Innovations in Flotation Technology

NATO ASI Series

Advanced SCience Institutes Series

A Series presenting the results of activities sponsored by the NA TO Science Committee,

which aims at the dissemination of advanced scientific and technological knowledge,
with a view to strengthening links between scientific communities.

The Series is published by an international board of publishers in conjunction with the

NATO Scientific Affairs Division

A Life Sciences Plenum Publishing Corporation

B Physics London and New York

C Mathematical Kluwer Academic Publishers

and Physical Sciences Dordrecht, Boston an<:J London
D Behavioural and Social Sciences
E Applied Sciences

F Computer and Systems Sciences Springer-Verlag

G Ecological Sciences Berlin, Heidelberg, New York, London,
H Cell Biology Paris and Tokyo
I Global Environmental Change

NATO-PCO-DATA BASE

The electronic index to the NATO ASI Series provides full bibliographical references
(with keywords and/or abstracts) to more than 30000 contributions from international
scientists published in all sections of the NATO ASI Series.
Access to the NATO-PCO-DATA BASE is possible in two ways:
- via online FILE 128 (NATO-PCO-DATA BASE) hosted by ESRIN,
Via Galileo Galilei, 1-00044 Frascati, Italy.
- via CD-ROM "NATO-PCO-DATA BASE" with user-friendly retrieval software in
English, French and German (©WTV GmbH and DATAWARE Technologies Inc.
1989).

The CD-ROM can be ordered through any member of the Board of Publishers or
through NATO-PCO, Overijse, Belgium.

Series E: Applied Sciences - Vol. 208

Innovations in
Flotation Technology
edited by

P. Mavros
and

K.A. Matis
Department of Chemistry,
Aristotle University, Thessaloniki, Greece

Springer-Science+Business Media, B.V.

Proceedings of the NATO Advanced Study Institute on
Innovations in Flotation Technology
Kallithea, Chalkidiki, Greece
12-25 May, 1991

ISBN 978-94-010-5175-0 ISBN 978-94-011-2658-8 (eBook)

DOI 10.1007/978-94-011-2658-8

AII Rights Reserved

©Springer Science+Business Media Dordrecht 1992
Originally published by Kluwer Academic Publishers in 1992
Softcover reprint of the hardcover 1st edition 1992
No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical, including photo-
copying, record ing or by any information storage and retrieval system, without written
permission from the copyright owner.
To Lina, Panayotis and Thanos
CONTENTS

Preface ix

1. Fundamentals

1.1. The Study of Thin Aqueous Films as Models for Froths and Flotation
R.J. Pugh and E. Manev
1.2. The Mechanisms of Collector Adsorption-Abstraction (Ionic and Non-Ionic
Surfactants) on Heterogeneous Surfaces 25
f.M. Cases and F. Villieras
1.3. Electrochemistry in the Plant 57
S.R. Rao, G. Labonte and f.A. Finch
1.4. Flocculation of Fine Particles 101
f. Gregory
1.5. Applied Mineralogy and Materials Balancing Procedures. Evaluations
of Flotation Concentrators 125
W.Petruk
1.6. Liberation by Size Reduction. Consequences and Improvements
on Flotation Kinetics 149
M.R. MachadfJ Leite
1.7. Elements of Physically-Based Modelling of the Flotation Process 171
HJ. Schulze

2. Chemical Technology Aspects

2. 1. Flotation Kinetics 183

C.Ek
2.2. Mixing and Hydrodynamics in Flotation Cells 211
P.Mavros
2.3. Case Studies of Flotation Control 235
D.J. McKee
2.4. Mineral Surface Complexation in Flotation 263
Z.-X. Sun and W. Forsling
2.5. Amide-Based Technologies in Flotation and Hydrometallurgy of Precious Metals 283
St. Gaidarjiev and S. Spassova
2.6. Determination of Floatability Data using the EmDee Microflot Method 291
M.W. Chudacek
2.7. Froth Flotation Kinetics - Theory and Model Experiments 295
B. Radoev and L. Alexandrova
viii

2.8 Electrolytic Flotation 301

KA. Matis

3. Mineral Processing

3.1. Features and Flotation of Complex Cu-Pb-Zn Sulphides 3fJ7

I. Bayraktar, UA. Ipekoglu and R. Tolun
3.2. Interactions of Anionic Collectors in Flotation of Semi-Soluble Salt Minerals 331
K. Hanumantha Rao and K.S.E. Forssberg
3.3. Flotation of Salt-Type Minerals 357
G.P. Gallios and KA. Matis
3.4. Vibroacoustic Improvements of Froth Flotation 383
St. Stoev, L. Kuzev, M. Metodiev and Sht. Djendova
3.5. The Institute of Geology and Mineral Exploration (IGME) of Greece and its
Research Activities on Flotation 400
D.MaUiaris
3.6. The Influence of Multivalent Cations on the Floatability of Scheelite, Fluorite
and Calcite 419
S. Raatz
3.7. Non-Cyanide Technology for the Flotation of Lead-Zinc Ores 427
V. Panayotov

4. Water and Wastewater Treatment

4.1. Flotation in Water Treatment 431

Th.F. Zabel
4.2. Mechanism of Selective Ion Flotation 455
WI. Walkowiak
4.3. Flotation Techniques in Waste Water Treatment 475
A.l. Zouboulis, KA. Matis and G.A. Stalidis
4.4. Flotation of Pigments and Inks from Waste Paper 499
B. Dobias, W. Klar and K. Schwinger
4.5. A Case Study of Flotation Applied to an Edible Oil Factory Wastewater 513
A. Tarkman, F. Sengiil and N. Ozel
4.6. The Need for Advanced Treatment of Storm water Run-<>ff 521
from Separate Sewer Systems
R. Pfeifer and H.H. Hahn

List of Contributors 527

List of Participants 528
Index 533
PREFACE

The present book is the outcome of an Advanced Study Institute meeting, which was held in
Kallithea, Chalkidiki, in Northern Greece, from 12-25 May 1991 and attended by 69 delegates
from 18 countries.
The Institute brought together scientists, engineers and technologists currently involved in
basic and applied research on the different aspects of flotation. The Institute covered subjects in
four major areas of flotation:
a) fundamentals;
b) chemical technology aspects;
c) mineral processing; and
d) water and wastewater treatment.
Apart from the papers reproduced in this volume, several short oral communications were also
presented. Participants also had the opportunity to visit the Hellenic Chemical Products &
Fertilizers Co. Ltd. mixed sulphides plant, in Chalkidiki.
Conference participants, whose interest and research projects are in this broad field of science
and engineering, provided a well-informed discussion of the problems encountered, as well as
possible directions of future technological developments. It is hoped that this book is not only a
good record of the presentations made (formal and informal), analyzing the state-of-the-art in
flotation, but will also be helpful for students, scientists and technologists working in the fields of
separation processes and in particular mineral processing and wastewater engineering.
All the invited speakers and the participants made this summer school possible, worthwhile
and enjoyable. The sponsorship by the NATO Scientific Affairs Division is gratefully ack-
nowledged. The Editors would like to thank the members of the Organizing Committee, Dr. B.A.
Wills for his advice at the early preparative stages, and Professor C. Ek for his help throughout
the Institute. Many thanks are also due to the members of the Laboratory of General & Inorganic
Chemical Technology (Department of Chemistry at the Aristotle University of Thessaloniki),
namely Ms. A. Daniilidou and Messrs. C.A. Kydros and N.K. Lazaridis for their tireless effort.
Last but certainly not least, the help of Ms. V. Paraschiakou and Ms. D. Zamboulis in the prepara-
tion of this book and the linguistic correlations of Ms. C. Scleparis are gratefully acknowledged.

Thessaloniki, September 1991

The Editors

ix
1. Fundamentals
1.1. THE STUDY OF THIN AQUEOUS FILMS AS MODELS FOR FROTHS
AND FLOTATION

R.I. PUGH and E. MANEyl

Institute for Surface Chemistry
Box 5607
S-1l486 Stockholm
Sweden

ABSTRACT. Thin liquid films playa significant role as basic structural elements of many
disperse systems (foams, froths, emulsions, etc.) as well as in processes of major industrial impor-
tance, such as flotation. The stability of thin liquid films from aqueous surfactant solutions is of
primary interest for froth generation as well as the contact formation between air bubbles and
mineral particles.
The importance of thin aqueous films as models for studying the fundamental behaviour of
these mineral processing operations is discussed. The characteristic properties of the aqueous film
is described and the experimental techniques developed to measure film thickness, stability and
intermolecular interactions are reviewed.

1. Introduction

Studies of thin aqueous films sandwiched between other phases which may be similar or dissimilar
(such as gases, bulk minerals or organic fluids) can contribute towards a fundamental understan-
ding of the interfacial behaviour occurring in mineral processing. These films are usually stabilised
or destabilised by surfactants. They are important since they model various types of dispersed
systems and can be utilised for systematic studies on stability.
They may be classified into the following groups:
(a) free aqueous films sandwiched on both sides by vapour. These films model froths and foam
systems and the stability is controlled by the adsorption of frother at the vapour/solution
interface.
(b) aqueous wetting films sandwiched between macroscopic mineral phases. This system models
minerals dispersions where the stability is controlled by the adsorption of collector or
dispersant at the mineral/solution interface.
(c) aqueous films sandwiched on both sides by an organic liquid. This system models oil in water
emulsions.
(d) aqueous wetting films surrounded on one side by a vapour phase and the other side by a
mineral phase. In this case the solid/aqueous solution and vapour/aqueous solution interfaces

1 Visiting Scientist from the Dept. of Physical Chemistry, Sofia University, Sofia, Bulgaria

P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 1-24.

© 1992 Kluwer Academic Publishers.
2

must be taken into consideration. This type of system is important in mineral flotation, since
the films model the selective attachment process, between an air bubble and a mineral particle.
In fact, following film rupture, the movement of the 3-phase contact across the mineral surface
generates a contact perimeter leading to an adhesive bond.
In a flotation cell, under turbulent flow conditions it is essential to control the drainage, diffusion
and coalescence of the various types of thin films in the system. Together they maintain the delicate
balance between attachment of the mineral particle to the froth and removal of the less valuable
hydrophilic gangue particles.
Essentially, all thin films or dispersed systems are thermodynamically unstable. This is caused
by the free surface energy associated with the large surface area which can be readily reduced by
rupture, coalescence or coagulation. However, the apparent "kinetic" stability or approach to
eqUilibrium can be increased by adsorbed surfactants and eliminating external vibrations, dust,
evaporation, mechanical shock, fluctuations in temperature, etc.

