical applications there are generally also extensional flows, i.e. flows where the veloci-
ty increases in the direction of flow, see Fig. 2.6. Extensional strain produces stretching of the ma-
terial due to changes in the flow geometry or acceleration of the flow, which reduces the thickness
of the elements and thus improves mixing quality (Edwards 1992).
Fig. 2.6 The thinning of fluid elements due to extensional flow (after Nienow 1992).

y
t0
t
l0 x
Fig. 2.7 Increase in interfacial area during idealized uniaxial extension (after Cheng 1979).
In uniaxial extension in the x direction, compression occurs in the y and z directions at half the
strain rate in the x direction, see Fig. 2.7. The ratio of final to initial length of an elemental volume
is given by (Edwards 1992):
γe
l / l0 = e (2.9)
where: l0 = initial length of elemental volume (m).

l = final length of elemental volume (m).
γe = strain in extensional movement (-).
The ratio of final to initial width is given by (Edwards 1992):
t / t 0 = e −γ e (2.10)
where: t0 = initial width of elemental volume (m).

t = final width of elemental volume (m).
From this it follows that (Cheng 1979):
S S 0 = eγ e / 2 (2.11)
Deep Mixing 25
For planar extensional flow the elongation in the x direction is accompanied by compression at the
same rate in the y direction with no compression in the z direction, see Fig. 2.8. In this case the ratio
of final to initial length of an elemental volume is (Cheng 1979):
γ pe
S S0 = e (2.12)
where γpe is the strain in the x direction upon planar extensional movement.
If a comparison is made by the change in interfacial area produced in unidirectional shear, uniaxial
elongation and planar extension, it is clear that effectiveness in generating new surface area increas-
es in the order (Cheng 1979):
1) Unidirectional shear
2) Uniaxial elongation
3) Planar extension
In practical applications it is not possible to continuously elongate material: an elongational region

must be followed by a compression zone. In mixing processes the material passes through different
zones (Edwards 1992). If a mixer has zones of high shear or elongation, the components should
pass through these zones as often as possible.
S0
S
Fig. 2.8 Increase in interfacial area during planar extensional movement (after Cheng 1979).

Distributive mixing
Blending by spatial redistribution due to slicing and replacing is illustrated by the idealized mecha-
nisms shown in Fig. 2.9. The laminar flow through a unit, e.g. an in-line static mixer, produces a
stream-splitting and hybridizing effect which reduces the striation thickness and increases the inter-
facial area between the components. This is called distributive mixing (Edwards 1992).
As Fig. 2.9 shows, the interfacial area between the two components increases and the striation
thickness decreases. The mixing quality improves with each application of the mechanism. In the
idealized situation shown in Fig. 2.9 the slicing and replacing mechanism is not accompanied by
shearing or extension. It is advantageous for a mixing process to be designed to have zones in
which the elements are subjected to shearing and extension, with the opportunity for reforming be-
tween these zones, i.e. a combination of laminar shear flow and distributive mixing.
slice
slice rotation
rotation
Fig. 2.9 Concept of distributive mixing by slicing and rotating (after Nienow et al. 1992).
Molecular diffusion
In high-viscosity fluids molecular diffusion is a slow process. Only when shear and strain have re-
duced the striation thickness to a low enough value does molecular diffusion become significant.
Solids
Solid particles range in size from fine powders to large particles varying in size, shape, density and
roughness. Fine powders can be difficult to mix due to cohesive forces and coarser particles due to
segregation. Mixing uniform, free-moving solids, a task intermediate between those aforemen-
tioned, is an easy matter.
The mixing of cohesive powders requires the mixture to be divided into smaller parts and each part
to move freely relative to the mixture. Mixers producing high shear forces are preferable for such
materials.
Deep Mixing 27
Three mechanisms account for the behaviour of solid particles during mixing (Perry et al. 1984):
1) Diffusion: random movements on a small scale. Individual particles move separately
2) Convection: random movements on a large scale. Lumps or groups of particles move together
3) Shearing
2.6 DISPERSION OF FINE PARTICLES IN A LIQUID

When fine solid particles are added to a liquid it is normally necessary to break down lumps of par-
ticles which are held together by either strong or weak internal forces. Suspensions are usually
highly viscous and have non-Newtonian flow characteristics. The lumps are broken down by means
of the shear stresses that are applied to the fluid.
A number of problems arise in connection with the dispersion of solid particles in a liquid when the
particles are not adequately wetted, when air is entrained in the fluid, when agglomerates are diffi-
cult to break down, and when flocculation occurs.
A detailed analysis of the stress fields in dispersive mixing is very complex (Parfitt & Barnes
1992). The flow can involve zones of high shear and probably elongation alongside zones in which
redistribution can take place. Although elongation is more effective, equipment which delivers high
shear forces is easier to design.
Solid particles must pass through four states during conversion into a dispersion (Parfitt & Barnes
1992). The particles must be incorporated, the particle surface wetted, agglomerates broken down,
and the particles stabilized against flocculation.
Incorporation
To obtain good mixing results with short mixing times it is important that incorporation takes place
so that the incorporated powder is distributed evenly throughout the volume. This is particularly
important when incorporating powders in materials having viscoelastic properties, high yield stress,
and high viscosity, since it is difficult and time-consuming to disperse the material in the mixture
during the distribution phase. Viscoelastic materials have high resistance to deformation and hence
to movement. Good distribution of the binding agent means that the scale of segregation (see Chap-
ter 4) is reduced at an early stage of the dispersion process.
Particles may be primary particles, aggregates or agglomerates. Aggregates are groups of primary
particles joined at their faces and having a surface area less than the sum of the areas of their con-
stituent particles. A collection of primary particles and aggregates which are joined at edges and
corners is called an agglomerate. Its surface area is not much less than the sum of the areas of its
constituents (Parfitt & Barnes 1992).
Interparticulate forces exist in dry powders. Van der Waals' forces are normally attractive and gen-
erally greater than electrostatic forces. At relatively high moisture contents a film forms which pro-
duces an attractive force between particles in contact. Larger amounts of water between the parti-

cles result in the formation of interparticulate water bridges which increase cohesion (Parfitt &
Barnes 1992). The magnitude of the strongest forces, van der Waals' forces and gravitation, can be
estimated. In general, the smaller the particles the greater is the effect of interparticulate forces.
Wetting
When a powder is mixed with a liquid, the lumps of powder contain enclosed air which must be
replaced with liquid (McDonagh & Heywood 1986). The liquid first engulfs the small particles,
then the larger ones. Larger aggregates can also be engulfed by the liquid. The liquid will penetrate
voids or pores in an engulfed agglomerate, but at a slower rate than the engulfing process. At a cer-
tain degree of penetration the bulk density of the agglomerate becomes greater than that of the me-
dium, causing the agglomerate to detach. If liquid penetrates an agglomerate from all sides, the air
pressure in the pores will increase until the agglomerate is disrupted. The characteristics of the wet-
ting process are dependent on the properties of the liquid phase, the character of the surface, the
dimension of the interstices in the agglomerates, and the pressure exerted on the components by the
mechanical system.
Fig. 2.10 shows a small cube being immersed in a liquid. The wetting process takes place in three
stages: adhesion, immersion and spreading (Parfitt & Barnes 1992).
No mechanical mixing is required during the wetting process if (1) the density of the powder is
greater than that of the liquid, (2) the liquid wets the powder, and (3) the powder is not strongly
agglomerated. The powder immerses in the liquid, the pores fill with liquid, and lumps fall apart
releasing the enclosed air. If all these three conditions are not fulfilled, the wetting process must be
assisted by mechanical mixing. Even with mechanical assistance, it is particularly difficult to re-
lease enclosed air when mixing a high-viscosity liquid. One way to avoid enclosed air is to premix
the powder with a liquid of lower viscosity.
Fig. 2.10 Three stages of the wetting process: (a) to (b) adhesion; (b) to (c) immersion;
(c) to (d) spreading (after Parfitt & Barnes 1992).
Deep Mixing 29
Breakdown of agglomerates
When the powder has been incorporated in the liquid and wetted or partially wetted, the next stage
is to break down the remaining agglomerates by mechanical mixing (Parfitt & Barnes 1992). The
disagglomeration is brought about by shearing, grinding, or impact. Disruption by shear relies on
viscous drag. The shear force on a particle is directly related to the mean viscosity of the mixture.
High shear stresses for this purpose can be achieved in e.g. roller mills or twin screw presses. High
shear rate and high viscosity of the liquid are advantageous for achieving high shear. It may be re-
marked that it takes a great deal of energy to achieve high shear rates in a high-viscosity liquid. In
viscous liquids where no mass transfer occurs between the solid and liquid phases the scale of seg-
regation (see Chapter 4.4) is of great importance. When the scale of segregation has reached a low
enough level, molecular diffusion can take over to reduce the intensity of segregation.
Depending on the rheological properties of the constituent materials and the type of mixing process,
the efficiency of the dispersion process while agglomerates are being broken down is controlled by
two mechanisms:
1 The effective strain and number of revolutions of the mixing tool or the number of circulations
of the material
2 The shear rate, the intensity of agitation
Which of these parameters has the greater influence on the dispersion process depends on a number
of factors, such as the rheological properties of the mixture and the mixing mechanisms active in
the mixing equipment used.
In complex mixtures, such as the suspension of solids in a liquid during a dissolution, the distribu-
tion of the dispersed phase may depend on the intensity of the agitation (Rielly et al. 1994). Thus it
is the agitation conditions and not the time of mixing that is important in determining the quality of
the mixture, i.e. the rotational speed of the impeller is of great importance for the mixing quality.
In the mixing of thick suspensions, according to Schofield & Stewart (1982), the efficiency of the
dispersion process depends on the effective strain and the number of revolutions of the impeller.
E.g. for a Z-Blade mixer the degree of deagglomeration was independent of the shear rate (speed of
rotation) and depended solely on the number of revolutions. This means that one can estimate the
number of revolutions required to break down agglomerates in small-scale experiments and then
use this relation for equivalent larger-scale geometries.
In concrete manufacturing, Padgett (1996) writes that shear rate has greater influence on the proper-
ties of cement slurry than the total mixing energy. Laboratory mixers were compared with field
equipment by measuring free water and thickening time at the same mixing energy. The results
showed that there is little correlation between the total mixing energy and the measured properties
of the cement slurry. On the other hand, the amount of free water is strongly dependent on the mix-
ing device. One of the biggest differences between the field and laboratory equipment is the shear
rate of the mixing process. The reason why the shear rate is more important than the total energy
input lies in the forces required to break down the agglomerates that form during the wetting of the
powder.

Schofield & Stewart (1980) in the course of a study of the breakdown of pigment in thick suspen-
sions formulated hypotheses on the influence of effective strain and intensity of agitation. Accord-
ing to these authors there are three mechanisms by which agglomerates are broken down. In the
first, the initial agglomerate breaks up into two or more smaller agglomerates. These smaller ag-
glomerates will break down into still smaller units if they are subjected to a shear force which ex-
ceeds their binding force. Given sufficient time for all of the agglomerates to experience sufficient-
ly large shear forces, under favourable conditions a final degree of deagglomeration will be
achieved. Thus the deagglomeration process depends on the maximum strength of the bonds within
the agglomerates and on the shear stress available. If the applied shear stress, or intensity of agita-
tion, is high enough, all the agglomerates will be broken down into their fundamental particles. An-
other possibility is that a number of the fundamental particles (the smallest particles which can exist
individually) which make up a large agglomerate become completely dissociated. The third possi-
bility is that the fundamental particles are rubbed off the outside of the agglomerates. Summing up
Schofield and Stewart's argument, agitation must be sufficiently intense and long-lasting for the
agglomerates to be broken down to the extent required. Once this has happened, improvement of
mixing quality depends on the effective strain.
We currently lack theories describing the basic mechanisms relating to effective strain and agitation
intensity and the influence in general of these parameters on mixing processes. It is difficult on the
basis of a literature review to draw general conclusions regarding the extent to which various pa-
rameters influence different mixing processes. Hence there is no alternative to carrying out trials for
each mixing process and application.
Distribution
Distribution is the process whereby the broken-down agglomerates are distributed randomly
through the liquid phase. This normally takes place simultaneously with deagglomeration. If the
mixing process has not been designed so that deagglomeration and distribution take place concur-
rently, long mixing times may be expected to be required. Thus it is important that the solid phase
is distributed through the largest possible volume at the incorporation stage. The distribution proc-
ess works best if the mixture has low viscosity.
Stability of dispersion
When the powder is wetted, the agglomerates broken down, the particles separated and evenly dis-
tributed through the liquid, this state must be maintained and flocculation avoided. In many mixing
processes this is a difficult problem. However, if the liquid is highly viscous and the powder of low
density, stability of the dispersion is not a major problem, and it is therefore not discussed further in
this thesis.
Deep Mixing 31
3. Mixing equipment
3.1 INTRODUCTION
According to the state of aggregation of materials, Šterbácek & Tausk (1965) distinguish between
the mixing of gases, liquids and solids. The mixing of liquids can be further divided according to
their consistency into the mixing of Newtonian liquids and the mixing of pastes, i.e. non-Newtonian
liquids with high solids content.
According to Šterbácek & Tausk (1965) mixing equipment can be divided into four groups:
1 equipment for mixing gases;
2 equipment for mixing mobile liquids (Newtonian liquids);
3 equipment for mixing pastes and other non-Newtonian substances;
4 equipment for mixing solid loose substances.
The wide range of mixing equipment available reflects the enormous variety of mixing duties re-
quired in the chemical, paint, food, pharmaceutical and other industries. The present chapter deals
with some of the types of equipment available, mainly in the process industries, for dispersion and
dissolving in liquids and for the mixing of pastes and plastic materials. The types of mixing equip-
ment discussed in this chapter span a rather broader range than those applicable to deep mixing.
Equipment types relevant to the latter application are discussed in Chapter 5.
The literature reviewed and the illustrations reproduced here were in some cases published over
30 years ago – in particular, Šterbácek & Tausk (1965). The reason is that this work contains an
excellent review of the basic designs of mixing equipment. More recent literature in many cases
adds nothing new, apart from the development of static mixers and special forms of e.g. propellers
and turbines.
3.2 DISPERSION AND DISSOLUTION IN LIQUIDS
General
There is an enormous variety of equipment for mixing different substances with liquids. Fields of
application include the chemical, food, paint and pharmaceutical industries. A rapid mixing process
requires effective motion achieved by convection. The motion of the material should be sufficiently
intense to generate turbulence and turbulent diffusion. Regions of high turbulence are suitable for
dispersive processes. The mixing of low-viscosity liquids is dependent on (Šterbácek & Tausk
1965):

1 degree of turbulence
2 rate of circulation
Most existing impellers transmit forces and momentum to the material by shear stress, i.e. at right
angles or parallel to the motion of the impeller (Šterbácek & Tausk 1965). Paddles, turbines and
propellers are examples of impellers belonging to this group.
There are various ways of imparting motion to the materials being mixed depending on their rheo-
logical properties. Concentrated suspensions generally have very complex rheology and may have
both elastic and yield properties. Impellers are operated at different speeds depending on the char-
acteristics of the materials to be mixed, see Table 3.1.
Table 3.1. The applications of different impeller types depend to a great extent on the
viscosity of the mixture (after Edwards & Baker 1992).
Propeller
Turbine
Paddle viscosity increases speed increases
Anchor
Helical ribbon
Helical screw
Propellers, turbines and paddles

Propellers, turbines and paddles are generally used for mixing relatively low-viscosity liquids and
are operated at high rotational speeds, but they can also be used for mixing rheologically complex
materials. When they operate in a tank they are usually of a diameter 1/4 to 2/3 that of the tank.
Impellers can be distinguished according to the flow pattern they produce: tangential (paddles),
radial (turbines), axial (propellers), and compound flow, see Fig. 3.1.
Predominantly tangential flow is produced by paddles with straight blades and by broad blade pad-
dles rotating at a speed such that no radial flow is caused by centrifugal force. Radial flow is pro-
duced by turbines and purely axial flow by propellers. Compound flow occurs, for example, with
paddle mixers having tilted blades.
Deep Mixing 33
Fig. 3.1 Tangential, radial and axial flow (Štìrbáèek & Tausk 1965).
Propellers
The commonest type of propeller is the marine type, often with three blades. Propellers work like
high-speed pumps. Mixing is dependent on the motion of the fluid and hence the propeller's pump-
ing capacity is critical.
With propellers mixing is achieved by the combination of two movements:

1. axial flow from the propeller;
2. helical turbulent flow.
Propellers can be used for mixing liquids with viscosities up to 2 Pa·s (Šterbácek & Tausk 1965).
They are suitable for rapid mixing, e.g. of low-viscosity emulsions. Propeller mixers are unsuitable
for the suspension of heavy or rapidly settling substances and for the absorption of gases.
Radial propellers or axial turbines have a compound axial-radial flow. They consist of 4 – 16 flat or
profiled blades mounted on a shaft at an angle of 10 – 25° with respect to the plane perpendicular to
the axis of the shaft. Unlike ordinary propellers, the blades end in plane faces, see Fig. 3.2, which
are the cause of the radial flow. These are high-speed mixers and are suitable for suspensions with
viscosities up to 20 Pa·s (Šterbácek & Tausk 1965).
Fig. 3.2 Radial propeller mixer (Štìrbáèek & Tausk 1965).

Turbines
Turbines have vertically mounted blades and are of numerous different designs. They have been
used as mixers for many years.
The turbine blades entrain and shed the fluid in a predominantly radial flow from the blades. Hence
the shape of the mixing vessel is of great importance for the performance of the turbine.
The main advantage of turbine mixers is that they can be used over a wide range of viscosities and
densities. There is no definite upper limit to the viscosity that can be mixed. Once turbulent flow
has been achieved in the fluid (Re>≈100), viscosity is of little significance for the power consump-
tion (Ulbrecht & Patterson 1985, Edwards et al. 1992).
The shape of turbine blades is dictated by the character of the flow and the purpose of mixing. For
mobile liquids, vertical, flat blades are used. The blades may be tilted to improving pumping effi-
ciency. When mixing high-viscosity substances it is advantageous to tilt the blades in the direction
opposite to the direction of rotation. If the blades are tilted or profiled, the initial momentum will be
less, which aids the transfer of energy from the blades to the fluid. Blades may be curved to reduce
the flow away from the impeller and make the turbine easier to start up.
Turbines have a wide range of application including suspension, dissolving, chemical reactions,
absorption of gases and heat transfer. They are less suitable, but can be used, for pastes and doughs.
Fig. 3.3 Typical propellers, turbines and paddles (King 1992).
Paddles
Rectangular impeller arms, perpendicular or oblique to the axis of the shaft and driven by mechani-
cal energy, fall under the heading of paddles. Paddles and turbines are distinguished from each oth-
er by the flow patterns they produce. With turbines, radial flow must dominate, with possibly some
tangential flow, see Fig. 3.1.
Deep Mixing 35
Paddles have been known as a mixing device for a long time. Their main advantage is their simplic-
ity and consequent low cost. Their main disadvantages are their small pumping capacity, their re-
striction to liquids with viscosities up to ~1 Pa·s, and the limited axial flow from the blades (Šter-
bácek & Tausk 1965). Good blending is attained only in thin strata of liquid in the immediate vicin-
ity of the impellers, where local eddies form on the upper and lower edge of the impeller.
When paddles are used for mixing a concentration gradient (stratification) often results in the fluid.
This is a disadvantage, particularly when one of the components is added during stirring. This can
be remedied by tilting the paddle blades by 30-45° to the axis of the shaft so that axial flow occurs.
A combination of paddle and propeller can take the form shown in Fig. 3.4. Having the blades
pitched in two directions produces both upward and downward axial flows, which is useful if circu-
lation of the fluid is desired.
Fig. 3.4 Impeller with paddles pitched in two directions to produce circulation of the fluid
(Ulbrecht & Patterson 1985)
Broad paddle blades do not have these disadvantages because they generate predominantly tangen-
tial flow. These mixers are especially suitable for processes requiring uniform concentration, e.g.
chemical reactions (Šterbácek & Tausk 1965). Holes may be drilled in the blades to improve shear-
ing action. Disadvantages of broad paddles are that they are unsuitable for suspensions and for mix-
ing viscous materials.
Since paddles predominantly act by tangential flow, problems may occur when the material and the
paddle take on the same velocity gradients, i.e. the material is dragged around by the paddle with-
out any appreciable degree of mixing taking place. This problem can be avoided by using several
paddles rotating in different directions and/or at different speeds. When mixing viscous materials in
vessels, the equipment may be designed as in Fig. 3.5. Two eccentrically arranged shafts, each with
several paddles, rotate in opposite directions with the paddles overlapping. The fluid is thus kept in
constant motion in continuously changing directions. This type of mixer places motion of the fluid
ahead of high shear stresses.

Fig. 3.5 Mixer for pastes and suspensions. Two eccentrically arranged shafts with
several paddles rotate in opposite directions. (Štìrbáèek & Tausk 1965)
Equipment for dispersing solid particles

The problems of dispersing solid particles in a liquid phase can vary greatly in difficulty and com-
plexity. The dispersion process and its problems are described in Chapter 2.6.
There are numerous types of mixing equipment that can be used for dispersing powders in liquids,
but only a few have been developed that are used specifically for dispersion (McDonagh & Hey-
wood 1986). These categories of equipment are described briefly below.
i Vertical shaft mixers. Mixers using impellers mounted on a vertical shaft can vary widely in
design and speed. Propellers, paddles, turbines etc. are used, and impellers can often be substi-
tuted and combined to obtain different mixing mechanisms at different stages of the process.
ii Horizontal shaft mixers. When mixing high-viscosity suspensions, horizontal mixers are often
used to produce movement in every part of the vessel. Paddles, screws, kneaders or Z-blades
may be used. Z-blades are described further in Chapter 3.3.
iii Mills. These are useful if the mixture contains agglomerates that are difficult to break down.
Mills such as colloid mills, roller mills and ball mills can force the material into motion and pro-
duce high shear stresses.
iv Homogenizers. These are often used for low-viscosity materials in which there is no need for
high shear stresses to break down agglomerates.
Deep Mixing 37
v Special powder-liquid contacting devices. A number of mixers have been developed which are
specially suited for dispersion duties where the solid is difficult to wet. They can be divided into
types where the liquid phase is sprayed into a falling stream of powder and types where the solid
falls into thin liquid films. Both types produce a large interfacial area between the materials, thus
avoiding large lumps of partially wetted particles. An example of such a mixer is shown in
Fig. 3.6.
Fig. 3.6 Wetting equipment. Powder is introduced into a fluid surface by e.g. a liquid.
(McDonagh & Heywood 1986)
When designing or improving a dispersion process it is important to be able to describe and under-
stand the individual stages of incorporation, wetting, disagglomeration, distribution and stabiliza-
tion, see Chapter 2.6. The stages of the dispersion process may take place separately or concurrent-
ly; some may not occur at all. It is important to evaluate the role that each stage plays in each spe-
cific mixing application. Based on a description of the predominant mixing mechanisms and the
difficulties of the dispersion process, effort can be focused on the most important stages of the proc-
ess. Because of the generally very complex mixing processes, laboratory tests may be necessary to
determine where the difficulties are (McDonagh & Heywood 1986).
Static mixers
Examples of static mixers are stationary or rotating units inserted in pipes, the components to be
mixed being forced through the pipe (Godfrey 1992).
There are numerous types of static mixers and the range of applications is very wide, including both
laminar and turbulent processes. Static mixers are often used to improve existing mixing processes,
a static mixer being added for pre- or post-mixing.

Comp. A
Comp. B
Scale of segregation of mixture
reduced to the required level
Fig. 3.7 Principle of an in-line static mixer.
The predominant mixing mechanism in laminar mixing is distributive, by splitting/slicing and fold-
ing/reorientation of the elements. This process reduces the volume of the individual regions of non-
uniformity. The number of elements that are mixed depends on the difficulty of the mixing process.
When dispersing solid particles in a liquid it is important that the mixer generates the shear forces
required to break down agglomerates.
As with other types of mixing equipment, mixing quality is difficult to determine, making it diffi-
cult to compare different types of device.
Jet mixers
Jet mixing is employed for both gases and liquids. The principle is the same in both cases. The use
of turbulent jets for mixing liquids is very common in the chemical industry.
Jet mixing is suitable for the mixing of mobile liquids, when it is generally combined with a pump.
Turbulent jet mixers can give very short mixing times of seconds or less. Several nozzles may be
used for more effective mixing, and the method is often combined with mechanical mixing.
The fluid discharging from a nozzle assumes the shape of a cone enlarging with the distance from
the orifice, see Fig. 3.8. This is caused by the following circumstances (Šterbácek & Tausk 1965):
Region of flow Region of

development developed flow
Fig. 3.8 A jet and its flow behaviour (after Štìrbáèek & Tausk 1965).
Deep Mixing 39
1. The flow of fluid displaces the fluid in the vessel in front of the orifice. The fluid adjacent to the
jet is entrained into a motion parallel to the flow through the transfer of momentum from the jet
to the surrounding fluid. This transfer of momentum is effected by the shear stresses between the
fast-moving and the stagnant layers and by the turbulent penetration of elementary particles
from the jet into the surrounding fluid. If we neglect the internal friction, the kinetic energy is
constant, i.e.:
∑M v = ∑M v
0 0 1 1
where M0 and v0 are respectively the mass and the velocity of the jet stream and M1 and v1 are
respectively the mass and the velocity of the entrained fluid. The velocity of the entrained fluid
is smaller than the velocity of the stream itself, and each layer of the fluid set in motion entrains
the adjacent layers. The cross section of the moving stream increases with distance from the
nozzle and its velocity decreases towards the periphery of the cone.
2. As the surrounding fluid is entrained into the stream a decrease in pressure is produced in its
original place. This decrease causes further fluid to be sucked into this space and likewise set in
motion by momentum transfer from the jet.
The entrainment of fluid into the moving stream results in a turbulent boundary layer at the periph-
ery of the cone in which active mixing takes place. The thickness of this turbulent boundary layer
increases proportionally to the distance from the nozzle, and this consumes some of the momentum
in the core of the jet. The gradual increase in the thickness of the boundary layer leads to an en-
largement of the cone and a constriction of the core, which finally disappears at a certain distance
from the nozzle.
3.3 MIXING OF PASTES AND PLASTIC MATERIALS
General
There are numerous types of mixers and processes for mixing solids and pastes. The mixing of plas-
tic materials forms a transition between liquid mixing and the mixing of solids. Commonly a mixer
is versatile and can be used not only for viscous liquids or thin pastes but also for low-viscosity
liquids. The same applies to certain mixers adapted for pastes which can also be used to mix solids.
Thus there are no clear boundaries between different types of mixing equipment. Other mixers
break down solid agglomerates in pastes and plastic masses and thus have the character of mills
(Šterbácek & Tausk 1965).
Mixers for pastes and plastic materials are found especially in the chemical industry (rubber prod-
ucts, plastics, ceramics, drugs, soap etc.) and in the foodstuff industry.
Mixing pastes and plastic materials is considerably more difficult and complex than mixing liquids.
The higher the consistency the slower is the flow of material, and hence the potential for turbulent

flow is limited. Mixers are generally designed so that the impeller passes through the entire volume
to be mixed, since it is difficult to produce mass flow throughout the volume.
Mixers for high-viscosity materials generally have very complex flow patterns. There are regions
with high shear forces (where the material is stretched) and regions where the material is redistrib-
uted (folded). It is generally very difficult to analyse these mixing processes quantitatively in terms
of idealized mixing mechanisms.
For most of the above equipment types, with the exception of mills and kneaders, the primary flow
pattern is tangential, i.e. the flow follows the mixing element as it rotates about its axis. Helical
mixing elements, steeply tilted paddles and propellers also generate an axial flow component, pro-
ducing a helical primary flow. Centrifugal forces generated by tangential flow create a secondary,
radial flow. In a viscoelastic material this secondary flow is counteracted by elastic forces.
Anchor agitators
The most common type of equipment for mixing pastes is paddle mixers designed so as to secure
movement in mixtures of high consistency. Anchor agitators are the most common mixer of this
type. The majority of these mixers have vertical arms in addition to horizontal arms to promote
vertical movement of the mixture, which is difficult to transmit e.g. in plastic materials. Anchor
agitators produce large circulation patterns. In general the impeller has a diameter of the same order
as that of the vessel and is situated close to the bottom. Fig. 3.9 shows some typical anchor agitators
for mixing in vessels.
In order to produce eddies throughout the fluid or a number of regions of shearing, the mixing ele-
ments may be formed as rakes. These are particularly suitable for mixing plastic materials. To in-
crease the shearing action they can be provided with two sets of arms. Either both impellers may
move in opposite directions or one impeller may be fixed while the other rotates. See Fig. 3.10.
Anchor agitators are particularly suited for combination with horizontal arms. One popular combi-
nation has an anchor and paddles with double motion, see Fig. 3.11. The paddles may be vertical or
inclined. This combination permits the two mixers to move in opposite directions and at different
speeds. This design is particularly suitable when the mixture has a relatively low initial viscosity
which develops into pastes or plastic materials on account of reactions. During the first stage mix-
ing is carried out by the paddles, and as thickening proceeds the anchor takes over.
Fig. 3.9 Some typical anchor agitators. (Štìrbáèek & Tausk 1965)
Deep Mixing 41
Fig. 3.10 Anchor agitator with two sets of arms, one rotating and the other stationary.
(Štìrbáèek & Tausk 1965).
Fig. 3.11 Anchor combined with paddles. (Štìrbáèek & Tausk 1965)

Helical mixers
To improve circulation in a mixture helical ribbons or screws may be used. These mixers operate at
a slow speed and do not generate high shear rates, relying instead on their pumping capacity to
reach every part of the vessel. Helical mixers are well suited for mixing materials with high yield
points and viscosities, since they produce axial and tangential movements while at the same time
cutting through the material. This is an advantage when mixing materials that tend to stick to the
mixer blades.
Helical impellers are considered more effective for mixing viscoelastic materials than anchor or
multi-paddle impellers (e.g. Nagata et al. 1972). The reason is that when e.g. paddles rotate in a
viscoelastic material the energy consumption is high in the immediate vicinity of the blades, but
forces cannot propagate through the mixture because of its viscoelastic behaviour. As a result, mix-
ing takes place near the blades, but just outside this region the mixture does not move at all. Viscoe-
lastic behaviour is discussed in Chapter 5, Rheology.
Fig. 3.12 Helical mixers (Parfitt & Barnes 1992, Silvester 1985)
Helical ribbons are particularly used in horizontal mixers where recirculation of the flow is desired.
This is achieved by combining two ribbons operating in opposite directions. These mixers are par-
ticularly suitable for mixing loose solids but are also used for mixing liquids with e.g. powders. The
mixing efficiency is increased by driving the material to the sides, after which it falls back into the
main flow. Helical screws are used to produce strong axial flow in high-viscosity materials.
Deep Mixing 43
Mixers can also be designed with paddles arranged in the form of a helix. The material is forced out
to the sides as well as along the tank. These mixer types are used e.g. for mixing paints (Šterbácek
& Tausk 1965). Fig. 3.13 shows a mixer of this type used for mixing soil and cement for sub-grade
soils.
Fig. 3.13 Helical mixer with paddles for mixing soil and cement (Mueller 1991).
Kneaders
Kneaders, Z-blade and sigma-blade units are generally used for mixing high-viscosity liquids,
pastes, rubbers, doughs etc. which cannot be handled by anchor impellers or helical ribbons. The
clearances between the blades and between blade and vessel are small in order to produce high
shear stresses in the material. The blades cut through the entire vessel volume. These mixing ma-
chines generally have the form of horizontally mounted tanks.
The mixing procedure for the dispersion of solids is usually as follows. First a part of the liquid is
put in the tank. With the mixing device running, the solid components are then added with suffi-
cient liquid to form a paste. In this first phase the liquid components are merely mixed with the
solids. During the course of the kneading the agglomerates in the paste are broken down. High
shear stresses occur between the blades and between the blades and the walls. It is important that
the paste consistency be sufficient to generate the necessary shear stresses. High shear stresses are
also necessary to aid wetting.
Two paddles may be used which rotate in opposite directions, arranged so that their paths overlap.
Common shapes of paddles are shown in Fig. 3.14.

Fig. 3.14 Various types of kneaders (Štìrbáèek & Tausk 1965)
a: Mainly used in the chemical industry
b: Used for thin pastes
c: Suitable for the dispersion of paints and pigments in plastic
materials and for the dispersion of water in rubber
d: Used for kneading very stiff materials in small quantities
e: Used for kneading very stiff materials in large quantities
f: Used for homogenizing fibrous materials
g: Suitable for kneading rubber
Deep Mixing 45
Mills
In many applications involving the dispersion of solids, conventional mixers are unable to achieve
high enough shear forces in the material to break down agglomerates. This problem may arise if
e.g. the components being mixed have widely differing rheological properties. Such applications
may be handled by means of mills, in which dispersion is achieved by shearing and crushing of the
agglomerates.
Fig. 3.15 shows the theoretical ranges of application of various types of mill and their principal
mixing mechanisms, i.e. whether the solids are dispersed by crushing or shearing.
Fig. 3.15 Crushing and shearing in dispersion equipment (Edwards & Baker 1992).