2 • Foams and Froths

The original thin liquid films extensively investigated were so-called soap films or dispersions of
gas bubbles in aqueous solutions. In such structures the term foam and froth are often used
interchangeably. However, it is more precise to consider foams as the unstable, imperfectly drained
liquid-bubble system whereas froths usually refer to the well drained film system, composed of
persistent bubbles. Kitchener [1] clearly demonstrated from a series of photographs the gradual
transition of structure from a nondrained foam (consisting of spherical bubbles) to a well drained
froth, consisting of polyhedral bubbles separated by thin lamellae.
Pure liquids of low viscosity do not produce stable 2-phase foams or froths. However, short
chain surfactants such as alcohols which only moderately reduce the surface tension, can produce
transient foams. To produce persistent froths with well drained thin walls (lamellae) it is necessary
to use long chain surfactants.
In the case of 3-phase mineralised froths then, the situation is more complex. In this case the
size, shape and surface chemistry of the mineral particle (hydrophobicity) has also to be taken into
consideration. These systems have recently been studied in some detail [2-5] and will not be
discussed in the present paper.
According to Leja [6], two major processes are responsible for the collapse of a 2-phase froth
system. Initially, the slow diffusion of gas bubbles to the surroundings and also the thinning,
drainage and rupture of the lamellae. To prevent this occurring, the systems can be kinetically
stabilised by adsorbed surfactant layers. Initially, drainage occurs hydrodynamically under gravity
until the spherical bubbles develop flat walls and form polyhedral structures. At this stage, drainage
occurs by Laplace capillary suction into the borders caused by the curvature of the meniscus in the
Plateau border regions.
Although an enormous amount of effort has been devoted in an attempt to isolate the important
parameter (such as surfactant type, structure, concentration, surface viscosity, etc.) which can
influence the rupture process, the overall phenomenon is still not fully understood. At present,
there is no single theory which can satisfactorily explain the mechanism of froth stability and it is
thought to occur due to a combination of several different mechanisms.
At least four separate types of theories have been proposed to explain the stability of foams and
froths. Early workers emphasised the importance of the mechanical-dynamical properties on the
kinetic stability of the surface film. The experimentally measurable parameters that characterise the
 3

mechanical-dynamical properties of monolayers are surface elasticity and surface viscosity.

Essentially, surface viscosity reflects the speed of the relaxation processes which restore the
equilibrium in the system after imposing stress on it. It is also a measure of the energy dissipation
in the surface layer. In contrast, the surface elasticity is a measure of the energy stored in the
surface layer as a result of an external stress.
Surface light scattering is suitable for determining the viscoelastic properties of monolayers. This
experimental technique may be used in order to gain information on the viscoelastic behaviour of
equilibrium and non-equilibrium monolayers at an air-water interface without disturbing the
original state of the monolayers.
Gibbs' early theory suggested that during stretching a foam lamella, a local increase in surface
tension occurs in thin layers due to depletion of surfactant in the stretched area. This caused a
change in elasticity of the thin film. Gibbs defined a coefficient of surface elasticity (E) as the
increase in tension (y) resulting from an increase in the area (A).

E - Zdy (1)
-dInA

This phenomenon basically takes into account local thinning, due to an increase in surface area,
leading to a decrease in surfactant content in the region and an increase in eqUilibrium surface
tension. However, the theory only applies to a static system where there is insufficient surfactant in
the film to diffuse to the surface and lower the surface tension. Under dynamic conditions, the
Marangoni effect becomes important and operates on both expanding or contracting film and
provide a restoring foree.
The Marangoni effect has been suggested as one of the most important mechanisms to explain
film stability with low molecular weight surface active frothers. The effect depends on both the
lowering of the surface tension and the rate of re-establishing of the eqUilibrium over the lamella
surface. The extension of the surface reduces the concentration of the surfactant in the surface
layer, providing the diffusion from the bulk of the solution to the surface region is relatively slow.
Under ideal conditions the film can rapidly heal itself if stretched locally, by flow of surfactant
from solution under a surface pressure gradient. As yet, there is no clear cut technique to measure
the magnitude of the Marangoni effect and the theoretical treatment is incomplete.
Generally, these early mechanical-dynamical theories did not explain the equilibrium stability of
thin metastable films. In these cases, the fluid film becomes so thin that surface forces are involved
and later Deljaguin [7] introduced the concept of a positive disjoining pressure.
It was recognised by Deljaguin [7] that at least three major types of forces contribute towards the
disjoining pressure (ll) in any thin layer: (a) the electrical double layer (dJ.) interaction, llel caused
by the overlap of the dJ. associated with the interfacial regions on each side of the thin film,
(b) long range van der Waals forces llvdW and (c) short range steric repulsive interactions lls
which depend on the molecular orientation of the hydration layers in the adsorbed film. Hence

II =llel + llvdW+ lls (Z)

These interactions can only occur on thin films (less than 100 nm) where the dJ. begins to
overlap and interplay with the van der Waals forees. The magnitude of II is a function of the film
thickness (h) and must have a positive value at certain thicknesses for the film to remain stable. In
cases where II < 0 then attraction prevails between interfaces which accelerates thinning, leading to
rupture.
4

A more detailed analysis of the individual forces contribution to stability will be discussed in the
following sections.

3. Thin Film Structure

In Figure la the general case of a thin aqueous film y',ofthickness t, is shown sandwiched
between bulk phases a and~. Both a and ~ contain surface phases ay' and ~y' with thickness
dUY' and df3y', respectively, where monolayer or multilayer adsorption can occur. If we consider the
interfacial plane corresponding to x= 0 then we can introduce a surface potential value of '\jJo and
charge <Jo . Within ay' and ~y' specifically adsorbed or hydrated ions may adsorb such that we can
introduce a Stem potential '\jJuJ' at x = dUY··
For equal charged surfaces, we can also construct an overlapping electrostatic double layer
according to Figure I b, which may contribute towards the electrostatic stability and positive
disjoining pressure of the system.

a b@
/

, , "'0 Y

a ay y/ 13v P a

~
day
- .... ~ f I~X
t

Figure 1. (a) Aqueous film y' sandwiched between two bulk phases a and ~; t is the film thickness and
day' and d~y' are the thicknesses of the interfacial regions ny' and ~y'. (b) Overlapping electrostatic double
layers cause distribution of potential.

For a mineral surface, the Stem layer potential can be taken as representing the zeta potential of
the system. However, in the case of an air bubble/aqueous solution or oil droplet/aqueous solution
the charging mechanism and location of '\jJd is less well characterised. Also for a well defined fIlm
system, the combined Hamaker function across the film can be calculated using accessible
refractive index/wavelength data for the various phases. Finally, in the case of adsorbed surfactant
molecules in the interfacial regions, consideration must be given to the short range steric
interactions.

4. Types of Thin Films

To achieve a stable thin fIlm a minimum concentration of surfactant is required. TIlin films can be
 5

classified into the following types (a) common thin films, (b) common black films and (c) Newton
black mms. The thickness and characteristic properties of each type of thin film differ substantially.
Usually common thin films (CF) occur in regions of low electrolyte concentration where IIeJ
dominates over IIvdW. The thicknesses of common thin films are in the region of 30 nm. In these
regions IIvdW is usually relatively small that IIeJ = Pc where Pc is the capillary pressure of the
meniscus. Rupture of the thin fIlm occurs at a critical thickness, her. Systematic studies on film
eqUilibrium, especially in the low electrolyte region, give the possibility of determining values of
\jJo and 0 0 , Common black films (CBF) occur during the thinning process when the conditions for
equilibrium are not completely fulfilled. During this thinning process on addition of electrolyte, a
semi-equilibrium state is reached when the fIlm either ruptures or an abrupt sudden transition to a
thinner stable film occurs through formation of black spots. The film formed at such low thickness
values (40-50 A) is called a Newton black film (NBF).
The conditions of eqUilibrium of the common films, common black films and the transition from
common black films to Newton black films have been studied in detail over the past 30 years. The
critical electrolyte concentration, Ccr , corresponding to the transition from common to Newton
black films has been shown to depend on the type of surfactant and is usually one order of
magnitude higher for ionic surfactants in comparison with non-ionics [8].
In Figure 2, a II versus h isotherm is shown, which can be conveniently used to explain
equilibrium conditions and transitions from common to Newton black films. According to this
diagram, two stable (equilibrium) thickness regions are possible, which causes the formation of the
two different types of black films. Common black films are formed at larger film thicknesses or
h> hmax where

II= IIeJ + IIvdW (3)

and for Newton black films at shon distances then

II = IIeJ + IIvdW+ II st (4)

In some cases, extremely thin NewlOn black films can be formed, where IIeJ is swamped but
there mechanical steric forces still remain due to adsorbed surfactant layers which retain an aqueous
core. The water distribution in the core in some cases could involve liquid crystals or bilayers. For
example, two-dimensional bimolecular leaflets of 4.2-4.5 nm have been reponed for sodium oleate
[9]. Foams, in which the stability of black films are prolonged, are characterised by an extremely
fragile structure at the top of the foam. The structure can often persist a long time after drainage of
the unbound water has occurred.

5. Experimental Studies on Free Aqueous Films

Most of the pioneering studies on thin films have been done in the Soviet Union and Bulgaria. Free
aqueous films are the most common and a method for studying these films has been developed by a
microinteferometric method as reported by Scheludko and Exerowa [IO]. Schematics of the
experimental set-up is presented in Figure 3. The film (0.05 to 1.0 mm radius) is formed between
the tips of the meniscii of a biconcave drop held in a cylindrical glass tube (of ca. 2 mm radius) by
sucking the liquids out of the tube. The amount of liquid in the biconcave drop is controlled by a
teflon microsyringe by which the radius of the film can be varied. The measuring cell (containing
6

IT

•
Q)
~

:l
(I)
(I)
Q)

c.
~

Cl ITmax
c: p
c:
'0
:~ hZ
c h

Figure 2. A disjoining pressure isotherm in 0.01-0.1 M electrolyte. II is the disjoining pressure, Pc is

the capillary pressure and h is the film thickness. h corresponds to a common film (CF) '" 300 A, hi to a
common black film (CBF) and h2 to a Newton black film (NBF) '" 40-50 A. The isotherm shape depends
on the type and concentration of surfactant and electrolyte.

the glass tube-holder of the film) is enclosed in a water jacket which is maintained at a constant
temperature (± 0.1 °C).
The film thickness is determined by interferometry. The reflected light intensity at its maximum
reflection and minimum reflection are measured when the optical distance corresponds to n).,/4,
where n is the refractive index. The equilibrium thicknesses of free films from aqueous solutions of
surfactants can be determined at a range of surfactant concentrations and the Pc can also be
measured by the method of capillary rise. Corrections at small film thicknesses due to optical non-
homogeneity (according to the model, taking into account the differences in the optical densities of
the adsorption layers and the aqueous core) can be made as in refs. [11.12].

5.1. THEORY

According to DLVO theory the eqUilibrium thickness of the film is governed by the condition that
the sum of the capillary pressure. pc. and the other additional pressures, IT (e.g. the various
components of the disjoining pressure) is zero.