4. Mixture quality
4.1 INTRODUCTION
The assessment of mixture quality is central to any investigation of mixing. The measurement of
degree of mixedness or required mixing time requires the measurement of mixture quality. Similar-
ly in the study of mixing mechanisms it is necessary to be able to describe the state of the mixture.
When a mixture has some industrial application and its properties depend on how well mixed it is,
the mixture quality is an essential part of the product specification (Schofield 1974).
When is a mixture well mixed? Traditionally the answer can be that a mixture is well mixed when it
is good enough for its purpose. This does not generally mean that it is homogenous. On a sufficient-
ly small scale, practically all mixtures are heterogeneous. Thus the term homogeneity is difficult to
define. Numerous researchers have discussed and written on the concept of homogeneity or the
perfect mixture. Poux et al. (1991) review some twenty articles on the subject. One example of a
definition is offered by Fan et al. (1970), who define a homogeneous mixture as one in which the
content of all constituents is uniform in every part of the mixture. The concept of homogeneity is
generally unsuitable and inadequate to characterise e.g. suspensions, mixtures of solids, pastes and
other complex mixtures. Indeed, it means different things in different process industries. Thus the
concept is only useful when associated with a suitable scale.
Many criteria have been developed to assess mixture quality (Nienow et al. 1992). This is under-
standable given the very broad application of mixing processes, such as the formation of emulsions,
solid dispersions, dry powder mixing etc. It is unfortunate that no universally applicable definition
of mixture quality exists.
The concept of mixture quality arises in many contexts. Among the simplest to consider are mix-
tures of two dry solids. Here mixture quality is used to evaluate the outcome of mixing (e.g. con-
centration variances). Mixture quality is more difficult to define when e.g. a chemical reaction is
involved. The mixing process is then often evaluated on the basis of factors, such as the product's
rheology, compressive strength, colour, taste, temperature etc.
When assessing mixture quality it is necessary to distinguish between the distribution of the compo-
nents in the mixture and special characteristics. Many factors are involved besides the efficiency of
the mixing process. When assessing the quality of a process it is important to focus on the most
important results for the product. To assess the effectiveness of the operation it may be necessary to
use a number of indirect measurement methods to assess a specific result, as mixing equipment may
perform several functions simultaneously, e.g. breaking up agglomerates and distributing solids.
The purpose of this chapter is to provide an overview of methods for assessing the homogeneity or
degree of mixedness of a mixture. The theories are mainly taken from scientific journals dealing
Deep Mixing 47
with the process industries. The types of mixture considered are mixtures of liquids, solids, and
combinations of different phases.
Chapter 7 presents a study which evaluates mixture quality in a Swedish ground improvement
project using the dry deep mixing method. In this study the dispersion of binder in the stabilized
soil is considered as the basis for the evaluation of mixture quality.
4.2 DESCRIBING THE QUALITY OF A MIXTURE

Mixing is a random process in which materials are systematically distributed and rearranged (Som-
mer 1981). The resulting distribution of particles or components can be described statistically. Gen-
erally an expression for the degree of mixedness is intended as an indication of the variations exist-
ing in the mixture. For example, the coefficient of variation can be used as an expression for mix-
ture quality. After a very long mixing time the concentration will be uniform throughout the entire
volume.
A sample taken from a mixture can only approximately reflect the distribution of the components of
the mixture. The more samples are analysed, the better the approximation will be. When the
number of samples analysed is small, the error assumed in the statistical analysis will be large. This
error is expressed as a confidence interval defining the bounds within which the statistical value
lies. The size of the confidence interval depends on the number of samples analysed and on the
statistical reliability.
An example from Rielly et al. (1994) is reproduced below to illustrate the difficulties of character-
izing a mixture. Fig. 4.1 shows the difference between a non-random and a random mixture of par-
ticles. Both mixtures are divided into 36 parts or sub-volumes each containing 16 white and 16
black particles. For a mixture to be homogeneous, the composition of each sub-volume must be the
same. Fig. 4.1b shows a homogeneous, non-random distribution of 50 % white and 50 % black par-
ticles. The probability of a mixing process distributing the particles in this way is very small. It is
more likely that the best a mixing process can achieve is a distribution such as shown in Fig. 4.1a,
where each sub-volume consists of different numbers of white and black particles. An analysis of
Fig. 4.1 a) Random mixture of particles. b) Non-random mixture of particles.

(After Lacey 1943 and Rielly 1994)

just one sample of this mixture would give a result indicating that the white and black particles
were not fully mixed. However, further random mixing would not result in any improvement.
Complete mixing could be defined as the state where all samples extracted from the mixture contain
the same proportions of components, or the same properties, as the whole mixture. On this basis,
the mixture in Fig. 4.1b shows the complete mixing of two components. Such a mixture is not pos-
sible to realize in practice. The mixture in Fig. 4.1a does not consist of eight white and eight black
particles in every sub-volume, although the particles are randomly distributed. This example shows
that it is not possible to define complete mixing without considering variations in composition or
properties and the size of the sample.
4.3 SCALE OF SCRUTINY

The quality of a mixture depends on the scale on which it is examined (Lloyd 1981). A mixture,
e.g. a suspension, may appear to be homogeneous when viewed by eye. When observed under the
microscope, however, it becomes apparent that the particles are not evenly distributed. Thus the
degree of homogeneity can only be determined once a suitable scale of scrutiny has been estab-
lished. The appropriate scale of scrutiny itself depends on the application of the product. In mix-
tures where molecular diffusion is involved as a mixing mechanism, it must also be remembered
that mixing may continue even though mechanical mixing has stopped. In such a process mixedness
improves with time, and hence the time at which the mixture quality is assessed will influence the
result. Thus the size of the samples taken from a mixture must be related to the mixing process and
to the end use of the product. In practice the appropriate scale of scrutiny has to be estimated, how-
ever the concept is a useful one for defining mixture quality. An established scale of scrutiny limits
the size or volume of samples to be analysed in order to assess the quality of a mixture.
4.4 SCALE AND INTENSITY OF SEGREGATION

An objective definition or measurement of mixture quality must describe the distribution of the
components in the mixture and how the interfacial area between the components has progressed
(Schofield 1974).
Danckwerts (1952, 1953) proposed two criteria for mixing, which can be discussed qualitatively
and quantitatively. The first is the scale of segregation, which is a measure of the size of unblended
lumps of pure consistency and is a measure of average size. As lumps are displaced and broken
down, the scale of segregation is reduced. The other criterion is intensity of segregation, which de-
scribes the effect of molecular diffusion in mixing. It is a measure of the difference in concentration
between neighbouring lumps in the flow. See Fig. 4.2.
Consider a mixture of two liquids A and B. The concentrations of the liquids at any point are denot-
ed by a and b respectively. The mean concentrations of the liquids in the whole mixture are a and
b respectively. Thus:
a + b = 1, a +b =1 (4.1)
Deep Mixing 49
Scale of segregation Decreasing →
Intensity of segregation Decreasing →
Fig. 4.2 Scale and intensity of segregation. (after Poux et al. 1991)
Thus for a mixture consisting of liquid A only, the concentrations will be a = 1 and b = 0 . The
coefficient of correlation can be defined as
(a1 − a )(a2 − a ) (b1 − b )(b2 − b )

R(r ) = ≡ (4.2)
(a − a ) 2 (b − b ) 2
where a1 and a2 are the concentrations of liquid A at two points, 1 and 2, a distance r apart. The
coefficient of correlation R(r) provides information about the average of the product of the concen-
tration differences from the mean, at position 1 and 2. The denominators are called the variance of
a (or b) and are sometimes written as σ a2 or σ b2 (which are equal). The coefficient of correlation
R(r) normally assumes values between 0 and 1.
According to Danckwerts (1952) there are two ways of measuring the scale of segregation in a mix-
ture, one based on distance and the other on volume. The length scale of segregation S is defined as
∞
S = ∫ R(r )dr (4.3)
0
The length scale of segregation S is the area under the curve R(r) in Fig. 4.3. The volume scale of
segregation V is defined as 2π times the area under the curve r2 R(r),
∞
V = 2π ∫ r 2 R(r )dr (4.4)
0

1
Coefficient of correlation R(r), (–)

0.8
0.6
0.4
0.2
0
0 0.5 1 1.5
Dimensionless distance r (–)
Fig. 4.3 Coefficient of correlation of a mixture. (after Danckwerts 1952)
The scale of segregation is a useful concept for mixtures of coarse materials in which individual
particles remain separate with no chemical bonding between them. However, the concept gives
little information about the intensity of segregation. In mixtures of fine-grained materials molecular
diffusion must be considered.
In order to allow for molecular diffusion in a mixture Danckwerts (1952) presented a second defini-
tion of degree of mixedness based on the concentration variance of a component of the mixture.
The mean concentration and the variance of component A of a mixture are written as
1 n
a= ∑ ai
n i =1
(4.5)
1 n
σ2 = ∑
n − 1 i =1
(ai − a ) 2 (4.6)
where ai is the concentration of A in sample i and n is the number of samples. The variance σ2 de-
scribes the variation of the concentration of A in the mixture, but it represents only an estimate of
the true variance. For a completely segregated mixture the variance is (Lacey 1943):
σ 02 = a b = a (1 − a ) (4.7)
The intensity of segregation is defined as (Danckwerts 1952)
σ2 σ2
I= 2 = (4.8)
σ 0 a (1 − a )
Deep Mixing 51
According to the concept of intensity of segregation, I = 1 for a completely segregated mixture and
I = 0 for a homogeneous mixture. The effects of changes in scale of segregation and in intensity of
segregation are shown in Fig. 4.2. As the intensity of segregation decreases the mixture becomes
more diffuse, while a decrease in the scale of segregation reduces the size of the mixed particles,
making the mixture more homogeneous.
In systems involving chemical reactions, mixing on the microscopic scale is important, as otherwise
the reaction will only take place on the surfaces of lumps in the mixture. In e.g. jet mixing the scale
of segregation is reduced by turbulent motion, while the scale of segregation is reduced by molecu-
lar diffusion. When solid particles are added, their particle size will influence the scale of segrega-
tion and how it is reduced.
In laminar mixing of two materials the striation thickness can be a measure of mixture quality
(scale of segregation) as shown in Fig. 4.4. The striation thickness can be related to the efficiency
of molecular diffusion between the two materials (intensity of segregation) so as to give an idea of
the requisite mixedness for the specific application.
Thus the outcome of mixing depends on the scale of scrutiny adopted. If a coarse mixture will give
satisfactory results, molecular diffusion will play a minor role. Molecular diffusion is a slow proc-
ess and difficult to bring about effectively in the laminar mixing of viscous media, e.g. thick sus-
pensions.
thickness
of streak
minor component
striation thickness
major component
Fig. 4.4 Striation thickness as a measure of mixture quality. (Edwards 1992)

4.5 MIXING INDICES
Based on the concepts of scale of segregation and intensity of segregation according to Danckwerts
(1953), numerous mixing indices have been developed. Most mixing indices are mainly useful for
the mixing of solids. Some forty different mixing indices have been tabulated by Poux et al. (1991)
and comparisons between them have been made by Fan and Wang (1975), Fan et al. (1979), and
Rielly (1994).
Danckwerts' definition of intensity of segregation assumes that a perfect mixture has a concentra-
tion variance equal to zero. However, a small concentration variance will still exist because the
number of particles in a sample taken from a mixture is finite. For a completely random (stochastic)
mixture of equally sized particles the concentration variance is (Lacey 1954):
a (1 − a )
σ R2 = (4.9)
np
where np is the number of particles in each sample and a is the mean fraction of component A in
the mixture.
In mixtures of any type (Poux et al. 1991)
σ 2R > σ 2 > σ 20 (4.10)
Assume that in an initial stage the binary components are completely segregated with a concentra-
tion variance σ 02 . Lacey (1954) here defines a mixing index as
The mixing that has occurred

M=
The mixing that can be achieved
σ 02 − σ 2 (4.11)
= 2
σ 0 − σ R2
According to Lacey's mixing index a fully segregated mixture has a value of zero and a fully ran-
dom mixture a value of one. When mixing liquids the concentration variance σ R2 of a fully random
mixture may be assumed to be zero and we obtain a mixing index defined as (Miles 1962):
σ2
M = 1− 2 (4.12)
σ0
Deep Mixing 53
Miles' mixing index is relatively insensitive to different mixture qualities since a poorly dispersed
mixture will have a concentration variance closer to σR2 than to σ02. That implies that this index
generally lies within the range 0.75 – 1.0 regardless of the degree of dispersion. A more sensitive
mixing index is defined as (Rose & Robinson 1965):
σ
M = 1− (4.13)
σ0
A number of similar mixing indices have been presented in the literature and are tabulated by Poux
et al. (1991). Some of these are shown in Table 4.1. The common feature of these mixing indices is
that they are mainly applied to mixtures of solids. The analysis is developed on the assumption that
the mixture is a binary system of particles of equal size and density. Practical mixtures consist of
particles that differ in size and density, which means that the concentration variance of a fully ran-
dom mixture σR2 will vary depending on whether the analysis is based on numbers of particles or
weights of particles.
Table 4.1 Mixing indices for binary mixtures based on statistical analysis.
(after Poux et al. 1991)
Author(s) Expression Fully segregated Fully mixed

σ = σ0 σ = σR
σ 0 −σ
Lacey (1943) M= 0 1
σ 0 −σ R
σ 02 − σ 2
Lacey (1954) M= 0 1
σ 02 − σ R2
σ2
Miles (1962) M = 1− 0 1
σ 02
σ
Rose & Robinson (1965) M = 1− 0 1
σ0
ln σ 02 − ln σ 2
Ashton & Valentin (1966) M= 0 1
ln σ 02 − ln σ R2

A number of mixing indices have been developed which are not based only on the definitions of σ,
σ0 and σR. Indices have also been presented which are based on experimental work. A number of
such mixing indices are presented in Table 4.2.
Table 4.2 Mixing indices (after Poux et al. 1991)
Author(s) Expression Comments
Legatt (1951) M=
∑ (x − x) i
2
xi − x
Izykowski (1956) M=
nx
Coulson & Maitra (1950) M = 1− X X : percentage unmixed
Lastovtsev et al. (1962)

M =
∑ (C i − C 0 ) 2 ni C0 : concentration of constituents for an ideal
mixture
C 02 (n − 1)
Ci : concentration of constituent i
n : number of samples
ni : number of samples at concentration Ci
Adams & Baker (1956) σ 2

=
∑ (x i − x )2
n −1
2
1 n  xi (t ) 
∑  − 1
Iinoya et al. (1985) xi(t) : volume fraction or concentration of
M (t ) =
n − 1 i =1  x∞ 
specified particles detected by the ith probe at
time t
x∞ : expected value in the final homogeneous
mixture
1
Gray (1957) σ2 =
N −1
∑ (I − I )2 I : probe meter reading
I : average probe meter reading.
In the context of chemical stabilization for sub-grade soils a measure of the efficiency of mixing is
used (Assarsson 1972, Sherwood 1993). Samples of field-mixed stabilized soil are divided into two
equal portions, where one is re-mixed in a laboratory mixer, and then preparing strength test speci-
mens from both halves. The mixing efficiency is given by
Deep Mixing 55
Strength field mixed material
Mixing efficiency = ×100% (4.14)
Strength re − mixed material
The efficiency of mixing can also be assessed by comparing the strength of laboratory prepared
samples with the strength of samples taken from the site. The efficiency of mixing is expressed as
Kh (Kézdi 1979):
σ u , field
Kh = × 100% (4.15)
σ u ,lab
where σu,field is the 7-day strength of samples stabilized and compacted in the field and σu,lab is the
7-day strength of samples stabilized and compacted in the laboratory under similar conditions. Ac-
cording to Baghdadi & Shihata (1999), Kh is usually in the range 60 – 80 %.
4.6 MEASUREMENT OF MIXTURE QUALITY

The theoretical methods presented in sections 4.4 and 4.5 provide a basis for assessing the homoge-
neity of a mixture. Such an assessment should be based on the concepts of scale and intensity of
segregation or on some defined mixing index. The measurement of homogeneity requires sampling
and the result is influenced by e.g. the number of samples, the sample size, the measurement meth-
od and the analytic method.
The number of samples taken from the mixture is related to the analytical method used. The fewer
samples are analysed, the more uncertain will be the estimate of the homogeneity. If the number of
samples is small, the confidence interval of the average can be estimated to a known probability by
using the Student t-distribution (Johnson 1994). Published tables give the t-distribution as a func-
tion for the error probability. With a larger number of samples the value of t decreases and the per-
mitted probability increases. However, statistical analyses of concentration variance require a rela-
tively large number of samples. Poux et al. (1991) recommend taking 20–40 samples from a mix-
ture to obtain correct information on the homogeneity.
Samples must be taken at different points of the mixture to give a global view of the mixing proc-
ess. To provide a satisfactory view of the homogeneity, samples must therefore be taken according
to an established method depending on the mixing process. Suitable sampling methods may thus
vary from one mixing process to another. Harnby (1972) presented a sampling method based on
random sampling which guarantees that samples will be taken from representative points of the
mixture.
What is the influence of particle size on mixing quality? Poole et al. (1964) showed experimentally
that particle size has an effect on the coefficient of variation of the mixture. The experiments indi-
cated a linear relationship for a variety of mixtures. Bourne (1968) showed that the variance of a
mixture depends not only on particle size but also on particle shape. However, for complex mix-

tures of solids in which chemical reactions take place there are no published theoretical relation-
ships concerning the effect of particle size.
An estimate of homogeneity utilizing concepts of mixing index is based on the ability to estimate
concentration variances. On the assumption that sampling errors and analytical errors are independ-
ent of each other, the concentration variance of a mixture can be written as (Yip and Hersey 1977,
Orr 1979):
σ = σ mix
2 2
+ σ analysis
2
+ σ sampling
2
+ σ 2purity (4.16)
where σ2mix = the variance due to the mix

σ2analysis = the variance due to the analytic method
σ2sampling = the variance due to the sampling method and sample size
σ2purity = the variance due to the sample purity
The contribution of sampling method and analytic method to the total error can be substantial (Yip
and Hersey 1977, Orr 1979). When studying e.g. the effect of a mixing process on homogeneity it
is therefore important to be able to estimate the influence of the analytic method and the sampling
method and to ensure that samples are not contaminated.
There are numerous ways of measuring homogeneity in a process. Some parameters often measured
and used in the process industries to assess homogeneity and process conditions are (AIChE 1987):
• mixing time
• density
• particle size distribution
• agglomeration
• yield properties
• chemical/physical properties
• temperature
• pressure
• flow rate
• volume
• phase proportions
Mixing time is the time required for the mixture of constituents to reach a specified degree of uni-
formity. The mixture is then said to be mixed (Edwards 1992). The criterion is used especially in
the mixing of miscible liquids. When assessing the result of mixing it is important to give careful
consideration to the technique used to assess homogeneity. Often the mixing time is given by the
process (e.g. a retrieval rate of a given number of mm/revolution in deep mixing) and it can be
highly subjective. A mixing time is only meaningful if the mixture quality achieved at the end of
the process is well-defined.
Deep Mixing 57
Because of the difficulty of measuring and estimating the scale of segregation in the mixing of liq-
uids, it is common to use very simple statistics, comparing the concentration variance at a number
of points with the mean (reference value). One way is to measure the concentration of e.g. a tracer
material and study e.g. the change in the coefficient of variation over time, Fig. 4.5. Practical mix-
ing times can be measured by a number of techniques, e.g. acid-base indicators, variations in elec-
trical conductivity, temperature variations and light absorption techniques (AIChE 1979, Edwards
et al. 1992).
Mixer performance is difficult to characterize because a mixer often performs several functions
simultaneously. Mixers are generally assessed on the basis of the finished product, often for reasons
of practicality. Numerous criteria are used to define a well-mixed product or a state of uniform dis-
persion, and often they have no direct relation to the distribution of concentration or phase propor-
tions in the mixture. Practical methods for measuring the scale and intensity of segregation of well-
mixed samples are not part of ordinary routine analysis. Too often the mixing process is assessed on
the basis of properties of the product which are dependent on other factors besides the efficiency of
mixing. An example of this is the practice of assessing degree of mixedness on the basis of strength
parameters.
Fig. 4.5
The coefficient of variation as a function
of mixing time for one test point in a
mixing process (after Ries 1978).

In e.g. mixtures of viscoelastic materials, concentrated suspensions, mixtures of complex substanc-
es, and/or mixtures involving relatively rapid chemical reactions, it may be difficult to measure
properties during the actual mixing process.
A commonly used way of assessing a mixing process is to investigate the energy input per unit vol-
ume. In viscous fluids the viscosity and rotation speed have little effect on the number of revolu-
tions per unit volume required in order to achieve mixing (Godfrey 1985). Godfrey writes that the
mixing process depends only on the number of revolutions of the impeller. The relationship be-
tween mixture quality and mixing time is linear. Homogeneity is independent of the viscosity, but it
can be reduced by either pseudoplastic or viscoelastic behaviour. According to Godfrey (1985),
both power consumption and mixing time are important for homogeneity. The energy required to
achieve a specified homogeneity can be expressed as a product of power consumption and mixing
time as follows (Lindley 1991b):
Ew = P0 × t m (4.17)
where: Ew = total energy, Nm

P0 = power input, W
tm = mixing time, s
In mixing processes where the properties of the product are a function of the energy input per unit
volume it is important to carry out small-scale trials with similar geometry and thus study the mix-
ing process to scale.
When studying mixing results the documentation of conditions is important. Apparently minor fac-
tors can have major effects, e.g. the measuring apparatus used and where it is located. Measure-
ments of mixing result should show data of following types (AIChE 1979):
• operating conditions
• description of equipment
• miscellaneous equipment performance characteristics
• feed characteristics
• product characteristics
• pressure, temperature, flow, and rotational speed data
• power data
In order to permit the development of methods for the measurement and analysis of the homogenei-
ty of a mixture, it is important that all practical trials are very carefully documented. A study of the
development of mixing indices and of methods for the measurement of various parameters in the
process industries shows the importance of studies being published and discussed. The present liter-
ature review shows the amount of work and time that lies behind the knowledge available in the
process industries. Thus it is important that a forum for discussion (written and open to scrutiny) be
set up in the near future for the purpose of building up a system of suitable criteria for mixing and
reliable measurement methods for estimating the state of a mixture.
Deep Mixing 59
5. Rheologi
5.1 INTRODUCTION
In all processes in which different materials are being mixed it is important to know or at least have
a good idea of the rheological behaviour of the input materials as a basic prerequisite to an under-
standing of the mixing process. There is a very wide range of processes, and they vary in the degree
of precision with which the rheological data of the feed must to be specified in order for the product
to meet specifications. In some processes the inputs must have the exact rheological properties for
e.g. a chemical reaction to take place or for the correct circulation patterns to occur. In other proc-
esses it may be sufficient for e.g. the viscosity to be below a certain limit in order to guarantee cir-
culation or turbulent flow.
With knowledge of the rheological properties of the inputs to a mixing process we can determine
what mixing mechanisms are necessary in order for the product to have the desired properties.
Knowing the required mixing mechanisms, suitable mixing processes can be determined. The rheol-
ogy of the materials is thus the starting-point for all mixing process design.
Loose clays present engineering challenges because of the variation in their composition and prop-
erties. It is very difficult to determine the rheological behaviour of a clay e.g. in deep mixing. The
mere fact that the geology and the water content can vary within the same locality makes the rheo-
logical behaviour of the soil during the mixing process difficult to predict. The present chapter re-
views some of the available knowledge of the rheology of clays in an approach to a better under-
standing of mixing mechanisms in deep mixing.
Our knowledge of the rheological properties of suspensions consisting of combinations of clay, air,
lime and cement is very limited. This is partly because these rheological properties are not studied
and measured as part of the routine laboratory tests carried out ahead of deep mixing. The method
in use today in Sweden (Carlsten & Ekström 1995) for laboratory mixing of soil offers no way of
investigating its rheological behaviour during the mixing process. Since the equipment associated
with this method is invariably used in R&D projects, the result is that the rheology is very poorly
known. Another reason is that it is difficult to make quantitative measurements of rheological prop-
erties with the technology and the equipment available today. These problems will be discussed in
the present chapter.
5.2 RHEOLOGICAL PROPERTIES
Viscoelasticity
Clays are heterogeneous materials consisting of elastic and brittle solid components separated by

surfaces of discontinuity from the liquid and air which occupies the void spaces. Clay has very
complex rheological properties which are influenced by (Gillott 1987):
1. the physical properties of the water in close proximity to mineral surfaces, the amount of water
and the compositions of the solutions present;
2. the extent of aggregation and cementation of the minerals;
3. the nature of the minerals;
4. the arrangement and packing density of the minerals;
5. mechanical, physical and chemical interaction of the minerals with each other and with en-
trapped pore solutions and air.
An ideal plastic material stressed beyond its yield value undergoes continuous and permanent de-
formation. Above a certain water content clays show plastic properties, while dry clays show little
or no plasticity (Gillott 1987). The plasticity of a clay is affected by its mineralogical composition,
the nature of the ions exchanged, the crystalline size, the size distribution of the minerals, the
amount of water, and the geological history of the clay. In soil mechanics a plasticity index is used
to describe the range of water content over which plastic properties are observed, Fig. 5.1.
Fig. 5.1 Relation between water content and volume change of a clay mineral
(after Gillott 1987).
All clay soils show time-dependent mechanical behaviour which may be described as viscoelastic
(e.g. Hvorslev 1960, Adeyeri et al. 1970). When a viscoelastic material is subjected to mechanical
stress, part of the energy is stored as elastic energy and the remainder is released as viscous energy,
i.e. heat. The release of the latter energy results in permanent deformation, while the stored energy
can contribute to damping and partial recovery from the deformation. A viscoelastic material thus
possesses both elastic (Hooke's law) and viscous (Newton's law) properties. The simplest case can
be described by the following relationship (Meyer 1874):
Deep Mixing 61
•
• τ τ
γ = + (5.1)
G µ0
(Hooke´s law) (Newton´s law)
•
where γ = shear rate (dγ / dt), s-1
τ = shear stress, Pa
•
τ = derivative of shear stress (dτ / dt), Pa·s
-1
G = shear modulus, Pa
•
µ0 = viscosity at γ = 0, Pa·s
Fig. 5.2 shows schematically the behaviour of an undisturbed viscoelasto-plastic material subjected
to gradually increasing shear stress. Initially, structural breakdown competes with remoulding
(Walters 1980, Nakajima 1994). The material deforms at an increasing rate as the stress increases.
At the maximum stress the material can no longer be remoulded. The structure is so degraded that
further deformation proceeds at an increasing rate even at low stresses. The properties of the mate-
rial become increasingly liquid-like. The maximum shear stress in the figure corresponds well to
the shear strength measured by e.g. a penetrometer or compression test.
Under laminar shearing the viscosity of a clay slurry is reduced due to the clay particles assuming a
more parallel orientation (Gillott 1987). The release of entrapped water from the pores assists the
structural breakdown under shearing. On the other hand, shearing reduces the size of the structures,
which can increase collisions between the particles. In some systems this can have a dominating
effect on the viscosity.
Fig: 5.2 Schematic representation of structural breakdown of a viscoelasto-plastic

material subjected to gradually increasing shear stress (after Walters 1980).

Granulo-viscous behaviour occurs when a mineral with high particle concentration has both solid
and viscous flow properties (Cheng & Richmond 1978). Under shearing, viscous behaviour is de-
scribed by a shear stress-shear rate relationship. This is independent of the normal stress unless the
pressure is extremely high. Viscous materials are normally taken to be incompressible. Dry solids
are characterized by the shear stress being dependent on normal stress and bulk density and inde-
pendent of shear rate. They are further characterized by compressibility. Under shear a material
having only viscous properties shows a continuous shear rate profile, while shearing of a dry pow-
der results in a shear plane of only a few particles' thickness, see Fig. 5.3. The shear strength of a
material with a high particle concentration can be expected to be dependent on both the shear rate
and the normal stress. Under shear the shear rate profile may show numerous fluctuations. In practi-
cal terms this means that in e.g. thick suspensions, the transfer of force from an impeller is inhibited
and it is difficult to produce the movements of materials that are necessary in order for mixing to
occur (Nagata et al. 1972).
Fig. 5.3 Velocity profile in shearing of a dilute suspension and a dry powder
(after Heywood & McDonagh 1987).
The development of slip planes and slip surfaces is also characteristic of clays undergoing rapid
shearing motion. Fig. 5.4 shows the laminar shear mixing mechanism operating in a dense suspen-
sion, e.g. a clay. The figure shows the velocity distribution in the material with the top surface in
motion while the bottom surface is stationary. Relative movement within the material, the precondi-
tion for mixing, takes place only at the slip surface itself and within a very restricted volume imme-
diately above it. Above this volume the material moves as a rigid body, while below the slip surface
no movement at all takes place.
Although in practice the movements and the mixing mechanisms around an impeller are more com-
plicated than indicated in Fig. 5.4, the figure does explain why it is so difficult to produce move-
Deep Mixing 63
Moving top surface
Velocity distribution
in material
Volume in which
mixing takes place Shear plane
Stationary surface
Fig. 5.4 Laminar shear of a yield stress material with high concentration of solid particles.
ment in soil when mixing it with a binding agent: a mixing tool rotating in the soil only causes
movement in the immediate vicinity of the blades. Because of the viscoelastic properties of the soil,
forces and movement cannot propagate away from the impeller.
A slip surface can also develop close to the moving surface. This phenomenon occurs when the top
surface is smooth or slippery. Influencing factors include the type of material, the particle size, the
geometry, the velocity, the shear stresses acting on the surface, the particle concentration, the yield
stress and the viscosity. In loosely consolidated soils in a plastic state the risk of slip is significant.
The theory of viscoelasticity is a well-established science, and there are numerous monographs
which summarize the field and describe the properties of various materials (among them Drozdov
1998, Haddad 1995, Lakes 1998, Tschoegl 1989). A very extensive review of the properties of
clays, rheological models and applications to engineering geology is that by Gillott (1987). Coussot
(1997) has written a monograph which sums up present-day knowledge of the rheology of clay
slurries with application to earth flows.
Factors affecting rheological behaviour

Even when a clay is dispersed to the point where it behaves as a liquid, it can still be difficult to set
it in motion. The material will still have a yield stress which must be exceeded before movement
can take place, Fig. 5.4.
Both hydrodynamic forces and non-hydrodynamic forces, such as the London-van der Waals forc-
es, act in a suspension (Giesekus 1983). Non-hydrodynamic interactions take place between neigh-
bouring particles as well as between particles and the liquid phase. At higher particle concentrations
the non-hydrodynamic forces tend to take over and the behaviour of the suspension becomes non-
Newtonian. The suspension has a yield stress which gives it the properties of a solid as long as the
stresses in the material do not exceed the yield stress.

In a completely dispersed state and at high water contents, clay slurries may be regarded as having
Bingham type flow behaviour (Nickel 1976; Locat 1992). Above the Bingham yield stress tB the
shear rate increases rapidly with increasing shear stress, see Fig. 5.5. In clay slurries the strength of
the interparticle bonds, the particles' mutual orientation and structure (Fig. 5.4), and the equilibrium
particle spacing will be functions of the van der Waals-London forces and of the strength and thick-
ness of the electric double layers on the edges and faces of the clay particles. These in turn are de-
termined by the mineral type, the ionic properties and the concentration of electrolytes in the sys-
tem (Nickel 1976). In most clay slurries the yield stress depends strongly on the clay concentration.
It also depends on the electrolyte concentration, but this dependence varies in different slurries. An
example of shear stress as a function of shear rate and particle concentration is shown in Fig. 5.5b.
a) b)
Fig. 5.5 a) Relation between shear stress and shear gradient for a Bingham system and
a typical clay slurry (after Nickel 1976)
b) Flow curves for kaolin slurries with different particle concentrations f (Cous-
sot 1997)
Clays show a wide range of shearing resistance depending on the water content. When the water
content is high, clays behave as fluid suspensions, even though the interactions between the parti-
cles, ions and molecules in the suspension make the properties of clay unusual. Lower water con-
tent gives a higher consistency. The non-viscous behaviour of clays at water contents below the
yield limit can be attributed to the interlocking of the clay minerals at contact faces (Allam & Srid-
haran 1984). In laboratory investigations of clay rheology it has been found that movement in the
clay slurry decreases as water content decreases (Torrance 1986), i.e. the failure zone is limited to a
part of the expected cross-sectional area.
Komamura & Huang (1974) studied the yield properties of clay soils with water contents both be-
low and above the yield limit. They showed that the viscosity of a clay can decrease dramatically as
the water content increases beyond the yield limit. This is attributed to an abrupt structural change
at the yield limit. A sensitive soil rapidly loses shear strength when it undergoes movement, and the
movement takes place at slip surfaces.
Deep Mixing 65
Fig. 5.6 Terminology and models of particle association in clay slurries
(after van Olphen 1977):
(a) “Dispersed” and “deflocculated”
(b) “Aggregated” but “deflocculated”
(c) Edge-to-face flocculated but “dispersed”
(d) Edge-to-edge flocculated but “dispersed”
(e) Edge-to-face flocculated and “aggregated”
(f) Edge-to-edge flocculated and “aggregated”
(g) Edge-to-face and edge-to-edge flocculated but “aggregated”
At the genesis of a clay the arrangement of the clay particles is assumed to be relatively open and
loose packed. Changes over time render the open fabric out of equilibrium. However, the particles
still cohere due to cementation, electrochemical bonding, pressure and interlocking caused by the
mutual interference of particles which prevents free rotation. When the clay is disturbed or re-
moulded, the particles assume a more close-packed configuration which is closer to equilibrium
with prevailing conditions. Since the water content is constant, the deposit becomes oversaturated
and may flow like a viscous liquid (Gillott 1987).
Changes in pH have direct effects on the state of flocculation and the solubility of the mineral com-
ponents (Bentley 1979). However, the influence of pH changes on rheology is small compared to
that of the water content of the slurry.
For equivalent yield index, coarser soils tend to have higher viscosity (Locat 1992). Thus the yield
stress may vary with the particle size at approximately the same yield index. Flocculation of clay
minerals results in a change in the particle size distribution.
Fig. 5.7 shows in principle how the viscosity of fluids is affected by different particle concentra-
tions and shapes.

Fig. 5.7 Effect of particle concentration and particle shape on rheological properties
(after Mork 1994).
Particle size and shape has a major effect on the rheological behaviour of a suspension. A system
will normally contain particles of many different sizes and hence the particle size distribution plays
a role. In addition, the particle size distribution can change over time, making the rheology of the
process extremely complicated. For a given solids content there are certain optimal particle size
distributions that give minimum and maximum viscosities.
Fine particles have a greater specific surface than coarse particles. Since the chemical reactions that
take place on stabilization with lime and cement are largely surface reactions with clay minerals
and water, this fact can have an influence on the outcome of stabilization. However, because molec-
ular diffusion is a slow process it is very important that the solid particles are completely dispersed
in the mixture.
In a range of applications additives, usually reactive chemicals, are used to change the rheological
properties in order to make mixing processes and mixture handling easier or more efficient. Exam-
ples of such applications are rubber mixing (Grossman 1997), concrete manufacturing (Betong-
handbok 1994), and industries in which materials are mixed in powder form (Kaye 1997).
When mixing binding agents with loose soils, additives may be used e.g. to disperse the soil so as to
avoid aggregation. This gives the soil a fluid character. Aggregating additives may be used to pro-
duce a more granular structure with a lower yield stress. The choice of additives depends on a
number of factors, but the main objective is to avoid plasticity of the mixture.
5.3 MEASUREMENT OF RHEOLOGICAL PROPERTIES

Quantitative measurements of the rheological properties of mixtures are a natural part of quality
control, the development of mixing equipment, and research and development on the composition
of materials for industrial applications in the process industries (Walters 1980). Rheological meas-
urement methods have contributed to the development of injecting agents and concrete types (e.g.
Håkansson 1993, Mork 1994). With regard to the rheology of loose soils there is a great deal of
(geotechnical) literature describing the rheological behaviour of undisturbed sensitive clays, partic-
Deep Mixing 67
ularly at slow deformation rates. However, no studies have been published on the rheological be-
haviour of loose soils in the undisturbed or partially remoulded state at high shear rates.
According to Mork (1994), suspensions can be divided into three regions, see Table 5.1. Rheologi-
cal properties within each region are measured with the instruments indicated in Table 5.2.
Table 5.1 Overview of particle systems (after Mork 1994).
Region I Suspensions of low to medium concentration

May be regarded as pseudo-homogeneous
Region II Yield properties controlled by packing effects
May exhibit granulo-viscous properties
Region III Unsaturated, compressible three-phase materials
Viscometers and rheometers often not usable
Table 5.2 Instruments used for different regions (after Mork 1994).
Rheology Geotechnics
Region I + II III ( II )
Instruments used Viscometer
Rheometer (1) Shear box
(2) Shear cell
(3) Triaxial compression apparatus
Fig. 5.8 shows how systems transition from pure liquid systems to powder systems. Cement paste
and concrete systems, for example, fall within regions II and III. In region II the system may dis-
play granulo-viscous behaviour, i.e. the suspension may behave as a mixture of powder and water,
leading to problems in the investigation of its rheological properties.
In order to study the rheological properties of systems in region III the usual instruments are the
triaxial compression apparatus, the shear box etc. In rapid deformation events, slip surfaces form
due to the material's viscoelastic properties e.g. as shown in Fig. 5.4.
Clay slurries have particle concentrations that usually place them in region II or III in Fig. 5.8.
They show non-Newtonian flow behaviour in general. Clays with water contents above the yield
limit and in a completely remoulded state can be studied with rheological instruments such as rhe-
ometers. However, it is uncertain whether data from such tests agree with the behaviour of clays
under natural conditions.

Fig. 5.8 Classification of systems based on particle concentration cv (after Mork 1994).
The rheological properties of sensitive clays have been studied by Canadian researchers (Locat &
Demers 1988, Locat 1992, Coussot 1997). This research was done in conjunction with studies of
remoulded material after major landslides. The velocity of landslides may be in the order of
10–20 m/s. Laboratory rheological tests were carried out in order to describe the relation between
the index parameters of soils and their flow properties. The tests reveal very interesting relations
between viscosity, yield stress, remoulded strength, and liquidity index. These relations are restrict-
ed to clay soils with water contents above the liquid limit (liquidity index >1). The relations can be
used e.g. to evaluate the mobility of flow slides.
A linear correlation exists between the remoulded shear strength measured by the fall-cone method
and the yield stress (Torrance 1987). A viscometer can be used to evaluate the strength of remould-
ed clays below the range of applicability of the fall cone (1070 Pa). The limitation is that the linear
relation between remoulded shear strength and yield stress must be determined for each material
and condition.
The formation of slip surfaces when a loose soil experiences rapid movement is one reason for the
difficulty of carrying out rheological measurements. The same soil may have quite different rheo-
logical properties depending on the degree of remoulding.
5.4 RHEOLOGICAL PROPERTIES IN PARTIALLY REMOULDED STATE

The effect of complete remoulding of a clay can be described by its sensitivity. The sensitivity is
defined as the ratio of the undisturbed shear strength to the shear strength in the fully remoulded
state. The shear strength is determined by the unconfined compression test, the vane shear test or
the fall-cone test. Nearly all normally consolidated and lightly overconsolidated clays show some
sensitivity (Gillott 1987).
Deep Mixing 69
Fig. 5.9 Relation between sensitivity and the ratio of the natural water content to the
liquid limit for some varved clays (Osterman 1965).
Söderblom (1974) proposed a new classification of quick clays. The background was that quick
clays exist with different mechanical properties, i.e. they behave differently under external mechan-
ical stress. Clays are usually termed quick clay if their sensitivity is greater than 50. Söderblom
argued that clays can break down at different rates and that this fact should be taken into account in
their classification.
An approximate classification of quick clays into groups was developed using a relative value
called the rapidity number. Slow breakdown gives a low rapidity number and rapid breakdown a
high one. To determine the rapidity number of a clay Söderblom used the Casagrande liquid limit
test. A clay sample of height 40 mm and diameter 50 mm is dropped 250 times a distance of 10
mm. The sample is inspected visually and the rapidity number determined on a scale from 1 to 10,
where Rn = 1 means that the sample is visually unaffected by the test and Rn = 10 means that the
sample is totally liquefied. The rapidity number is thus subjectively determined and cannot be relat-
ed to any soil parameter.
In an investigation of soil movements due to pile driving, Massarsch (1976) expressed the loss of
shear strength after partial remoulding by a disturbance ratio. The disturbance ratio DR is defined
as
τ undisturbed soil
DR = (5.2)
τ partially remolded soil
where τundisturbed soil is the undrained shear strength of the undisturbed soil and τpartially remoulded soil the
undrained shear strength of the clay after a certain degree of mechanical disturbance. Massarsch

(1976) recommended using the disturbance ratio rather than the sensitivity in the analysis of piling
problems in clay.
The disturbance ratio can be used in relatively simple rheological studies of soils and their behav-
iour under the influence of mechanical stress, e.g. different degrees of remoulding. The effective-
ness of different types of impeller can be studied in this way, since any significant mixing operation
will increase the disturbance ratio of a soil. It is easy, for example, to carry out fall-cone tests or
vane shear tests using miniature vanes in conjunction with laboratory tests in which soils are sub-
jected to various amounts of mechanical work.
Appendix B reports a study of the effect of mixing energy on the disturbance ratio. This study com-
prised a series of trials in which the mixing tool was rotated in the earth at different retrieval rates
without adding any binding agent. Vane shear tests were carried out on a number of column cross
sections using a pocket vane of diameter 25 mm and height 50 mm in order to evaluate the distur-
bance ratio produced by the impeller. The trials were done on a grey, somewhat muddy clay with
sulphide spots. This soil had very low rapidity, implying that it is relatively insensitive to remould-
ing. The results in Fig. 5.10 show that with the mixing tools currently in use for dry deep mixing in
Sweden, it is difficult to break down the structure of a soil and thus change its rheological condi-
tions during the mixing process. Notwithstanding a relatively large mixing energy, a retrieval rate
of 5 mm/revolution, the disturbance ratio was only of the order of DR = 4. The disturbance ratio of
a completely disrupted structure, which is equal to the sensitivity of the soil in question, is 12 – 13.
According to this study, considerably greater effort is required to break down the soil structure.
The study of the influence of the mixing tool on the disturbance ratio, Appendix B was supplement-
ed with laboratory trials. In the course of an investigation of the influence of the impeller on com-
pressive strength (Larsson et al. 1999, Appendix C) a series of trials was carried out in which soil
samples were mixed without binding agent. The study was done using the equipment and the soil
Fig. 5.10 Disturbance ratio and coefficient of variation as functions of the retrieval rate of
an impeller (author's experiments, Appendix B).
Deep Mixing 71
reported in Appendix C. The disturbance ratio was measured by the fall-cone test and can range
from 1 to the value of the sensitivity of the soil, which was measured at 13. The soil studied is a
sulphide clay with water content 107 %, liquid limit 100 % and fall-cone shear strength 14.7 kPa. A
more detailed description of the soil appears in Walter (1998).
Fig. 5.11 shows the results of the laboratory tests, which again show that a large amount of mixing
energy is required to break down the structure of the soil studied. However, the results show that
the disturbance ratio after a relatively small amount of mixing is in the order of 7.
Fig. 5.11 Disturbance ratio as a function of mixing energy input (author's experiments).
5.5 RHEOLOGICAL PROPERTIES ON INCORPORATION OF LIME AND CEMENT

The rheology of suspensions is normally very complex. Solid particles such as lime and cement
suspended in a clay slurry are a good example of a highly complex system. In dry deep mixing the
rheological properties during the mixing process are also affected by chemical reactions.
There are no published scientific studies of the rheological properties of a soil while it is being
mixed with cement and lime in deep mixing. Current knowledge of the rheology of such systems is
empirical and all too often anecdotal in character.
We describe below very briefly the chemical reactions that take place when loose soils are mixed
with lime and with cement, and effects of the reactions on the rheological properties of the mixture.
Mixing with lime

The use of lime to stabilize soils dates back over 5000 years (Greaves 1996). The development of
the surface stabilization technique took off in the 1940s in the USA, and it has since become widely
used, particularly in road building. The use of lime in other foundation contexts has been described
by e.g. Rogers & Glendinning (1996), Sherwood (1993), and the British Lime Association (1990).
Numerous articles have discussed the chemical reactions and changes in properties that occur on

mixing lime with soil. Recent reviews of this work include Locat et al. (1990), Bell (1988), Sher-
wood (1993), West & Carder (1997). The chemical reactions are:
1 ion exchange;
2 flocculation and aggregation;
3 carbonation;
4 pozzolanic reactions.
The first two reactions cause an immediate change in the plasticity of the soil and can have a major
effect on rheological behaviour during the mechanical mixing stage. When quicklime, CaO, is
mixed with a soil it absorbs moisture, forming calcium hydroxide, Ca(OH)2. Since one kilogram of
lime takes up 320 g of water the water content of the soil is reduced. The heat evolved by this reac-
tion can also affect the mixing process.
The dissolved calcium hydroxide is then available for ion exchange with ions in the clay minerals
(Na+, K+, H+). This causes flocculation of the clay minerals and the plastic limit increases. De-
pending on its initial water content, the properties of the soil may change dramatically.
When cohesive soils are mixed they do not disintegrate into elementary particles as granular materi-
als do, but form striations or lumps whose size depends on the effectiveness of the mixing. It is
well-known in surface stabilization that it is difficult to mix lime and cement with loose plastic soils
(Petry & Wohlgemuth 1988). The addition of lime causes the structure to change to a more granular
one. However, the immediate structural change in the lumps is confined to their surfaces (Lindh &
Ydrevik 1978). The remoulded clay tends to form pellets with a surface coating of binding agent,
and breaking down this structure can be difficult. A study by van Ganse (1974) showed that the
initial reactions take place very quickly, within a few minutes after mixing in the case of laboratory
samples. Another observation of interest from this study is that the lime-soil lumps that are formed
retain their individuality to a large extent when the mixture is subjected to kneading and compac-
tion. An aggregate-forming effect results in the soil absorbing more water, which may render mix-
ing inefficient but can also promote compaction. The effect depends on the initial water content of
the soil.
It is difficult to mix crusted clay effectively with lime. Experience from surface stabilization indi-
cates that on mixing, the dry clay breaks up into large, hard lumps which are very difficult to pul-
verize (Lindh 1977). Pulverization is only possible with the addition of water.
Mixing with cement

Cement has a very positive effect on strength gain, particularly in loose soils, and its use has there-
fore been increasing in e.g. surface stabilization and deep mixing. The use of cement in surface
stabilization is described by e.g. Sherwood 1993 and Portland Cement Association (1990).
As in the case of soil stabilization with lime, numerous articles have been published on the chemi-
cal reactions and changes in properties produced by the treatment of soil with cement. Recent re-
Deep Mixing 73
views of this literature include Petry & Wohlgemuth (1988) and Sherwood (1993). Two main proc-
esses occur when cement is mixed with a loose soil. The primary process is hydration of the ce-
ment, causing hardening of the soil. The cement particles bind neighbouring lumps of soil together
into a more or less continuous skeleton of hard material enclosing a matrix of unmodified soil. The
hydration of cement results in the formation of calcium hydroxide, and the calcium ions react with
the clay minerals. This secondary process also forms hardening materials.
The chemical reactions that take place on mixing cement with soil are much slower than the reac-
tions with quicklime. It is therefore assumed that cement does not drastically alter the rheology of
the system during the mixing process. However, the cement does affect the rheology inasmuch as
adding a dry powder to the soil increases the particle concentration. The result thus depends very
much on the effectiveness of the mixing process in disaggregating the soil and distributing the ce-
ment through it.
In surface stabilization projects in which clayey soils are to be treated with cement it is advisable to
treat the soil with 1.5 – 2.5 % lime prior to cement stabilization (Cementa 1973). This is done be-
cause it is difficult to disaggregate clayey soils adequately in-situ. The addition of lime gives the
soil a granular structure which can be more easily disaggregated with less machine effort. The
drawback of this method is that it necessitates two distinct mixing steps and hence greater expense.

6. Deep mixing
6.1 INTRODUCTION
Deep mixing is a generic term for ground improvement methods in which binding agents, often
lime and/or cement, are mechanically mixed with the soil. A range of methods have been developed
which today go under names such as "deep mixing (method)" (e.g. Terashi 1998, Porbaha 1998).
Some Scandinavian literature uses the term "lime-cement columns", "deep stabilization", "dry jet
mixing method" or "column stabilization". The mechanical mixing is generally done by rotating
impellers of the paddle or helical type. The binder is either in powder form or premixed with water
to form a slurry. As a rule, the columns formed in the soil have better strength and deformation
properties than the surrounding untreated soil.
The review of present-day mixing procedures in the present chapter sets out from a division of the
process into stages as in Fig. 6.1. The discussion is based on the principal method employed in the
Scandinavian countries, using columns of lime, lime-cement or cement. The phases of the mixing
process are described and discussed. Factors that influence the mixing process are surveyed. A spe-
cial section is devoted to mixing tools, mainly those used in Japan and Sweden. A concluding sec-
tion examines some characteristic defects found in columns as a result of the mixing process.
6.2 HISTORICAL SKETCH

Deep mixing, in its present form, was first presented in an international forum in 1975, when two
papers were submitted to the same conference. One came from Sweden, Broms & Boman (1975b),
and the other from Japan, Okumura & Terashi (1975). The method had been under development
concurrently in the two countries (Okumura 1997). The first journal articles, by Broms & Boman
(1979a and 1979b), focussed on applications, properties and design.
Development of the lime column method in Sweden began in 1967 under the leadership of Kjeld
Paus. It was based on the in-situ mixing of soil with quicklime to form hardened columns of cohe-
sive material (Boman & Broms 1975a). Binding agents were investigated by laboratory tests at the
Swedish Geotechnical Institute (SGI), and after development of a suitable installation procedure
some trial columns were installed at the SGI test field at Skå-Edeby. The first equipment consisted
of a drill rig mounted on a Volvo tractor, see Fig. 6.24a. The number of projects carried out in
Scandinavia was fairly small until the end of the 1980s, when machinery development got properly
under way. The development of the equipment currently used in Scandinavia has described by
Bredenberg (1999).
Concurrently with its development in Sweden, research and development on deep mixing had been
in progress in Japan since 1967 in the form of laboratory and model scale tests (Schaefer et al.
1997). These tests were run by the Port and Harbor Institute of the Japanese Ministry of Transporta-
Deep Mixing 75
tion with the aim of developing a method for the deep stabilization of marine clays under the sea
floor. In Japan there has been a huge growth in the reclaiming of land from the sea by filling. The
first Japanese project, using lime columns, was completed in 1974. The Japanese also developed the
"wet method", in which the binder, usually cement, is premixed with water and injected into the soil
as a slurry. The first project using the wet method was completed in 1975 (Kawasaki et al. 1981),
and in 1977 the CDM Association was set up to coordinate the development of the method in col-
laboration between industry and research institutes (Porbaha 1998).
In 1975 Japanese engineers began research and development into deep mixing using dry binder
agents (Chida 1982). Development of dry deep mixing was led by the Ministry of Construction.
The first project was carried out in 1981 (Schaefer et al. 1997). The method was very similar to the
Swedish lime column method.
Deep mixing was highly successful in Japan and extensive research and development took place in
the 1980s in the form of laboratory and field work. The method thus underwent rapid evolution and
a relatively large number of organisations and companies have developed their own mixing proc-
esses, resulting in a variety of installation techniques.
According to Bruce et al. (1998a, 1998b) the original concept comes from the US, where in 1954 a
single auger was used to mix binders with soil. Handy & Williams (1967) reported the results of
stabilizing loose soil at the "Aurora Avenue landslide" in Iowa, US. Five hundred drill holes of di-
ameter 150 mm were filled with quicklime and water. A similar method had been used in 1960 in a
highway project in Oklahoma, where drill holes of diameter 230 mm were filled with hydrated lime
and water (Handy & Williams 1967). However, the method in its present form derives from devel-
opment in Japan and Sweden. In 1986 the method was reintroduced in the form of a Japanese meth-
od (Bruce 1996), and Swedish practice has been represented in the US since 1997 (Bruce et al.
1998a).
Development of the method has also taken place in Italy (Paviani & Pagotto 1991) and in England
(Harnan 1993). An overview of practice worldwide can be found in Bruce et al. (1998b, 1999) and
Porbaha (1998).
6.3 THE MIXING PROCESS

The mixing process in deep mixing is very complex, comprising many phases, and many factors
influence the process and its result. It is difficult to clearly separate the different mechanisms in-
volved in the mixing process, however it is important to understand how the mechanisms affect
each other.
The purpose of the mixing process is to disperse the binder in the soil so as to provide the best pos-
sible conditions for the chemical reactions to take place. If all of the binder is to contribute actively
to the improvement of the soil, the particles of binder must all be evenly dispersed throughout the
volume of the column. Moreover, the binder should be evenly distributed over the column cross
section in order to limit the dispersion of strength and deformation properties. Wide dispersion of

the properties not only reduces the predictive value of laboratory tests but also complicates produc-
tion control. The consequence may be a loss of control of the process and its results.
The manufacturing process in deep mixing is illustrated in Fig. 6.1. It can be divided into three
principal phases:
1) penetration of mixing tool to the required depth;
2) dispersion of binder;
3) molecular diffusion.
1) Penetration of mixing 2) Dispersion of binder agent in the soil. 3) Manufacturing completed.

tool to desired depth Sub-phases: Mixing continues by
a) incorporation and spreading molecular diffusion
of binder agent;
b) wetting of dry binder particles;
c) breakdown of agglomerates;
d) distribution (and stabilizing of dispersion).
k/c k/c k/c
Unstabilized Stabilized soil

loose soil
More solid soil, e.g. moraine
Fig. 6.1 The mixing process in dry deep mixing (e.g. lime-cement columns).
Penetration of mixing tool

In the first phase of the mixing process the rotating mixing tool is driven into the soil to the desired
depth. The insertion process can be executed in such a manner that the resulting remolding and dis-
aggregation of the soil changes the conditions for subsequent phases. Total disaggregation (dis-
persed and feflocculated structure) can have a positive effect on the active mixing mechanisms by
making it easier to produce the necessary movements in the soil for the mixing of the materials. The
mixing mechanisms that may be affected are:
a) molecular diffusion. Disaggregation of the clay renders the clay minerals and the liquid phase
more accessible for chemical reactions;
Deep Mixing 77
b) laminar mixing. The conditions for laminar mixing are improved by reducing the elastic drag of
the clay suspension. Laminar mixing is an effective mechanism for disagglomeration and also
promotes molecular diffusion.
In the current mixing procedure there is a risk that much of the energy used in the dispersion proc-
ess is expended on remolding the clay. Laminar mixing is likely to be less effective as the clay slur-
ry can withstand deformation thanks to elastic drag and the high yield point.
However, it is not certain that complete disaggregation of the soil will exclusively positive effects
on the strength gain. Complete disaggregation combined with poor mixing efficiency (high concen-
tration variances) may result in the low shear strength of relatively massive striations and/or lumps
impairing the overall strength properties of the column.
Existing mixing tools are equipped with paddles which are generally set at a small angle to the hori-
zontal. This facilitates insertion of the tool as it requires little energy to penetrate the soil. The inev-
itable consequence, however, is that relatively little energy is expended on remolding the soil. It
takes movement of the soil to produce the shear forces necessary for disaggregation. Large shear
rates around the paddles can be produced by increasing their angle to the horizontal plane, causing
forced axial and tangential movement, or "kneading", of the soil. High rotation speeds can also pro-
mote disaggregation. However, it may be difficult to drive the tool at a high speed when using
steeply tilted paddles because the resistance and hence the shear forces are higher. It is also possible
to increase the number of shear zones by e.g. drilling holes in the paddles, increasing the number of
paddles, angling the blades in opposite directions, or using a tool in the form of a rake (i.e. with
teeth).
An important issue is whether increasing the agitation energy during penetration of the tool actually
improves conditions for the mixing process sufficiently to justify the extra cost it entails. It is uncer-
tain today how much the result of mixing is affected by the input of agitation energy during pene-
tration as compared with other factors. The tool is currently inserted at the rate of 100 mm/rev, and
it is doubtful whether the soil structure is significantly affected. It is also doubtful whether it is eco-
nomical to remold and disaggregate the soil before adding the binding agent.
The dispersion process

The process by which the binder is dispersed in the soil can be divided into four steps (basically
based on a division of the mixing process by Parfitt & Barnes 1992):
a) incorporation and spreading of the binder;
b) wetting of solid particles;
c) breakdown of agglomerates;
d) distribution.
In practice the stages overlap and it can therefore be difficult to distinguish them by visual observa-
tion.

a) Incorporation and spreading of the binder
It is most important that the mixing tool is designed in such a way that the binder is spread as even-
ly as possible over the column cross section during the incorporation phase. This avoids major con-
centration variances, makes the breakdown of agglomerates easier, and avoids long mixing times,
since the whole dispersion process take place during a short time of mixing. Due to the complex
rheological behaviour of a clay slurry it is difficult to distribute a binder over a column cross sec-
tion in a short time by mechanical mixing, see Fig. 5.4. It is therefore important that the binder is
evenly spread over the cross section at an early stage of the mixing process.
In the Scandinavian countries binder is delivered in powder form from tanks via hoses to the mix-
ing tool using compressed air as a transport medium. The binder is expelled through a hole in the
Kelly rod just above the mixing tool, usually in the manner shown in Fig. 6.20. The binder is spread
through the cavities formed in the soil by the rotating mixing tool, and to some extent by pneumatic
fracturing caused by the air pressure . The size and shape of the cavities formed depends on the
geometry of the mixing tool and the intensity of mixing, the air pressure, stress conditions in the
soil, the rheological properties of the soil and the binder. In order to avoid large agglomerates
(loosely bound lumps) and accumulations of binder it is very important that the binder is spread
evenly over the entire column cross section in this phase of the mixing process.
The incorporation and spreading of the binder by compressed air is a mixing mechanism in itself,
since it produces relative movement of the soil and the binder. A relatively large energy input takes
place as the compressed air fractures the soil. Unfortunately our knowledge of this mixing mecha-
nism is very limited. Better knowledge is highly desirable, as this mechanism can be a significant
one for the mixing process. If the binder is spread evenly over the cross section at incorporation, the
main task of the mixing tool will be breaking up the agglomerates that form, and the risk that the
mixing tool will be unable to distribute the binder is reduced.
b) Wetting of solid particles

When a powdered binding agent is mixed with a soil, the lumps of powder contain entrapped air
which must be replaced with liquid. In e.g. calcareous and cementitious reactions liquid is drawn
from the surrounding soil by diffusion. If the air is not to remain entrapped mechanical work to be
done, but even with mechanical assistance the release of the air can be difficult if agglomerates or
aggregates have formed. The wetting process is assisted by mixing tools which generate high com-
pressive and shear stresses in the soil.
The wetting process cannot occur spontaneously because the lime and cement particles are not
dense enough to sink into the clay slurry. The high particle concentration, the complex spreading
process, the presence of air as a component, and the chemical reactions proceeding during the mix-
ing process make the behaviour of the mixture extremely complex. As mentioned in Chapter 2.6,
the wetting process is influenced by the properties of the liquid phase, the character of the surface,
the dimensions of the interstices in the agglomerates, and the compressive forces exerted by the
mechanical system on the components. The rheological behaviour of the clay (dependent on e.g.
clay and water content) is also highly significant for the wetting process. Complete remolding of
the clay releases water which is available for wetting the lime and cement particles. It also gives the
clay particles a larger active interfacial area with the binder, which promotes molecular diffusion.
Deep Mixing 79
Fig. 6.2 shows an in-principle design of an impeller paddle shaped so that a film of clay is forced
against the inside of the paddle while binder contacts the soil in the same region. The principle is
adapted from equipment specifically designed for wetting solid particles, see Fig. 3.6. The binder is
spread along the cross section of a cavity created by a forward paddle. The stretching of the film of
clay along the blade increases the interfacial area between the components being mixed. Below the
paddle, large shear stresses and movements are generated in the material, producing the necessary
compressive stresses to force liquid into the pores and release the entrapped air from the agglomer-
ates.
Expulsion of binder A film of clay is pressed and stretched against the inside
of the paddle, increasing the interfacial area between
the clay and the binder.
Forward paddle creates a cavity
through which the binder is Laminar shearing produces high shear stresses at the
spread over the cross section lower edge of the paddle, generating the compressive
stresses necessary for liquid to penetrate the pores and
release entrapped air from the agglomerates.
Fig. 6.2 In-principle design of an impeller paddle in which the binder is wetted in contact
with a film of clay.
In dry deep mixing, particularly in soils with water contents below the liquid limit (liquidity index
< 1), wetting problems in combination with inadequate mechanical mixing can lead to agglomera-
tion and entrapped air in the columns. A way of avoiding air and increasing the efficiency of wet-
ting is to premix the stabilizer with water, i.e. carry out mixing in two steps. This method, known as
the "wet method", is mainly used with cement as binder agent. In applications where satisfactory
performance depends on a high degree of homogeneity and where the liquidity index is less than
unity wet deep mixing may be appropriate.
In the improvement of loose soils with a liquidity index greater than unity the wet method may
have a negative effect on the mixing process, since lumps or portions of the soil may be entrained
by the mixture without undergoing significant action by the mixing tool. It is important that the
mixing tool is designed to produce high-intensity and if possible turbulent movement.
The wetting process is closely linked to the process of compacting the mixture of soil and binding
agent. The capacity of the mixing tool to compact the mixture is all the more important when air is
injected into the soil. The influence of the compaction process on mixing is discussed at the end of
section 6.4.

c) Breakdown of agglomerates
Once the binder is incorporated and spread and the particles are wetted, the agglomerates that have
formed should be broken down before the chemical reactions begin. Unless the particles are evenly
dispersed in the mixture, large aggregates will form, resulting in concentration variances and poor
mixture quality. Furthermore, if the particles are not dispersed, not all the particles of binder will
make their full contribution to even strength gain throughout the stabilized volume.
Agglomerates are broken down by shearing the soil or by generating large compressive forces. To
produce sufficiently large shearing or compressive forces the mixture must be set in motion. For the
effective breakdown of agglomerates in a mixture with dough-like properties, experience from
process industries indicates that equipment with a kneading or grinding action should be used. A
kneading action in the mixture can be produced by paddles mounted at a relatively large angle to
the horizontal, ~45°, thus generating both axial and tangential motions. It is advantageous to gener-
ate movements in a number of directions because clayey soils have viscoelastic properties, entailing
a risk that the mixing tool will only generate movement in the immediate vicinity of the blades, i.e.
that the soil will "yield" in a thin shear plane close to the tool as illustrated in Fig. 5.4.
Depending on the rheological properties of the soil and on which mixing mechanisms are active,
the efficiency of the mixing process is influenced mainly by the following two factors:
1) the effective strain in the mixture. This factor can be expressed as a function of the number of
revolutions per metre (rev/m) or the retrieval rate (mm/rev) of the mixing tool;
2) the intensity of mixing or agitation. This factor can be expressed as a function of the rotation
speed of the mixing tool (rev/min).
A hypothesis in this connection is that when the mixing intensity is sufficient to break down ag-
glomerates, the effective strain is the most significant factor for better mixture quality. If on the
other hand the mixing intensity is too low to break down the agglomerates, the effective strain is of
little importance as the agglomerates are merely entrained with the mixture. According to this hy-
pothesis it is the rheology of the soil that determines how far the retrieval rate and rotation speed of
the tool influence the process.
In order to bring about the necessary movements and shear stresses in the mixture the mixing tool
may be designed on the principle of an anchor agitator having two sets of arms operating in oppo-
site directions, or as an anchor agitator combined with paddles operating in opposite directions (cf.
Figs 3.9 and 3.10). Such a tool could be of the in-principle design shown in Fig. 6.3. To produce
large shear stresses in the material the paddles may run at different speeds, with the smaller paddle
rotating at the faster speed (a higher shear rate results in higher shear stresses in a suspension with
viscous properties). The tool produces both high shear stresses and movement, enabling the materi-
al to remould between successive shearings. The dominant mixing mechanisms are laminar shear
and distributive mixing.
If a completely remoulded loose soil has a liquidity index > 1, under favourable conditions turbu-
lent flow can be generated in the mixture. Turbulent flows result in effective circulation and high
shear forces. If the viscous drag in the fluid is small, a larger proportion of the mixing energy will
Deep Mixing 81
Fig. 6.3 In-principle mixing tool design combining an anchor agitator with a paddle
operating in the opposite direction.
go into breaking down agglomerates. It takes relatively high rotation speeds to produce turbulent
flow. At low speeds there is a risk that the tool will merely move the agglomerates around without
breaking them up.
In Chapter 5 we discussed the difficulty of producing movements in materials with viscoelastic

properties, due to the fact that the intermediate forces around the impeller rapidly die out. The im-
peller generates no significant movement of the mixture, with the result that if the mixing time is
too short, layers of solid additive will be left in the soil. Hence it is important to design the mixing
tool to produce forced movement of the materials and generate laminar mixing within a sufficient
volume of soil around the tool. When a paddle is tilted at a small angle to the horizontal, one may
expect it to cut through the clay without imparting movement to a significant volume of soil. The
risk of poor results with a given tool is thus particularly great if the water content is close to the
liquid limit and the soil has plastic properties, or if the clay is not sufficiently disaggregated.
In a process in which agglomerates are being broken down the fluid elements need to remould
themselves between successive shear stresses. In a turbulent flow regime this is not a problem,
since the elements are moving randomly in continually changing directions. In a viscoelastic soil,
however, it can be difficult for the fluid elements to remould themselves if movement in the mix-
ture is limited. This problem can be overcome by subjecting the mixture to continually changing
directions of shear, with each shear stress operating in a different plane to the previous one. An in-
principle design for such a mixing tool is shown in Fig. 6.4. This tool is moreover designed so that
the blades generate alternate inward and outward radial movements in the mixture.
Shearing forces operating in different planes can also be achieved by equipping the paddles with
teeth. These can also facilitate penetration of the dry surface crust during insertion. Yet another way
of achieving a desirable shearing action is to provide the paddles with holes in order to obtain ex-
tensional flow with increased shearing capacity.

Fig 6.4 In-principle design of a mixing tool which generates shear forces in continually
changing planes.
d) Distribution
Distribution is the process by which the disaggregated agglomerates are randomly scattered through
the mixture. This normally takes place concurrently with the preceding process.
If the binder has not been adequately spread during the previous phases of the process, long mixing
times may be expected, as it is difficult to generate movement in all soil types. Distribution of the
binder is promoted by complete disaggregation of the soil, high water content, and low viscosity. It
will only be successful with mixing tools that produce large movements in the soil.
Molecular diffusion
After the manufacturing process is completed and the column is installed in the soil, the mixing
process continues by molecular diffusion. When concentration variances exist in an otherwise uni-
form solution, they tend to decrease over time and finally disappear. In a homogeneous system this
process, known as diffusion, must ultimately result in uniform concentrations of all materials
throughout the system.
The binders most commonly used today, lime and cement, differ in their diffusion properties. When
lime reacts with water it produces calcium hydroxide, which diffuses in dissolved state and thus
improves the degree of mixedness. Cement on the other hand reacts with water to form products
which harden and fill out the pores between the soil particles. Molecular diffusion is a minor mix-
ing mechanism for cement, which thus requires a higher degree of dispersion than lime. Petry &
Wohlgemuth (1988) write that cement is not a suitable binder for shallow subgrade stabilization of
loose plastic soils unless the mixture quality is very good.
The diffusion of calcium from the column into the unstabilized surrounding soil, or from regions of
the column with a high lime concentration into lumps of soil with lower concentration, depends on
the water content, the mineral content, the permeability, the clay content and the void ratio (Thomp-
son 1966).
Deep Mixing 83
The molecular diffusion of calcium ions in soil as a mixing mechanism was demonstrated and dis-
cussed in the journal literature as long ago as 1939 by Jenny & Overstreet (1939). However, re-
search did not take off until the 1960s, when lime stabilization began to come into more general use
for the surface stabilization of terraces in road building. Perhaps the most comprehensive study
from that time is Davidson et al. (1965). Notwithstanding numerous studies published since the
1960s, it remains unclear how molecular diffusion works as a mixing mechanism and how much
effect it has in deep mixing. A survey of some of the studies of the molecular diffusion of lime in
loose soil follows below.
Shen et al. (1997) presented analyses of samples taken from the loose soil in between installed ce-
ment columns (wet method, 140 kg/m3). The columns were of diameter 1140 mm and were in-
stalled in a grid at 1500 mm centres. The results showed that the disturbance in the soil caused by
the installation process is quickly repaired by cation diffusion. One month after installation the
strength of the surrounding soil was somewhat higher than that of the original soil.
The deep mixing of loose soil can causes vertical fracturing of the surrounding soil. The fractures
can be explained by the excess pore pressure in the soil produced by lateral pressure from the injec-
tion of binder, and by the shearing of the soil caused by the mixing tool (Shen & Miura 1999). The
fractures form channels for binder migration and may account for the measurably elevated concen-
trations of cations well out in the surrounding unstabilized soil.
Mathew & Narasimha Rao (1997) studied the diffusion of calcium ions in a marine clay from India
with a liquid limit of 63 % and a plastic limit of 22 %. In this laboratory study holes of diameter
50 mm were filled with quicklime. Columns were installed at 400 mm centres (triangular mech).
Measurement of pH and lime content at points in between between the columns showed that lime
had migrated throughout the soil volume. In 60 days the calcium oxide content had increased from
1 % to 2.5 % and the pH from 7.5 to 10.
Rajasekaran & Narasimha Rao (1996 & 1997) and Narasimha Rao & Rajasekaran (1996) studied
the diffusion of lime from a central lime column (a hole filled with quicklime) of diameter 50 mm
in a tank of diameter 600 mm. A marine clay from India was used having a liquid limit of 85 % and
a plastic limit of 53 %. The calcium content of the soil was 0.2 %. Forty-five days after installation
the calcium content ranged from an average value of 2.5 % 25 mm outside the column to 1.65 % at
a distance equal to five times the column diameter. The authors conclude that the strength and com-
pressibility behaviour were improved over a radial distance of 4 to 6 column diameters.
In the course of investigations into the suitability of lime columns for slope stabilization in
England, Rogers & Glendinning (1994) presented results of a laboratory investigation into the
diffusion of lime in a lower Lias clay. The soil was compacted in layers in a box of dimensions
1000×500×500 mm. A hole of diameter 50 mm was filled with quicklime. Samples were taken at
various distances from the column on successive occasions. The results showed that migration of
lime took place to a maximum distance of 20 – 30 mm from the column. In subsequent field trials,
holes of depth 2.5 – 3 metres and diameters 63 mm and 200 mm were filled with quicklime (Rogers
& Glendinning 1997). The findings were similar: a zone up to 30 mm from the periphery of the

column was noticeably stiffer and drier in character. Radial cracks of length ~50 mm and width
2 – 5 mm had appeared around the columns.
A number of studies of molecular diffusion have been done in connection with research on the sur-
face stabilization of loose soils with lime. Fohs & Kinter (1972) carried out a study in which a lime
slurry was placed in contact with small blocks of compacted soil. The results of this study showed
that that migration of lime took place over a distance of 20 – 30 mm, i.e. in the same order of mag-
nitude as the findings by Rogers & Glendinning (1994, 1997).
A very extensive study of the diffusion of lime and cement in small lumps of loose soil was report-
ed by Stocker (1975). The study shows that lime can migrate through lumps 40 mm thick. Seepage
flow in the material presumably has a positive effect on lime migration, since water is a medium for
molecular diffusion (Lindh & Ydrevik 1978).
In the course of an extensive investigation of the feasibility of stabilizing loose soils with polymet-
ric hydroxy-aluminium, OH-Al, a study was made of changes in soil properties and pore-water
composition outside stabilized columns (Bryhn et al. 1988). The columns were installed by the
Swedish lime column method. Diffusion of cations, either added in the binder or resulting from ion
exchange, increased the remoulded shear strength within a zone extending up to 86 cm from the
column. On the other hand, a laboratory study of the diffusion of lime showed that OH-Al binder
only affected a zone within 100 mm of the stabilized column (Bryhn et al. 1984). Similar tests using
lime as binder showed that the zone affected by diffusion was only ~30 mm in thickness.
Åhnberg et al. (1995) carried out a series of laboratory tests in which lime-stabilized and cement-
stabilized soil was placed in cylinders with unstabilized soil on either side. The results showed that
calcium ion transfer was greater for the lime-stabilized than the cement-stabilized soil. An uncer-
tainty in the results made it difficult to assess the extent of diffusion, but it was estimated at no
more than a few centimetres. A thin layer of looser clay observed between the stabilized and the
unstabilized soil was attributed to shrinkage of the stabilized material. The tests were followed up
with field trials in which a Ljungskile clay was stabilized with 50 % lime and 50 % cement. Looser
clay was observed in the immediate vicinity of the column and was considered to be due to a reduc-
tion in the salt content of the clay.
Joshi et al. (1981) presented a study in which a slurry of slaked lime was injected into slurries of
bentonite and kaolin. Calcium ion migration was studied by strength measurements, scanning elec-
tron microscopy and X-ray studies. The results indicated an effect on strength in the soil up to
~30 mm from the calcium source. However, an elevation in the calcium ion concentration was de-
tectable up to 75 mm from the source, indicating that the calcium ion migration front lies considera-
bly further out in the soil than the front at which geotechnical parameters are affected.
In a study of the mixture quality of a number of lime-cement columns, reported in Chapter 7, the
water content of the soil was measured just beyond the column periphery. A number of columns
were recovered by means of a split sample tube of diameter 900 mm. Since the diameter of the col-
umns was substantially smaller, 600 mm, an amount of relatively undisturbed soil was recovered
Deep Mixing 85
along with the columns. The water content was measured in samples taken every 5 mm to a dis-
tance of 100 mm from the column. The results of four sampling sections are shown in Fig. 6.5. The
properties of the soil and the installation data of the columns are reported in Chapter 7. The results
show that the water content is affected within a zone up to 30 mm from the column, which agrees
fairly well with the majority of the studies reviewed above. The surface of the clay that was in di-
rect contact with the column, a zone of about 5 mm in thickness, was very stiff. The zone from
5 mm to 30 mm was also noticeably stiffer than the surrounding loose clay, Fig. 6.6. An interesting
observation was that the water content was somewhat lower in the soil just outside the column than
inside the column (0 – 5 mm). This may have been caused by the increase in pore pressure due to
mechanical mixing, and by consolidation due to water transport from the soil into the column dur-
ing hydration of the binder. The hardening material of the column itself increased in volume, sub-
jecting the surrounding soil to radial consolidation. Thus the material of the column was consolidat-
ed to a lesser degree than the surrounding soil, resulting in the column having a somewhat higher
water content.
95
90
Water contemt (%)
85
Column 32, depth 3.3m
80
75
70 Column 16, depth 2.4m
65
-10 0 10 20 30 40 50 60 70 80 90 100 110
Distance from column periphery (mm)
Fig. 6.5 Water content measured in surrounding clay at various distances from column
periphery (author's work, 1988).
Column periphery
≈ 30 mm
Soil stabilized with

lime and cement
Fig. 6.6 Boundary zone between column and surrounding soil (author´s work, 1998).