(5)

Here Pc originates from the curvature of the liquid meniscus in which the film is formed. The
 7

procedure of accurate detennination of the capillary pressure is described in detail by Exerowa et al.
[13). In the case of small films and small contact angles. its value can be sufficiently accurately
detennined from the approximate fonnula <Rc is the radius of curvature of the liquid surface)

Pc =2y/R;

The van der Waals (molecular) component of the disjoining pressure TIvdW can be represented
by the well-known expression

A(h)
TIvdW =- 6nh3 (6)

where A(h) is the "Hamaker function" of the molecular interactions weakly dependent on the film
thickness h. Values of TIvdW for the thin surfactant films can be calculated following Nir and
Vassilieff [14).

SI-concave
menlscus\

Film-forming I 1\.\
glass holder 1--II~1
Lower portion of the cell

Inverted Photo Amplifier Recorder

microscope detector
for reflected
light

Figure 3. Experimental apparatus for studying thin aqueous free films.

TIel can be detennined from the basic expressions of the DLVO theory. derived from the
Poisson-Boltzmann equation and can be calculated from the approximation

TIel =64 RT tanh2 Cf!o exp(-xh) (7)

where
8

zeo'l/1o
QJo = 4kT
z is ion valency, eo is elementary charge, '1/10 is potential of the diffuse electric layer at the plane of
charge, x is the inverse Debye length).
The charge of the diffuse double layer can be calculated using the relationship derived from the
Grahame equation at T = 293 K:

00= 0.116..JCci sinh ('1/10150.5) (8)

where 0 0 is the electric surface charge per unit area (C m· 2), Cel is the electrolyte concentration in
mols per dm 3 , '1/10 is expressed in millivolts.

125 100

---E
,...
...... 100 80
E
..s z
I = 10.4 M
§.
75 60
!II
!II
Q) I:
I: 0
.:.::

-
"iii

-
"~ 50 40 I:
J: 'Y Q)

E Q)
0
ii: 25 20 co
't:
:J
CMC en
0
10- 7 10- 6 10-s 10- 4 10- 3 10- 2 10- 1
Surfactant concentration (M)

Figure 4_ Equivalent equilibrium film thickness as a function of C 12(EO)s bulk concentration. Open
circles, 10-4 M KCl; closed circles: 10-3 M KCI; Y indicates the surface tension isotherm at 20°C. pH =
5.7 ± 0.1 [16].

5.2. EXPERIMENTAL

The stability of free aqueous films in the presence of nonionic surfactant (representing model
frothers) are of particular interest in mineral processing. Values of the diffuse electric layer ('1/10) at
the air/aqueous solution interface in the presence of adsorbed non-ionies have been determined by
Exerowa et al. [13] from eqUilibrium film thickness studies. Experimental measurements were
carried out at constant ionie strength at pH - 5.5 at a range of surfactant concentrations (from 10-6
to 10-2 M solutions). From these studies it was found that the eqUilibrium film thicknesses reached
a plateau at surfactant concentrations before c.m.c. and the potential of the diffuse electric layer
 9

corresponding to this region was estimated. In the work of Exerowa and Zacharieva [15] the
equilibrium thickness of films containing non-ionics of the C12(EO)n type was found to be
sensitive to the pH and electrolyte concentration and an isoelectric point at pH =3.7 was reported.
In a more recent study, Manev and Pugh [16] considered electrostatic effects in aqueous foam
films containing a nonionic surfactant and discussed some of the early results. They determined the
equilibrium thickness and stability of films stabilised by pentaethyleneglycol-n-monododecylether
of high purity. The influence of pH and ionic strength on the values of the "1\10 and a o and the value
of pHiep at the air/surfactant interface were determined from the DLVO theory.
Figure 4 shows the equivalent equilibrium film thicknesses of the aqueous foam films as a
function of C12(EO)s bulk concentration at natural pH (pH = 5.7 ± 0.1) at two ionic strength
values (10-4 M and 10.3 M KCI added, respectively) and temperature 20 ± 0.1 °C. The surface
tension of the C12(EO)s aqueous solutions are also shown in this figure.
The film thicknesses determined experimentally as a function of nonionic surfactant concentra-
tion in the concentration range from 10-7 M to 4.5xlO- 2 M exhibit several features of interest.
Equilibrium films of varying thickness form throughout this concentration range. Exceptions are
the unstable (rupturing) films which occur in all cases at the lowest studied surfactant concentration
(= 10- 7 M) and the occasional ruptures at the other concentrations below the c.m.c. At a higher
surfactant concentration (~ 10- 3 M) critical phenomena are also observed. In this case, however,
they do not lead to the destruction of the films. Instead stable, so-called "black films" of smaller
thickness « 10 nm) are formed.
These plots, relating the dependence of equilibrium film thickness to C12(EO)s concentration,
suggest that the films are stabilised by an electrostatic disjoining pressure, despite the fact that the
surfactant in the system is nonionic. Similar behaviour trends have been well established for ionic
surfactants, where the eqUilibrium film thickness decreases as the ionic strength increases, caused
by the increase in surfactant concentration. Moreover, the films are extremely sensitive to pH in the
system. This is clearly illustrated from the results presented in Figure 5 showing the film
thicknesses for two C12(EO)S concentrations (10-S M and 10-4 M) as a function of pH. The curves
reach a maximum and are assymmetrical with respect to the natural pH (-5.7), although the
equilibrium thicknesses decrease in either direction from the maximum.
However, at low pH values, critical thicknesses and rupture (at lO- s M) is observed at pH $; 3,
while in the basic range the reduction in equilibrium thickness is more gradual and equilibrium film
thicknesses are obtained at the ionic strength corresponding to the critical thicknesses in the acidic
range. The gradual decrease of the eqUilibrium thickness in the basic range, should be mainly
attributed to the suppression of the electrical double layer repulsion upon increasing the ionic
strength, while the sharper decrease in the acidic range suggests reduction of the surface charge.
These findings indicate negatively charged film surfaces above pH 3 with a pHiep = 3.
From the experimental data the values of the potential of the diffuse electric layer and the specific
charge of the boundary were calculated as a function of C 12(EO)S concentration, pH and ionic
strength. These studies showed the influence of surfactant concentration on the "I\Io-potential as a
function of inert electrolyte concentration of KG and the results clearly indicated a decrease in "1\10,
beginning to occur above 5xlO-6 M C !2CEO)s. In addition, although the fall in "1\10 started below
the c.m.c. (by an order of magnitude) it continued beyond the c.m.c. From these studies it was
concluded that the decrease in 1/Jo-potential could be attributed to the displacement (desorption) of
ionic species (H+ and OH- ions) from the solutiOn/air interface at high surfactant adsorption
densities.
It was also found that the diffuse electric layer potential preserved its plateau value upon
decreasing surfactant content, to as low concentrations as possible to form a film. This data
10

appeared to support the idea that the charge is not due to the adsorption of the surfactant but is a
property of the air/water interface itself under the studied experimental conditions. Although the
charge (and potential) starts decreasing below c.m.c., this occurs only at very high adsorption
levels, near saturation. To explain these results, it was suggested that a change of state in the surfa-
ctant adsorption layer occurs. Such a change, actually condensation or transition from gaseous to
liquid-expanded condensed monolayer, has been established with ionic surfactants. This behaviour
appeared to be confirmed from studies on black spots which can be considered as being closely
related to the phase transition and the so-called Cb)' or the minimum bulk surfactant concentrations
sufficient to produce black spots in microscopic films. This characteristic concentration, associated
with structural changes at the film surfaces, was introduced by Scheludko [17].

200

-
--
E 160
t:

120
Ul
Ul
Q)
t:
80

-
~
0
.t:

E 40
:c
U.
0
0 2 4 6 8 10 12 14

pH

Figure 5. Equilibrium film thickness of free films from C12(EOh aqueous solutions as a function of pH
for two surfactant concentrations: Open circles 10-5 M C12(EO)s; closed circles 10-4 M CI2(EO)S [16].

The influence of frother/collector interactions on the stability of free aqueous films has also been
studied. The thickness of eqUilibrium common thin films and black films from aqueous solutions
of the nonionic frother C12(EO)5 were established also in the presence of potassium ethylxanthate
(KEX) of varying concentrations [18]
The established equivalent eqUilibrium film thicknesses, are presented in Figure 6 as a function
of the studied frother (C12(EO)5) and collector (KEX) concentrations. Several characteristic featu-
res of these films should be considered. Firstly, eqUilibrium films form at all frother and most
collector concentrations in the studied range but the addition of the collector makes the films
increasingly unstable. It is not possible at all to form films at the lowest frother concentrations in
the presence of xanthate: instant rupture of the films occurs at great thicknesses (above 100 run),
even before reaching the critical thickness of rupture. Secondly, the effect of addition of the colle-
ctor appears to be analogous to that of a common electrolyte. Upon increasing KEX concentration
 II

the equilibrium film thicknesses are reduced significantly, apparently because of the suppression of
the electrostatic repulsion, although the frother in the system is nonionic.
However, it was shown that the xanthate played an additional role in controlling the thickness
and stability of the film. Upon increasing KEX and/or C12(EO)s concentration, the eqUilibrium
thicknesses of the film containing KEX remain larger than the values obtained with common
electrolyte (see broken line in Figure 6). In terms of the electrostatic interactions, this means !hat
the reduction in the surface charge, in increasing C12(EO)s concentration, is substantially delayed
in the presence of KEX.

150

I = 10-4 M
125 ~--o.
CY-" .....,
E --0--- '&
.s 100
",~,
Ul
Ul
Q)
s:::
75
1= 10-3 M
'\ 0
..:.::
----
~_~_Jl __!
",,, • •
•

-
.2 ....... 0
.c 50
E
u::: 25
", • • 0

\\
CMC
'------- ------
\

0 t
10- 7 10- 6 10- 5 10- 4 10- 3 10- 2 10- 1

Frother concentration / (M)

Figure ii. Equilibrium film thickness as a function of C12(EO)s concentration at two different ionic
strengths. controlled by addition of KEX. Open circles: 1()4 M KEX; closed circles: 10-3 M KEX. Broken
lines indicate the respective dependences obtained with KCI as an electrolyte [18].

The natural explanation is that the negatively charged xanthate species interact with the non-ionic
frother molecules in the adsorption layers on the film surfaces and contribute to building the electric
charge and negative potential there. This conclusion is supported by the observed effect of the
xanthate on the film stability, that is, the increasing fragility of the films upon increasing xanthate
concentration. This "antifoaming" action can readily be linked with the presence of the short-
chained xanthate at the film surface.