6.4 FACTORS AFFECTING THE MIXING PROCESS
The mixing process in dry deep mixing is a very complex one. A variety of factors influence the
process and its results:
• the rheology of the soil and the binder and the amount of binder;
• pressure conditions in the soil during column installation;
• the delivery pressure and the amount of air used;
• the geometry of the mixing tool;
• the mixing energy: the retrieval rate and rotation speed of the mixing tool;
• the consolidation stress, the compaction energy, the temperature and the availability of water,
which affect molecular diffusion.
This section reports and discusses studies which focus mainly on factors relating to the mixing tool,
i.e. the retrieval rate and rotation speed. We also discuss the influence of the binder, the consolida-
tion stress and the compaction energy. The soil rheology was discussed in Chapter 5.
Mixing energy
It is well-known today that the effective strain in the mixture (the retrieval rate of the mixing tool)
has an influence on the mixing process and its results. A longer mixing time increases the effective
strain and the dispersion of the binder in the mixture. Many studies have drawn this conclusion on
the basis of laboratory-prepared specimens without subjecting this factor to any closer examination,
e.g. Locat et al. (1990), Åhnberg et al. (1995). However, few studies have been carried out with the
explicit purpose of investigating the influence of individual factors on the mixing process.
In the Scandinavian countries the retrieval rate (mm/rev) of the mixing tool is used as a measure of
the mixing time. The effective strain in the mixture is a function of the retrieval rate. The intensity
of mixing (agitation) is a function of the rotation speed. In Japan, on the other hand, the mixing
time is measured in terms of the penetration speed (m/min), the retrieval speed, and the rotation
speed. To provide a measure of the mixing time these parameters are combined as mixing cycles
per meter T, given by (Yoshizawa 1997):
T = ∑ M × {(N d Vd ) + ( N u Vu )} (6.1)
where ΣM = number of mixing tool blades

Nd = rotation speed of mixing tool during penetration (rev/min)
Vd = mixing tool penetration velocity (m/min)
Nu = rotation speed of mixing tool during retrieval (rev/min)
Vu = mixing tool retrieval velocity (m/min)
In Japan part of the binder is often delivered as the mixing tool is being inserted into the soil, partic-
ularly when binder in liquid form is used. The binder thus assists the remoulding of soil by the tool.
Deep Mixing 87
Fig. 6.7 Relation between degree of mixedness a and total number of revolution R.
(Muro et al. 1987b).
a) b)
Fig. 6.8 a) Relation between compressive strength s and degree of mixedness a

b) Relation between compressive strength s and the number of revolution R
(Muro et al. 1987b).
(Note: 1 kgf/cm2 = 100 kPa)

Dry binder may also be delivered in this way in order to utilize the mixing tool more efficiently. To
take account of the fact that only part of the binder is delivered during insertion of the mixing tool,
the following expression may be used to calculate the number of mixing cycles per column metre T
(Hayashi & Nishikawa 1999):
T = ∑ M × {( N d Vd )× (Wi / W ) + ( N u Vu )} (6.2)
where Wi = quantity of binder delivered during penetration (kg/m3)

W = total quantity of binder (kg/m3)
In Sweden the retrieval rate (mm/rev) of the mixing tool is used as a measure of the mixing time.
The number of cycles per column metre T can be calculated as:
1
T = ∑ M × × 1000 (6.3)
s
where ∑ M = number of mixing tool blades

s = retrieval rate of mixing tool during withdrawal (mm/rev)
It will be noted that when investigating the effect of different rotation speeds in Japan, both the ef-
fective strain and the mixing intensity are varied. It is therefore difficult to separate these two pa-
rameters in Japanese studies, since the rotation speed is increased in order to obtain longer mixing
times.
Muro et al. (1987a, 1987b) studied the effect of mixing time and rotation speed in a series of labo-
ratory tests in which the mixing process was simulated by means of two types of equipment. The
tests were carried out using kaolin and cement as binders. Fig. 6.7 shows the results, in which the
degree of mixedness a is evaluated as a function of total number of revolution R. The degree of
mixedness a in the figure is equivalent to the mixing index defined by equation 4.13 in Chapter 4.
The water content of the mixture was used as a variable for evaluating the degree of mixedness.
The initial water content of the kaolin was 70 %. The tests were carried out at three rotation speeds,
20, 40 and 60 rev/min. Mixing took place in a circular tank of diameter 160 mm and height
160 mm. The mixing tool was rotated in a horizontal plane, i.e. at zero pitch.
The results show clearly that the mixing time or the total number of revolution has a major impact
on the degree of mixedness, particularly at short mixing times.
Unconfined compression tests were carried out to investigate the effect of mixing energy on the
strength of the laboratory-mixed soil. Fig. 6.8 shows the relations between degree of mixedness,
mixing cycles, and unconfined compressive strength.
Deep Mixing 89
The results of this study show that the degree of mixedness and the number of mixing cycles have a
major impact on the compressive strength. The degree of mixedness is proportional to the logarithm
of the number of cycles. Similar results were reported by Baker (1954) from laboratory mixes of a
clayey silt stabilized with cement.
Muro et al. (1987a, 1987b) also carried out mixing tests in order to investigate the relation between
mixing energy and unconfined compressive strength when using a vertically reciprocating mixing
tool in a container of diameter 157 mm and height 400 mm. Fig. 6.9 shows the relation between
unconfined compressive strength, the modulus of deformation at half the ultimate stress, and the
retrieval rate of the mixing tool.
Fig. 6.9 Relation between unconfined compressive strength s, modulus of deformation

at half the ultimate stress E50, and the retrieval rate of the mixing tool P.
(Muro et al. 1987a)

Fig. 6.9 shows some results which are more difficult to interpret than those in Fig. 6.8. As the re-
trieval rate is reduced from 10 mm/rev. to 2.5 mm/rev the strength decreases slightly, only to in-
crease as the number of mixing cycles is increased (the retrieval rate is decreased) further. There is
also a significant increase in strength as the number of mixing cycles increases from one to six cy-
cles. It is of course difficult to draw general conclusions from the tests, and the authors write that
further studies are needed in order to understand the mechanical relations between degree of mixed-
ness and the compressibility and strength parameters. That the degree of mixedness does have an
effect on these parameters is clear, however.
Another interesting study of the effect of mixing energy on the degree of mixedness was reported
by Nishida et al. (1996). Fig. 6.10 shows the results of mixing tests in which small cylindrical plas-
tic beads were mixed with three clays differing in sensitivity. The degree of mixedness was deter-
mined by counting the concentration of beads in different parts of the mixture. In this study, too, the
degree of mixedness was evaluated by the mixing index of equation 4.13. The tests showed that
mixing energy has a major impact on mixedness at short mixing times, and that a higher degree of
mixedness could be expected in clays of high sensitivity.
Fig. 6.10 Relation between mixing time and degree of mixedness. (Nishida et al. 1996)
Yoshizawa et al. (1997) report on a number of interesting studies carried out in Japan, mainly in the
1980s. Fig. 6.11a shows the results of a comparison between different rotation speeds and their ef-
fect on the compressive strength of the stabilized soil. Fig. 6.11b shows the results of a comparison
of different penetration velocities and their effect on compressive strength. These two studies show
that the mixing time or number of mixing cycles have an impact on the strength of the stabilized
soil. However, they do not show whether the rotation speed affects the result when the number of
mixing cycles is kept constant.
Deep Mixing 91
a) b)
Fig. 6.11 a) Relation between rotation speed and compressive strength.

(Nishibayashi et al. 1985, Yoshizawa et al. 1997)
b) Relation between penetration velocity and compressive strength.
(Enami et al. 1986a, Yoshizawa et al. 1997)
(Note: 1 kgf/cm2 = 100 kPa)
Yoshizawa et al. (1997) also report results of strength measurements on samples from of stabilized
soil which show that the coefficient of variation tends to fall as the mixing time increases,
Fig. 6.12a. Hayashi & Nishikawa (1999) reported results from both laboratory tests and field trials
in which the effect of mixing time on strength was studied in cement-stabilized peat. The investiga-
tion showed that mixing time had a significant effect on both strength and mixedness. As the
number of mixing cycles increased from 300 to 600 rev/metre, the coefficient of variation of the
compressive strength fell from 40 – 60 % to 15 – 40 %, Fig. 6.12b.
Appendix C presents a laboratory study of the effect of mixing energy on unconfined compressive
strength when mixing cement and lime-cement (Larsson et al. 1999, Walter 1998). In this study
compressive strength was used as a measure of mixture quality, on the assumption that strength will
increase as the binder is better dispersed in the mixture. This study too indicates that mixing energy
has a major impact on compressive strength at short mixing times, Fig. 6.13.
Omine et al. (1998) treated cement-treated soil as a two-phase mixture consisting of improved and
unimproved parts. In a series of laboratory tests the improvement rate (mixedness) was evaluated as
the ratio of improved surface to total surface. The improved area was measured by means of a plan-
imeter. A kaolin clay was treated with a cement slurry. Columns were manufactured in a cylinder of
diameter 150 mm and height 150 mm. The results of this study showed that the improvement rate
had a major effect on the compressive strength. A linear relation was obtained between compressive
strength and the number of mixing cycles.

a) b)
Fig. 6.12 Relation between number of mixing cycles and the coefficient of variation of the
strength achieved.
a) Mizuno et al. 1988, Yoshizawa et al. (1997).
b) Hayashi & Nishikawa (1999).
Fig 6.13 Compressive strength as a function of mixing energy. Test specimens treated
with lime-cement and cement. (Larsson et al. 1999, Appendix C)
Deep Mixing 93
In the Scandinavian countries binder is almost exclusively used in dry form and is delivered to the
soil as the mixing tool is withdrawn (see section 6.3). The mixing energy could be increased by
delivering the binder while the mixing tool is penetrating the soil. This would increase the mixing
energy (cycles per column metre) at the cost of only a slight increase in the time for column manu-
facture. One reason why this technique is not used in Scandinavia is that it is difficult and energy-
intensive to mix the soil and the binder once they have had time to react together. This problem is
particularly acute in the case of lime, as the mixture very rapidly becomes granular. A way of over-
coming the problem, not yet tested in Sweden, is to use a peptizing agent to change the rheological
properties of the mixture. An in-principle design for such a mixing process is shown in Fig. 6.14.
The peptizing agent could retard the chemical reactions, keeping the mixture loose and easily
mixed. Dry peptizing agents exist which could be mixed with powdered binding agents. A further
effect is that it could be considerably easier to evacuate the air that is injected into the soil when the
mixture behaves like a liquid. The problem of evacuating air is discussed in section 6.7. Problems
that have been much discussed are the uncertain quality of the base of columns, and whether the
columns are supported on solid underlying soil. Injecting the binder while inserting the mixing tool
would avoid the latter problem as the binder could be delivered from the base of the impeller.
The mixing tool mixes and loosens

the soil and the binder + peptizing
agent. The peptizing agent retards
the chemical reactions and loosens
the soil
Powder delivery takes place

behind the bottom paddle
The powdered mixture of binder

and peptizing agent is spread in
the surrounding soil during The tool mixes and compacts
insertion of the tool the soil and the binder as it is
withdrawn from the soil
Fig. 6.14 Principle of a mixing process in which binder and powdered peptizing agent are
incorporated in the soil as the rotating tool is inserted into the soil.

Type and quantity of binder
Walter (1998) and Larsson et al. (1999) presented a very simple laboratory device in which binder
was mixed directly with soil samples (in standard 50×170 mm sample tubes). Mixing binder with
undisturbed soil permits the rheological properties during the mixing process to be investigated.
The samples were mixed using two simple mixing tool geometries. The total mixing energy was
evaluated by varying the retrieval rate of the mixing tool and measuring the torque during mixing.
The results showed that in the order of two times more energy was consumed when the binder con-
sisted of a lime-cement mixture (powder) than when it consisted of cement alone. Since mixture
quality depends strongly on the mixing energy when the binder consists of cement, the implication
is that mixing tools can and should be used which are better able to transfer energy to the mixture.
Chida (1982) also concluded that the type of binder has an effect on the mixing process and its out-
come. Chida studied the effect of the type of binder, powder or slurry, on the strength of laboratory-
prepared specimens. The results showed that when powdered cement was used as binder, the
strength increased as a function of the mixing time during preparation of the specimens. With ce-
ment binder the compressive strength increased up to mixing times of around 3 minutes. With long-
er mixing times the compressive strength remained relatively constant, Fig. 6.15b. The results in
Fig. 6.15a showed that using cement powder and cement slurry, the dispersion of the results was
greater for short mixing times. The compressive strength is less dependent on mixing time when
quicklime is used as binder, a finding also reported by Larsson et al. (1999).
There is no question that the quantity of binder affects the strength of stabilized soil, however it can
also affect the distribution of the binder and the dispersion of the strength values. Asano et al.
(1996) studied the compressive strength and coefficient of variation while varying binder content
for three types of binder. The results showed that when using a cement slurry, the coefficient of
variation fell from ~40 % to ~20 % when the binder quantity was doubled. Matsuo et al. (1996)
found a similar drop in the coefficient of variation with a doubling of the binder (cement slurry)
content, Fig. 6.16a. However, their results suggest that the effect on the dispersion is greatest at low
binder contents. Matsuo et al. (1996) also report results of a study showing that the coefficient of
variation increases with the water:cement ratio, suggesting that a greater quantity of mixing water
impairs the mixing efficiency, Fig. 6.16b. The effect of a greater quantity of mixing water on the
mixing process depends on the initial rheological properties of the soil. When water is mixed with a
loose soil with a high water content, the soil breaks up into lumps which will be entrained by the
mixture while undergoing relatively little mixing themselves.
Deep Mixing 95
a) (Chida 1982) b) (Chida 1982)
Fig. 6.15 Relation between compressive strength and mixing time with different binders.
a) b)
Fig. 6.16 Change in strength and coefficient of variation of stabilized soil as:
a) the binder quantity is varied; b) the water:cement ratio in the binder is varied
(Nishibayashi et al. 1988, Matsuo et al. 1996).

Mixing tool design
No fundamental scientific studies have been published on the geometry of the mixing tool and its
effect on the mixing process in the context of deep mixing. The few studies on tool design that have
been reported are somewhat of a “sales” character, often presented by private companies at confer-
ences or in periodicals specializing in equipment.
Dong et al. (1996) studied the effect of tool geometry, mixing time and rotation speed in a series of
laboratory tests. Columns of diameter 400 mm and height 1000 mm were installed in a clayey sand
of compressive strength ~29 kPa. The binder was a cement slurry. Two different types of mixing
tool were used: type A, with two levels of paddles, and type B, with one paddle and a frame-like
blade configuration rotating in opposite directions. Fig. 6.17 shows the 7-day compressive strength
as a function of the retrieval rate for each tool.
Tool A
Tool B
Fig. 6.17 Relation between mixing tool retrieval rate and compressive strength
(after Dong et al. 1996, modified)
The results show that the retrieval rate or number of cycles of the mixing tool do have an effect on
the strength and that tool B is probably more efficient than tool A. The authors also write that the
degree of mixedness, based on visual observations, is better at higher rotation speeds. Unfortunately
it is difficult to separate the rotation speed from the number of mixing cycles, since lower speeds
were used with large retrieval rates and higher speeds with small retrieval rates.
Yoshizawa et al. (1997) reported a number of studies in which the effect of mixing tool geometry
was investigated. Fig. 6.18 shows a comparison of the achieved compressive strength when a tool
has one or four shafts. An example of a tool with multiple shafts appears in Fig. 6.31. The results
show that smaller dispersion and higher strength are achieved with four shafts than with a single
shaft. The reason for this, according to the authors, is that neighbouring shafts rotate in opposite
directions, producing interacting mixing effects in the overlap. The improvement in mixing effi-
ciency may be explained by the kneading action between the blades mounted on different shafts.
Deep Mixing 97
Fig. 6.18 Comparison of compressive strengths obtained with a tool having one or four
shafts (Nishibayashi et al. 1985, Yoshizawa et al. 1997)
Yoshizawa et al. (1997) further report results from a series of tests comparing two different config-
urations of six blades on a dual-shaft mixing tool, Fig. 6.19. The tool labelled “H” has six horizon-
tal paddles. In the tool labelled “0 – 1” the two middle paddles are replaced with two stirrups
(termed Oval DM according to Nishibayashi et al. 1985). The stirrups shear the earth in different
directions (compare with Fig. 2.5, Chapter 2.5). The authors write that the stirrups are particularly
effective in highly viscous (plastic) soils. However, it must be noted that the replacement of blades
with stirrups entailed fitting the mixing tool in this study with more blades (compared to the H-
tool).
Fig. 6.19 Comparison of compressive strengths obtained with two different designs of a
mixing tool (Nishibayashi et al. 1985, Yoshizawa et al. 1997)

Al-Tabbaa & Evans (1999) and Al-Tabbaa et al. (1999) compared three types of mixing tool in a
series of laboratory tests and field trials. However, no quantitative comparison of the tools was
made, the comparison being based on visual observation of the mixing results. The three tools stud-
ied in the laboratory are shown in Fig. 6.20. Tool 2 is a simplified version of a number of commer-
cially available tools. Tool 3 is similar to Tool 2 but has fewer blades and has cutting teeth at the
bottom. The tools were of diameter 6 – 9 cm. Tests were carried out with both water-based and
powdered binders. The results indicated that the tools achieved equally good results in pure sand. In
sandy clay and cohesive material Tool 2 was significantly more effective than Tool 1. Better mix-
ing results were also obtained when the soil was premixed before binder was added.
Fig. 6.20 Three mixing tools investigated in the laboratory. (Al-Tabbaa et al. 1999)
In model experiments it is important to study the impact of scale effects, such as small vertical and
horizontal stresses in the mixture. There is a risk of the mixture clinging to the rotating mixing tool
during withdrawal. The mixture may also clump together between the blades, particularly when
lime is used as a binder. Such effects were not investigated or remarked upon in the studies reported
above.
Workers in agricultural research have studied the behaviour of discs, e.g. harrows, in the soil. The
principle is to produce maximum movement in the soil with the least possible expenditure of ener-
gy, i.e. the same principle as in mixing loose soils with binders. Ali Al-Ghazal (1989) studied the
effect of different angles, diameters, depths and spacings of different circular discs. Simple models
of interaction between soil and discs were developed based on the Mohr-Coulomb failure criterion.
The concept of specific resistance (= force/area) was used to rate the efficiency of the discs. The
effects of different factors were studied in large-scale model tests, and the results showed that in-
creases in the factors angle, diameter, depth and spacing of the circular discs caused an increase in
the area of soil disturbance and a decrease in the specific resistance.
Deep Mixing 99
Rheological properties of the soil
The mixing of fine soils with binders is particularly difficult when the soil is extremely cohesive
and sticky at moderate to high water contents. The type of soil and its rheological behaviour have a
considerable impact on the efficiency of a mixing process. It is relatively easy, for example, and
takes only limited effort to mix cement and dry sand to produce a mixture with small concentration
variances. A considerably greater and more intensive mixing effort is necessary to mix cement with
a cohesive clay. Unfortunately, no extensive investigations have been made into the effect of soil
rheology on the mixing process in deep mixing.
Extensive studies of technical properties have been carried out in connection with shallow subgrade
stabilization with lime and cement (reviewed by e.g. Sherwood 1993). The properties of deep-stabi-
lized loose soils are similar to those obtained in surface-stabilized soils. The main difference be-
tween deep and surface stabilization is that most soils that are candidates for deep stabilization are
water-saturated and therefore not compacted. The laboratory procedures for manufacturing stabi-
lized soil are also very similar. For these reasons we shall also discuss experience of soil treatment
with lime and cement for the purpose of surface stabilization.
In the late 1950 extensive studies were carried out at MIT (Massachusetts Institute of Technology,
USA) on e.g. the effect of water content on the mixing process when mixing fine soils with lime or
cement (Freitag et al. 1961). Laboratory studies showed that the water content was one of the most
significant factors affecting both the power consumption and the time necessary to achieve a given
degree of mixedness. The tests showed that for a given degree of mixedness, the most favourable
water content in order to obtain the fastest possible mixing was around the plastic limit. However, it
was around this same water content that the greatest mixing energy was required. Table 6.1 shows
the effect of water content on a number of mixing characteristics. Note that it is easiest to mix a
completely dry powder or a liquid. The tests also showed that the effect of the rotation speed of the
mixing tool varied depending on the water content. At low water contents, below the plastic limit,
the most effective mixing was at low rotation speeds. At high water contents the most effective
mixing was obtained with high rotation speeds.
Table 6.1 Effect of water content on mixing process parameters. (Freitag et al. 1961)
Increasing water content →
State Powder Pellets and Pellets Plastic Sticky Liquid

powder
Energy consumption Low Low to Moderate High Moderate Low

moderate
Homogenization rate Rapid Slow Very slow Rapid Slow Rapid
Energy efficiency Very high Favourable Favourable Poor Favourable High

to poor to poor

Freitag et al. (1961) also reported results of experiments with the use of additives to change the
rheological properties of the soil. The tests showed that the energy consumption and efficiency
could be significantly impacted by the treatment with dispersing or aggregating agents. When the
water content of the soil was above the plastic limit it was found that aggregating agents were most
effective for reducing the energy consumption and improving mixing efficiency. Dispersing agents
did not increase efficiency or mixedness but did reduce the energy consumption in soils with high
water contents.
When powdered quicklime, CaO, is used as binder the rheological properties change very rapidly
once the lime comes into contact with the soil. The soil granulates and may become difficult to
work. If this rapid chemical reaction is not desirable, a peptizing agent may be used to retard and
delay the reaction process. A possible application, as mentioned above, is when binding agent is
delivered while the mixing tool is being inserted into the soil. However, no studies have been pub-
lished which show whether peptizing is an effective method for changing the rheological properties
of a mixture when using dry quicklime as binder.
Nishida et al. (1996) reported results illustratind the effect of the sensitivity of the soil, Fig. 6.10.
The results showed that a higher degree of mixing may be expected in clays with high sensitivity.
In Japan the possibility has been investigated of using the mixing tool as a rheological measuring
instrument during the mixing process (Hata et al. 1987, Aoi & Tsujii 1996). According to the au-
thors, the method may be developed by e.g. measuring the energy input during the mixing process
and relating it to soil properties.
Compaction energy
There has been discussion of how soon and how heavily the ground surface may be loaded after
deep mixing. It may be important to impose the greatest possible consolidation stress at an early
stage while the chemical reactions are at their most active. In Sweden normally no compaction is
done apart from that brought about by the mixing tool. Since it is customary in Scandinavia to use
air as the transport medium for a powdered binder, large amounts of air are injected into the soil.
Unless the air is evacuated, the compacting capacity of the mixing tool may assume great impor-
tance for the mixing process and the strength gain. Even surface compaction may have a significant
impact. The interfacial area between the binder and the soil is a measure of the mixture quality and
the efficiency of molecular diffusion. The compaction energy will thus have an impact on the mix-
ing process, since increased density reduces the distance between the binder particles and the soil
particles. In Japan the binder is generally premixed to a slurry, with the result that compaction is of
little consequence provided the soil to be stabilized is water-saturated.
In shallow subgrade stabilization, compaction of the stabilized material is one of the most important
operations (Assarsson 1972). It is well-known that the compaction energy and the type of compac-
tion have an effect on the density and strength of non-saturated cohesive soils (e.g. Bell 1977). An
important factor in compaction is the delay between the incorporation of the binder and the com-
paction operation (“aging” or “mellowing”). This delay may affect the compaction properties, pro-
ducing a loss of strength in stabilized soil (e.g. Uppal & Bhasin 1979, Sweeney et al. 1988, Sivapul-
Deep Mixing 101

laiah et al. 1998a, Holt & Freer-Hewish 1998). Sivapullaiah et al. (1998a) conclude from their own
laboratory tests that the delay between binder incorporation and compaction has varying effects on
strength depending on the binder (slaked lime) content. The longer the delay, the greater is the
strength loss and the lower is the density attainable by compaction. With lower quantities of binder
the delay has negligible effect on density and strength. With larger binder quantities, on the other
hand, there is a considerable effect on the strength and compaction parameters, and therefore, ac-
cording to Sivapullaiah et al. (1998b), a long delay between mixing and compaction should be
avoided.
Lindh & Hjalmarsson (1975) also concluded from laboratory tests that a long delay impairs strength
at high cement contents. The delay has a particularly large impact on cement-stabilized soil (Kézdi
1979). A study reported by Sivapullaiah et al. (1998b) showed that impact of delay is affected by
the composition of the binder, making it difficult to draw general conclusions.
With a rotating tool equipped with paddles or similar arrangements, the effectiveness of compaction
depends on the kneading capacity of the tool. It is important that the energy input from the mixing
tool is directed axially downwards in the mixture. It is further important that no part of the tool
tends to lift the material or cause it to catch in the tool and thus be pulled upwards as the tool is
withdrawn.
6.5 LABORATORY PHILOSOPHY

Field studies of the mixing process are difficult and costly to perform. It is therefore appropriate for
fundamental studies to be carried out in the laboratory environment using materials with controlla-
ble rheological properties and mixing tools with simple geometries. Such work may shed light on
which phenomena are likely to be the most relevant for field trials. It should not aim to provide
directly applicable data but to yield results that will suggest new approaches to the mechanism be-
hind the real problem. To ensure that new findings are relevant, laboratory tests must keep in close
contact with real-world problems. This will also ensure that problems are not over-simplified. Un-
fortunately, current laboratory methods for preparing test specimens tell us little about the mixing
process. Some of the parameters reported in section 6.4 lie far outside the range of application of
deep mixing. Examples are the laboratory preparation of specimens consisting of completely
remolded soil and the mixing of specimens many times more than it is reasonable to do in the field.
Such methods can teach us nothing about the soil's rheological properties, which are very important
parameters of the mixing process. Current laboratory methods are fixated on the factors binder type
and binder quantity. The primary purpose of laboratory tests is to tell us whether it is possible to
stabilize a soil (Carlsten 1991).
It is an established opinion in the deep mixing industry that the mixing process cannot be simulated
in the laboratory (e.g. Terashi 1998, Bruce et al. 1998a). The only variables can be simulated are
the type and quantity of binder. However, there is an aspiration to discover an empirical relation,
based on accumulated experience, between the strength achieved in the laboratory and that
achieved in the field. But such empirical relationships can only be found if the process parameters
of the laboratory tests are within the range of application of the method. Since the factors affecting
mixing as discussed in section 6.4 are not shown to be mutually independent, there is a serious risk

that incorrect conclusions will be drawn if one or more of these parameters are kept constant far
outside the application range.
Fig. 6.21 presents some of the reported relations between strength measured in the field or on sam-
ples taken from in-situ stabilized soils, and the measured strength of laboratory-prepared speci-
mens. The purpose of presenting this data is to illustrate the difficulty of attempting to apply experi-
ence drawn from comparisons between lab and field. It is obviously difficult to apply such experi-
ence when one is comparing results of tests done by different methods or based on irreconcilable
assumptions. The relation between field and lab strength measurements also varies with the number
of measurements.
Fig. 6.21 Relation of field results to results from laboratory-stabilized soils in a number of
reported studies.
A comparison between results obtained in laboratory tests and results from field trials or from
measurements on samples of in-situ stabilized soils is further complicated by the variation in the
mixture quality, and hence the strength, of stabilized soils. In Japan compression tests are carried
out on drill cores from stabilized soils as a means of quality control. Fig. 6.22 shows results from a
number of reported studies in which the coefficient of variation of compression tests was evaluated.
These results show that while the coefficient of variation is normally in the range 0.2 – 0.4, larger
dispersions of results can be observed. According to Suzuki (1982), Honjo (1982) and Kawasaki et
al. (1984), the dispersion in plan over column cross sections is of the same order. The wide varia-
tions in the coefficient of variation and in the relation between laboratory tests and field trials re-
ported by Hayashi & Nishikawa (1999) is caused by the variations in the mixing time (number of
mixing cycles) in their study.
Deep Mixing 103

0.7 1. Suzuki (1982)
Honjo (1982)
0.6
Kawasaki et al. (1984)
Coefficient of variation
2. Mori et al. (1997)

0.5
3. Babasaki & Suzuki (1996)
0.4 4. Saitoh et al. (1996)
5. Hosomi et al. (1996)
0.3 6. Unami & Shima (1996)
7. Hayashi & Nishikawa (1999)
0.2 8. Futaki et al. (1996)
9. As ano et al. (1996)
0.1
10. Axelss on & Rehnman (1999)
0 11. Braaten et al. (1999)
0 1 2 3 4 5 6 7 8 9 10 11
Reference
Fig. 6.22 Coefficient of variation evaluated from compression tests in a number
of reported studies.
It is unusual that the number of samples of stabilized soil extracted in a Swedish project is sufficient
to permit evaluation of the coefficient of variation in e.g. compression tests. However, Axelsson &
Rehnman (1999) presented a large number of results from lime-cement columns. Fig. 6.22 also
presents the coefficient of variation from a large number of compression tests on specimens from
lime-cement columns in Norway (Braaten et al. 1999).
With column penetration tests, which test a large part of the cross-sectional area of a column
in-situ, large variations can be obtained between different columns. Kujala et al. (1985) report
standard deviations in the order of 15 – 60 % of the mean.
One purpose of today's standards for the laboratory preparation of specimens is that they should
result in a best improvement effect in respect to the degree of mixing which should be considered as
a sort of index for the soil in concern (Babasaki et al. 1997). The result is generally assessed on the
basis of strength measurements. This maximum degree of mixedness is to be regarded as an index
of the soil in question and is to be correlated against field data. However, a number of factors indi-
cate that it is not possible to regard laboratory measurements as the best improvement effect. In
section 6.4 we discussed a number of factors affecting the mixing process and its result. It should be
beyond question that the mixing process is extremely complex and that the same factor may have
different effects on it under different conditions. An example of this is the observation that different
laboratories may return different results even when stabilizing the same soil with the same binder
(Edstam & Carlsten 1999). There are often one or more parameters in which laboratories differ.
Thus, having a standardized laboratory method is not the same thing as achieving best or optimum
improvement effectiveness. Another factor that should be kept in mind is that a lengthy mixing pro-
cedure can totally alter the soil structure and in the worst case result in “over-mixing”. Pousette et

al. (1999), reporting on laboratory mixing of specimens of peat, cement and slag, noted that long
mixing times had a negative effect on stabilization, which was attributed to the breaking down of
the peat fibres.
Temperature and water content during storage have a major effect on the properties of laboratory
specimens. A major factor in the discrepancies between laboratory tests and field trials is the differ-
ence in temperature during hardening (Viatek 1997). Hence there is discussion of whether the
in- situ temperature regime can be simulated in the laboratory, and Viatek (1997) present examples
of temperature curves for different binders.
When quicklime is used as a binder the mixture dries out, and the sample may require wetting in
order to allow the lime to hydrate and restore the natural water content (Brookes et al. 1997). The
procedure for simulating seepage of water from the surrounding soil into the stabilized soil is rela-
tively complex. Trials by Brookes et al. (1997) showed that the wetting of specimens over a long
period resulted in softening of the stabilized soil, which is not characteristic of in-situ behaviour.
In a variety of other applications and sciences there has been a growing trend towards simulating,
wherever possible, the processes that occur in the field. Four examples of such applications in the
construction industry are:
1) Shallow subgrade lime stabilization. British Lime Association (1990) states that when manufac-
turing laboratory specimens it is important that the laboratory preparation does not differ signifi-
cantly from field conditions;
2) Paving works (VTI 1999). Specimen preparation by the Marshall method is being increasingly
abandoned because the properties of specimens produced by this method do not conform to
those in the field;
3) Soil compaction (Venkatarama Reddy & Jagadish 1993, Kouassi et al. 2000). The laboratory
compaction of soil samples by the Proctor method is increasingly criticized because the method
does not resemble field compaction in terms of energy input and compaction mechanism;
4) Block stabilization (Hoikkala et al. 1997). In Finland a procedure has been developed for the
laboratory mixing of peat samples which simulates the actual field loads.
The present review of the literature on laboratory tests indicates that efforts are being made to
standardize laboratory mixing and testing, although the prevailing opinion is that laboratory simula-
tion of mixing processes is not possible. There are also efforts directed at laboratory simulation of
certain parameters of the mixing process. In the literature, however, one parameter is generally con-
sidered at a time. Nobody has yet attempted to simulate the whole mixing process.
Unless laboratory tests aspire to add to our knowledge of the mixing process in deep mixing, our
understanding of the parameters affecting this process is likely to remain limited for a long time to
come. This may well impede the further development of deep mixing methods.
Deep Mixing 105

6.6 MIXING TOOLS
Mixing tools in Scandinavia
Classification of existing tools

The most widely used mixing tool in Sweden at present is that referred to as the “standard tool”,
Fig. 6.23. A number of variations exist, but the differences in terms of the basic mixing mechanisms
are slight.
Fig. 6.23 The standard mixing tool. (Carlsten & Ekström 1995)
In its basic geometry the standard tool resembles impellers of the anchor agitator type, see Chapter
3. However, there are essential differences between them as a result of which their dominant mixing
mechanisms differ.
Anchor agitators are used for mixing high-viscosity liquids, pastes and plastic materials. They are
less well suited for mixing viscoelastic materials in which forces producing movement rapidly die
out due to elastic drag and shear resistance. In a mixture with these properties it can be difficult for
an anchor agitator to produce movement in every part of the mixture. Anchor agitators are equipped
with paddles which are generally vertical, round, or tilted at ~45° to the horizontal in order to force
the largest possible volume of feed into movement around the blades.
The blades of the “standard tool” are generally tilted at a very small angle to the horizontal
(~10°–20°). This means it is difficult for the tool to produce the tangential movements in the mix-
ture which are the dominant flow pattern for paddles. The small angle of the blades to the horizon-
tal rather places the tool in the propeller category. Propellers generate mainly axial flow, see Fig.
3.1. (They may be compared with ceiling fans, which even at low speeds generate axial flow in the
room air; but a propeller operating in a liquid must run at high speed to produce sufficiently large
movements.) On the basis of this classification the “standard tool” may be classed as being interme-
diate between a paddle and a propeller.