5.3. THE RELATIONSHIP BETWEEN THE STABILITY OF A FREE AQUEOUS SINGLE

FILM AND A FROTH

Many fundamental experiments have been made on single aqueous films and it is important to
discuss the relevance of these studies to a froth. Essentially a froth system cannot be considered as
12

a simple combination of independent aqueous films and is obviously a much more complex
system, consisting of many interrelating elements such as Plateau-Gibbs borders and junctions.
However, the destruction of a single thin film in a froth system can lead to the destruction of an
adjacent film, causing local fluctuations in pressure and surface tension. This can lead to a so-called
"collective" effect [19] and this behaviour has been investigated in some detail. From these studies
it was concluded that changes in thickness and disjoining pressures in thin independent films playa
decisive role in the stability of a froth system [20].
Froths are also formed in a gravitation field and under these conditions it must be recognised that
bubble growth, coalescence and drainage can change at different heights due to pressure differen-
ces etc. This can produce differences in froth structure, particularly under practical conditions such
as column flotation where a downward wash flow has also to be taken into consideration
(Figure 7).

WQsh
water
c:onceontra1e-

draining
froth
E;> o.a 0

pack~d
bubble bed
!
E,> 0.74

expande-d \
bubble bed

..
E9<O.74

bubbling
o 0
zone- o 0
• 0
00 • 0

Figure 7. Suggested two-phase froth structures in froth flotation column [21].

If a pressure difference L'lP is applied to a foam system (with L'lP '" Pc) then in cases where L'lP
has a fairly high value it will be reasonably constant for the entire foam system and equal to TI.
From experimental studies the foam lifetime (t p ) can be determined at a range of L'lP values
using different surfactants. By changing the electrolyte concentration it was found possible to study
different thin film systems such as common, common black and Newton black films (Figure 8).
Results are shown in Figure 8 for sodium dodecylsulphate and show a range of stability
behaviour for different systems under a range of pressures. A series of experimcnts has also been
carried out to compare the effect of Pc on the stability and thickness of free films produced in a
porous plate cell with the stability of a froth system. The critical disjoining pressure (TIcr)
 13

corresponding to the point of rupture of a single film was compared to the critical pressure to
destroy the froth (~Pcr)' Generally, it was found that the froths gave lower ~Pcr values in
comparison to TIer values for the films. This was explained by the collective effect and the larger
size of the foam films.

8r--------------------------------.

4 (c) 8 ;:
Newton black films
4 x 10- 1 M NaCI
'E
Fairly stable 6 .!::.
E 41
'"o
u.
2 4 E

2
~
E
'"
0
u.
0 0
10 3 10' 10 5

Pressure difference in Plat'!au-Gibbs borders I [Nm"]

Figure 8. Dependence of the foam lifetime (t p ) on the pressure difference (~P) in the Plateau-Gibbs
borders (a) minutes (b,c) hours. Sodium dodecyl sulphate stabilised foams [19].

6. Experimental Studies on Aqueous Films Sandwiched between Solid Surfaces

(the Surface Force Apparatus)

The forces acting on a thin film sandwiched between smooth solid surfaces can be measured as a
function of surface separation using a surface force apparatus developed by Israelachvili [22]. This
apparatus (Figure 9) is thoroughly described in detail and a brief description is given below.
The ideal surface to use in these measurements is muscovite mica (a layered, negatively charged
aluminosilicate mineral) which can be cleaved into macroscopic molecularly smooth sheets. Two
such sheets are glued onto half-cylindrical silica discs (radius R = 1-2 cm). One of them is mounted
on a piezo-electric crystal, which is employed in order to regulate the surface separation. The
distance between the two surfaces is measured interferometrically to ±O.l nm, using fringes of
equal chromatic order. The second surface is mounted on a spring. Interferometric measurements
of the deflection of this spring makes it possible to detennine the interaction forces with a detection
limit of about 10-1 N. This is sufficient for a precise detennination of all important surface forces
(electrostatic double-layer forces, van der Waals forces, steric forces, hydration forces and hydro-
phobic interactions) and critical testing of existing theories of interparticle forces.
Hence, by using this method we can directly detennine how the adsorption of surfactants
14

influences the interaction between solid particles separated by an aqueous film. From these
measurements, conclusions about the stability, structure and orientation of adsorbed surfactant or
polymer layers in a wide range of electrolyte concentrations are easily obtained (monolayers or
bilayers, headgroup or tail against the solution). By further analysing the measured forces within
the framework: of the DLVO theory, the effective surface potential can be determined as well as the
presence of forces not accounted for in that theory.

Iighl 10
speclromeler

upper rod

movable clamp

sliff double-
canlilever spring

variable
stiffness "'='":1':+'+-- helical spring
force-
measuring
spring
lower rod
o cm 5
while lighl

Figure 9. The surface force apparatus for measuring the interaction between mica surfaces.

The surface force apparatus has also been used to study the interaction between mica coated with
hydrophobic surfaces and more recently thin sapphire crystals have been used instead of mica [23].
The surface force apparatus was originally used to test DL VO theory by studying the interaction
curve for mica surfaces in water. It was found possible to fit this curve as an electrostatic repulsion
using the non-linear Poisson-Boltzmann equation with constant charge boundary conditions. Expe-
riments have also been carried out to study the interactions of ethylene oxide hydrophilic tails [24).
These types of hydration forces can be expected from hydrophilic flotation depressants or disper-
sants such as starches and tannins.
In such studies it was necessary to adsorb the ethylene oxide surfactant molecule with the
hydrocarbon tails towards the substrate and the ethylene oxide tails extending into the aqueous
solution. In order to obtain this configuration, the mica was made hydrophobic by depositing a
compressed molecular film of dioctadecyldimethylammonium bromide (DDOA) on the naked mica
surface using a standard Langmuir-Blodgett (L.B.) technique. Such a monolayer is electro-
 IS

statically attached to the naked mica surface and is very close-packed, since each molecule has two
hydrocarbon chains. The advancing contact angle of the water on the hydrophobed mica surface
was 94°. The nonionic surfactant was allowed to adsorb from solution onto this hydrophobic
substrate. (Figure 10 shows a sketch of the surface after adsorption of C12(EO)s surfactant).

Mica

Figure 10. Sketch of the C 12 (EO)s coated surfaces. The mica surfaces are first covered with a double-
chain cationic surfactant (dioctadecyldimethylammonium bromide - [Meh [C 18h NBr) oriented outwards,
oxyethylene groups of C12(EO)5 are oriented towards the aqueous phase.

On introduction of C 12(EO)5 into the measuring chamber (6x 10- 5 M concentration), the
surfactant adsorbed on the hydrophobic surface and surface force measurements were performed as
a function of separation distance at different temperatures. The force measurements at 30°C
between the C12(EO)S coated surfaces is shown in Figure 11. At large separation (d > 5 nm) there
is a repulsive double layer interaction which is in agreement with a theoretical CUNe calculated from
the Poisson-Boltzmann equation assuming a surface potential of 46 mY. This is a long range
repulsion independent of temperature. There is also a force maximum at 4.5 nm and a steep, short-
range repulsion at separations below 2.1 nm. The short-range components of the force show a
dramatic change with temperature as indicated (see Figure 12). At 15°C the repulsion continues to
increase with decreasing separation and becomes larger than the double layer interaction below 3.5
nm. At 20°C the short range repulsion decreases in range, while at 30°C the secondary minimum
has moved in and the repulsions is still further reduced. At higher temperatures the minimum
becomes more attractive and shifts towards smaller separation distances and finally at 37°C reaches
a separation distance value of < 1 nm.
These experiments show that a decreased repulsion occurs with rise in temperature for the
C12(EO)s coated surfaces. The temperature dependence on the long-range repulsive double layer
16

force is negligible over the range Sludied and most probably arises from imperfect neutralisation of
the charges on the mica surfaces, by adsorbed dioctadecyldimethylammonium bromide (DDOA).

·1.0
SHORT RANGE
REPULSION

E 0.5 I-,
........
z
..s.
..' _
........ ......... .. - ... _-_ _-
.. ..
LONG RANGE
REPULSION

I~
0: /
........... 0 I
lLu ~
ATTRACTION

I I I

5 10 15 20
d (nm)
Figure 11. The measured force, FJR (where R is the mean radius of curvature of the surfaces) between
two C12(EO)s-coated surfaces at 30 °C as a function of surface separation. The dashed-dotted line is a force
calculated from solutions to the non-linear Poisson-Boltzmann equation with constant charge boundary
conditions and the following parameters: x-I = 100 nm, 0 0 =0.4 mC/m2, corresponding to a 1:1 electrolyte
concentration of9xlO-6 M and 1j!~ = 46 mV [24).

The very strong short range repulsion forces can be explained by a steric effect from the interaction
of the ethoxy head groups. It can be assumed that the repulsion decreases with increasing
temperature due to a "dehydration" of the head groups, the water in the hydration shell becoming
more like bulk water (a break-down of the hydration-shell water structure).
Since hydration forces give a repulsive potential, it is not surprising that a hydrophobic surface
produces an attractive potential and this has also been studied in some detail. Experiments were
carned out by adsorbing dodecylammonium ions onto the clean mica surface [25]. In Figure 13,
the distance dependence of the forces acting between mica surfaces in solutions of dodecyl-
ammonium chloride are displayed. For each concentration, theoretically predicted surface force
curves are shown which were obtained by solving the numerically exact solution of the non-linear
Poisson-Bolzmann equations and adding an estimated van der Waals force. In pure water the
interaction is dominated by a long-range repulsive double-layer force which is overcome by an
attractive van der Waals force at a separation of 3.5 nm. This agrees well with previous
measurements on the mica-water system. The surface jump into adhesive contact at a separation of
7.0 nrn.
In the concentration range 9xlO· 6 - 4xlO- 5 molll the double layer force is very weak or
undetectable. The data is shown for I x 10. 5 mol/l. Obviously, the adsorption of DAH+ and H+ ions
completely neutralises the negative charge on the mica. However, after about one to two days of
equilibration, at the higher concentration (4x I 0-5 molll) a small surface charge was built up. 1he
 17

very slow build-up of this small charge was slightly faster at the higher concentrations. The jump
into adhesive contact took place at 10.0 nm for the uncharged cases. This is further out than the
theoretically predicted jump distance of 7 nm for a nonretarded van der Waals force.
At higher concentrations (> Ix 104 mol/l) repulsive double-layer forces are again observed. The
surfaces now attain a positive charge due to an excess ofDAH+ ions on the surfaces. The surface
charge then increases at higher concentrations and at the same time the surface separation at the
jump into adhesive contact decreases. At concentrations> 6xlO- 3 mol/l no jump into an adhesive
contact was observed and it could be suggested that a bilayer is formed on each surface.

1.0

~ 0.5
~

--
0::

~
0 t-It--t---r~----.,t-+---!

-0.5

234 567
d (nm)

Figure 12. The force between the C12(EO)s coated surfaces at various temperatures for separations in the
region of the minimum (linear scale). Symbols: • 15°C; 0 20°C; A 30°C; A 37°C [24].

In contrast to the hydrophobic interaction in DAH+ monolayers (single chained) a much stronger
hydrophobic interaction force has been detected between DDOA monolayers (double chain) in
water (deposited by Langmuir-Blodgett type depoSition as described earlier in this section), parti-
cularly in cases where the mica surface has been completely neutralised by the adsorbed species.
The area per hydrocarbon chain is about 0.25 nm 2 in the DDOA case and 0.7 nm 2 in the DAHCl
case. This indicates that the hydrophobic force is strongly dependent on the density of hydrocarbon
on the surface, although this is not obvious from the contact angle measurements since a value of
93° was obtained on both surfaces.
It has also been shown that the interactions between mica surfaces coated with a thin layer of a
fluorocarbon based surfactant (adsorbed from solution) give rise to weak hydrophobic forces,
while if thicker layers (2.0 nm) are adsorbed from solution the hydrophobic forces become long-
18

range strong and comparable to what is found for DDOA coated surfaces [25].
The surface force apparatus has also been used to study steric interactions between polymers
such as high molecular weight polyethylene oxide systems which showed long range (80 nm)
hydration or solvation forces. The precise mechanism of hydration and hydrophobic interactions is
not known. In flotation, electrostatic, van der Waals interaction, hydrophobic and hydrodynamic
forces must be considered. Possible ideal conditions occur when both the particle and bubble are
hydrophobic with a relatively low electrostatic repulsive force.

100r-----~10000r=========~

10

.--.. 1
.§
z
E
'-"
0.1
a:
U: •
0.0 • • •••• • •
(')( lO's /'\

-0.1

-0.2
0 20 40 60 80 100
DISTANCE (nm)

Figure 13. The force between dodecylammonium chloride adsorbed at various concentrations from
solution on clean .mica. The curves are theoretically calculated from DLVO theory. The arrows in the figure
indicate the jump into contact adhesion [25].

7. Experimental Studies on Aqueous Thin Films Sandwiched Between an Air

Bubble and a Flat Mineral Substrate

A method used to study this system is a microscopy technique developed by Scheludko et aI. [26]
similar to the apparatus described in section §5. The method allows the kinetic stability of the thin
film to be measured by pressing an air bubble against a mineral substrate in dilute solution of the
flotation surfactants (such as model collectors and depressants). The method was first developed in
Bulgaria, where the air/solution interface in dilute surfactant solutions was brought in close contact
with a polished flat quartz surface. Later studies were carried out in Sweden using a mica substrate
 19

[27]. The mica substrate offers the additional advantage that the state of the surfactant adsorption
layer has been characterised more completely, through previous ESCA and surface force investiga-
tions. Moreover, freshly cleaved mica is a molecularly smooth substrate.
Using this experimental technique (Figure 14), the formation ofthe thin film and the expansion

GAUGE

.
Wk
AFTER CONTACT

,I
!
iGAS !GAS
, R I
I
~ ; I
. . oj ••• " un
I
,
I '
, , ,

a..~FILMjHlN LIQUID
~HR.EPHASE
~ONTACT

Figure 14. Scheludko et al. [26] apparatus for studying the contact between an air bubble.

of the meniscus perimeter (in the case of unstable films) is monitored by means of a microscope
with an attached video tape recorder. The lifetime of the film or induction time (t) is the time for the
film to drain, before rupture and form the solid/liquid/gas interface. Also, the rate of expansion of
the three-phase boundary (dr/dt) (where r is the radius of the circular three-phase contact TPC area
and t is the time) can be measured and in most cases the values of the receding (El r ) and advancing
(El,.) contact angle can also be determined.

7.1. THEORY

To summarise, essentially two kinds of rate processes can be distinguished: (i) the thinning of the
liquid film to its critical thickness which is controlled by hydrodynamic and capillary forces, and
(ii) the formation and expansion of the three-phase contact after film rupture which is controlled by
surface forces.
The receding contact angle can be determined from
20

sin Or =r!R (9)

Thus the stable radius, r, of the wetting perimeter and R, the radius of the capillary tubing, which
can easily be measured, give the contact angle Or against the solid substrate. Slowly decreasing the
capillary pressure by ~p causes the wetting perimeter to move slightly inwards. The decrease in
pressure, ~p, needed to bring about the movement of the perimeter yields information about the
advancing contact angle Oa

. ~p(R2 - r 2) .
sm Oa = 2yr + sm Or (10)

Thus, when ~p = 0 there is no contact angle hysteresis.

7.2. EXPERIMENT WITH INITIALLY CLEAN MICA

In Table 1 results are summarised for studies in dodecylammonium chloride solution (DAC) [27].
In pure water a stable rather thick ~- film was formed, as might be expected from the hydrophilic
nature of the mica surface, with an advancing contact angle of about 7°. At very low DAC concen-
trations, around 0.1 !1Jl1ol/l, the films broke up at some points (or lines) indicating a patchwise
hydrophobisation. The hydrophobicity increased, i.e., additional and more extensive patches
occurred, after each passage of the three-phase boundary across the mica surface.

TABLE 1 Mica in dodecylammonium chloride solution [27]

Conc. of Lifetime tu/l3.t Receding Advancing Comment
surfactant before Perimeter contact contact
CDAC rupture, t expansion angle re angle e.
rate
moll·! s cm s·! <leg <leg
0 00 Stable film
10.7 PaIChwise film rupture
upon moving the
perimeter
10-6 -"-
3xlO·6 0.21 19 52 - "-
10.5 0.9 2.4 70 90 -"-
10-4 5 2.3 -90 -90 No hysterisis, occa-
sionally stable film
10.3 10 3.8 -90 -90 - "-
10.2 00 Stable film

The concentration range between 1 and 10 !1Jl1ol/l is intermediate in behaviour and represents a
gradual transition where adsorption directly from the DAC solution starts to become of importance.
The primary event here would be a condensation of DA + surfactant onto the negative mica surface
sites. This process takes place at about CDAC = 3x 10.6 mol/l, as indicated by the relatively high
contact angles observed of Or = 19° and Oa = 52°. The hydrophobicity of the surface was further
increased at CDAC of about 10 !1Jl1olfl, and the lifetime of the thin film before its rupture was short.
 21

At higher surfactant concentration, > 0.1 mmol/l, the film became more stable, and finally at
10 mmol/l it was infinitely stable.
The stability of the thin liquid film on the negatively charged mica surface, which is obtained in
pure water, can be accounted for by assuming a negative surface charge at the air/solution interface,
giving rise to repulsive electrostatic double-layer forces. It may also be noted that the Lifshitz
theory predicts a negative Hamaker constant for the system mica/water/air, likewise yielding a
positive disjoining pressure. The stability of the thin liquid film on the negatively charged mica
surface, which is obtained in pure water, can be accounted for by assuming a negative surface
charge at the air/solution interface giving rise to repulsive electrostatic double-layer forces.
The patch wise hydrophobisation at concentrations of about 0.1-1 Ilmol/l indicates that
adsorption of DA+ ions takes place on the surface whereby protons are exchanged for dodecyl-
ammonium ions. However, there is strong evidence that an additional, rather unusual, adsorption
mechanism is operative in this concentrations range here, which seems to involve a kind of L.B.
depoSition from the three-phase boundary region.
This idea is largely based on the observation that hydrophobic patches form on the surface which
grow in size each time the wetting perimeter traverses the surface. The main cause of such a
patch wise hydrophobisation, at these very low surfactant concentrations, is probably the formation
of a condensed surfactant monolayer at the three-phase contact line due to the strong electrostatic
attraction between the cationic surfactant and the negatively charged mica surface and, in addition,
the hydrophobic effect which generally promotes aggregation. Accordingly, a kind of L.B.
deposition may be envisaged to occur from a thin precursor film. Note that a rapid transport of
surfactant molecules along the solution/air interface would be essential for this mechanism.
Otherwise, it could hardly compete with adsorption directly from the solution.
It is also of interest to note that in an earlier fundamental study [28] the expansion velocity of the
TPC in a dodecylamine hydrochloride solution on silica surfaces showed three different types of
mobility depending on the solution pH :
(a) at low pH-values (pH ~ 5) the expansion velocity is very high and roughly constant until
equilbrium is reached. The thin film recedes spontaneously after the rupture;
(b) in the neutral and weakly alkaline range (pH = 7.5 - 10.1), dr/dt is initially similar to low pH
values but then it decreases abruptly; and
(c) at high pH values (pH = 11.8), dr/dt is very small and an approximately linear function of
In (t).

8. Experimental Studies on Aqueous Thin Film Between Air Bubble and Mineral
Particles; Induction Time ('t) Measurements

Instead of using a smooth flat surface, several workers have studied induction time by pressing an
air bubble against a bed of particles. However, it must be stressed that "[; determined by these static
test methods gives considerably greater values than the dynamic contact times which occur
following collision between a particle and bubble in a flotation cell. As already discussed, since the
thinning and rupture process contains both hydrodynamic and capillary contributions, then this is
not altogether surprising. Under kinetic conditions, the particles will have a considerable amount of
momentum which will reduce the capillary contributions. Eigeles and Volova [29] initially carried
out systematic studies using a series of different minerals (such as quartz, calcite, barite, galena
etc.) with different sizes and collector concentrations.
The basic apparatus used by these workers has recently been made more sensitive by attachment
22

of a microcomputer (Figure 15).

Eige1es and Vo10va [29] reported that induction times decreased with collector concentration and
decreasing particle size. In fact, it varied from infinity for strongly hydrophilic minerals to about
0.01 to 0.001 secs with collector. In addition,"(; decreased exponentially with increasing tempera-
ture. From this data, a value of the activation energy for the bubble-particle adhesion process was
estimated.

Figure 15. Schematic diagram of the induction time apparatus [30]. 1, CRT; 2, Computer; 3, Interface;
4, Electronic pulse amplifier; 5, electromechanical actuator; 6, microscope.

Measurements of "(; on the quartz-dodecylamine system have been extensively investigated in

recent years. Yoon and Yordan [30] recently reported a minimum at pH 10.5 which corresponded
to a flotation maximum. Also "(; was found to increase linearly with particle size.