When mixing loose soils, no significant axial movement of the mixture is likely to occur with exist-
ing tools. The tangential and axial movements generated by the ”standard tool“ are limited to a very
small volume close to the blades, as illustrated in Fig. 5.6. Dispersion of the lime and cement parti-
cles requires movement of the ingredients to occur so that the necessary shear forces may be pro-
duced, the materials deformed, agglomerates broken down, the interfacial area between the compo-
nents increased, and the scale of segregation reduced. Without movements in the material the added
components cannot be distributed in the mixture.
With each cut that the blades of the tool make through an mixture element, the thickness of the ele-
ment is reduced, i.e. the mixing process takes place through the distributive mixing mechanism.
Unfortunately the “standard tool” does not give the sliced elements much opportunity to redistribute
themselves between successive slicing actions. It is also probable that the existing tool does not
slice and remould the elements a sufficient number of times for the distributive mixing mechanism
to be effective. Distributive mixing is generally unsuitable as the dominant mixing mechanism for
the dispersion of solid particles in plastic and viscoelastic materials, where high shear stresses are
necessary in order to break down agglomerates.
The dominant mixing mechanism with the “standard tool” is distributive mixing, plus laminar
shearing within a very restricted volume close to the blades. For its primary purpose – dispersing
lime and cement particles in the soil – we must unfortunately conclude that the “standard tool” is
not a very efficient or appropriate mixing tool for all the applications in which deep mixing is used.
Particularly in the case of very loose clays with a high water content, its limited capacity to gener-
ate movement in the mixture may give rise to problems.
Tool development in Scandinavia

Development of equipment for dry deep mixing in Sweden was begun in the early 1970s by
Lindén-Alimak AB, and the first prototype may be seen in Fig. 6.24a. The method was adapted for
the market by BPA Byggnadsproduktion AB (Boman & Broms 1975). During the early phase of
development a variety of types of mixing tool were tried, and the tool shown in Fig. 6.25a was
found to give the best results in loose clay. In coarser and more solid soils an auger type tool gave
better results. The aim in the early stages of development was a device of high production capacity.
In the years 1975 – 1979 many types of mixing tool were tested (Wikström 1979). A tool with a
number of tilted vanes (paddles) was trialled, but it was found that the clay stuck to the tool, imped-
ing mixing. In the course of development the feed orifice was moved from below the top vane as in
Fig. 6.25a to a position above the vane as in Fig. 6.23 and Fig 6.25b. The trials of the early tools are
unfortunately unpublished and it is therefore not possible to review and comment on the results.
Since the initial development of mixing tools in the 1970s, most projects have been carried out with
tools of the type in Fig. 6.23. Slight variations appear in Figs 6.25b-d. Further tool development has
been very limited. On the machinery side, however, development has been dramatic, as Fig. 6.24
may illustrate. How does it come about that the mixing equipment and the in-situ mixing process
have remained practically unchanged? The answer may be that there is a lack of basic knowledge of
the mixing process and the fundamental mixing mechanisms. Why is this knowledge still lacking
after nearly thirty years' experience?
Deep Mixing 107

a) Lime column machine.
(Broms & Boman 1975)
b) Lime column machine, Junttan.
(Stabilatören 1999)
Fig. 6.24 Machinery development from first prototype to today's machine type.
a) Mixing tool from early 1970s b) Mixing tool from late 1990s
(Boman & Broms 1975) (Photo: Morgan Axelsson)
c) Standard tool modified for winter conditions d) Standard tool, Hercules

(Rathmayer 1997) (Johansson & Jons 1995)
Fig. 6.25 Some variations on the “standard tool.”

Until the 1990s the main use of the method was for reducing settlement and increasing stability for
relatively low embankments. From the late 1980s, however, it began to be applied to stabilize exca-
vations, cuttings and increasingly high embankments. In response to increasing demands for con-
trol, columns began to be exposed for visual inspection and even whole columns were recovered.
Some columns exhibited a binder distribution that was not acceptable to the client. An answer to the
above question may thus be that problems with column quality did not appear on the agenda until
the end of the 1980s.
The development of mixing tools has been slow, and one reason for this may be that there is cur-
rently no incentive for contractors to develop new tools. Today's tendering procedures, with general
contracts and conditions “recommending” a tool of the type in Fig. 23, make it economically unat-
tractive for contractors to develop new tools. Another explanation may be that wear on the tools is
severe and more complicated designs are prone to break down (Bredenberg 1999).
One problem is that improved column quality may make deep mixing more expensive. It is well-
known today that mixing energy affects the quality and the strength gain of the column. It may be
difficult to develop a tool that produces better results but does not require a higher energy input or
longer mixing times. If the column quality normally achieved today is no longer accepted, a higher
production cost may unfortunately be unavoidable due to longer mixing times and higher energy
consumption.
There is a lack of methods for assessing mixture quality. As discussed in Chapter 4, mixture quality
is often assessed and compared on the basis of the finished product, sometimes using methods that
have no direct relation to the performance of the mixing tool. Thus the difficulty of comparing dif-
ferent mixing tools in the field is a further reason for the slow pace of development.
The efficiency of different mixing tools may vary depending on the rheological parameters of the
soil. There is very little knowledge of these parameters under conditions of extremely rapid defor-
mation. This is be partly because there may have been little need for such knowledge, but it is also
because the necessary studies are difficult to perform. For plastic soils and “slow” deformations,
rheological instruments such as shear boxes and triaxial compression apparatus are used. For liquids
and fluids instruments such as the cone and plate viscometer are suitable. In consolidated undis-
turbed or partially remoulded loose soils, slip faces easily develop (see Fig. 5.6), making rheologi-
cal parameters difficult to evaluate. It is important to note, however, that the mixing tool and mix-
ing process themselves are a rheological measuring instrument in deep mixing. By measuring the
force applied to the system in the form of torque, it may be possible to study the rheological proper-
ties of the soil.
A number of attempts have in fact been made to develop new mixing tools. However, testing is
generally very limited and the results are rarely reported. There is therefore a severe lack of experi-
ence in Scandinavia of the performance of different types of tool under different conditions.
Fig. 6.26a shows a mixing tool that has been used in many projects by Sweden's largest deep mix-
ing contractor. However, it has not been possible to demonstrate that this tool differs in efficiency
Deep Mixing 109

b) Modified KC mixing tool used
for cell stabilization. (Jelisic 1999)
a) ”Pinnborr“, LC-Markteknik.
(Larsson 1999)
c) Tool developed by Hercules-Norén. d) Double-vaned tool.

(Johansson & Jons 1995) (Johansson & Jons 1995)
Fig. 6.26 Mixing tools that are used or have been tested in Scandinavia.

from the “standard tool”. An examination of the shapes of the tools shows that they are unlikely to
differ significantly, as both are characterized by paddles set at a very shallow angle to the horizontal.
Fig. 6.26b shows a mixing tool that has been used for cell stabilization (Jelisic 1999, Rogbeck et al.
1999). It has curved blades, making the soil less likely to cling to the tool during mixing. This phe-
nomenon can be particularly troublesome when stabilizing peat. However, the tool has a negligible
angle to the horizontal, making it difficult to generate movements in the mixture. The same feature
also impairs its compacting capacity. Still, it would be interesting to study the effectiveness of this
tool with blades mounted at a steeper angle.
A number of mixing tool designs were studied in the course of field trials reported in a thesis pre-
sented at Chalmers University of Technology, Sweden (Johansson & Jons 1995). Fig. 6.26c shows
a tool in which some of the blades have been fitted with upward-pointing combs. The concept is
that the mixture will be compressed between the combs and simultaneously pressed downwards and
outwards away from the blades. Unfortunately the blades are not inclined to the horizontal, causing
the tool to lift the mixture during mixing. Moreover, the combs on the blades are very closely
spaced, which may cause mixture to get stuck between them. The result was, as Johansson & Jons
observed, that large quantities of mixture clung to the tool as it was withdrawn, leaving large holes
in the ground. Another consequence was that the remaining material in the column was poorly com-
pacted.
Fig. 6.26d shows another tool studied by Johansson & Jons (1995). It had blades angled different
ways, causing shearing of the mixture in various directions. The blades are rather closely spaced,
however, which may result in material sticking between them.
Mixing tools in Japan

Numerous Japanese firms have developed variations of deep mixing, resulting in a large number of
mixing tools. Some of these are illustrated and commented upon briefly in this section. It should be
noted that numerous patents exist on devices developed in Japan.
The first method to be developed in Japan, concurrently with its development in Sweden, was the
“dry method”, using quicklime in powdered form as binder. Subsequently Portland cement
premixed with water was introduced. According to Saitoh et al. (1985) this development took place
because injecting the binder into the soil in liquid form yielded a more uniform mixture with the
soil and increased strength. The method is preferable in “more difficult conditions” because it
makes the mechanical mixing of the soil easier (Bruce et al. 1998b). Deep mixing by the “wet
method” has been extensively used in Japan, particularly in marine projects.
Fig. 6.27 shows mixing tools used in deep mixing by the “dry method”, based on the principle that
the binder is spread over the column cross section in the cavity formed behind the blades of the tool
as it rotates in the soil. Faster rotation speeds increase the volume of the cavity and create a vacuum
which facilitates the process when manufacturing large-diameter columns (Chida 1982). Further,
according to Chida (1982), this process causes the injected air to move towards the outer parts of
the cavity. The air is then evacuated from the column via the shaft of the mixing tool.
Deep Mixing 111

Fig. 6.27 Mixing tools in which the binder is distributed in a cavity formed behind a paddle
(Tateyama et al.1996, Porbaha 1999, Chida 1982)
Fig. 6.28 shows a tool that was trialled in the early 1980s (Nishibayashi et al. 1984). The tool and
the method were known as “Oval-DM”. The tool is fitted with combinations of screws, thick and
thin paddles, teeth and stirrups. A variety of other basic designs of mixing tool and different ways
of incorporating binders into the soil were tested at this time, see Fig. 6.29. One of these was the
method of allowing the binder to spread over the cross section of the column in the cavity formed
behind the paddle of a mixing tool.

Fig. 6.28 The Oval-DM. (Nishibayashi et al. 1984)
Fig. 6.29 Mixing tools and ways of spreading binder into the soil. (Nishibayashi et al. 1984)
There are a variety of ways of injecting water-based binders into the soil. A common approach is to
inject part of the binder as the mixing tool is penetrating the soil, e.g. from the tip of a screw or
paddle. As the mixing tool penetrates, it disaggregates the soil and at the same time lifts it slightly
in order to facilitate incorporation. The remainder of the binder is injected as the tool is withdrawn
from the soil. Some typical tools are shown in Fig. 6.30.
Deep Mixing 113

Fig. 6.30 Mixing tool for the "wet method". (Taki & Bell 1998, Yoshida 1996)
It is common in Japan to install a number columns simultaneously using multiple shafts and mixing
tools. It is possible by this method to efficiently construct configurations such as blocks, lattices,
walls etc. (e.g. Kawasaki et al. 1981, Nicholson 1998). It is common to reinforce these columns
with I-beams (Schaefer et al. 1997). Fig. 6.31 shows some examples of mixing tools used in combi-
nation.
Fig. 6.32a shows a mixing tool for manufacturing columns of diameters between 1.4 and 2.0 m. By
providing the tool with an outer and an inner rod rotating in opposite directions it is possible to mix
frictional materials. Fig. 6.32b shows a tool with an auger to compensate for the expansion of the
soil that takes place when the binder is injected. This expansion can cause horizontal displacement
of the ground.
In Japan it has also been a common practice to install rectangular columns. This technique was de-
veloped in the early 1980s (Khono 1984). Examples of equipment for the purpose appear in
Fig. 6.33.
Fig. 6.34 shows the effect of fitting the mixing tool with “anti-rotation vanes” which prevent the
materials being mixed from rotating along with the tool. These vanes do not rotate in the soil and
thus provide resistance to the rotating blades as they pass. The mixing mechanisms with this tool

Fig. 6.31 Mixing tools used to manufacture a number of columns simultaneously.
Deep Mixing 115

a) b)
Fig. 6.32 a) Mixing tool for large column diameters in frictional soil. (Isobe 1996)
b) Mixing tool which compensates for expansion due to binder injection.
(Hirai et al. 1996)
Fig. 6.33 Mixing equipment for rectangular columns. (Watanabe et al. 1996)

Fig. 6.34 Effect of “anti-rotation vanes.” (Yoshizawa 1997)
are thus similar to those of the mixers shown in Fig. 3.9. The tool in Fig. 6.30 is also fitted with
blades which do not rotate during mixing. These vanes are somewhat longer than the rotating
blades and thus cut into the surrounding soil for support. However, the technique was previously
presented in 1986 by Enami et al. (1986b).
Figs 6.35–6.37 illustrate methods which are combinations of deep mixing using mechanical mixing
and jet mixing. The main advantage of these methods is that they can produce large-diameter col-
umns without large, bulky mixing equipment. The oldest of the methods is SWING, developed at
the beginning of the 1980s (Kawasaki et al. 1996, Yang et al. 1998, Ogawa 1990). New methods
have subsequently been developed incorporated evolved forms of jet mixing (e.g. Miyoshi 1996,
Mori et al. 1997, Ueki et al. 1996).
Fig. 6.38 illustrates a new method that has been trialled in Japan. A continuous auger is used to mix
the soil and the binder. In an initial step soil is transported to the ground surface as the rotating heli-
cal mixer is inserted into the earth. The soil is mixed in a mixing plant with a water-based binder
and then pumped back into the ground as the tool is withdrawn. This method yields a controllable
product, comparable to concrete, whose strength and deformation properties can be varied.
Deep Mixing 117

Fig. 6.35 Combination of deep mixing and jet mixing: the SWING method.
(Kawasaki et al. 1996, Ogawa et al. 1990)
Fig. 6.36 Combination of deep mixing and jet mixing: JACSMAN (Miyoshi 1996)

Fig. 6.37 Combination of deep mixing and jet mixing: the LDis-method (Ueki et al. 1996)
Fig. 6.38 Deep mixing with factory mixing (Mori et al. 1997)
Deep Mixing 119

Other countries
Deep mixing soil equipment has been developed in other countries such as England and Italy (Har-
nan 1993; Paviani & Pagotto 1991). Fig. 6.39 shows two such devices which have been used in
Europe.
The Colmix method, developed by Bachy in the late 1980s, involves mixing the soil with a water-
based or dry binder by means of a helical tool. The binder is injected as the tool penetrates the soil.
Mixing and compaction take place as the tool is withdrawn.
The Trevimix method was developed in Italy and has many similarities to the Japanese method
shown in Fig. 6.27.
a) b)
Fig. 6.39 a) Colmix. (Bachy 1999)

b) Trevimix. (Paviani, A. and Pagotto, G. 1991)

6.7 DISCUSSION OF SOME CHARACTERISTIC DEFECTS
Graininess
On visual inspection after exposure of the top of a column, a grainy, porous structure is often ob-
servable. This phenomenon is probably attributable to insufficient compaction. On incorporation of
binder into the soil, the water content drops while the plastic limit increases, and the soil assumes a
more granular structure. Lumps of material form showing a localized structural change on their
surface. Unless the stabilized material is compacted within a relatively short time (a few days), it
will have low density and low strength.
This phenomenon is particularly clear in dry surface clay but can also occur in the looser layers
further down the column. A grainy, lumpy texture at this depth may be due to failure to evacuate all
the air from the column. Once again the phenomenon is due to inadequate compaction.
Craters
Craters are a frequent phenomenon in dry deep mixing. They range in depth from a few tens of
centimetres to a couple of metres. The phenomenon is most easily illustrated in model scale studies
of the mixing process for column manufacturing (Larsson S.1999). Crater formation was clearly
observed when manufacturing lime-cement columns of diameter 50 mm and length 500 mm in a
shear box. The manufacturing process is shown in Fig. 6.40. A tube was driven into the soil, its
lower end being sealed by means of a plug fitted to the end of a rod. As the tube was inserted it
displaced soil, and slight surface heave took place corresponding to the volume of the tube. The rod
was withdrawn from the tube, which was then filled with dry binder. The tube was withdrawn from
the soil, leaving a pile of air and binder in its place. A rotating mixing tool of diameter 50 mm was
inserted to the desired depth. The direction of rotation was reversed and the mixing tool withdrawn.
The angled blades of the mixing tool compacted the soil and simultaneously dispersed the binder.
During this process the air is extracted along with the mixer. When the air reaches the surface, a
crater forms. Under field conditions the mixing tool may also pull the top layer of dry clay up with
it, exacerbating the crater effect.
The depth of the crater depends primarily upon the amount of air injected, the rheological proper-
ties of the soil, and the compaction characteristics of the mixing tool. According to one hypothesis,
when the soil is very loose and behaves like a liquid, the air is easier to evacuate and hence a deeper
crater forms. When the soil is stiffer, the compaction characteristics of the mixing tool have a major
effect on the amount of air evacuated with the tool. If these characteristics are poor, much of the
injected air will be left in the column and hence the crater will be shallow, while the column may be
porous and grainy in character.
Deep Mixing 121

Fig. 6.40 Crater formation in model-scale column manufacture.

Tapered or tubular columns
Uneven spreading can result in an accumulation of binder around the column periphery or around
its centre, Fig. 6.41. This may be the result of delivering the binder into the soil by means of a jet.
The upper paddle on the mixing tool passes through the soil and creates a cavity where the binder
can be spread over the column cross-section. The spreading of binder can be uneven since the bind-
er flow is not necessarily even over the section. A high air pressure can contribute to an uneven
spreading and furthermore cause pneumatic fracturing of the clay. Under unfavourable conditions
the high air pressure can fracture the soil and lift the soil mass above the mixing tool. This process
can result in large variances in binder concentration across the column cross section. The relatively
short mixing time is then insufficient for the tool to disperse and distribute the binder over the cross
section. For similar reasons, binder may concentrate in the central part of the column. Tapered col-
umns can occur when the paddle on the mixing tool is unable to form the requisite kavity and the jet
is unable to spread the binder over the whole column cross section. The problem is more severe at
depth, since the confined in-situ pressure is high and the strength properties of the soil often im-
prove with depth, i.e. the rheological properties of the soil change and its yield point increases at
higher pressures. Greater input of energy may thus be required to form a kavity and for adequate
disaggregation of the soil. Mixing tools of existing designs do not generate the movements neces-
sary to distribute the components of the mixture throughout the volume being mixed. The short
mixing time may be another reason for the poor distribution.
Short mixing times are obviously desirable in any type of mixing process. In many applications,
however, satisfactory results cannot be achieved with short mixing times because of limitations of
the equipment, usually lack of power.
Fig. 6.41 Example of concentration of binder around the column periphery.

(Axelsson & Larsson 1994)
Deep Mixing 123

7. Assessment of mixture quality:
Arboga field trial
7.1 INTRODUCTION
Deep mixing by means of lime-cement columns is very extensively used in Swedish infrastructure
construction. Newer and bolder applications, such as slope stabilization and the stabilization of high
embankments, impose ever increasing demands on the quality of columns, in particular the disper-
sion of the binder, in order to secure predictable and verifiable properties. The functions of deep
mixing, differing with the application, entail different requirements on column properties. Thus the
demand for uniform distribution of binder in the columns may vary depending on the requirements
of the final product. New applications may demand better mixture quality and better control of the
state of the mixture. Both clients and contractors need methods for assessing mixture quality.
No standard procedure exists today for the verification and assessment of mixture quality in respect
of the dispersion of binder through the column volume. Routine checks such as column penetration
tests, which are done to assess column strength and continuity, are not normally suitable for assess-
ing mixture quality, since the strength parameters of a stabilized soil depend on other factors be-
sides the efficiency of the mixing process.
In Chapter 4 we discussed statistical models and a range of mixing indices for the quantitative as-
sessment of the quality of various types of mixture. The experience and the methods referred to
there are mainly drawn from the process industries. In the context of deep mixing, mixing indices
and statistical models have only been used in a few Japanese studies. Muro et al. (1987a,b) used a
mixing index to assess the degree of mixedness of laboratory-prepared specimens. The degree of
mixedness was used as a quantitative estimate of the effect of mixing process parameters such as
the mixing time (number of mixing cycles) and rotation speed. The results of this study are reported
in Chapter 6, Figs 6.7-6.9. Nishida et al. (1996) also used a mixing index to assess the degree of
mixedness of laboratory-prepared specimens. The degree of mixedness was used to investigate the
effect of mixing time and soil sensitivity on mixture quality. The results of this study are reported in
Chapter 6, Fig. 6.10.
In order to assess mixture quality as a function of the mixing time of stabilized soil in the laborato-
ry, Omine et al. (1998) measured colour variations in the stabilized material and defined an im-
provement rate as the ratio of improved area to total area. This measure of the improvement rate
was used to validate a proposed model for predicting stress-strain relationships in mixtures of dif-
ferent materials.
In quality control of the strength properties of stabilized soil, statistical analyses are increasingly
utilized for the evaluation of results. A quality index, expressed as the coefficient of variation, was
used by Porbaha et al. (1999) to evaluate the results of CPT probes in a cement-stabilized soil (sta-

bilized fill). In this study the quality index is proposed to form the basis for the quality control of
stabilized soil. A statistical model from Finland uses the mean and the standard deviation to evalu-
ate the results of column probes (Halkola 1999). The evaluation is done on strength parameters.
In a number of Swedish studies workers have attempted to assess mixture quality on the basis of
subjective visual inspection. Such studies have been reported by e.g. Larsson & Axelsson (1994),
Johansson & Jons (1995) and Holm et al. (1999).
However, no comprehensive studies of mixture quality with respect to binder dispersion have ap-
peared in which direct measurements of binder content have been done. The aim of the present
study is to investigate the dispersion of binder in lime-cement columns by measuring binder con-
centrations in samples taken from in-situ production columns. Samples were taken from different
columns, at different depths, and from different parts of the column cross section in order to study
the effects of these parameters. Samples were also taken of different sizes in order to study the ef-
fect of sample size.
Another aim of the present study is to investigate whether statistical analyses and mixing indices
can be used for assessing the mixture quality of lime-cement columns. However, the study does not
attempt to pre-empt judgements of how good the mixture quality needs to be in order for the col-
umns in a structure to meet specifications of e.g. strength and deformation parameters.
Binder content is expressed as calcium oxide content, and it is assumed that all the calcium ions
come from the lime and the cement in the binder. In addition to the chemical analyses and measure-
ments of calcium oxide content, the water content, pH and electrical conductivity of selected sam-
ples was measured.
7.2 METHODS AND MATERIALS
Sampling site and sample columns

In the course of the research project “Production control methods in deep mixing with lime-cement
columns” a number of test columns were installed (Axelsson & Rehnman 1999). These were used
to study various methods of control, primarily by means of column penetration tests. For purposes
of sampling and visual inspection, a number of whole columns were recovered by means of a split
sampling tube, Fig. 7.3. The recovery method and the full extent of the field trials are reported by
Axelsson & Rehnman (1999). Their field investigation and the present study were coordinated in
conjunction with the recovery of the whole columns.
Deep Mixing 125

The test site is located just outside the town of Arboga near the Arboga – Örebro E18 highway
project. The soil on the site consists of a very soft clay to a depth of 8 metres. Table 7.1 and Fig. 7.1
show analysis results from soil samples taken from the test site.
Table 7.1 Some geotechnical properties of the test soils.
Depth Soil type Density Liquid Plasticity Sensitivitya Undrained Rapidityb

limita index shear strengtha
(t/m3) (%) (%) (-) (kPa)
2.0 Brown-grey clay 1.49 102.8 62 7 35.5 3-4

(some dry surface clay)
3.0 Grey clay 1.49 84.2 54 15 22.5
4.0 Grey sulphide clay 1.49 65.9 35 25 14.0 5
5.0 Grey sulphide clay 1.53 60.7 32 13.7 6
6.0 Grey clay 1.56 56.6 30 36 14.3 6
7.0 Grey varved clay 1.49 79.6 52 19 9.0
a By cone test to Swedish Standard.

b Defined by Söderblom (1974).
Water content (%) CaO (wt-%) pH Electrical conductivity

(mS/m)
60% 80% 100% 0,0% 0,2% 0,4% 0 2 4 6 8 0 100 200 300
0 0 0 0
1 1 1 1
2 2 2 2
3 3 3 3
Depth (m)
4 4 4 4
5 5 5 5
6 6 6 6
7 7 7 7
8 8 8 8
Fig. 7.1 Results of laboratory analyses of soil samples from test site, Arboga.

The test columns were manufactured in the same way as the production columns in the highway
project. Installation data appear in Table 7.2.
Table 7.2 Test column installation data.
Parameter Nominal value
Column diameter φ600 and φ800 mm

Column length 6–8 m
Binder: type, ratio Lime/cement 50/50
Binder quantity 82 kg/m3 (23kg/m, φ600 mm column)
Mixing tool Pinnborr
Rotation speed 180 rev/min
Retrieval rate 20 mm/rev
During the manufacturing process the quantity of binder and the rotation speed of the mixing tool
were recorded. The retrieval rate of the mixing tool on the type of machine used is a fixed parame-
ter, i.e. it cannot be varied while manufacturing a column. The plot from the machine showed very
small variations in the nominal values of the binder quantity injected and the rotation speed.
The mixing tool used is referred to as the Pinnborr, see Fig. 6.24a. The tool has six paddles ar-
ranged in three tiers.
Sampling procedure
Samples were taken from a number of columns in three sections at three different levels, see
Figs 7.2 and 7.3. Samples from four columns are analysed and reported in the present study. Two of
the columns were of diameter 600 mm (columns 16 and 26) and two of diameter 800 mm (columns
43 and 53). The samples were not taken according to a randomized pattern but on the pattern illus-
trated in Fig. 7.2. This pattern was chosen in order to be able to demonstrate the radial dispersion of
binder from the centre to the periphery of the column. Mixing tools used in Sweden today have the
outlet hole for the binder in the Kelly rod, i.e. the binder is blown from the centre of the column
towards its periphery. Based on this method of incorporating and spreading the binder, the hypothe-
sis was framed that the initial stage of the dispersion process as shown in Fig. 6.1 is the stage hav-
ing the greatest impact on the quality of a column.
Samples were taken in three sizes, referred to as the small, medium and large scale, in order to in-
vestigate the effect of sample size on the evaluated mixture quality. As discussed in Chapter 4, the
scale of scrutiny has a major impact on the evaluation of column quality. Further, the samples were
taken on three radii from the column centre, on bearings 0°, 120° and 240°, in order to study the
dispersion of the binder over the cross section. This sampling method results essentially in a two-
dimensional scrutiny of mixture quality at three levels in each column.
Deep Mixing 127

φ 600 or
φ 800 mm
Level 1
A A
Level 2
6-8 m Column section on A-A
Level 3 Symbol Sample φ600 column φ800 column

size Diam. Weight Diam. Weight
(mm) (g) (mm) (g)
Large 150 220 200 220
Medium 75 90 100 90
Small 38 25 67 25
S
Fig. 7.2 Sampling procedure.
Fig. 7.3 Sample collection in the field. (Photo: Morgan Axelsson)

Laboratory analyses
At the laboratory the water content and binder quantity of all 432 samples were determined. Water
content was measured by the method of Swedish Standard SS 02 71 16. The dry samples were then
ground in a mill to a powder. A 0.5 g quantity of each powdered, homogenized sample was dis-
solved in 50% hydrochloric acid. The soil was removed by filtration and the solution diluted 1000
times. The diluted solution was then analysed for calcium ion concentration, Ca2+, by OES-ICP
(optical emission spectroscopy in inductive coupled plasma). The Ca2+ content (wt-%) is the dry
weight concentration of calcium ions in the dry sample, expressed as a percentage. The equivalent
calcium oxide (CaO) content is obtained by multiplying the Ca2+ content by the molar mass of CaO
divided by the molar mass of Ca. However, the analytic method does not show whether the calcium
ions come from the added lime or cement. The nominal binder quantity, 82 kg/m3, corresponds
roughly to a CaO content of 7 % (CaO content calculated from binder quantity according to Tränk
& Johnson 1997). The CaO content of the unstabilized soil is low, 0.1 – 0.4 %, and is therefore
ignored in evaluating the binder dispersion in the columns. All the CaO is thus assumed to come
from the binder.
A number of triple tests were carried out in order to verify the reproducibility of sample preparation
and of the analytic method. The results showed that the CaO content is evaluated to an accuracy of
± 0.2 % in this study. This error interval applies to samples having a relatively high CaO content,
> 5 %.
For 40 samples from the four columns, pH and electrical conductivity were determined in accord-
ance with Swedish Standard SS-ISO 10 390 and SS-ISO 11 265 respectively. All of these determi-
nations were made on samples of the medium size.
A number of double tests of the pH analysis showed that pH could be determined to an accuracy of
one decimal place. A large number of pH measurements were made at various times after agitation.
This study too showed that pH could be determined to an accuracy of one decimal place. Calibra-
tion was done using buffer solutions of pH = 7 and pH = 10, which may result in lower accuracy for
samples with low pH. The samples were dried at 105 °C, which is considerably higher than the
maximum temperature of 40 °C specified in the standard.
A number of triple tests showed that the variation in the determination of electrical conductivity
was somewhat greater than specified in the standard, ~15 % as against < 10 %. Investigation of the
cause of this showed that the conductivity was reduced by contact with the air during filtration.
However, the Swedish Standard does not mention this.
Statistical analysis
Based on the results of the chemical analyses and the evaluation of the CaO content of the samples,
a number of statistical analyses were carried out to provide a quantitative assessment of mixture
quality in the four test columns.
The sampling procedure of Fig. 7.2 has the effect that the inner parts of the columns are over-repre-
sented and the outer parts under-represented. Since sampling was not done according to a random
Deep Mixing 129

test procedure, a method was adopted for weighting the samples according to their position. The
mean and the variance for each series of samples taken from the column cross sections were evalu-
ated as
∑ ∑ (a × α )
3 nS
j =1 i =1 ij i
a= (7.1)
3× ∑ α
nS
i =1 i
∑ ∑
3 nS
j =1 i =1
((aij − a ) 2 × α i )
σ 2
= (7.2)
3 × ∑i =S1α i − 1
n
where a is the measured calcium oxide content of each sample. The samples are numbered out-
wards from the column centre from i = 1 till i = ns, where ns = 6 for the large scale, ns = 4 for the
medium scale and ns = 2 for the small scale. The samples were taken on three radii numbered from
j = 1 to j = 3. The concentration aij is assumed to be valid for the entire area Ai/3 according to
Fig. 7.4. The coefficient a represents the area quotients represented by the samples according to
Fig. 7.4.
A1
A2
A3
Area quotient, αi = Ai/A1
A6
A5
A4
Column section
Fig. 7.4 Area relations between sampling positions.