9. References

1. Kitchener, I.A., and Cooper, A. (1959) Current concepts in the theory offoaming, Q. Rev.
(London) 13.,71-97.
2. Dippenaar, A. (1982) The destabilisation of froths by solids. I. The mechanism of film
rupture, Int. I. Min. Process. 2,1-14.
3. Dippenaar, A. (1982) The destabilisation offroths by solids. II. The rate determining step,
Int. I. Min. Process. 2, 15-22.
4. Harris, P.I. (1982) Frothing Phenomena and Frothers, in R.P. King (ed.), Principles of
Flotation, S. Afr. Inst. Min. and Metal., p. 237.
5. Iohnasson, G., and Pugh, R.1. (1991) The influence of particle size and hydrophobicity on
the stability of mineralised froths, Int. 1. Min. Process. (in press).
6. Leja, I. (1981) The Surface Chemistry of Flotation, Plenum Press, New York.
7. DeIjaguin, B.V., Churaev, N.V., and Muller, V.M. (1987) Surface Forces, Nauka, Moscow
(Consultants Bureau, New York).
8. Exerowa, D. (I978) Problems of black foam film stability, Commun. Dept. Chern,. Bulg.
Acad. Sci. 11, 739-763.
9. Ross, S., and Morrison I. (1988) Colloid Systems and Interfaces, Wiley, New York., p. 307.
 23

10. Scheludko, A., and Exerowa, D. (1959) Uber den electrostatischen und van der Waalsschen
Druck in Schaumfilmen aus Wiisserigen Elektrolytwsungen, Kolloid Zh.~, 148-151.
11. Van den Boomsgaard, A.S. (1985) Ph.D. Thesis, Dept. of Physical and Colloid Chemistry,
Agricultural University, Wagenigen, The Netherlands.
12. Barneveld, P., Scheutjens, J.M.H.M., and Lyklema, J. (1991) Ionic features offree liquid
films ofnon-ionics, Colloids and Surfaces 52, 107-121.
13. Exerowa, D., Zacharieva, M., Cohen, R., and Platikanov, D. (1979) Dependence of the
equilibrium thickness and double layer potential offoamfilms on the surfactant concentration,
Colloid and Polymer Science ill, 1089-1098.
14. Nir, S., and Vassilieff, e.S. (1988) Van der Waals Interactions in Thin Films, in LB. Ivanov
(ed.), Thin Uquid Films, Marcel Dekker, New York, Chapter 4.
15. Exerowa, D., and Zacharieva, M. (1972) Research in Surface Forces, Moscow, p. 234.
16. Manev, E., and Pugh, RJ. (1991) The diffuse layer electrostatic potential and stability of thin
aqueous films containing a non-ionic surfactant, Langmuir (in press).
17. Scheludko, A. (1967) Thin liquidfilms, Adv. Colloid Interf. ScLl, 391-494.
18. Manev, E., and Pugh, RJ. (1991) Stability of thin liquid films from aqueous solutions of
nonionic frother in the presence of short chain xanthate, Surf. and Interf. Analysis (in press).
19. Kristov, Kh., Kruglyakov, P.M., and Exerowa, D. (1979) Influence of the pressure in the
Plateau - Gibbs borders on the drainage and the foam stability, Colloid Polym. ScL 252,
505-511.
20. Kruglyakov, P.M. (1988) Equilibrium properties of free films and stability of foams and
emulsions in thin liquid films, in LB. Ivanov (ed.), Thin Uquid Films, Marcel Dekker, New
York.
21. Finch, J.A., Yianatos, J., and Dobby G.S. (1989) Column froths, in J. Laskowski (ed.),
Mineral Processing and Extractive Metallurgy Reviews, Gordon and Breach., vol. 5, p. 281-
305.
22. Israelachvili, J.N. (1985) Intermolecular and Surface Forces, New York.
23. Hom, R.G., Clarke, D.R., and Clarkson, M.T. (1988) Direct measurement of forces
between mica surfaces in aqueous electrolytic solution, Materials Res. Soc. 3.(3), 413-416.
24. Claesson, O.M., Kjellander, R, Stenius, P., and Christenson, H.K. (1986) Direct measure-
ment of temperature-dependent interactions between non-ionic surfactant layers, J. Chern.
Soc. Farad. Trans. I, 82, 2735-2746.
25. Herder, e. (1988) Ph.D. Thesis, Dept. Phys. Chern., The Royal Institute of Technology,
Stockholm.
26. Scheludko, A., Tchaliovska, S., and Fabrikant, A. (1970) Contact between a gas bubble and
a solid surface and froth flotation, Spec. Discuss. Farad. Soc. I, 112-117.
27. Tchaliovska, S., Herder, P., Pugh, R, Stenius, P., and Eriksson, J.e. (1990) Studies of the
contact interaction between an air bubble and a mica substrate in dodecylammonium chloride
solution, Langmuir fI, 1535-1543.
28. Schulze, H.J. (1984) Physico-Chemical Elementary Processes in Flotation, Elsevier,
Amsterdam.
29. Eigeles, M.A., and Volova, M.L. (1960) Kinetic investigations of effect of contact time,
temperature and surface conditions on the adhesion of bubbles to mineral surfaces,
Proceedings 5th Int. Min. Process. Congress (IMM, London), pp. 271-284.
30. Yoon, R,H., and Yordon, J.L. (1991) Induction time measurements for quartz-amine
flotation, J. Colloid Interf. ScLlli, 374-383.
Attending the lectures of the Institute ...
1.2. THE MECHANISMS OF COLLECTOR ADSORPTION-ABSTRACTION
(IONIC AND NON-IONIC SURFACTANTS) ON HETEROGENEOUS
SURFACES

J.M. CASES and F. VILLIERAS

LA. 235 "Environnement el Mineralurgie" el CRVM-ENSG
B.P.40
F-54500 Vandoeuvre-Cedex
France

ABSTRACT. The mechanisms of adsorption-abstraction of surfactants on heterogeneous surfaces

can be described by taking into account: I) the imponance oflateral bonds that favour the formation
of aggregates through a cooperative process on surfaces, 2) the intensity of normal adsorbate-
adsorbent bonds that favour either the formation of two-dimensional aggregates in the monolayer
range and at higher concentration the formation of a bilayer, if the interaction with the solid is
strong or if the temperature is lower than the Krafft point, or the formation of micelles in the case
of weak interaction with the solid surface (T > TK), 3) the importance of energetic surface hetero-
geneity which controls the size of two-dimensional aggregates, and 4) the possible association of
ionic surfactants with ions in solution that leads to bulk or surface precipitation (three-dimensional
condensation on substrate). The importance of the selection of a reference phase and the specific
methods to determine the nature of the adsorbed-abstracted phase are discussed.

1. Introduction

The adsorption of surfactants from an aqueous solution is a phenomenon of central importance in

applications ranging from ore flotation, lubrication, paint technology, enhanced oil recovery to
water treatment. It is also a process of such scientific interest and such considerable complexity that
a general tendency to present a new model of surfactant adsorption for each considered application
can be observed in the literature. The manifold uses of surfactants in life and industry are
dependent in part upon their physicochemislry and in part upon the behaviour of the molecule, the
structure of which is responsible for that of the surfactant solutions.
Authors dealing with surfactant adsorption followed two opposite aims in the past: some of
them were looking for selectivity improvement in flotation processes. Thus, they were using
adsorbate-adsorbent systems in which the surfactant has a strong affinity for the surface, i.e. a
strong normal bond generally of electrostatic origin. These systems lead to the formation of a
monolayer for high undersaturation values (concentrations much lower than the saturation
concentration or the CMC). Particles are hydrophobic well before the air-solution interface is
saturated with surfactant molecules. These systems can be called "strong normal adsorbate-
adsorbent bond systems" [I) and lead mainly to two theories: the "hemimicelle" theory [2,3),

25
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 25--55.
© 1992 Kluwer Academic Publishers.
26

which assumes that the filling of the adsorbed layer occurs through the formation of patches due to
strong lateral bonds and limited to one layer, and the "two-dimensional condensation on
heterogeneous surface" theory [1,4-7]. This theory takes into account both potential energies
(normal and lateral) and the entropic terms in the adsorbed layer as well as the surface heterogeneity
to predict the formation of patches, the aggregation number (size of the aggregates) of which are
controlled by the surface heterogeneity, the conditions of bilayer formation and the state of the
adsorbed layer.
Other researchers dealing, for instance, with studies aiming at enhanced oil recovery, used ionic
or non-ionic surfactants with little affinity towards mineral surface in order to keep the detergent
properties of the solutions. Such adsorbate-adsorbent systems can be called "weak normal
adsorbate-adsorbent bond systems". From these works emerges the "pseudophase separation
model" for surfactant adsorption of Harwell et al. [8) that predicts the formation on the surface of
bilayered aggregates called "admicelles" (its very restricted validity will be discussed below) and
the noticeable studies of Levitz [7,9-13) on the adsorption of non-ionic surfactants at the solid-
solution interface. Using the fluorescence decay method developed by others for the determination
of micellar aggregation numbers and based on the exclusive solubilisation of a strongly
hydrophobic fluorescent probe molecule (pyrene) within the hydrophobic region of the condensed
molecular assemblies of the adsorbed layer, Levitz and co-workers found in a broad range of
surface coverage that the adsortJed layer can be viewed as an assembly of aggregates very similar to
regular micelles. Recently, Bohmer and Koopal [14] presented an extension of the self-consistent-
field theory for chain molecules of Scheutjens and Fleer, and used it to study the micellization of
nonionic surfactants and their adsorption from aqueous solution on surfaces.
But what are the mechanisms driving the formation of such different aggregates? Must tre
models insert, for instance, the origin of the normal bond and change with the importance of tre
electrostatic contribution? Can the mechanisms in spite of their complexity be explained with a
reduced number of parameters? Is the introduction of electrical double layer concepts to adsorption
a more convenient method of dealing with the interaction of charged surfactants with hydrophilic
mineral surfaces? We do not think so in a first approach.
In the present work, the principal attention has been focused upon the surfactants themselves in
relation with the general laws of adsorption, taking into account the influence of potential energies
(normal and lateral bonds), surface energetic heterogeneity, and the possibility of surface
precipitation. The necessity, in the absence of a direct means of investigation, to estimate the state
of the adsorbed layer by a judicious choice of a reference phase, the structure of which is well
known, will also be discussed.

2. Physico-chemical properties of ionic surfactants in solution

It is dangerous to treat the adsorption of surfactants without an elementary knowledge of the

amphiphile-water systems, since the structure of the molecule and the conformation of the chain
impose the organization of the adsorbed layer. Amphiphilic substances have a hydrophobic part
(for instance, a hydrocarbon chain) and a hydrophilic part (a polar group, for example, a charged
group). It is generally considered that the rich polymorphism of the so-called lyotropic liquid
crystals is associated with the conformational polymorphism of the paraffinic chains of amphiphilic
molecules. For instance, in the low temperature range, the chains are predominantly all trans:
therefore, they can be regarded as rigid cylindrical rods and they aggregate in lamellar sheet only.
At high temperatures, as they are deformed and disordered by isomeric rotations around their
 27

solubility curve
lyotropic phases < Lamellar phase

hexagonal phase

micellar solution <asymetric micelles

spherical micelles

TOC

Figure 1. Schematic representation of the solubility of ionic surfactants versus temperature: phase
diagram with the Krafft point.

carbon-carbon bonds. the chains can aggregate in various shapes.