The coefficient of variation was then calculated as the standard deviation divided by the mean.
Two mixing indices were evaluated in the study, those defined by Miles (1962) and by Rose &
Robinson (1965):
σ2
M 1 = 1− (7.3)
σ 02
and
σ
M 2 = 1− (7.4)
σ0
where the variance σ2 is evaluated according to equation 7.2 above and the variance for a complete-
ly segregated mixture σ02 is evaluated according to Lacey (1943):
σ 02 = a (1 − a ) (7.5)
7.3 RESULTS
CaO content and statistical analysis

The results of the chemical analyses and the evaluation of CaO content are reported in Figs 7.5–7.8.
The results of the statistical evaluation are presented in Tables 7.3–7.6.
The CaO contents and the statistical analyses are reported for each column separately, i.e. the first
page shows the results for column 16, the second page those for column 26, etc.
Deep Mixing 131

Legend: 0°, 120°, 240°
Depth 0.9 m 20 20 20
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 0 100 200 300 0 100 200 300
CaO (% dry weight)
Depth 2.7 m
20 20 20
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 0 100 200 300 0 100 200 300
Depth 4.5 m
20 20 20
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 0 100 200 300 0 100 200 300
Distance from column centre (mm)
Small size Medium size Large size
Fig. 7.5 Results of chemical analysis of samples from column 16.
Table 7.3 Statistical analysis for column 16.
Column Sample size Depth Mean Variance Coeff. of M1 M2

(m) (wt-%) variation (-) (-)
(-)
16 Small -0.9 7.28 32.5 0.78 0.952 0.781

-2.7 5.80 9.03 0.52 0.983 0.871
-4.5 7.46 1.94 0.19 0.997 0.947
Medium -0.9 5.17 5.89 0.47 0.988 0.890

-2.7 4.80 2.45 0.33 0.995 0.927
-4.5 6.64 0.88 0.14 0.999 0.962
Large -0.9 5.32 6.30 0.47 0.988 0.888

-2.7 5.73 3.05 0.31 0.994 0.925
-4.5 6.48 0.84 0.14 0.999 0.963
1 Mixing index defined by equation 7.3

Legend: 0°, 120°, 240°
20 20 20
Depth 0.4 m
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 0 100 200 300 0 100 200 300
CaO (% dry weight)
20 20 20
Depth 2.2 m
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 0 100 200 300 0 100 200 300
20 20 20
Depth 4.4 m
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 0 100 200 300 0 100 200 300

(-)
26 Small -0.4 6.33 3.02 0.27 0.995 0.929

-2.4 8.03 3.18 0.22 0.996 0.934
-4.4 5.43 4.70 0.40 0.991 0.904
Medium -0.4 6.06 0.64 0.13 0.999 0.966

-2.4 7.78 1.91 0.18 0.997 0.948
-4.4 5.78 3.63 0.33 0.993 0.918
Large -0.4 6.16 0.52 0.12 0.999 0.970

-2.4 8.05 1.64 0.16 0.998 0.953
-4.4 6.15 2.18 0.24 0.996 0.939
Deep Mixing 133

Legend: 0°, 120°, 240°
Depth 1.4 m
35 35 35
30 30 30
25 25 25
20 20 20
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 400 0 100 200 300 400 0 100 200 300 400
CaO (% dry weight)
Depth 3.5 m
35 35 35
30 30 30
25 25 25
20 20 20
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 400 0 100 200 300 400 0 100 200 300 400
35
Depth 5.2 m
35 35
30 30 30
25 25 25
20 20 20
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 400 0 100 200 300 400 0 100 200 300 400

(-)
43 Small -1.4 6.79 24.89 0.73 0.961 0.802

-3.5 10.39 20.16 0.43 0.978 0.853
-5.2 7.28 25.66 0.69 0.962 0.805
Medium -1.4 7.11 30.48 0.78 0.954 0.785

-3.5 11.67 15.17 0.33 0.985 0.879
-5.2 8.92 27.30 0.59 0.966 0.817
Large -1.4 8.49 25.48 0.59 0.967 0.819

-3.5 12.44 7.07 0.21 0.994 0.919
-5.2 8.82 14.74 0.44 0.982 0.865

Legend: 0°, 120°, 240°
20
Depth 0.5 m
20 20
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 400 0 100 200 300 400 0 100 200 300 400
CaO (% dry weight)
Depth 3.1 m
20 20 20
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 400 0 100 200 300 400 0 100 200 300 400
20 20
Depth 6.0 m
20
15 15 15
10 10 10
5 5 5
0 0 0
0 100 200 300 400 0 100 200 300 400 0 100 200 300 400


(-)
53 Small -0.5 10.70 17.82 0.39 0.981 0.863

-3.1 7.36 10.71 0.44 0.984 0.875
-6.0 8.39 24.55 0.59 0.968 0.821
Medium -0.5 9.70 17.23 0.43 0.980 0.860

-3.1 8.11 12.48 0.44 0.983 0.871
-6.0 7.95 13.70 0.47 0.981 0.863
Large -0.5 10.91 10.64 0.30 0.989 0.895

-3.1 7.80 5.29 0.29 0.993 0.914
-6.0 7.93 3.50 0.24 0.995 0.931
Deep Mixing 135

pH and electrical conductivity
Fig. 7.9 shows the results of the determination of CaO content, pH and electrical conductivity for
40 samples. The results are shown as relations between CaO, pH and electrical conductivity.
1400 14 14
pH
1200 12 12
Conductivity (mS/m)
1000 10 10
800 8 8
pH
pH
600 6 6
Electrical
400 conductivity 4 4
200 2 2
0 0 0
0 10 20 30 40 0 500 1000
CaO (weight-%) Conductivity (mS/m)
Fig. 7.9 Relations between CaO, pH and electrical conductivity
Water content
Fig. 7.10 shows the relation between calcium oxide content and water content determined from
samples taken from the four test columns.
Column 16 Column 26
20 20
-0.9m -0.4m
15 -2.7m 15 -2.2m
CaO (wt-%)
-4.5m -4.4m
10 10
5 5
0 0
60 70 80 90 100 60 70 80 90 100
20 Column 43 20 Column 53
-1.4m -0.5 m
15 15
CaO (wt-%)
-3.5m -3.1 m
10 -5.2 m 10 -6.0 m
5 5
0 0
60 70 80 90 100 60 70 80 90 100
Water content (%) Water content (%)
Fig. 7.10 Relation between CaO content and water content.

7.4 DISCUSSION
CaO content and statistical analysis

As mentioned in the introduction to this chapter, it has not been the practice to use statistical ex-
pressions or mixing indices for the quantitative assessment of mixture quality in lime-cement col-
umns. In Swedish practice (Carlsten & Ekström 1995) there are no methods for assessing mixing
quality. Far too often, subjective methods are used to assess the dispersion of the binder in columns.
An example is visual inspection, where there is a risk of incorrect evaluation. As discussed in Chap-
ter 5, Mixture Quality, it is common to judge a mixing process by properties of the final product.
There is a risk attached to assessing a mixing process on the basis of e.g. column penetration tests
of lime columns, since strength properties are related to other parameters besides the efficiency of
the mechanical mixing. An assessment of mixture quality must distinguish between the distribution
of binder and a particular property of the stabilized soil. If it is the mixture quality or the distribu-
tion of binder that is to be assessed, measurements of the quantity of binder should be performed.
Figs 7.5–7.8 show the results of chemical analysis of the CaO content of samples taken from the
four test columns. A number of comments need to be made on the results.
The samples were taken on three angularly separated radii, and the results show that the dispersion
of binder is relatively similar on all three radii in the sections studied. An exception is e.g. column
53, level 6.0 m, where the dispersion patterns in the three directions differ. However, the results do
show that the dispersion of binder is radial in character, i.e. the process of incorporation and distri-
bution by means of an air jet has a major impact on the character of the binder dispersion.
The binder dispersion varies between the column sections, most clearly in columns 16 and 26. In
column 16, for example, there is a heavy concentration of binder in the outer regions of the column
down to about the 3 metre level. There was a clearly visible ring of harder material all the way
around the column periphery. At level 4.5 m the binder is evenly dispersed over the entire cross
section. In column 26 there is even dispersion of binder down to a depth of about 4 metres. At level
4.4 m the binder concentration is very low in the outer parts of the column. Consequently it is diffi-
cult to draw conclusions about the distribution of binder from a few samples taken from a few col-
umns and levels, since the binder dispersion can vary with depth.
Tables 7.3–7.6 show the results of the statistical analysis, which is an attempt to quantify the results
of the chemical analysis presented in Figs 7.5–7.8. The statistical analysis will now be discussed.
As mentioned in Chapter 4, the mixing index defined by equation 7.3 is relatively insensitive to
mixture quality. This index may therefore be inappropriate to use, as the results of the present in-
vestigation confirm. The mixing index defined by equation 7.4 is more sensitive to different binder
distributions and can be used as a quantitative rating of mixture quality. The difference between the
mixing index defined by equation 7.4 and the coefficient of variation is that the mixing index takes
into account the mixing that can be achieved according to equation 4.11. The relation between coef-
ficient of variation and the mixing index defined by equation 7.4 is illustrated in Fig. 7.11.
Deep Mixing 137

1 CaO content
4%
Coefficient of variation
0.8
8%
0.6 16%
0.4
0.2
0
0.8 0.85 0.9 0.95 1
Mixing index
Fig. 7.11 Relation between coefficient of variation and mixing index

defined by equation 7.4.
Fig. 7.12 shows the relation between coefficient of variation and sample size. The results show that
the sample size has significant impact when the mixture quality is poor and less impact when the
mixture quality is good. Thus in order to avoid overestimating or underestimating the mixture quali-
ty, the “scale of scrutiny” must first be defined, after which an appropriate sample size can be es-
tablished.
Fig, 7.12 Relation between coefficient of variation and sample size.

The different sample sizes display generally similar binder dispersion pattern, even though the mag-
nitude of the mixing index varies with the different sample sizes. There are a number of exceptions,
however, where a larger sample does not indicate the mean of a smaller sample. An example is col-
umn 43, level 5.2, radius 240°, Fig. 7.7, where the small scale shows a considerably smaller quanti-
ty of binder than do the medium and large scales. One cause of this is that the small samples are not
sub-samples of the medium samples, nor are the medium samples sub-samples of the large samples.
Moreover, the samples differ in their tangential extension, which may have an effect on the results.
The mean values for the different sample sizes vary. The differences lie within the range 0.25 –
2.1 wt-% CaO. The largest differences are found in the sections where the binder dispersion was
poor. This shows that there are major uncertainties in the evaluation of the mean and the coefficient
of variation when column quality is poor.
Because the sampling procedure was not based on a random selection procedure of sampling points
and because the number of samples in each series was small, it is difficult to estimate a confidence
interval for the means, variances and coefficients of variation. As equation 4.16 shows, the meas-
ured concentration variance of the results depends on the sampling method and the analytic method.
In the present study the variance due to the analytic method and to sample purity may be assumed
to be small. However, the variance due to the sampling method is difficult to evaluate quantitative-
ly. The mean may vary between the sample sizes even when the mixing quality is relatively good,
indicating that the sampling method has some influence. The statistical analyses reported in Tables
7.3–7.6 must thus be regarded as rough estimates of the mixture quality of the columns.
Our current knowledge of binder dispersion in lime-cement columns is very limited and conse-
quently it is difficult to draw conclusions from a small number of samples. Thus there is a need for
greater knowledge of the character of binder dispersion. In general, a poor quality mixture will not
be normally distributed. As the mixture quality improves the distribution approaches a normal dis-
tribution (Harnby 1992). This makes it difficult to compare poor quality mixtures. Given greater
knowledge, it may be possible to use statistical analysis to assess mixture quality and thus provide a
tool for the development of the mixing process.
Measurements of mixing quality can facilitate and improve our understanding of the mixing proc-
ess. Its experimental determination is a major problem, however, since the collection of samples
from columns and their chemical analysis is difficult and costly. Extensive sampling should there-
fore be carried out in conjunction with major construction or R&D projects in which whole col-
umns are recovered. The sampling method should be based on random selection, e.g. the method
presented by Harnby (1972). Determining the nature of binder dispersion will make it possible to
use statistical methods for rating mixture quality. Given greater knowledge, it may be possible to
rate mixture quality from a small number of samples. Lacking this knowledge, the present situation
will continue in which we cannot assess mixture quality from a few samples. The present study
shows that there is a major risk that a small number of samples will not be representative of the
binder dispersion in the columns.
Deep Mixing 139

pH, electrical conductivity and water content
Fig. 7.9 shows the relation between pH, electrical conductivity and CaO content based on analyses
of 40 samples. The results show that pH increases very rapidly even at low contents of lime and
cement. In the CaO content range from 0 % to 15 % the pH increases very rapidly to pH ~11.5 at a
CaO content around 4 %. At a nominal binder content of 7 % the pH is >12. These results make it
doubtful whether pH can be used to assess column quality with reference to binder dispersion.
For electrical conductivity the picture is reversed. Conductivity is relatively constant in the range
50 – 200 mS/m up to a CaO content of 7 %. It then increases very rapidly to 600 – 700 mS/m at
CaO content 12 %. For CaO contents between 12 % and 25 % there is a moderate increase of con-
ductivity to 700-1000 mS/m. Again it is doubtful whether it is appropriate to use electrical conduc-
tivity for assessing column quality with reference to binder dispersion.
The results of the analysis and evaluation of the water content of the samples show that the disper-
sion of the results in relation to the CaO content is very large, Fig. 7.10. Causes of the large disper-
sion include edge effects, differing degrees of compaction, the sampling method and the analytic
method. A hole had been drilled down the centre of all four columns studied, permitting the free
flow of water through the column centre. Differences in hydraulic gradient across the column sec-
tions result in water flows, with differing effects on the samples depending on the part of the sec-
tion from which they were taken. A high CaO content does not necessarily mean that the water con-
tent will fall, since the degree of compaction may vary. An accumulation of binder may be very
porous, with a high water content, due to poor compaction or poor consolidation. The sampling and
analytic methods can also affect the results. The analysis was carried out over a relatively long peri-
od of time.
7.5 CONCLUSIONS
The most important conclusions of the present study are:
• The difference in binder contents is relatively large over the cross-section and the character of
the dispersion of the binder varies between the columns in the study.
• Sample size has a significant influence on the evaluated mixture quality when the mixing quality
is poor. It has a smaller effect when the mixture quality is good.
• The pattern of binder dispersion in the columns is of a radial character, implying that the incor-
poration and distribution process using compressed air has a major impact on mixture quality.
• Statistical methods and mixing indices can be used for the quantitative assessment of the mixture
quality of lime-cement columns. However, the precision of the assessment depends on the mix-
ture quality being relatively good.
• Since the sampling and chemical analysis of samples is difficult and costly, extensive sampling
by a random selection method should be carried out in conjunction with major construction and
R&D projects in which whole columns are recovered. A greater knowledge of binder dispersion
in lime-cement columns is essential in order for reliable estimates of column quality to be made
on the basis of small numbers of samples taken for production control.

8. Further research
Because the mixing process in deep mixing is very complex and the method is relatively young,
there are many areas in which further research is needed. In most industrial applications develop-
ment is based on existing equipment. A common problem with this approach is the risk of repeating
mistakes. It is difficult to formulate exact recommendations for research and the development of
new technologies, even though it is clear that the purpose is to understand and solve problems with
the current mixing procedure in deep mixing. The suggestions for further research and technologi-
cal development in this thesis are intended to serve as an approach for further investigation.
There are two principal difficulties associated with studies of mixing processes:
1) deficient theoretical background;
2) lack of usable mixing criteria.
There is a danger in producing voluminous experimental data which remain specific to the condi-
tions studied. Hence there is a need for a conceptual framework to assist in the delimitation of ex-
periments. There are a number of potentially valuable concepts based on experience from research
in the process industries, and there is a need for applications to practical problems. Experience may
decide whether new concepts are necessary.
The deep mixing industry is small compared with the process industry. It is to be expected that it is
difficult to obtain funding for relevant basic research, e.g. for studies of the rheological behaviour
of a soil–binder mixture. The mixing process is very complex and therefore it is important to keep
abreast of research and development in other industries.
There is a need for investigations into the mixing process and the key factors affecting it. Qualita-
tive and quantitative studies should include investigations of the influence of the following factors:
1. physical properties of the mixture and its components;
2. mixing tool size and geometry;
3. variables in the mixing process: rotation speed, retrieval rate, air pressure, torque etc.;
4. sampling procedures: sampling method, sample size, number and position;
5. aims of the mixing process. Criteria used to determine whether the process is successful, based
on specific or general mixing indices, flow patterns, power consumption, costs etc.
There are a number of properties that are particularly suitable for laboratory (and possibly field)
investigations:
Deep Mixing 141

1. soil properties. The influence of clay content, water content, and disturbance on the mixing
process and its results;
2. the level of energy input per unit volume for a given degree of mixedness. The effect of the en-
ergy input and strain rate (function of retrieval rate and rotation speed) is best investigated ini-
tially in the laboratory, where extreme values of these factors can be studied;
3. the initial orientation and distribution of the components to be mixed can have a major impact
on mixing time and dispersion.
Theoretical and experimental investigations of mixing mechanisms should be carried out in a labo-
ratory environment with materials having controllable rheological properties and with simple mix-
ing tool geometries. Such work would shed light on phenomena likely to be important in more com-
plex mixing processes and suggest the most relevant variables to study in field trials. The work
should not aim to provide data that are directly applicable to complex industrial problems, but to
yield results that will suggest new approaches to the mechanism behind the real problem. To ensure
that new findings are relevant, laboratory tests must keep in close contact with real-world problems.
This will also ensure that problems are not over-simplified.
Problems suitable for this approach are discussed below:

i There is a need for appropriate, rapid methods for sampling and analysis of the mixing proc-
ess. Consideration must be given to the number of samples that needs to be taken to assess
the state of a process and to the question whether sampling should be random or selective.
ii Experiments to study mixing mechanisms in viscoelastic clayey soils are likely to be difficult
to interpret and generalize until there is an established understanding of viscoelastic behav-
iour. The range of validity of experiments should be determined by isolating individual
mechanisms. If individual mixing mechanisms can be isolated, can these be used for a classi-
fication of mixing equipment?
iii There is a need for the investigation of air pressure as a mixing mechanism. The distribution
of binder over the cross section must also be studied. Is there a need for better, more even
distribution of the binder than that produced by the air pressure?
iv Although many industrial products are viscoelastic in their properties, there is little under-
standing of the distribution of velocity and shear stresses in these rheologically very complex
materials. There is a need for a better understanding of phenomena that can occur during
mixing. Particularly important is the investigation of the volume of material around a mixing
tool that is set in motion during mixing.
v The theories of laminar mixing need to be studied. The theories are based on the deformation
or reduction of fluid elements of the material in which the orientation of the interfacial sur-
faces between the components is known. It is uncertain how these theories can be applied to
three-dimensional patterns of movement in a mixing process.

vi The energy input to the mixture from the mixing tool can be consumed in a number of differ-
ent ways, e.g. to overcome elastic and cohesive forces in a clay suspension. How much of the
energy input is expended on disaggregating a clayey soil and how much goes into deforming
and reducing the constituent elements of the mixture?
vii The structure of a clayey soil can be important for the dominating mechanisms in a mixing
process. How much effect does the disturbance ratio of a soil have on the mixing process and
on its results? The importance of the disturbance ratio of a clayey soil is likely to be different
for different binder types.
viii Compaction energy has a major effect on the mixing process when air is mixed with the soil.
How much effect does the compaction energy have? Is it important to design tools to be ca-
pable of evacuating all the air from the soil?
ix The possibility of changing the rheological properties of the mixture by means of peptizing
agents should be investigated.
Deep Mixing 143

References
Adams, J.F.E. and Baker, A.G. 1956. An assessment of dry blending equipment. Transactions of
the Institution of Chemical Engineers, Vol. 34, pp. 91-107.
Adeyeri, J.B., Krizek, R.J. and Achenbach, J.D. 1970. Multiple integral description of nonlinear
viscoelastic behaviour of a clay soil. Transactions of the Society of Rheology. Vol. 14, Issue 3,
pp. 375-392.
Åhnberg, H., Johansson, S-E., Retelius, A., Ljungkrantz, C., Holmqvist, L. och Holm, G. 1995.
Cement och kalk för djupstabilisering av jord, en kemisk-fysikalisk studie av stabiliseringseffek-
ter (Cement and lime for deep stabilization of soil – a chemical physical investigation of im-
provement effects). Swedish Geotechnical Institute, Report 48, Linköping. 170+43 pp. (in
Swedish).
AIChE, 1979. Equipment Testing Procedure. Dry Solids Paste and Dough Mixing Equipment. Sec-
ond Edition. American Institute of Chemical Engineers, New York. 29 pp.
AIChE, 1987. Equipment Testing Procedure. Mixing Equipment (Impellar Type). Second Edition.
American Institute of Chemical Engineers, New York. 40 pp.
Ali Al-Ghazal, A. 1989. An investigation into the mechanics of agricultural discs. Cranfield Insti-
tute of Technology, United Kingdom. 318 pp.
Allam, M.M. and Sridharan, A. 1984. The shearing resistance of saturated clays. Geotechnique,
Vol. 34, No. 1, pp. 119-122.
Al-Tabbaa, A. and Evans, C.W. 1999. Laboratory-scale soil mixing of contaminated site. Ground
Improvement, Vol. 3, pp. 119-134.
Al-Tabbaa, A., Al-Tabbaa, A.M.B. and Ayotamuno, J.M. 1999. Laboratory-scale dry soil mixing of
a sand. Proc. Int. Conf. on Dry Mix Methods for Deep Soil Stabilization, Stockholm, pp. 73-80.
Aoi, M. & Tsujii, T. 1996. Mechanism of machine for Dry Jet Mixing Method. Grouting and Deep
Mixing. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosystems, Tokyo,
pp. 579-584.
Asano, J., Ban, K., Azuma, K. and Takahashi, K. 1996. Deep mixing method of soil stabilization
using coal ash. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosystems,
Tokyo, pp. 393-398.
Ashton, M.D. and Valentin, F.H.H. 1966. The mixing of powders and particles in industrial mixers.
Trans. Inst. Chem. Engrs., Vol. 44, No. 5, pp. T166-T188.
Assarsson, K.G. 1972. Stabilisering och jordförstärkning med kalk (Stabilization and soil improve-
ment using lime). Cement och betong, 1972:1, pp. 55-71. (in Swedish).

Axelsson, A. & Larsson, S. 1994. Provningsmetoder på kalkcementpelare, Svealandsbanan (Test
methods for lime-cement columns – The Svealand rail link). MSc thesis 94/10, Department of
Soil and Rock Mechanics, Royal Institute of Technology, Stockholm, 62+10 pp. (in Swedish).
Axelsson, M. and Rehnman, S-E. 1999. Field methods for quality control at the dry jet mixing
method. Proc. of the Int. Conf. on Dry Mix Methods for Deep Soil Stabilization, Stockholm, pp.
303-310.
Babasaki, R. & Suzuki, K. 1996. Open cut excavation of soft ground using the DCM method. Proc.
of the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosystems, Tokyo, pp. 469-473.
Babasaki, R., Terashi, M., Suzuki, T. Maekawa, A., Kawamura, M. and Fukazawa, E. 1997. JGS
TC Report: Factors influencing the strength of improved soil. Proc. of the IS-Tokyo´96, 2nd Int.
Conf. on Ground Improvement Geosystems, Tokyo, 14-17 May, 1996. Vol. 2, pp. 913-918.
Bachy Limited. 1996. Colmix brochure. Bachy Limited, Godalming.
Baghdadi, Z.A. and Shihata, S.A. 1999. On the durability and strength of soil-cement. Ground Im-
provement, Vol. 3, pp. 1-6.
Baker, C.N.Jr. 1954. Strength of soil-cement as a function of the degree of mixing. Stabilization of
Soils, Highway Research Board Bulletin 98, 33d Annual Meeting, pp. 33-52.
Barnes, H.A., Hutton, J.F. and Walters, K. 1989. An Introduction to Rheology. Rheology Series,
Vol 3. Elsevier. 199 pp.
Bell, F.G. 1988. Stabilsation and treatment of clay soils with lime. Part 1. Basic principles. Ground
Engineering, Vol. 21, No. 1, pp. 10-15.
Bell, J.R. 1977. Compaction energy relationship of cohesive soils. Stabilization of soils. Transpor-
tation Research Record 641. pp. 29-34.
Bentley, S.P. 1979. Viscometric assessment of remoulded sensitive clays. Canadian Geotechnical
Journal, Vol. 16, pp. 414-419.
Betonghandbok Material. 1994. (Concrete design handbook – material). (Ljungkrantz, C., Möller,
G. and Peterson, N.; eds.) AB Svensk Byggtjänst, Stockholm, 1127 pp. (in Swedish).
Boman, P. och Broms, B.B. 1975. Stabilisering av kohesionsjord med kalkpelare (Stabilization of
cohesive soils with lime columns). Nordic Geotechnical Meeting, Copenhagen, pp. 265-279. (in
Swedish).
Bourne, J.R. 1964. The mixing of powder, pastes, and non- Newtonian fluids. Industrial Research
Fellow, Report No. 2. The Chemical Engineer, September, 1964, pp. CE202-CE216.
Bourne, J.R. 1968. Statistical relations for square samples. Chemical Engineering Science, Vol. 23,
pp. 339-343.
Braaten, A., Aaboe, R. and Oset, F. 1999. Development of in-situ control methods for lime/cement
columns. Proc. Int. Conf. on Dry Mix Methods for Deep Soil Stabilization, Stockholm, pp. 295-
301.
Deep Mixing 145

Bredenberg, H. 1999. Keynote lecture: Equipment for deep soil mixing with the dry jet mixing
method. Proc. Int. Conf. on Dry Mix Methods for Deep Soil Stabilization, Stockholm, pp. 323-
331.
British Lime Association. 1990. Lime Stabilisation Manual. Second edition. British Lime Associa-
tion, London. 41 pp.
Brodkey, R.S. 1966. Fluid motion and mixing, Chapter 2. In: Mixing Theory and Practice, (Uhl,
V.W. and Gray, J.B.; eds) Volume 1. New York: Academic Press, pp. 7-110.
Broms, B.B. and Boman, P. 1975a. Kalkstabiliserade pelare—en grundförstärkningsmetod vid väg-
byggnad (Columns stabilized with lime – a new foundation method at road construction). Väg-
och vattenbyggaren, No. 5, pp. 40-43. (in Swedish).
Broms, B.B. and Boman, P. 1975b. Lime stabilized columns. Proc. 5th Asian Regional Conf. on
Soil Mechanics and Foundation Engineering, Bangladore, India, Vol. 1, pp. 227-234.
Broms, B.B. and Boman, P. 1979a. Stabilisation of soil with lime columns. Ground Engineering,
Vol. 12, No. 4, pp. 23-32.
Broms, B.B. and Boman, P. 1979b. Lime columns—a new foundation method. ASCE, Journal of
Geotechnical Engineering Division, Vol. 105, NoGT4, pp. 539-556.
Brookes, A.H. West, G. and Carder, D.R. 1997. Laboratory trial mixes for lime-stabilised soil col-
umns and lime piles. TRL Report 306. 16 pp.
Bruce, D.A. 1996. The return of deep soil mixing. Civil Engineering, ASCE, Dec. pp. 44-46.
Bruce, D.A., Bruce, M.E.C. and DiMillio, A.F. 1998a. Deep mixing methods: A global perspective.
Civil Engineering- ASCE, Vol. 68, No. 12, December 1998. pp. 38-41.
Bruce, D.A., Bruce, M.E.C. and DiMillio, A.F. 1998b. Deep mixing methods: A global perspective.
Geotechnical Special Publication No. 81. Soil Improvement for Big Digs. Proc. of sessions of
Geo-congress 98. Boston, Massachusetts, pp. 1-26.
Bruce, D.A., Bruce, M.E.C. and DiMillio, A.F. 1999. Dry mix methods: A brief overview of inter-
national practice. Proc. Int. Conf. on Dry Mix Methods for Deep Soil Stabilization, Stockholm,
pp. 15-25.
Bryhn, O.R., Løken, T. and Ass, G. 1984. Stabilisation of sensitive clays with hydroxy-aluminium
compared with unslaked lime. Norwegian Geotechnical Institute, Oslo, Publication 151. 14 pp.
Bryhn, O.R., Løken, T. and Reed, M.G. 1988. Stabilization of sensitive clays (quick clays) using
A1(OH)2.5 Cl0.5. Norwegian Geotechnical Institute. Publication 170. 9 pp.
Carlsten, P. 1991. Jordförstärkning med kalk-, cement- och kalk/cementpelare (Ground improve-
ment by lime-, cement- and lime/cement columns). Allmän teknisk beskrivning, metoder. Swed-
ish National Road Administration, Publ. 1991:42. (in Swedish).
Carlsten, P. och Ekström, J. 1995. Lime and lime cement columns. Guide for project planning, con-
struction and inspection. Swedish Geotechnical Society, Report 4:95E, 111 pp.
Cementa. 1973. Jordstabilisering med cement och cementstabiliserat grus vid europeiskt vägbyg-
gande. Cementa AB, Malmö. 157 pp. (in Swedish).

Cheng, D.C-H. 1979. The role of extensional flows in fluid mixing and dispersion. Proc. 3rd Euro-
pean Conference on Mixing, BHRA, Fluid Engineering, Cranfield, Bedford, Vol. 1, pp. 73-104.
Cheng, D.C-H. and Richmond, R.A. 1978. Some observations on the rheological behaviour of
dense suspensions. Rheological Acta, Vol. 17, pp. 446-453.
Chida, S. 1982. Development of dry jet mixing methods. Public Works Research Institute, Ministry
of Construction, Japan. pp. 29-35.
Coulson, J.M. and Maitra, N.K. 1950. The mixing of solid particles. Ind. Chem., Vol. 26, p. 55.
Coulson, J.M., Richardson, J.F. and Peacock, D.G. 1979. Chemical Engineering. Volume Three
(2nd Edn) Oxford: Pergamon Press. 648 pp.
Coussot, P. 1997. Mudflow Rheology and Dynamics. International Association for Hydraulic Re-
search. Monograph Series. Balkema, Rotterdam, Brookfield. 255 pp.
Danckwerts, P.V. 1952. The definition and measurement of some characteristics of mixtures. Ap-
plied Scientific Research. Section A, Vol. 3, pp. 279–296.
Danckwerts, P.V. 1953. Theory of mixtures and mixing. Research: Science and its Application in
Industry. Vol. 6, pp. 355-361.
Davidson, L.K., Demirel, T. and Handy, R.L. 1965. Soil pulverization and lime migration in soil-
lime stabilization. Highway Research Record 92. HRB, National Research Council, Washington
D.C., pp. 103-118.
Dong, J., Hiroi, K. and Nakamura, K. 1996. Experimental study on behaviour of composite ground
improved by deep mixing method under lateral earth pressure. Proc. of the IS-Tokyo´96, 2nd Int.
Conf. on Ground Improvement Geosystems, Tokyo, pp. 585-590.
Drozdov, D.D. 1998. Mechanics of Viscoelastic Solids. John Wiley & sons. 472 pp.
Edstam, T. 1997. Erfarenhetsbank för kc-pelare (Compilation of experiences of lime-cement col-
umns). Swedish Deep Stabilisation Research Center, Report No. 1, Linköping, 154 pp.
Edstam, T. and Carlsten, P. 1999. A new method for laboratory preparation of stabilised clay. Proc.
Int. Conf. on Dry Mix Methods for Deep Soil Stabilization, Stockholm, pp. 315-319.
Edwards, M.F. 1992. Laminar flow and distributive mixing, Chapter 11. In: Mixing in the Process
Industries, (Harnby, N., Edwards, M.F. and Nienow, A.W.; eds) Second Edition. London: But-
terworth Heinemann, pp. 200-224.
Edwards, M.F. and Baker, M.R. 1992. A review of liquid mixing equipment, Chapter 7. In: Mixing
in the Process Industries, (Harnby, N., Edwards, M.F. and Nienow, A.W.; eds) Second Edition.
London: Butterworth Heinemann, pp. 118-136.
Edwards, M.F., Baker, M.R. and Godfrey, J.C. 1992. Mixing of liquids in stirred tanks, Chapter 8.
In: Mixing in the Process Industries, (Harnby, N., Edwards, M.F. and Nienow, A.W.; eds) Sec-
ond Edition. London: Butterworth Heinemann, pp. 137-158.
Enami, A., Hibino, S., Takahashi, M., Akiya, K. and Yamada, M. 1986a. Properties of soil cement
produced by compact machine system for tenocolumn method. 21st Annual Meeting, Japanese
Society of Soil Mech. and Found. Engng., pp. 1987-1990. (in Japanese).
Deep Mixing 147

Enami, A., Hibino, S., Akiya, K. and Yamada, M. 1986b. Subsoil stabilization by a newly devel-
oped compact deep mixing machine. Tsuchi to Kiso. Vol. 34, No. 6, pp. 37-43. (in Japanese).
Fan, L.T., Chen, S.J. and Watson, C.A. 1970. Solids mixing. Industrial Engineering Chemistry,
Vol. 62, No. 7, pp. 53-69.
Fan, L.T., Rubison, R.M. and Lai, F.S. 1979. Studies of multicomponent solids mixing and mix-
tures. Part III. Mixing indices. Powder Technology, Vol. 24, pp. 73-89.
Fan, L.T., Wang, R.H. 1975. On mixing indices. Powder Technology, Vol. 11, pp. 27-32.
Fohs, D.G. and Kinter, E.B. 1972. Migration of lime in compacted soil. Public Roads, Vol. 37,
pp. 1-8.
Freitag, D.R., Kozan, G.R. and Stouffer, J.D. 1961. Summary reviews of soil stabilization process-
es. Mixing principles, techniques and equipment. Miscellaneous paper no. 3-122, Report 6. May
1961. U.S. Army Engineer Waterways Experiment Station. Corps of Engineers. Vicksburg, Mis-
sissippi. 33 pp.
Futaki, M., Nakano, K. and Hagino, Y. 1996. Design strength of soil-cement Columns as founda-
tion ground for structures. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement
Geosystems, Tokyo, pp. 481-484.
Geankoplis, C.J. 1993. Transport Processes and Unit Operations, 3rd Edition, Prentice Hall. 921 pp.
Giesekus, H. 1983. Disperse systems: Dependence on rheological properties on the type of flow
with implication for food rheology. In: Physical Properties of Foods (Jowitt, R., Escher, F., Hall-
strom, B., Meffert, H.F.T., Spies, W.E.L. and Vos, G. Eds.). London: Applied Science Publish-
ers, pp. 205-220.
Gillott, J.E. 1987. Clay in Engineering Geology. Developments in Geotechnical Engineering; 41.
Elsevier. 468 pp.
Godfrey, J.C. 1985. Mixing of high-viscosity fluids, Chapter 11. In: Mixing in the Process Indus-
tries, (Harnby, N., Edwards, M.F. and Nienow, A.W.; eds) London: Butterworth Heinemann,
pp. 185-201.
Godfrey, J.C. 1992. Static mixers, Chapter 12. In: Mixing in the Process Industries, (Harnby, N.,
Edwards, M.F. and Nienow, A.W.; eds) 2nd Ed. London: Butterworth Heinemann, pp. 225-249.
Gray, J.B. 1957. Performance of dry solids mixing equipment. Chemical Engineering Progress,
Vol. 53, No. 1, pp. 25-32.
Gray, J.B. 1963. Batch mixing of viscous liquids. Chemical Engineering Progress, Vol. 59, No. 3,
pp. 55-59.
Greaves, H.M. 1996. An introduction ti lime stabilisation. Lime Stabilisation. Proc. of the Seminar
Held at Loughbo Rough University, Civil and Building Engineering Dept on 25 sept., pp. 5-12.
Grossman, R.F. 1997. Mixing cycles and procedures. Chapter 2, In :The Mixing of Rubber (Gross-
man, R.F.; eds). Chapman & Hall. pp. 25-48.
Haddad, Y.M. 1995. Viscoelasticity of Engineering Materials. Chapman & Hall. 378 pp.

Håkansson, U. 1993. Rheology of Fresh Cement- Based Grouts. PhD thesis, Dept. of Soil and Rock
Mechanics, Royal Institute of Technology, Stockholm. 154 pp.
Halkola, H. 1983. In-situ investigations of deepstabilized soil. European Conf. on Soil Mech. and
Found. Eng., 8, Helsinki. Proc., Vol. 2. pp. 33-36.
Halkola, H. 1999. Keynote lecture: Quality control for dry mix methods. Proc. of the Int. Conf. on
Dry Mix Methods for Deep Soil Stabilization, Stockholm, pp. 285-294.
Handy, R.L. and Williams, W.W. 1967. Chemical stabilization of an active landslide. Civil Engi-
neering, ASCE. Aug. 1967, pp. 62-65.
Harnan, C.N. 1993. Colmix. The process and its applications. Lime Column Day ´93, Stockholm,
5 pp.
Harnby, N. 1972. The statistical analysis of particulate mixtures. Part 1. The sampling of mixtures
and the resultant prediction of estimates based on the sample. Powder Technology, Vol. 5, pp.
81-86.
Harnby, N. 1992. Characterization of powder mixtures. In N. Harnby, M.F., Edwards and A.W.
Nienow (ed.), Mixing in the Process Industries. Second Edition: Chapter 2. Butterworth Heine-
mann. pp 25-41.
Harnby, N., Edwards, M.F. and Nienow, A.W. (Eds), 1992. Mixing in the Process Industries. Sec-
ond Edition. London: Butterworth Heinemann, 414 pp.
Hata, S., Higaki, Y., Aoi, M. and Ashida, S. 1987. Study on the monotoring in deep mixing meth-
od. Doboku Gakkai Ronbushu, No. 382, pp. 35-43
Hayashi, H. and Nishikawa, J. 1999. Mixing efficiency of dry jet mixing methods applied to peat
soft ground. Proc. Int. Conf. on Dry Mix Methods for Deep Soil Stabilization, Stockholm, pp.
333-338.
Herschel, W.H. and Bulkley, R. 1926. Measurement of consistency as applied to rubber-benzene
solutions. Proc. Am. Soc. Test. Matls, 26:621–633.
Heywood, N.I. and McDonagh, M. 1987. Rheology of high solids content materials. Cours on Fluid
Rheology held at Uppsala, Sweden, 29 Sept – 2 Oct.
Hirai, T., Ise, J., Kusakara, T., Gotou, M. and Hibi, Y. 1996. Development and application of deep
mixing soil stabilization method to control displacement of adjacent ground. Proc. of the IS-
Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosystems, Tokyo, pp. 485-490.
Hoikkala, S., Leppänen, M. and Tanska, H. 1997. Blockstabilization of peat in road construction.
Int. Conf. on Soil Mechanics and Foundation Engineering, 14, Hamburg, Proc., Vol. 3. pp.
1693-1696.
Hold, P. 1982. Mixing of polymers—an overview, Part II. In: Advances in Polymer Technology,
2(3), pp. 197-228.
Holm, G., Bredenberg, H. and Broms, B. 1981. Lime columns as foundation for light structures.
Proc. 10th Int. Conf. on Soil Mechanics and Foundation Engineering, Stockholm, Vol. 3,
pp. 687-694.
Deep Mixing 149

Holm, G., Ruin, M. and Håkansson, S. 1999. Column penetration tests and extraction of lime/ce-
ment columns. Proc. of the Int. Conf. on Dry Mix Methods for Deep Soil Stabilization, Stock-
holm, pp. 311-314.
Holt, C.C. and Freer-Hewish, R.J. 1998. The use of lime-treated British clays in pavement con-
struction. Part 1: The effect of mellowing on the modification process. Proc. Inst Civ. Engrs
Transp., 129, pp. 228-239.
Honjo, Y. 1982. A Probabilistic approach to evaluate shear strength of heterogeneous stabilized
ground by deep mixing method. Soils and Foundations, Vol. 22, Nr. 1. pp. 23-38.
Hosomi, H., Nishioka, S., Takei, S. and Qing, C.C. 1996. Method of deep mixing at Tianjin Port,
People´s Republic of China. Proc. of IS-Tokyo´96, Second Int. Conf. on Ground Improvement
Geosystems, Tokyo, pp. 491-494.
Hvorslew, M.J. 1960. Physical components of the shear strength of saturated clays. American Soci-
ety of Civil Engineers, Res. Conf. Shear Strength Cohesive Soils, Boulder, Colorado, pp. 169-
273.
Iinoya, K., Masuda, H. and Watanabe, K. 1985. Powder and Bulk Solids Handling Process – Intro-
duction and Control, M. Dekker, New York.
Isobe, K., Samaru, Y., Aoki, C., Sogo, K. and Murakami, T. 1996. Large scales deep soil mixing
and quality control. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosys-
tems, Tokyo, pp. 619-624.
Izykowski, Z. 1956. Przegl. Odl., Vol. 6, p.175.
Jelisic, N. 1999. Masstabilisering (Mass stabilization). Lic. thesis, Lund Technical University, De-
partment of Geotechnical Engineering, Lund, 154 pp. (in Swedish).
Jenny, H. and Overstreet, R. 1939. Surface migration of ions and contact exchange. Journal of
Physical Chemistry, Vol. 43, pp. 1185- 1196.
Johansson, C. & Jons, A-K. 1995. Kalkcementpelartillverkning. En studie av hur olika produktions-
faktorer påverkar pelarnas homogenitet och hållfasthet (A study of different factors affecting the
homogeneity and strength of lime/cement columns). MSc thesis 1995:5. Department of Geotech-
nical Engineering, Chalmers University of Technology, Gothenburg. 69+90 pp. (in Swedish).
Johnson, R.A. 1994. Miller and Freund’s Probability and Statistics for Engineers. Fifth Edition.
Prentice Hall, Englewood Cliffs, New Jersey. 630 pp.
Joshi, R.C., Natt, G.S. and Wright, P.J. 1981. Soil improvement by lime-fly ash slurry injection.
Proc. 10th Int. Conf. on Soil Mechanics and Foundation Engineering, Stockholm, Vol. 3, pp.
707-712.
Kamon, M. 1992. Recent developments of soil improvement. Int. Symp. on Soil Improvment and
Pile Found., Nanjing, China, Vol. 1. pp. 26.
Kawasaki, K., Kotera, H., Nishida, K. and Murase, T. 1996. Deep mixing by spreadable wing
method. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosystems, Tokyo,
pp. 631-636.

Kawasaki, K.,Niina, A., Saitoh, S., Suzuki, Y., Honjyo, Y. 1981. Deep mixing method using ce-
ment hardening agent. Proc. 10th Int. Conf. on Soil Mechanics and Foundation Engineering,
Stockholm, Vol. 3, pp. 721–724.
Kawasaki, T., Saitoh, S., Suzuki, Y. and Babasaki, T. 1984. Deep mixing method using cement
slurry as hardening agent. Seminar on Soil Improvment and Construction Techniques in Soft
Ground, Singapore, pp. 17-38.
Kaye, B.H. 1997. Powder Mixing. Chapman & Hall. 263 pp.
Kézdi, A. 1979. Stabilized Earth Roads. Elsevier, Amsterdam.
Khono, I. 1984. Special issue. Soil cement work method. Stirring impellar, auger system. SBW
work method, simultaneous stirring mechanism in horizontal and vertical directions. Kisoko
(Foundation Engineering and Equipment, Monthly), Vol.12, No.11, pp. 48-53. (in Japanese).
King, R. 1992. Mechanical aspects of mixing, Chapter 13. In: Mixing in the Process Industries,
(Harnby, N., Edwards, M.F. and Nienow, A.W.; eds) Second Edition. London: Butterworth
Heinemann, pp. 250-293.
Komamura, F. and Huang, R.J. 1974. New rheological model for soil behaviour. Journal of the
Geotechnical Engineering Division, Vol. 100, No. GT7, pp. 807-824.
Korkiala- Tanttu, L. 1997. The design instruction for deep stabilized structures. Finnish National
Road Administration. Report 18/1997, 33+9 pp. (in Finnish).
Kouassi, P., Breysse, D., Girard, H. and Poulain, D. 2000. A new technique of kneading compac-
tion in the laboratory. Geotechnical Testing Journal, Marsch, pp. 72-82.
Kujala, K., Halkola, H. and Lahtinen, P. 1985. Design parameters for deep stabilized soil evaluated
from in-situ and laboratory tests. 11th Int. Conf. on Soil Mechanics and Foundation Engineering,
San Francisco, Proc., Vol. 3, pp. 1717-1720.
Lacey, P.M.C. 1943. The mixing of solid particles. Transactions of the Institution of Chemical En-
gineers, Vol. 21, pp. 53-59.
Lacey, P.M.C. 1954. Developments in the theory of particles mixing. Journal of Applied Chemis-
try, Vol. 4, pp. 257-268.
Lakes, R.S. 1998. Viscoelastic Solids. CRS Press. 476 pp.
Larsson, R. 1999. Provbank på kalk/cementpelarförstärkt gyttja och sulfidhaltig lera i Norrland
(Test embankment on gyttja and sulphide clay reinforced with lime/cement columns in Norrala).
Swedish Deep Stabilisation Research Center, Report No. 4, Linköping, 80 pp. (in Swedish).
Larsson, S. 1999. Shear box apparatus for modelling chemical stabilised soil – introductory tests.
Proc. Int. Conf. on Dry Mix Methods for Deep Soil Stabilization, Stockholm, pp. 115-121.
Larsson, S., Rehnman, S-E. & Walter, M. 1999. Laboratory method for design and development of
the dry jet mixing method. Proc. of the XIIth ECSMGE, Amsterdam, Vol. 3, pp. 1533-1538.
Lastovtsev, A.M., Khyalnova, M. and Makarov, J.I. 1962. Khim. Prom. SSSR, Vol. 11, p. 815.
Legatt, C.W. 1951. Method of making the homogeneity test. Assoc. Official Seed Analysis News
Letter, Vol. 25, No. 4, p. 3.
Deep Mixing 151

Lindh, E och Ydrevik, K. 1978. Inverkan av jordmaterialets sönderdelningsgrad vid kalkstabiliser-
ing. Statens väg- och trafikinstitut. VTI Meddelande; 132. 37 pp. (in Swedish).
Lindh, E. 1977. Kalkstabilisering av lera med hög vattenhalt på undergrund med extremt låg
bärighet. Fältstudie vid Linköping 1976. 45 pp. (in Swedish).
Lindh, E. och Hjalmarsson, S-O. 1975. Inverkan av fördröjd packning på vissa egenskaper hos ce-
mentstabiliserade grusmaterial. En laboratorieundersökning. Statens Väg- och trafikinstitut. VTI
Internrapport nr 200. 10+5 pp. (in Swedish).
Lindley, J.A. 1991a. Mixing processes for agricultural and food materials: 1, Fundamentals of mix-
ing. Journal of Agricultural Engineering Research, Vol. 48, pp. 153-170.
Lindley, J.A. 1991b. Mixing processes for agricultural and food materials: 2, Highly viscous liquids
and cohesive materials. Journal of Agricultural Engineering Research, Vol. 48, pp. 229-247.
Lloyd, P.J. 1981. The evaluation of powder mixtures. In: Mixing of Particulate Solids. Proc. of the
2nd European Symposium. Institution of Chemical Engineers Symposium Series No. 65, pp. S3/
K/1–S3/K/8.
Locat, J. 1992. Viscosity, yield strength, and mudflow mobility for sensitive clays and other fine
sediments. Canadian Symposium on Geotechnique and Natural Hazards, GeoHazards ´92, Van-
cover, BC, Proc., pp. 389-396.
Locat, J., and Demers, D. 1988. Viscosity, yield stress, remoulded strength, and liquidity index
relationship for sensitive clays. Canadian Geotechnical Journal, Vol. 25, pp. 799-806.
Locat, J., Bérubé, M-A. and Choquette, M. 1990. Laboratory investigations on the lime stabiliza-
tion of sensitive clays: shear strength development. Canadian geotechnical journal, Vol. 27, pp.
294-304.
Mahesh, V. Bhatia, P.E., Paul, N, and Cheremisinoff, P.E.1979. Solids Separation and Mixing.
Process Equipment Series, Volume 1. Technomic Publication Co., Inc.
Massarsch, K.R. 1976. Soil Movements Caused by Pile Driving in Clay. Rapporter JoB No. 6. Ph D
Thesis, Dept. of Soil and Rock Mechanics, Royal Institute of Technology, Stockholm, 261 pp.
Mathew, P.K. and Narasimha Rao, S. 1997. Effect of lime cation exchange capacity of marine clay.
Journal of Geotechnical and Geoenvironmental Eng., Vol. 123, No. 2, pp. 183-185.
Matsuo, T., Nisibayashi, K. and Hosoya, Y. 1996. Studies on soil improvement adjusted at low
compressive strength in deep mixing method. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on
Ground Improvement Geosystems, Tokyo, pp. 421-424.
McDonagh, M. & Heywood, N.I. 1986. Design of systems for dispersing powders into liquids. In:
Eurochem ´86—Process Engineering Today, pp. 43-49.
Metzner & Otto. 1957. Agitation of non-Newtonian fluids. American Institute of Chemical Engi-
neering, Vol. 3, No. 1. pp. 3-10.
Meyer. 1874. J. Rein. und Angew. Math. 78:130.

Middleton, J.C., Edwards, M.F. and Stewart, I. 1994. Recommended standard terminology and no-
menclature for mixing. Eight European Conference on Mixing Held at the University of Cam-
bridge. Institution of Chemical Engineers, Rugby, UK, pp. vii-xiii.
Miles, S.R. 1962. Heterogeneity of seed lots. Proc. of the International Seed Testing Association,
Vol. 27, No. 2, pp. 407-413.
Miyoshi, A. and Hirayama, K. 1996. Test of solidified columns using a combined system of me-
chanical churning. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosys-
Mizuno et al. 1988. Experiment on the improvement of sandy soil using the deep mixing method.
23rd Annual Meeting, Japanese Soc. of Soil Mech. and Found. Engng., pp. 2301-2304. (in Japa-
nese).
Mori, F., Furukawa, S., Kikuchi, Y. and Matsumoto, K. 1997. Analysis and field loading tests of
building foundations treated by new soil improvement methods. In: Ground Improvement Geo-
systems: Proc. 3rd Int. Conf. on Ground Improvement Geosystems, Densification and Rein-
forcement. Thomas Telford, London, pp. 246-252.
Mork, J.H. 1994. Effekt av sementens forhold mellom gips og hemihydrat på den ferske betongens
reologi. PhD thesis 1994:04. Department of Building Materials. Norwegian University of Sci-
ence and Technology. Trondheim, 285 pp. (in Norwegian).
Mueller, P.E. 1991. Mixing procedures for soil cement. Concrete International, Jan., pp. 53-55.
Muro, T., Fukagawa, R. and Mukaihata, K. 1987a. Fundamental study of mixing conditions affect-
ing the mechanical properties of soil-cement piles. Doboku Gakkai Ronbunshu, 1987, No. 385,
pp. 98-195. (in Japanese).
Muro, T., Fukagawa, R. and Mukaihata, K. 1987b. Effect of mixing condition on deformation and
strength characteristics of cement treated soil. Ehime Daigaku Kogakubu Kiyo, Vol. 11, No. 2,
Nagata, S., Nishikawa, M., Katsube, T. and Takaish, K. 1972. Mixing of highly viscous non-New-
tonian liquids. International Chemical Engineering, Vol. 12, No. 1, pp. 175-182.
Nakajima, N. 1994. Mixing and viscoelasticity of rubber (part 1). International Polymer Science
and Technology, Vol. 21, No. 11, pp. T47-T67.
Narasimha Rao, S. and Rajasekaran, G. 1996. Reaction products formed in lime-stabilised marine
clays. Journal of Geotechnical Engineerin, Vol. 122, No. 5, pp. 329-336.
Nelson, R.D. 1988. Dispersing Powders in Liquids. Edited by Williams, J.C. and Allen, T. Hand-
book of Powder Technology, Volume 7. Elsevier.
Nguyen, Q.D. & Boger, D.V. 1992. Measuring the flow properties of yield stress fluids. Annual
Review of Fluid Mechanics. Volume 24, pp. 47-88.
Nicholson, P.J., Mitchell, J.K., Bahner, E.W. and Moriwaki, Y. 1998. Design of a soil mixed com-
posite gravity wall. Geotechnical Special Publication No. 81. Soil Improvement for Big Digs.
Proc. of Sessions of Geo-congress 98. Oct 18-21, Boston, Massachusetts, pp. 27-58.
Deep Mixing 153

Nickel, S.H. 1976. A rheological approach to diapersive clays. ASTM Special Technical Publica-
tion 623, pp. 303-312.
Nienow, A.W., Edwards, M.F., Harnby, N. 1992. Introduction to mixing problems, Chapter 1. In:
Mixing in the Process Industries, (Harnby, N., Edwards, M.F. and Nienow, A.W.; eds) Second
Edition. London: Butterworth Heinemann, pp. 1-24.
Nishibayashi, K., Matsuo, T. and Hosoya, Y. 1984. Special issue. Soil cement work method. Stir-
ring impeller, auger system. Oval-DM work method. Kisoko (Foundation Engineering and
Equipment, Monthly), Vol.12, No.11, pp. 54-60. (in Japanese).
Nishibayashi, K., Matsuo, T., Hosoya, Y. and Kohinata, T. 1988. Studies of getting low improving
strength of deep mixing method for latter excavation (Part 2) – Investigation the method by
model mixing tests. 23rd Annual Meeting, Japanese Society of Soil Mech. and Found. Engng.,
Nishibayashi, K., Matsuo, T., Hosoya, Y., Hirai, Y. and Shima, M. 1985. Experimental research
into mixing devices used for the deep mixing method of soil improvent (Part 4). 20th Annual
Meeting, Japanese Society of Soil Mech. and Found. Engng., pp. 1747-1750. (in Japanese).
Nishida, K., Koga, Y. & Miura, N. 1996. Energy consideration of the dry jet mixing method. Proc.
of the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosystems, Tokyo, pp. 643-648.
Ogawa, M, Terashima, T. and Orita, S. 1990. Special issue: Latest foundation work equipment.
Implement of soil foundation, SWING. Kisoko (Foundation Engineering & Equipment, Month-
ly), Vol. 18, No. 2, pp. 84-87.
Okumura, R. and Terashi, M. 1975. Deep lime-mixing method of stabilisation for marine clays.
Proc. 5th Asian Regional Conf. on Soil Mechanics and Foundation Engineering, Bangladore,
Vol.1, pp. 69-75.
Okumura, T. 1997. Deep mixing method in Japan. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on
Ground Improvement Geosystems, Tokyo, 14-17 May, 1996. Vol. 2, pp. 879-887.
Omine, K., Ochiai, H. and Yoshida, N. 1998. Estimation of in-situ strength of cement-treated soils
based on a two-phase mixture model. Soils and Foundations, Vol. 38, No. 4, pp. 17-29.
Orr, N. 1979. Assessment of an ordered mix. Powder Technology, Vol. 24, pp. 105-107.
Osterman, J. 1965. Studies on the properties and formation of quick clays. Swedish Geotechnical
Institute, Stockholm. Reprints and Preliminary Reports, No 8, pp. 87-108.
Padgett, P. 1996. Shear rate has greater influence on cement slurry properties than total mixing en-
ergy. Oil & Gas Journal, October 7, pp. 84-91,
Parfitt, G.D. and Barnes, H.A. 1992. The dispersion of fine particles in liquid media, Chapter 6. In:
Mixing in the Process Industries, (Harnby, N., Edwards, M.F. and Nienow, A.W.; eds) Second
Edition. London: Butterworth Heinemann, 1992, pp. 99-117.
Paviani, A. and Pagotto, G. 1991. New technological developments in soil consolidation by means
of mechanical mixing, implemented in Italy for the ENEL Power Plant at Pietrafitta. 10th Euro-
pean Conf. on Soil Mech. and Found. Engng., Florence, Proc., Vol. 2, pp. 511-516.

Perry, R.H., Green, D.W. and Maloney, J.O. 1984. Perry´s Chemical Engineer´s Handbook, (6th
Edn). New York : McGraw-Hill. 2300 pp.
Petry, T.M. and Wohlgemuth, S.K. 1988. Effects of pulverization on the strength and durability of
highly active clay soils stabilized with lime and portland cement. Transportation Research
Record 1190, pp. 38-45.
Poole, K.R., Taylor, R.F. and Wall, G.P. 1964. Mixing powders to fine-scale homogeneity: studies
of batch mixing. Transactions of the Institution of Chemical Engineers, Vol. 42, pp. T305-T315.
Porbaha, A. 1998. State of the art in deep mixing technology: Part 1. Basic concepts and overview.
Ground Improvement, Vol. 2, pp. 81-92.
Porbaha, A. 1999. Dry jet mixing method for liquefaction remediation. Proc. Int. Conf. on Dry Mix
Methods for Deep Soil Stabilization, Stockholm, Sweden, pp. 27-32.
Porbaha, A., Tsuchida, T. and Kishida, T. 1999. Technology of air-transported stabilized dredged
fill: Part 2; Quality assessment. Ground Improvement, Vol. 3. pp. 59-66.
Portland Cement Association. 1995. Soil-cement Construction Handbook. EP003.10S. Portland
Cement Association, Skokie, III. 59 pp.
Pousette, K., Mácsik, J., Jacobsson, A., Andersson, R. and Lahtinen, P. 1999. Peat soil samples
stabilised in laboratory—Experiences from manufacturing and testing. Proc. Int. Conf. on Dry
Mix Methods for Deep Soil Stabilization, Stockholm, pp. 85-92.
Poux, M., Fayolle, P., Bertrand, J., Bridoux, D. and Bousquet, J. 1991. Powder mixing: some prac-
tical rules applied to agitated systems. Powder Technology, Vol. 68, pp. 213-234.
Rajasekaran, G. and Narasimha Rao, S. 1996. Lime column technique for the improvement of soft
marine clay. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosystems, To-
kyo, pp. 443-446.
Rajasekaran, G. and Narasimha Rao, S. 1997. Lime stabilization technique for the improvement of
marine clay. Soils and Foundations, Vol.37, No.2, pp. 97-104.
Rathmayer, H. 1997. Deep mixing methods for soft subsoil improvement in the Nordic countries.
Proc. of IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosystems, Tokyo, 14-17 May,
1996, Vol. 2, pp. 869-877.
Rielly, C.D., Smith, D.L.O., Lindley, L.A., Niranjan, K. and Phillips, V.R. 1994. Mixing processes
for agricultural and food materials: Part 4, Assessment and monitoring of mixing systems. Jour-
nal of Agricultural Engineering Research, Vol. 59, pp. 1-18.
Ries, H.B. 1978. Mixing quality: problems, test methods, and results. International Chemical Engi-
neering. Vol. 18, No. 3, pp. 426-442.
Rogbeck, Y. 1997. Lime-cement columns on the ‘Svealand’ rail link: Performance observations.
Int. Conf. on Soil Mechanics and Foundation Engineering, 14, Hamburg, Proc., Vol. 3, pp.
1705-1710.
Rogbeck, Y., Jelisic, N. and Säfström, L. 1999. Properties of mass- and cell stabilization: Two case
studies in Sweden. Proc. Int. Conf. on Dry Mix Methods for Deep Soil Stabilization, Stockholm,
pp. 269-274.
Deep Mixing 155

Rogers, C.D.F. and Glendinning, S. 1994. Deep-slope stabilisation using lime piles. Transportation
Research Record. No. 1440, pp. 63-70.
Rogers, C.D.F. and Glendinning, S. 1996. Lime Stabilisation. Proc. of the Seminar Held at Lough-
bo Rough Univ., Civil and Building Eng. Dept. Thomas Telford. 183 pp.
Rogers, C.D.F. and Glendinning, S. 1997. Stabilization of shallow slope failures with lime piles.
Transportation Research Record. No. 1589, pp. 83-91.
Rose, H.E. and Robinson, D.J. 1965. The application of digital computer to the study of some prob-
lems in the mixing of powders. American Institute of Chemical Engineers Symposium Series,
Vol. 10, p. 61.
Saitoh, S., Suzuki, Y. and Shirai, K. 1985. Hardening of soil improved by deep mixing. Proc. 11th
Int. Conf. on Soil Mech. and Found: Eng., San Francisco, pp. 1745-1748.
Schaefer, V.R., Abramson, L.W. Drumheller, J.C. and Sharp, K.D. 1997. Ground Improvement,
Ground Reinforcement and Ground Treatment: Developments 1987-1997. 616 pp.
Schofield, C. & Stewart, I.W. 1982. Pigment deagglomeration in paste mixers. Papers presented at
the 4th European Conf. on Mixing Held at Noordwijkerhout, Netherlands; BHRA, Cranfield,
UK, pp. 157-171.
Schofield, C. 1974. The definition and assessment of mixture quality in liquids and pastes. First
European Conference on Mixing and Centrifugal Separation, Churchill College, Cambridge,
England. Paper C1, pp. 1-13.
Schofield, C. and Stewart, I.W. 1980. Pigment deagglomeration in a model mixer. The Chemical
Engineer, july, pp. 486-489.
Shen, S.L., and Miura, M.N. 1999. Soil fracturing of the surrounding clay during deep mixing col-
umn installation. Soils and Foundations, Vol. 39, No. 5, pp. 13-22.
Shen, S.L., Miura, M.N., Koga, K. and Nakamura, M.R. 1997. Geotechnical engineering properties
of clay around soil-cement columns. Jiban Kogaku Kenkyu Happyokai Happyo Koenshu, Vol.
32, No. 2-2, pp. 2387-2388.
Sherwood, P.T. 1993. Soil Stabilisation with Cement and Lime. Transport Research Laboratory.
State-of the Art-Review. HM Stationery Office, London. 152 pp.
Short, D.G.R. and Etchells, A.W. 1982. Unsolved problems of industrial mixing. Papers presented
at the. 4th European Conf. on Mixing Held at Noordwijkerhout, Netherlands; BHRA, Cranfield,
UK, pp. 1-10.
Silvester, R.S. 1985. Mixing of Non-Newtonian media – A technical review. Volume 12 in the
BHRA Fluid Engineering Series. British Hydromechanics Research Assosiation, Cranford, Bed-
ford, England, 92 pp.
Sivapullaiah, P.V., Prashanth, J.P. and Sridharan, A. 1998a. Delay in compaction and importance of
the lime fixation point on the strength and compaction characteristics of soil. Ground Improve-
ment, Vol. 2, pp. 27-32.

Sivapullaiah, P.V., Prashanth, J.P. and Sridharan, A. 1998b. Effect of delay between mixing and
compaction on strength and compaction parameters of fly ash. Geotechnical Engineering Bulle-
tin, Vol. 7, No. 4, pp. 277-285.
Söderblom, R. 1974. A new approach to the classification of quick clay research. Swedish Geotech-
nical Institute, Stockholm. Reprints and Preliminary Reports, No. 55, pp. 13-17.
Sommer, K. 1981. Mechanisms of powder mixing and demixing. In: Mixing of Particulate Solids,
Proceedings 2nd European Symposium, Institution of Chemical Engineers Symposium Series
No. 65, pp. S1/A/1-S1/A/19.
Spencer, R.S. and Wiley, R.M. The mixing of very viscous liquids. Journal of Colloid Science,
Vol 6, No 2, April, pp. 133-145.
Stabilatören. 1999. Nyhetsmagasin från Stabilator & Internordisk Spännarmering, Dec. 1999.
(in Swedish).
Štìrbáèek, Z. & Tausk, P. 1965. Mixing in the Chemical Industry (English edition translated by K.
Mayer). Oxford: Pergamon Press, 442 pp.
Stocker, P.T. 1975. Diffusion and diffuse cementation in lime and cement stabilised soils—chemi-
cal aspects. Australian Road Research, Vol. 5, No. 9, pp. 6-47.
Suzuki, Y. 1982. Deep chemical mixing method using cement as hardening agent. Symposium on
Soil and Rock Improvement, Techniques Including Geotextiles, Reinforced Earth and Modern
Piling Methods, Bangkok, pp. 1.1-1.24.
Sweeney, D.A., Wong, D.K.H. and Fredlund, D.G. 1988. Effect of lime on highly plastic clay with
special emphasis on aging. Transportation Research Record 1190. pp. 13-23.
Taki, O. and Bell, R.A. 1998. Soil-cement pile/column—A system of deep mixing. Geotechnical
Special Publication No. 81. Soil Improvement for Big Digs. Proc. of Sessions of Geo-congress
98. Boston, Massachusetts, pp. 59-71.
Taki, O. and Yang, D.S. 1990. The emergence of soil-cement mixed wall technique and its applica-
tion in the 1990´s. Deep Foundations Institute Annual Meeting, Settle, WA.
Tateyama, K., Fukagawa, R. & Tsujii, T. 1996. Development of a ground monitoring system in
Deep Mixing Method. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosys-
Terashi, M. 1998. Theme lecture: Deep mixing method—Brief state of the art. International Confer-
ence on Soil Mechanics and Foundation Engineering, 14, Hamburg. Proc., Vol. 4, pp. 2475-
2478.
Thompson, M.R. 1966. Lime reactivity of Illinois soils. Journal of Soil Mechanics and Foundation
Division, ASCE, Vol. 92, No. SM5.
Torrance, J.K. 1987. Shear resistance of remoulded soils by viscometric and fall-cone methods: A
comparison for the Canadian sensitive marine clays. Canadian Geotechnical Journal, Vol. 24,
pp. 318-322.
Deep Mixing 157

Tränk, R. & Johnson, U. 1997. Undersökning av KC-pelare med avseende på dess ”homogenitet”
(Investigation of lime/cement column regarding the homogeneity). Arbetsrapport 4, Swedish
Deep Stabilization Research Centre, Linköping. 6+60 pp. (in Swedish).
Tschoegl, N.W. 1989. The Phenomenological Theory of Linear Viscoelastic Behaviour. An Intro-
duction. Springer- Verlag Berlin Heidelberg. 769 pp.
Ueki, H., Hasegawa, K., Suzuki, K. and Bessho, M. 1996. Development of a high pressure jet mix-
ing method for displacement- reducing. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on Ground Im-
provement Geosystems, Tokyo, pp. 767-772.
Uhl, V.W. & Gray, J.B. 1966. Introduction, Chapter 1. In: Mixing: Theory and Practice, Volume 1.
(Uhl, V.W. & Gray, J.B., Eds), New York: Academic Press, pp. 1-5.
Ulbrecht, J. J. & Patterson, G. K. 1985. Mixing of Liquids by Mechanical Agitation. Chemical En-
gineering: Concepts and Reviews. Vol. 1. Gordon and Breach Science Publishers, 345 pp.
Unami, K. and Shima, M. 1996. Deep mixing method at Ukishima site of the Trans-Tokyo bay
highway. Proc. of the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosystems, Tokyo,
pp. 777-782.
Uppal, M.L. and Bhasin, N.K. 1979. A laboratory study on the effect of delayed compaction on the
strength of soil-lime mixes. Road Research Papers, Vol. 160, pp. 1-17.
van Ganse, R.F. 1974. Immediate amelioration of wet cohesive soils by quicklime. Transportation
Research Record; 501, pp. 42-53.
van Olphen, H. 1977. An Introduction to Clay Colloid Chemistry. 2nd Edition. J. Wiley & Sons,
New York. 318 pp.
Venkatarama Reddy, B.V. and Jagadish, K.S. 1993. The static compaction of soils. Geotechnique,
Vol. 43, No. 2. pp. 337-341.
Viatek. 1997. Deep Stabilization Guidebook. Viatek oy. 55+2 pp.
Vilkenas, A., Bredenberg, H. & Persson, O. 1984. Kalkpelarprovtagare. BFR-Rapport. Proj. nr.
821340-0. 49 pp. (in Swedish).
VTI. 1999. Gyratorisk packning. VTI aktuellt 5/99, p. 25. (in Swedish).
Walter, M. 1998. Kemisk stabilisering av jord – en studie av inblandningsprocessen (Chemical sta-
bilization of soil – a study of the installation process). MSc thesis 98/9. Department of Soil and
Rock Mechanics, Royal Institute of Technology, Stockholm, 62 pp. (in Swedish).
Walters, K. 1980. Rheometry: Industrial Applications. Materials Science Research Studies Series.
John Wiley & Sons. 418 pp.
Watanabe, T., Nishimura, S., Moriya, M. and Hirai, T. 1996. Development and application of a
new deep mixing soil improvement method to form a rectangular stabilized soil mass. Proc. of
the IS-Tokyo´96, 2nd Int. Conf. on Ground Improvement Geosystems, Tokyo, pp. 783-786.
West, G. and Carder, D.R. 1997. Review of lime piles and lime-stabilised soil columns. TRL Re-
port 305. 16 pp.

Wikström, A. 1979. Maskinutveckling gällande kalkpelarmaskin LPS-4 (The development of the
lime column machine LPS-4). One-day Seminar on the Practical Application of the lime column
method. Royal Institute of Technology, Stockholm. (in Swedish).
Williams, J.C. 1983. The mechanisms of segregation. In: Mixing in the Process Industries, Volume
One: Particulate Mixing, Proceedings of a Course by The Institution of Chemical Engineers at
University of Bradford.
Yang, D.S., Yagihashi, J.N. and Yoshizawa, S.S. 1998. Swing method for deep mixing. Geotechni-
cal Special Publication No. 81. Soil Improvement for Big Digs. Proc. of Seccion of Geo-con-
gress 98. Oct 18-21, Boston, Massachusetts, pp. 111-121.
Yip, C.W. and Hersey, J.A. 1977. Perfect powder mixtures. Powder Technology, Vol. 16, pp. 189-
192.
Yoshida, S. 1996. Shear strength of improved soil at lap-joint face. Proc. of the IS-Tokyo´96, 2nd
Int. Conf. on Ground Improvement Geosystems, Tokyo, pp. 461-466.
Yoshizawa, H., Okumura, R., Hosya, Y., Sumi, M. and Yamada, T. 1997. JGS TC Report: Factors
affecting the quality of treated soil during execution of DMM. Proc. of the IS-Tokyo´96, 2nd Int.
Conf. on Ground Improvement Geosystems, Tokyo, 14-17 May, 1996. Vol. 2, pp. 931-937.
Deep Mixing 159

Appendix A
NEWTONIAN FLUIDS
Fluids in which the shear stress is proportional to the shear rate are classified as Newtonian fluids.
Their resistance to deformation is dependent on the dynamic viscosity m and is synonymous with
the internal friction under laminar flow conditions. In Fig. A1 the fluid is moving across the solid
surface AB in the direction of the arrow. As a result of internal friction the layers are in motion
relative to each other, the velocity distribution increasing linearly with the distance from the sur-
face. The layer A“B” pulls layer A'B' along with it due to viscous drag. The relation between the
shear rate γ [ s −1 ] and the shear stress τ in a Newtonian fluid is shown in Fig. A2 and is defined by
•
the equation
dv •
τ =µ = µγ [kPa]
dy
v (m/s) dv
A" B"
A´ B´ dy
A B
Fig. A1 Newton's postulate.