For an ionic surfactant. an examination of the solubility curve as a function of the temperature
reveals two large domains (Figure 1). At temperatures below the Krafft point ('fiU. the solubility
curve describes the saturation concentration of the hydrated crystal in equilibrium with the
monomers (singly dispersed surfactants) in solution. In the solid phase. the conformational entropy
of the chain is nil because all the chains are extended in the trans "zig-zag" poSition and are parallel
to one another as in the alkane crystal. The distance between chains is 4.9 A corresponding to a
chain area of 20.5 A2 [IS].
This phase has a lamellar structure with the polar groups lying along the interface with the water.
The arrangement is such that the hydrocarbon chains are perpendicular to the interface and each
sheet is approximately two hydrocarbon chain lengths thick. The length of the alkyl chain in the
extended state can be approximately given by the following relation:

lmax (A) = 1.265n + 0.2

where n is the number of CH3 and CH2 groups. The space between sheets is mainly occupied by
water and counter ions (Figure 2). For an homologous series of surfactants. the free energy of
transfer of a methyl group from solution to the hydrated crystal. ~g'. can be calculated using a
relation similar to Traube's empirical law (the saturation concentration decreases logarithmically
with an increase in the number of carbon atoms). Values higher than 1.4 kT are usually found [16].
The solubility of paraffin chain salts increases enormously above the Krafft point. In sufficiently
dilute solutions. there are only surfactant monomers. As the concentration is increased. these
molecules associate to form micellar aggregates (spherical. ellipSOidal. rod-like particles without
persistence of long range intermicellar order) which still exist in an isotropic solution. At still
higher amphiphile concentrations. lyotropic liquid crystal phases are formed. For simple soaps. for
example. the phase formed first is the normal hexagonal phase. consisting of cylindrical aggregates
packed in an hexagonal array. At very high amphiphile concentrations a lamellar phase can appear
(the neat phase) [17].
The concentration at which spherical micelles appear in solution is referred to as the Critical
Micelle Concentration (CMC). which should not be confused with the concentration of saturation
28

Figure 2. Schematic representalion of the hydrated crysLal state. Molecules A and B are in a kink position
(half crysLal position). One of the reticular planes can be taken into account as a reference phase for
adsorption thermodynamics treatment ifT < TK.

of the hydrated crystal named Cs. The simplest way to understand this rich polymorphism is to
assume that the paraffinic chains are flexible, in order to uniformly fill the irregularly shaped
volume available to them. The conformations of the paraffinic chain can be obtained by rotation
through the positions trans and gauche (g+, g') about each C-C bond. In the "liquid like" state of
the lamellar phase, the mean distance between chains is 6 Aand the chain area is 33 A2 [18], From
X-ray scattering studies the chains appear to be in a chaotic state which might be referred to a
liquid, However, this comparison cannot be drawn too far because the two states are not
analogous: in the liquid state the chain ends, which are a source of disorder, are uniformly
distributed, whereas in the liquid crystal state half of them, the polar heads, are localised at the
interface. Indeed, NMR shows that the two states are drastically different; the orientational order of
the chain segments is never in a liquid crystal lamellar phase as it is in the true liquid.
Curves representing the order parameter [19,20] or the rotation-diffusion profile obtained from
carbon-13 relaxation data [21] as a function of the carbon poSition in the alkyl chain give an
accurate representation of chain flexibility in the parafinic medium. They display a striking feature :
excluding the first (upper value) and the three last segments (lower values) of the chain, all the
others CH2 groups have nearly the same order parameter or the same correlation time if the number
n of methyl groups is higher than 8. For n S 8 correlation times monotonically decrease from the
polar head to the chain extremity and it is impossible with such surfactants to obtain lyotropic liquid
crystals.
The measurements of correlation time also suggest that the results are independent of both
concentration and chain length if 8 < n < 14 . Apparently, the local molecular environment is rather
independent of aggregate shape and/or phase structure, This shows that there is a strong similarity
between the different phases and that the formation of micelles is only the first step in a continuing
association process. It also indicates that the changes in the stability of different phase structures
for a given amphiphile is not primarily caused by changes in local molecular properties of chains,
 29

but rather by interactions associated with aggregates themselves or changes between aggregates and
the aqueous medium where the counterions reside. Above the Krafft point, the aliphatic chains
exhibit two apparently opposed molecular properties: 1) numerous conformations which allow an
important flexibility of the chain and then a spatial expansion, and 2) strong dispersion forces
between chains with a tendency to dense packing.
For an homologous series of surfactants, the free energy of transfer of a methyl group from
solution to a liquid crystal phase, Ag', can be calculated from the values of CMC : values between
1.0 and 1.3 kT [ll are usually found. In conclusion, it is possible to distinguish the different states
of the chain from the chain area and from the free energy of transfer of a methyl group from
solution to a condensed phase :

~g' Chain area

I Hydrated crystal state > 1.4 kT 20.5 A2
I Liquid crystal state 1.0 - 1.3 kT 33 - 35 A2 [1]

3. Adsorption of surfactant molecules on a homogeneous surface

3.1. THE GENERAL EQUILIBRIUM EQUATION

This case will be treated as a theoretical one because homogeneous surfaces are extremely rare in
nature but it is necessary for a good understanding of the adsorption process. The way to obtain a
general equation valid for localised adsorption without dissociation of the adsorbed molecules onto
a homogeneous surface was presented in details in earlier papers [5,6]. The equilibrium adsorption
equation can be presented in the form :

-
kT Log Ce,n = - <Pa - kT Log Wa + kT Log ""I-=-e
e (1)

where :
- Ce,n is the equilibrium molar concentration of the adsorbed layer; k, the Boltzmann constant;
T, the absolute temperature; CPa, the differential work of desorption at fractional surface covera-
ge e (-CPa represents the corresponding potential energy per molecule). This parameter is not
referred to zero, as for the solid/gas interface, but to a reference state taking into account the fact
that during adsorption the adsorbed molecules displace the solvent (in which the adsorbent has
previously been immersed) [6,22]. The expression for CPa is :

(2)

where w is the sum per molecule of lateral interactions in the adsorbed layer if the fractional
surface coverage is e and <p~ is the normal binding interaction between one adsorbed molecule
and a site of the surface (the origin of this bond will not be discussed in this paper, it can be an
electrostatic or a chemical bond with an ionic surfactant or a hydrogen bond with some nonionic
surfactants. Adsorption originates from this interaction).
- k LogWa is the sum per molecule of all entropic terms except the external configurational entropy
for a molecule in the adsorbed state k Log (8/ 1 - 8), this latter parameter being calculated taking
30

into account the Bragg-Williams approximation (mean field approach):

where NA is the Avogadro number, Za is, for an adsorbed molecule, the product of the partition
functions relative to internal vibrations za', rotation za,rot (external rotation and conformation due
to the isomerism of rotation), external vibration of the centre of gravity, za,ext.vib. and Zs is, for
a molecule of the bulk, the product of the molecular partition functions relative to the internal
vibrations zs', rotation zs.rot and translation of the centre of gravity, Zuans. In fact, Zuans is
presented on a reduced form to express Ce in mole per unit of volume [6). If the adsorption does
not alter the internal vibrations. then. za' = Zs' and ifT<TK • the entropy of conformation is nil
because the chains are elongated.
If it is assumed that w = O. then qJa is independent of the surface coverage, the curve
corresponds to an ideal localised monolayer and eqn. 1 can be rearranged to give the familiar
Langmuir isotherm and. for low surface coverage. the Henry law or the Stem equation [5,6). This
possibility has to be ruled out for surfactants because in the condensed state the lateral bonds are
never nil.
If w < 4 kT, then eqn. 1 gives the Frumkin-Fowler isotherm, valid below the critical value of
lateral bonds, also called the Temkin isotherm.This possibility can be applied for surfactants if the
number n of carbon atoms in the alkyl chain is < g [4).
Finally, if w > 4 kT, then eqn. 1 allows the well known Frumkin-Fowler isotherm to be
obtained, which describes the two-dimensional condensation on a homogeneous surface.
With surfactants, it is very difficult to know the entropic terms in the adsorbed layer. Therefore,
it is very useful to compare the organization (the state) of the adsorbed layer with that of a reference
phase, for instance if T < TK, a reticular plane of the hydrated crystal phase or, if T > TK, a plane
of the lamellar lyotropic phase. For instance, ifT < TK, the equilibrium equation of the hydrated
crystal with the bulk can be written:

kT Log Cs.n =- qJo - kT Log Wo (3)

where Cs.n is the saturation concentration for a surfactant with n atoms of carbon in the alkyl chain,
k.Log W0 is the sum of all entropic terms for a molecule in the kink position (molecule A or B in
Figure 2) and -qJo is the corresponding molecular potential energy. (jJo can be written in the form:

(4)

where w' is the molecular lateral interaction, qJ ~ is half of the normal work of desorption of a
molecule in the half-crystal position, determined from the following equation:

(5)

where <l>°A represents the normal work required to dissolve the molecule A (i.e. to break a tail-tail
bond and a head-head bond through water and counter ions) and <I> Bthe normal work to dissolve
the molecule B (i.e to break a head-head bond through water and counter ions).
Comparing the state of the adsorbed layer with that of the reference phase is possible: 1) by sub-
 31

stracting eqn. 3 from eqn. I, which is always possible, and 2) making the assumption that in the
adsorbed layer the number of coordination and the lateral interactions are the same as in the reticular
plane of the reference phase (the surface field does not influence the organization of the layer and
the properties of alkyl chains are the same in the adsorbed layer as those in the reference phase),
such that:

00'=00
and
Log Wo = Log Wa

With these hypotheses, it is possible to obtain the reduced Cases-Mutaftshiev's equation used for
thermodynamic eqUilibrium between an adsorbed layer and a dilute solution of the same surfactant
that may be written as :

kT Log Ce,n - kT Log Cs.n =i\fl =(cp ~ - qfal + I (1- 26) + kT Log I ~ e (6)

where i\fl is the undersaturation, i.e the difference in chemical potential between the surfactant
molecule in the adsorbed layer or at equilibrium in the bulk solution and the same molecule in the
reference phase or in the saturated solution.

1,0 ---------------N--=::--;;:---=---:::,.......- ,,~

.' I
//,,--'"

I f
\
0,5
'--

4 ...
4 ... '

Figure 3. (a) (left) Schematic Frumkin-Fowler isotherms on a homogeneous surface above (curve I)
and below (curve II) the critical value of lateral bonds. (b) (right) Variation of the differential work of
desorption versus the fractional surface coverage for the Frumkin-Fowler isotherm (curve I). The portion
between eM and eN corresponds to the domain described by the first order change (kink position of the
condensed adsorbed layer).

e
The plot of vs i\fl for two cases, viz. 00 < 4 kT (Temkin isotherm) (i.e.for instance n ~ 8 [I])
and 00 > 4 kT is given in Figure 3a. The isotherm presents a point of inflexion for = 0.5. At this e
point, the slope is negative if 00 > 4 kT and the isotherm presents the mathematical shape of the
Frumkin-Fowler isotherm (curve I) with its real (continuous line) and theoretical (dashed line)
parts. The step characterising the two-dimensional condensation, i.e the first order change, is
located at a panicular value of i\fl named i\fl· and is defined as :

(7)
32

The dotted portions of the curve between M and N represent unstable phases in relation to Ire
condensation of the layer. On the left of M, the layer is in a dilute state and on the right of N, in a
condensed one. On this portion, Le. on the step of the two-dimensional condensation, eqn. 2
becomes (Figure 3b) :

(2')

and the molecular potential energy is again independent of surface coverage. On the other hand,
taking into account lateral interactions and writing a function of state relative to adsorption
independent of surface coverage is equivalent to implicitly assuming the two-dimensional
condensation on the homogeneous surface.