Shear stress (Pa)
Vis cos ity (Pa*s )
She ar r ate (1/s )
•
Fig. A2 The relation between shear stress τ and shear rate γ [ s ] in a Newtonian fluid.
−1
Deep Mixing 161

For a Newtonian fluid the dynamic viscosity m is constant. The SI unit for viscosity is the pascal
second [Pa·s]. Table A1 shows approximate viscosities of some common materials at room temper-
ature. Another frequently used unit of viscosity is the poise [P], named after the Frenchman Poi-
seuille (1799-1869). The relation between these units is:
1 Pa·s = 10 P or 1 mPa·s = 1 cP
Another quantity used to characterize a material's rheological properties is the kinematic viscosity
υ. This is defined as the dynamic viscosity divided by the density of the material. The unit is m2/s,
known as the stokes after the British physicist Stokes (1819–1903).
The kinematic viscosity is used to define the Reynolds number, named after Reynolds (1842–
1912):
v⋅d
Re = [–]
υ
where ν is the mean flow velocity of the liquid through a pipe of diameter d and ν is the kinematic
viscosity.
If the velocity becomes too great, the viscosity of the fluid can no longer be described by Newton's
equation. At high velocities the flow ceases to be laminar and becomes turbulent.
Non-Newtonian viscosity is a function of a material that is independent of the flow geometry (Ul-
brecht & Patterson 1985). In most fluids viscosity is a function of shear rate, temperature and pres-
sure. The effect of these factors is discussed briefly below.
Table A1. The viscosity of some familiar materials at room temperature.

(after Barnes et al. 1989)
Liquid Approximate viscosity [ Pa·s ]
Glass 1040
Molten glass (500oC) 1012
Bitumen 108
Molten polymers 103
Golden syrup 102
Liquid honey 101
Glycerol 100
Olive oil 10-1
Bicycle oil 10-2
Water 10-3
Air 10-5

Variation with shear rate
Shear rate can be estimated by dividing the average flow velocity by a characteristic dimension of
the flow geometry, e.g. the radius of a tube or the thickness of a sheared layer. Some typical shear
rates are shown in Table A2.
Table A2. Shear rates typical of some familiar materials and processes (Barnes et al. 1989).
Situation Typical range of Application

shear rates [s-1]
Sedimentation of fine powders in 10-6–10-4 Medicines, paints

a suspending liquid
Levelling due to surface tension 10-2–10-1 Paints, printing ink
Draining under gravity 10-1–101 Painting and coating. Toilet bleaches
Extruders 100–102 Polymers
Chewing and swallowing 101–102 Foods
Dip coating 101–102 Paints, confectionary
Mixing and stirring 101–103 Manufacturing liquids
Pipe flow 100–103 Pumping. Blood flow
Spraying and brushing 103–104 Spray-drying, painting, fuel atomization
Rubbing 104–105 Application of creams and lotions to the skin
Milling pigments in fluid bases 103–105 Paints, printing inks
High speed coating 105–106 Paper
Lubrication 103–107 Gasoline engines
Variation with temperature

The viscosity of Newtonian liquids decreases with increase in temperature approximately according
to the following relationship:
µ = Ae B / T
where T is the absolute temperature and A and B are constants of the liquid. In general, for Newto-
nian liquids, the greater the viscosity, the stronger is the temperature dependence (Barnes et al.
1989). Hence it is important to measure rheological properties at the same temperature as that exist-
ing in the application.
Variation with pressure

The viscosity of liquids increases exponentially with isotropic pressure. However, the effect on
viscosity is relatively small for pressure differences less than one atmosphere. It is therefore usual
to ignore the effect of pressure difference in rheological measurements (Barnes et al. 1989).
Deep Mixing 163

NON-NEWTONIAN LIQUIDS
There are a number of theoretical and empirical models for describing the rheological behaviour of
time-independent fluids. They describe the relation between shear stress and shear rate. The general
equations describing the constitutive relation for simple liquids are:
•
γ =0 τ <τ0
•
τ =τ0 + µ γ τ >τ0
where τ is the shear stress, τ0 is the yield stress that must be exceeded to start flow, γ is the shear
rate, and µ is the characteristic viscosity, see Fig. A3. The characteristic viscosity normally varies
with the shear rate and temperature and as such is known as the “apparent viscosity”.
If the characteristic viscosity is constant the liquid is Bingham plastic and is described by the equa-
tion (Bingham 1922):
•
τ =τB + µ γ τ ≥τB
The model assumes that when the Bingham yield stress τB is exceeded the material flows like a
liquid with constant viscosity. The model describes an ideal plastic fluid, see Fig. A3. The model is
normally used to describe the flow of cement-based grouts (Håkansson 1993).
Shear stress
A: Bingham plastic
B: Yield-dilatant
C: Yield-pseudoplastic
D: Newtonian
Shear rate
Fig. A3 Typical flow curves for yield stress fluids (Nguyen 1992).

Normally the relationship between the shear rate and the shear stress is only linear at high shear
rates. Most fluids show non-linear viscous behaviour, being shear-thinning or shear-thickening, see
curve C and B respectively in Fig. A3. Non-linear viscoplastic behaviour may be described by the
following empirical equation (Herschel & Bulkley 1926):
•
τ = τ ΗΒ + k γ m τ ≥ τ HB
where τHB is the yield stress, and k and m are constants defined as the “consistency index” and the
“flow behaviour index”. For m < 1 the fluid is yield-pseudoplastic and for m > 1 it is yield-dilatant,
see Fig. A3.
A number of additional models for describing fluids are not considered in this report.
Certain pastes, foods, paints, clays etc. exhibit changes in rheology as the time of shearing increas-
es. These fluids are classified as time-dependent and are of two types. In thixotropic fluids, e.g.
paints, tomato ketchup, clays, the apparent viscosity decreases with time at constant shear rate. In
rheopectic fluids, which are very unusual, the apparent viscosity gradually increases with time as
the fluid is subjected to shear. Time-dependent effects are of little importance is rapid mixing proc-
esses.
Deep Mixing 165

Appendix B
The Effect of Mixing Effort on Mixture Quality in Dry Deep Mixing.
Field Trials at Highway E18, Arboga
CONTENTS
Summary --------------------------------------------------------------------------------------------------- 168
1 Introduction ---------------------------------------------------------------------------------------------- 170

1.1 General ----------------------------------------------------------------------------------------- 170
1.2 Statement of problem ------------------------------------------------------------------------- 170
1.3 Aim --------------------------------------------------------------------------------------------- 171
2 Method ------------------------------------------------------------------------------------------------ 171
3 Results ------------------------------------------------------------------------------------------------ 176

3.1 Lab results, column group 1 ----------------------------------------------------------------- 176
3.4 Calcium oxide content ----------------------------------------------------------------------- 182
3.5 Disturbance ratio, column groups 4-6 ------------------------------------------------------ 182
4 Discussion ------------------------------------------------------------------------------------------ 184
References -------------------------------------------------------------------------------------------------- 187
Deep Mixing 167

SUMMARY
The present report is a progress report on field trials carried out within the project “Production con-
trol methods in deep mixing with lime-cement columns”, Part 1: "Mixing methods in dry deep mix-
ing with lime-cement columns". The project is part of an ongoing industry-wide research pro-
gramme run by the Swedish Deep Stabilization Research Centre. This progress report deals with
field trials carried out in connection with the construction of a new section of Highway E18 be-
tween Arboga and Örebro. The field trials took place on 4 November 1997 and the laboratory work
was done between December 1997 and February 1998.
The aim of the trials was to investigate the effect of mixing effort on mixture quality. This was done
by the investigation of:
1 the effect of mixing effort on the distribution of binder over the column section by unconfined
compression tests on samples taken from columns;
2 the relation between the input of mixing effort and (a) the resulting disturbance ratio of the soil
and (b) the distribution of strength over the column section.
For the investigation of the effect of mixing effort on mixture quality, nine columns were manufac-
tured using cement as binder. The columns were divided into three groups based on the retrieval
rate of the mixing tool, the retrieval rates used being 5, 10 and 25 mm/cycle. Other parameters were
kept as nearly constant as possible. Immediately after manufacture, column sections at the –2 m
level were exposed by excavation and samples punched out from different parts of the sections.
The samples were stored in a humidity chamber and unconfined compression tests carried out after
32 – 35 days. A total of 66 unconfined compression tests were performed. In addition, determina-
tions of water content and density were made.
For the investigation of the relation between the input of mixing effort and the resulting soil distur-
bance ratio, a further nine columns were manufactured, but without binder. Mixing tool retrieval
rates of 5, 15 and 30 mm/cycle were used. Immediately after manufacture, column sections at the
–2 m level were exposed and measurements performed on various parts of the sections by means of
a pocket vane test apparatus. A total of 108 vane tests were made on 9 sections.
Some conclusions from the tests are:
• No conclusions can be drawn from the present investigation about the effect of mixing effort on
the distribution of binder over the column section. The results show that no strength gain oc-
curred in the majority of the test columns. Binder distribution can only be investigated in the
columns installed at the lowest mixing tool retrieval rate, 5 mm/cycle. Only trace amounts of
binder were found in the other columns. Thus it is not possible to compare binder distribution at
different retrieval rates on the basis of strength gain. Supplementary sampling and CaO determi-
nations indicate low cement contents at the level investigated, –2 m.

• An important experience gained from the trials is that it is very difficult to do parameter studies
in the field today because such studies often require the relevant parameters to be varied over a
wide range. Because the existing mixing process is designed for a relatively narrow range of
mixing tool retrieval rates, problems arise if the system is pushed to extremes within a narrow
time frame. Field parameter studies should be preceded by laboratory investigations which can
identify in a simpler and more controllable way the parameters most relevant to mixing quality
and the magnitude of their effect on results.
• The investigation of the effect of mixing effort on the disturbance ratio of the soil shows that a
very large mixing effort is required for complete remolding of the soil. From the results of this
investigation it may be inferred that a six-fold increase in the mixing time, corresponding to a
reduction in the retrieval rate of the tool from 30 mm/rev to 5 mm/rev, does not markedly
change the rheological conditions of the mixing process. The most probable active mixing
mechanism under the prevailing conditions is distributive mixing. Conditions may vary in other
soil types, e.g. quick clays.
Deep Mixing 169

1 INTRODUCTION
1.1 General
The present report is a progress report on field trials carried out within the project “Production con-
trol methods in deep mixing with lime-cement columns”, Part 1: “Mixing methods in dry deep mix-
ing lime-cement columns”. The project is part of an ongoing industry-wide research programme run
by the Swedish Deep Stabilization Research Centre. This progress report deals with field trials car-
ried out in connection with the construction of a new section of Highway E18 between Arboga and
Örebro.
The field trials took place on 4 November 1997 and the laboratory work was done between Decem-
ber 1997 and February 1998.
1.2 Statement of problem

In dry deep mixing of soft soils by means of lime-cement columns, there is currently no routine
verification of the achieved binder distribution across the column section. Our knowledge of how
different degrees of mixing effort affect the distribution of the binder across the section (the degree
of mixedness), and the strength gain, is very limited.
The usual practice today is for the client to set up specific requirements for certain parameters of
the mixing process, e.g. binder quantity, retrieval rate, and rotation speed. These “recipes” may be
based on laboratory work, experience, or in the case of larger projects on a number of test columns
installed at the project site. A good indication of the defective nature of the experience involved is
that all of the “recipes” are closely similar no matter what the properties of the local soil may be.
We know, for example, that there are large differences in the effort required to remold two different
clayey soils having the same sensitivity (clay from the Swedish west coast vs. clay from the
Mälaren valley). A given input of mixing effort may result in better or poorer distribution of binder
through the soil depending on the capacity of the soil to resist disturbance. Hence investigations of
the effect of mixing effort on degree of mixedness should take account of the properties of the soil,
in particular its sensitivity and above all the disturbance ratio produced by a given mixing effort.
The disturbance ratio (DR) of a soil is the ratio of its undisturbed shear strength to its shear strength
after a certain amount of disturbance (Massarsch 1976):
τ undisturbed soil
DR =
τ partially disturbed soli
The disturbance ratio of a soil may thus range from unity to the value of the sensitivity.

1.3 Aim
The aim of the tests is to investigate:
1 the effect of mixing effort on the distribution of binder across the column section by means of
unconfined compression tests on samples from columns;
2 the relation between the input of mixing effort and (a) the resulting disturbance ratio of the soil
and (b) the distribution of strength over the column section.
2 METHOD
Since the primary aim of the tests was to study the effect on the distribution of binder over the col-
umn section, the retrieval rate of the mixing tool was varied during column installation. To study
the actual value of the mixing effort, it is necessary to measure the input of energy to the system by
measuring e.g. the torque and the work done by the air jet. In the present study no parameter was
measured which provides an estimate of the energy input. The retrieval rate of the mixing tool was
therefore used as an indirect measure of mixing effort.
For the purpose of the investigation of the effect of mixing effort on binder distribution, nine test
columns were installed in three groups as shown in Figs 1 and 2, using cement as binder. For the
investigation of the relation between the input of mixing effort and the resulting soil disturbance
ratio, a further nine test columns were installed in three groups, without any admixture of binder.
Installation data for the 18 test columns may be found in Tables 1 and 2.
Each group consists of 3

Group 1 Group 4 columns.
Group 2 Group 5
Installation data for each group can
Group 3 Group 6 found in Tables 1 and 2.
Column 1
Column 2
Column 3
Columns with binder Columns without binder
Fig. 1 Plan of column installation. Six groups of three columns were installed.
Deep Mixing 171

Group 1
Group 4
Group 2 Group 5
Group 3 Group 6
A A B B
-2.0 m
≈4m
Section
Extraction of samples for Measurement of disturbance ratio (DR)

unconfined compression by pocket vane apparatus.
tests. 12 readings/column
6 samples/column
Section A-A Section B-B
Fig. 2 Section of column installaton.
Table 1 Installation data for test columns. Parameters common to all test columns.
· Column length 4m
· Column diameter 600 mm
· Mixing tool “Standard tool”
· Penetration rate of mixing tool 100 mm/cycle
· Rotation speed 200 rev/min
· Binder Cement

Table 2 Installation data for test columns. Values of cement addition and tool retrieval
rate according to plot from column machine.
Group Column Cement quantity Retrieval rate

kg/m mm/rev
1 11 25 5
12 22 5
13 22 5
2 21 20 10
22 18 9
23 21 15
3 31 18 24
32 17 22
33 17 25
4 41 0 5
42 0 5
43 0 5
5 51 0 15
52 0 15
53 0 15
6 61 0 30
62 0 30
63 0 30
Shortly after installation of the test columns, the sections at level –2m were exposed by excavation.
The investigation was limited to a study of the sections at one level. This made it possible to carry
out an extensive investigation of a relatively large number of columns and samples. The earth at
level –2m was a gyttja bearing clay with sulphide flecks. The results of laboratory investigation of
the soil are presented in Table 3.
Table 3 Properties of soil at level –2m.
Density 1.45 t/m3

Water content 107 %
Cone liquid limit 100 %
Shear strength, cone test 15–17 kPa
Shear strength, unconfined compression test 15 kPa
Sensitivity 12
Rapidity * 2–3
* A soil sample 50×40 mm is shaken 250× in a Casagrande apparatus. The sample is then rated on a ten-point scale (1–10)
according to the effect of agitation on the soil. A rapidity of 2 means that the sample is not significantly affected. A rapidity of 3
means that a zone of 1 mm thickness at the bottom of the sample was affected.
Deep Mixing 173

The distribution of binder over the column sections was investigated indirectly by means of uncon-
fined compression tests carried out on samples from the test columns. The distribution of binder
across the section is assumed to correspond to the strength gain shown by the samples. The
50×170 mm sampling tubes were punched into the column and exposed by hand as shown in sec-
tion A–A in Fig. 2. A total of 66 samples were taken from the nine test columns as listed in Table 4.
The test columns of groups 1–3 were manufactured using cement alone as binder. The reasons for
using cement alone were:
1 to reduce the risk of problems during sampling;
2 to eliminate uncertainties regarding the lime:cement ratio;
3 that cement requires better mixing.
The samples were stored in a humidity chamber at a temperature of +20 °C. After 32 – 35 days,
unconfined compression tests were carried out with a deformation rate of 0.5 mm/min. The bulk
density and the water content were determined on the same occasion.
As a complement to the determination of the strength distribution across the section, determinations
of CaO content were made for six of the test columns as shown in Table 4. The samples for CaO
determination were prepared by taking pieces from different parts of the cross section, such that
each sample represented an average for the section. This investigation was done to obtain a back-
wards estimate of the quantity of cement injected at the level sampled.
The second aim of the tests was to investigate the relation between input of mixing effort and the
resulting disturbance ratio of the soil. This investigation is most important for linking the mixing
effort-mixedness relation to the soil properties, i.e. the soil's sensitivity to disturbance. For this pur-
pose, nine columns were manufactured without binder, forming groups 4 to 6 in Fig. 1 and Table 2.
The disturbance ratio was determined by vane apparatus measurements carried out on different
parts of the column cross sections according to section B–B (Fig. 2) and on the surrounding soil.
The numbers of vane apparatus measurements made are shown in Table 4.

Table 4 Extent of tests.
Group Column Sampling
Sampling tubes Vane test readings CaO content

50×170mm Lomme pocket vane
test apparatus
1 11 10 1
12 6 1
13 6
2 21 10 1
22 6 1
23 6
3 31 10 1
32 6 1
33 6
4 41 12
42 12
43 12
5 51 12
52 12
53 12
6 61 12
62 12
63 12
Total 66 108 6
Deep Mixing 175

3 RESULTS
3.1 Lab results, column group 1
Column 11 Column 11
300 1700 100

Comressive strength (kPa)
250
Water content (%)

ρ wn 90
Density (kg/m3)
1600
200
80
150 1500
100 70
1400
50 60
0 1300 50
0 100 200 300 0 100 200 300
Distance from colum n centre (m m ) Distance from colum n centre (m m )
Fig. 3 Strength, density and water content of samples from column 11. The circles in
the graph on the right indicate the density and the squares the water content.
Column 12 Column 12
300 1700 100

250
Water content (%)
90
Density (kg/m3)
1600
200 wn
80
150 1500
ρ 70
100
1400
50 60
0 1300 50
0 50 100 150 200 250 300 0 100 200 300
Distance from column centre (mm) Distance from column centre(mm)
Fig. 4 Strength, density and water content of samples from column 12.

Column 13 Column 13
250 1700 100

Comressive strength(kPa)
200
Water content (%)

90
Density (kg/m3)
1600 wn
150 80
1500
100 70
1400 ρ
50 60
0 1300 50
0 100 200 300 0 100 200 300
Distance from column centre (mm) Distance from column centre (mm)
300 1700
250
1600
Density (kg/m3)
R2 = 0,8711
200
150 1500
100
1400
50
0 1300
50 70 90 110 50 70 90 110
Fig. 6 Strength and density as functions of water content in samples from group 1,
columns 11, 12 and 13.
Deep Mixing 177

Column 21 Column 21
10
1700 110
9
8 wn 100
Water content ((%)

Density (kg/m3)
7 1600
6 90
5 1500 80
4
3 70
1400
ρ
2
1 60
0
1300 50
0 100 200 300
0 100 200 300
Column 22 Column 22
14
1700 110
12
100
Water content (%)
10 wn
Density (kg/m3)
1600
90
8
6 1500 80
4 70
1400 ρ
2 60
0
1300 50
0 100 200 300
0 100 200 300
Distance from colum n centre (m m )

Column 23 Column 23
35 1700 110
30
Water content (%)

Densitet (kg/m3)
25 1600
wn 90
20
1500
15
70
10
1400
ρ
5
0 1300 50
0 50 100 150 200 250 300 0 100 200 300
100 1700
80
Densitet (kg/m3)
1600
60
1500
40
20 1400
0
1300
50 70 90 110
50 70 90 110
Fig. 10 Strength and density as functions of water content in samples from group 2,
Deep Mixing 179

Column 31
Column 31
120
1700 110
100
100
Water content (%)

Density (kg/m3)
80 1600
w n
90
60
1500 80
40
ρ 70
20 1400
60
0
1300 50
0 100 200 300
0 100 200 300
Column 32 Column 32
16
1700 110
14
100
Water content (%)
12 wn
Density (kg/m3)
1600
10 90
8
1500 80
6
4 ρ 70
1400
2 60
0 1300 50
0 100 200 300 0 100 200 300
Distance from colum n centre (m m )

Column 33 Column 33
14 1700 110
12 100
Water content (%)

Density (kg/m3)
1600 wn
10 90
8
1500 80
6
ρ 70
4 1400
2 60
0 1300 50
0 100 200 300 0 100 200 300
100 1700
80
Densitet (kg/m3)
1600
60
1500
40
20 1400
0
1300
50 70 90 110
50 70 90 110
Fig, 14 Strength and density as functions of water content in samples from group 3,
Deep Mixing 181

3.4 Calcium oxide content
Table 5 Results of determination of calcium content in six test columns at level –2m.
Sample from CaO Equiv. cement Structure

column content addition *
wt-% kg/column metre
11 3.62 12.1 Tuffic

12 2.18 7.1 Tuffic
21 1.02 3.3 Bivalve-rich
31 2.45 8.0 Between tuffic and bivalve-rich
* Calculated on the following assumptions: soil bulk density ρj = 1.45 ton/m3; soil water content wn = 107 %;
CaO content of Std P cement = 63 %.
3.5 Disturbance ratio, column groups 4-6
Group 4
Rotational speed 200 rev/min
12
11
10
9
Disturbance ratio (-)
8 Column 1
Column 2
7
Column 3
6
5
0
0 0.1 0.2 0.3 0.4 0.5
Distance from column centre (m)
Fig. 15 Disturbance ratio, group 4. Columns installed without binder.

Group 5
12
11
10
9
Column 1
8
Column 2
7 Column 3
6
5
0
0 0.1 0.2 0.3 0.4 0.5
Group 6
12
11
10
9 Column 1
8 Column 2
7 Column 3
6
5
4
3
2
1
0
0 0.1 0.2 0.3 0.4 0.5
Deep Mixing 183

Table 6 Tabulation of disturbance ratios of 9 columns, groups 4-6.
Group Column Retrieval rate Mixing effort* Disturbance ratio

mm/cycle rev/metre min. max. mean Mean for group
4 1 5 210 2.9 5.7 4.2

2 5 210 2.1 5.3 3.6
3 5 210 2.0 8.0 5.0 4.3
5 1 15 78 2.2 5.0 3.6

2 15 78 1.3 2.6 2.0
3 15 78 3.6 4.2 3.2 2.9
6 1 30 43 2.2 4.0 3.3

2 30 43 1.5 2.6 2.0
3 30 43 1.6 4.0 2.9 2.8
* Total of number of revolutions per column metre during penetration and withdrawal of mixing tool.
4 DISCUSSION
In Figs 3–14 we present the results of the laboratory tests performed on the 50×170 mm samples
taken from the nine test columns of groups 1–3. The unconfined compression tests carried out on
the samples from group 1 show a clear trend towards higher strength in the central regions of the
columns, Figs 3–5. The compressive strength is 150 – 280 kPa in the zone 50 mm from the column
centre and 20 – 100 kPa in the zone 250 mm from the column centre. Similarly we can see that the
water content varies over the column section. In the zone 50 mm from the column centre water con-
tents in the region 58 – 72 % were measured, while in the zone 250 mm from the column centre the
water content is 75 – 88 %. Fig. 6 presents compressive strength as a function of water content for
the columns in group 1. It can be seen that there is a good correlation between these parameters.
Fig. 6 also shows density as a function of water content.
A study of the compressive strength of the test columns in group 2 shows that no strength gain has
occurred, Figs 7–9. With isolated exceptions, the results of unconfined compression tests show that
the compressive strength is lower than that of the surrounding soil, Fig. 10. The results of uncon-
fined compression tests and of measurements of the bulk density and water content of the test col-
umns of group 3 show similar results, Figs 11–14. However, it can be seen that one sample from
column 31 attained an unconfined compressive strength of 100 kPa, Fig. 11. Visual inspection of
the samples from groups 2 and 3 reveal only isolated particles of cement gel in some of the sam-
ples. In a number of samples no traces of cement could be found at all. A plausible explanation of
the absence of strength gain in groups 2 and 3 is that not enough cement was injected. However, the
feed plots show that cement was injected at level –2 m in all columns of groups 2 and 3, see Ta-
ble 2.

As a complement to the extraction of samples for unconfined compression tests, determinations of
CaO content were carried out for two test columns per group, Table 5. The CaO determinations
were done in order to see how much cement had been injected in each column. The results after
conversion from CaO content (wt-%) to the equivalent quantity of injected cement indicate a wide
dispersion of values. The largest cement quantity is 12.1 kg per column metre, found in column 11.
The recorded cement quantity for this column was 25 kg/m. In the other columns considerably larg-
er deviations were found between the measured/calculated cement quantity and the recorded quanti-
ty. It is of interest to compare the measured CaO contents of column 12 and column 31. According
to the analysis, these columns contain cement quantities of 7.1 kg/m and 8.0 kg/m respectively. In
column 12 fairly good strength gain was achieved, with compressive strengths up to 250 kPa, while
in column 31 no strength gain was achieved at all (except for one sample whose compressive
strength was 100 kPa). However, traces of cement were found in most of the samples from column
31, although the material showed a very grainy structure. It may be further noted that column 12
attained a relatively good strength gain even though the measured/calculated cement quantity was
only 7.1 kg/m. If it is not the case that mixing effort has a very large effect on strength gain, this
investigation of the CaO content of the columns is open to question.
The primary aim of the present investigation was to study the effect of mixing effort on the distribu-
tion of binder over the column section. However, the results of the laboratory tests on the samples
from groups 1–3 show that it is not possible to analyse them with respect to this primary aim be-
cause no strength gain at all took place in the majority of the test columns. The binder distribution
can only be studied in the columns installed with the lowest retrieval rate, 5 mm/cycle. Since no
strength gain occurred (at the –2 m level) in the columns installed with retrieval rates 10 and
25 mm/rev, the binder distribution, as measured by strength gain, at different retrieval rates cannot
be compared.
The results of the tests on the columns in groups 1–3 illustrate the difficulties of field parameter
studies. There is always a risk attached to carrying out field trials without previous laboratory in-
vestigation to identify what are likely to be the most important phenomena and the most relevant
variables in more complex mixing processes. Moreover, it is difficult to carry out parameter studies
of mixing processes in which the parameters are sensitive to mutual variations. For example, it is
difficult to quickly rejig the whole machine process so that a variable can be tested over a wide
range. In normal production the retrieval rate is kept within a narrow range, about 15–30 mm/rev.
An investigation of the effect of mixing effort on the distribution of binder across the column sec-
tion is likely to require a considerably wider range than this. Based on the experience of the present
investigation, it is very difficult to obtain a satisfactory dispersion process when pushing the system
to extremes in a limited time frame.
The other aim of the investigation was to study the relation between the input of mixing effort and
the resulting disturbance ratio of the soil. The results of measurements of the disturbance ratio for
different mixing efforts are presented in Figs 15–17 and in Table 6. The retrieval rate was varied for
the three groups of test columns-groups 4–6-manufactured without binder. The smallest retrieval
rate differed from the largest by a factor of 6. The pocket vane test measurements show a mean
disturbance ratio of 4.3 for the columns installed with retrieval rate 5 mm/rev (210 revolutions per
Deep Mixing 185

column metre including penetration). The scatter of values was large, however, with readings rang-
ing from 2.0 to 8.0. The mean disturbance ratio for the columns installed with a retrieval rate of
15 mm/rev (78 revolutions per column metre including penetration) was 2.9. The scatter of the dis-
turbance ratio readings was less in these columns, 1.3–5.0. The three columns of group 6, installed
with a retrieval rate of 30 mm/rev (43 revolutions per column metre including penetration) showed
a mean disturbance ratio of 2.8. The scatter of the readings for group 6 was 1.5–4.0, which is less
than those for groups 4 and 5. The sensitivity, which is a measure of the maximum disturbance ratio
attainable, was measured with the pocket vane test at 12.
We can see from the pocket vane test results that a relatively large mixing effort is required in order
to fully remold the soil. The largest disturbance ratio measured was 8.0 in a column in group 4. It is
of interest to observe that many readings around 2 – 3 were obtained regardless of the retrieval rate.
It seems that the shear strength of the soil is quickly reduced by a factor of 2 – 3 upon first being
disturbed, but the soil retains most of its strength during continued mixing notwithstanding substan-
tial mixing effort. Reducing the retrieval rate by a factor of 2, from 30 mm/rev to 15 mm/rev, makes
no significant difference to the measured disturbance ratio. Based on the results of this investiga-
tion, we may conclude that a reduction of the retrieval rate from 30 to 5 mm/rev does not greatly
change the conditions of the mixing process. It would be interesting to investigate whether a mixing
tool with a significantly larger area normal to its direction of motion (e.g. a number of blades tilted
at 45°–90° to the horizontal) would give different results.
Similar results were obtained in an investigation at Norrala, Söderhamn (Larsson 1997). As in the
present study, a number of columns were installed without binder and with varying mixing effort.
Instead of varying the retrieval rate, the penetration rate of the mixing tool was varied between
100 mm/rev and 22 mm/rev. The aim of the investigation was to study whether increased mixing
effort during tool penetration changes the conditions of the mixing process. The disturbance ratio of
the soil at a mixing effort of 55 rev/metre ranged between 1.2 and 2.5. With a mixing effort of
91 rev/metre the disturbance ratio was around 2.5 with relatively little dispersion of the results.
Thus an increase in the total mixing effort from 55 rev/metre to 91 rev/metre did not significantly
affect the conditions of the mixing process. It is of interest to note that the shear strength of the soil
was quickly reduced by a factor of 2 – 2.5, after which it took a large mixing effort to further
remold the soil.
Clayey soils in particular have very pronounced viscoelastic properties which cause the movements
in the material around the mixing tool to be quickly damped. Unless the soil is remolded and takes
on viscous properties, the dominating mixing mechanism will surely be distributive mixing. This
means, in simpler terms, that particles in direct contact with parts of the mixing tool are displaced
by the movement of the tool. Other mixing mechanisms, such as laminar shear, can play only a
minor role since the soil is not being continually deformed by the movements of the tool.

REFERENCES
Larsson, S., 1997. Undersökning av effekten av ett ökat blandningsarbete vid neddrivning av
inblandningsverktyget (Investigation of an increasing mixing work during penetration of the
mixing device). Report, Ostkustbanan, Norrala. 5 pp. (in Swedish).
Massarsch, K.R., 1976. Soil Movements Caused by Pile Driving in Clay. Ph D thesis. Report
No. 6., Dept. of Soil and Rock Mechanics, Royal Institute of Technology, Stockholm. 261 pp.
Söderblom, R., 1974. A New Approach to the Classification of Quick Clay Research.
Swedish Geotechnical Institute, Stockholm. Reprints and Preliminary Reports, No 55, pp. 13-17.
Deep Mixing 187

Appendix C
Deep Stabilisation of Soft Soils
– Laboratory Method for Design and Development
of the Dry Jet Mixing Method
Deep Mixing 189

Deep Mixing 191
Deep Mixing 193
Deep Mixing 195
Swedish Deep Stabilization research Centre
c/o Swedish Geotechnical Institute, SE-581 93 Linkoping, Sweden
Phone: +46 13 20 18 61, Fax: +46 13 20 19 14.
Internet: www.swedgeo.se/sd

SD R6e

Uploaded by

Copyright:

Available Formats

SD R6e

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SD R6e

Uploaded by

Copyright:

Available Formats

Svensk Djupstabilisering

Swedish Deep Stabilization Research Centre

Report 6 Mixing Processes for

English Translation July 2002

2 Swedish Deep Stabilization Research Centre, Report 9

Göran Holm, Linkoping, Sweden Ali Porbaha, California, USA

My warmest thanks go to my supervisor Associate professor Sven-Erik Rehnman, to Professor

Stockholm, May 2000

4 Swedish Deep Stabilization Research Centre, Report 6

Foreword – English Translation ................................................................................................ 3

2. Mixing mechanisms .......................................................................................................... 16

3. Mixing equipment .............................................................................................................. 32

4. Mixture quality .................................................................................................................... 47

6. Deep mixing ........................................................................................................................ 75

7. Assessment of mixture quality: Arboga field trial .................................................. 124

8. Further research .............................................................................................................. 141

References ......................................................................................................................... 144

Appendix A ......................................................................................................................... 161

6 Swedish Deep Stabilization Research Centre, Report 6

8 Swedish Deep Stabilization Research Centre, Report 6

Agglomerate A collection of primary particles and aggregates which are

Aggregate Aggregates are groups of primary particles connected by their

Dough A non-Newtonian material having elastic and viscous flow

Diffusive mixing Natural or spontaneous mixing. The components blend by

Disperse To distribute (become distributed) through a liquid in the form

Dispersion A chemical mixture of fine particles dispersed in a liquid.

Distributive mixing Mixing by redistribution through slicing and replacing.

Flocculation Suspended particles bound together by weak structures in a liquid

Fluid (adj. and noun) Liquid.

Intensity of segregation The intensity of segregation (IS) is a measure of the difference in

Laminar mixing Reduction of the scale of segregation or layer thickness by laminar

Layer thickness The average distance between adjacent boundary layers of

Macromixing Mixing by turbulent or laminar motion on a scale larger than the

Micromixing Mixing by molecular diffusion on a scale smaller than the smallest

Molecular diffusion Reduction in the intensity of segregation. The transport or

Paste A non-Newtonian dispersion with a high concentration of solid

10 Swedish Deep Stabilization Research Centre, Report 6

Shear rate The velocity gradient of a fluid under shear or elongation.

Suspension Fine particles of an insoluble substance uniformly dispersed in a

Turbulent mixing Reduction of the scale of segregation by random turbulent

Viscosity Internal friction, resistance to flow (from Latin viscosus ‘sticky’).

Viscous fluid A fluid having high viscosity (sticky, thick-flowing)

Yield stress The minimum stress required to break up structure of a material

1.1 THE PURPOSE OF MIXING PROCESSES

The objective of mixing is homogenization, manifesting itself in a reduction of concentration or

1.2 DEEP MIXING

12 Swedish Deep Stabilization Research Centre, Report 6

1.4 PURPOSE AND EXECUTION

14 Swedish Deep Stabilization Research Centre, Report 6

Chapter 8, Further research, discusses further research needs in the field.

1.5 IMPORTANT LITERATURE

2.2 CLASSIFICATION OF MIXING MECHANISMS

16 Swedish Deep Stabilization Research Centre, Report 6

Some simple models of flow are discussed in Appendix A.

Additional materials classifications might be (Lindley 1991a):

2.3 CLASSIFICATION OF FLOW TYPES

Two classes of flow types are defined as:

Fig. 2.1 Types of plane flow (after Giesekus 1983).

2.4 AGGLOMERATION AND SEGREGATION