A B c

Figure 4. Schematic adsorption isotherms of a given surfactant on homogeneous surfaces of three

different solids.

3.2. THE PHYSICAL SIGNIFICANCE OF THE 8 vs ilIA- PLOT

Equation 7 predicts:
1) tltat as {j)"o is a constant for an homologous series of surfactants, the higher (j)~ is, the greater
the undersaturation is (low eqUilibrium concentrations in regard to the concentration of
saturation of the hydrated crystal taken as the reference phase). In Figure 4, schematic
adsorption iso-therms of a given surfactant on homogeneous surfaces of three different solids
are presented, i.e. (j) ~ is different for each of them;
2) the superimposition for an homologous series of surfactants taken in the same range in regard
to the Krafft point, of all the isotherms presented in the 8 vs ilIA- plot;
3) the non superimposition of the isotherms obtained with an homologous series of surfactants if
"* "*
the state of the adsorbed layer is different to the reference one (ro' ro and LogWo LogW..)
or if the normal bond is dependent on the concentration of counterions Le. the chain length.
This case was studied during the investigation of systems such as sodium alkylsulfonate-
natural kaolinite: at low concentrations surfactants adsorb through calcium ions and through
sodium ions at high concentrations, due to the cationic exchanges when increasing surfactant
 33

concentmtion [23.24]. The exchange phenomena is displaced with the chain length.
1fT> TK. isothenns also superimpose in the e vs ~Il' plot where ~IL' is the difference in chemi-
cal potential between the surfactant molecule in the bulk in equilibrium with the adsorbed layer and
the same surfactant molecule at the critical micelle concentration. This is possible because the
confonnational tenns of the alkyl chains are the same in a plane of the lamellar lyotropic phase and
in spherical micelles. This is possible if for each tenn of an homologous series of surfactant. the
following equality can be written:

CoLn -A (8)
CMCn -

where CoLn is the satumted concentration of the lamellar lyotropic phase. Under this condition.
eqn. 7 becomes :

A
ull *. = kT Log CMC
C en = Cjlo+
0 A -<jla=
0 B -Cjla0 (7')
n

where B is a constant for an homologous series of surfactants.

The position of the step characterising the two-dimensional condensation on the X-axis depends
only on the value of the nonnal bond.

3.3. THE LATERAL INTERACI10NS

The combination of eq. 1 and 2' taking into account:

1) the rough approximation that w can be considered as a linear function of the chain length. n.
i.e. such that:

w =n ~WCH2 - Wo (9)

where ~WCH2 represents the attractive contribution of one CH2 group in the lateral interactions
and Wo the repulsive contributions due to the heads or the polar groups;
2) the fact that the step characterising the two-dimensional condensation is located for a particular
value oce equal to 0.5,
allows the following fonnula to be obtained:

* n ~WCH2 0 Wo
kTLogC en=- 2 -kTLogWa+ (-Cjla+ 2 ) (10)

On melting crystals of compounds containing n-alkane chains. the total entropy change will
include increases in confonnational. positional and orientational disorder. Of these. the confonna-
tional tenn alone was found proportional to the number (n) of carbon atoms in the hydrocarbon
chain (where 11 :0; n :0; 20) [25]. It is then possible to rewrite eqn. 10 in the alternative fonn :

kTLogC* e,n = -n ~ + E' (11)

where ~g represents the Gibbs energy change of transfer of one CH 2 group from the bulk to the
34

adsorbed layer and E' a constant including all the entropic terms not related linearly to n, CPa and
roo·
Equations 10 and 11 predict that for an homologous series of surfactants the isotherms shift
toward low equilibrium concentrations as the number of carbon atoms in the aliphatic chain
increases. Figure 5 shows schematically such a variation. It can be observed that, at lower
concentrations, eM is still nearer to zero and eN is still nearer to unity, thus justifying the usual
sentence on the 0-1 approximation of the two-dimensional condensation.

1 -------- ,.-------r
C14 C12 / C1D
(

0.5 - - - - - - - ---- --- ~,

IA) \:(8)

/
/
0"ltr:;o.-r--+--tr~5T-r-
2-"'T'M'---b·,_C. ImoLl")
C.,14 C•.12 ~..to

Figure 5. Schematic representation of the adsorption isotherms of an homologous series of surfactants

with 10, 12, 14 carbon atoms in the aliphatic chain onto a homogeneous surface.

3.4. ON THE POSSIBILITIES OF MULTILA YERS

Finally, it is possible to discuss the possibility to obtain surfactant mUltilayers in the domain of
validity of the Frumkin-Fowler isotherm, i.e. if ~Il or ~Il' < O. Taking into account only potential
energies, it is possible to rewrite eqn. 7 and by substituting the value of -cPg given by eqn. 5, we
obtain the value of the undersaturation of the first layer:

(5')

For the second layer, we must have

so that

cf>~-cf>~
2

as cf>~ > cf>~, t.1l~ is negative and the second layer can be formed in the undersaturation region. For
the third layer, it is possible to write CP~.rrI = cf>~, then ~Iltn > O. The third layer can form only at
 3S

supersaturation, i.e in the domain ofthe three-dimensional condensation (precipitation ifT < TK,
fonnation oflamellar lyotropic phase if T > TJ().
It is possible to conclude that in the adsorption domain, the adsorption of surfactants is restricted
to the bilayer.

4. Adsorption of surfactant molecules on heterogeneous surface

It is possible to apply the previous considerations to heterogeneous surfaces if they are viewed as a
combination of homogeneous ones. Each homogeneous domain, i, is characterised by the value of
its area Sj and the value of the nonnal molecular potential energy -qJ ~;. A homogeneous domain
mayor may not be fonned of several separate domains. The fractional surface coverage, e, can be
expressed as in the case of the 0-1 approximation of the two-dimensional condensation (a
homogeneous domain i is empty or full and in this latter case a patch of a condensed layer is
fonned on the surface) by the relation:

e = r (Sj/S) (12)

All the equations obtained for a homogeneous surface can be used for each homogeneous domain.
In particular, eqns. 7 or 7' can be rewritten as :

(7")
or
(7"')

Based on this concept, surfactant isothenns then become discontinuous stepwise curves with
those patches of the heterogeneous surface, which are most energetic undergoing phase transition
for the smaller values of the solution concentration, followed by successive phase transitions on
less energetic patches. As the area Sj of a given domain is smaller, the more energetic the
considered i-th domain is. The heights of the different steps corresponding to the filling of the
different homogeneous domains are higher as the eqUilibrium concentrations increase (Figure 6,
curve I, for instance). For this last reason, the isothenns on heterogeneous surfaces always begin
parallel to the X-axis.
e
The shape of an isothenn in the plot vs t1!-t or t1!-t' or Log Ce is independent of the lateral
bonds (n> 8). It depends, however, on a double distribution:
- on the e axis, taking into account eqn. 12, it depends on the Sj area distribution. The height of a
vertical step, corresponding to the two-dimensional condensation on a domain i, depends on the
SJS ratio.
- on the t1!-t or t1!-t' or Log Ce axis, taking into account eqns. 7" or 7''' or for instance a modified
eqn. 10' in order to be applicable onto heterogeneous surface, it depends on the qJ~i distribution
corresponding to the different homogeneous domains.
On heterogeneous surfaces, the experimental isothenns present no precise shape. The latter
mainly depends on the energetic heterogeneity of the real solid surface.
The discrete steps mayor may not be observable, depending on the distribution of the surfaces
Si and the energy of patches. If the surface of the mineral is very heterogeneous, the size of the
different homogeneous domains is small and if the nonnal molecular potential energies are
36

e
2,

1,0

l"/-------

Cl0 I
-Y/
10'3 " Ce (moleXI)

Figure 6. Schematic representation of adsorption isotherms on heterogeneous surfaces,

• Cl0 H21 NHJ Cl +

2
• C12 H7s NHJ Cl +

+ C14 H29 t+lJ Cl

• C16 Hn NHJ Cl
o C18 HJ7 NHJ Cl
1.

Figure 7. Adsorption isotherms of alkylammonium chlorides of different chain lengths on biotite at

pH 5.5 and 25°C [41.
 37

distributed continuously. the isothenn appears as a continuous cUlVe. In this case. the slope of the
isothenns is independent of the number of carbon atoms in the chain and each point of the isotherm
represents a homogeneous domain i filled with one or numerous condensed patches. where it is
possible to apply the thermodynamics of infinite adsorbed phases - if this point is 0.01 in e
extension it corresponds to 2.9xlO 16 molecules. the layer being in the liquid crystal state: this is
quite large compared to the aggregation number of micelles. Another factor can lead to smoothing
of the cUlVes [4] : the determination of experimental isotherms at the solid-solution interface
imposes for convenience. to change the sample for each run. As all the samples are not strictly
identical from the energetic adsorption sites distribution point of view. the statistical distribution of
experimental results around mean values leads to a smoothing of the cUlVe. A good example is
given in Figure 7 : the isothenn of tetradecylammonium ions was constructed with the same sample
and the others with different samples of adsorbents. Only this isotherm presents the micro steps
characterising the two-dimensional condensation on heterogeneous surface.
However. the best way to characterize the two-dimensional condensation without ambiguity is to
consider the second layer. If the surface field does not influence the formation of the bilayer and if
this formation through tail-tail bonds is possible. the second layer is always characterised by a
vertical step on the isotherm (Figures 6 and 7). because adsorption now occurs on a surface ener-
getically homogeneous. due to the first layer fonnation.

4. 1. A METHOD OF EV ALVA TING THE ADSORPTION ENERGY DISTRIBUTION

In the case of two-dimensional condensation. a function fj can be defined as the ratio of site fraction
by energy unit [5] :

dO
fj = dl\fJ. (13)

that allows the determination of the energy distribution function. The ionic surfactants can be used
as tracers of surface heterogeneity [4] to determine. for instance. the aspect ratio of phyllosilicates
[26].

4.2. THE e vs ~I! PLOT

In the plot e vs ~~ or ~~. as predicted by eqns. 7" or 7'''. if the isotherms of an homologous
series of surfactants superimpose. it means that the state of the adsorbed layer is the same as in the
reference phase chosen. This was observed in numerous systems as for instance alkylammonium
chloride-biotite systems (Figure 8) (liquid crystal state). sodium salts of fatty acid-magnesite
systems [6] etc. (see also for other examples ref. [6])

4.3. THE EXPERIMENTAL DETERMINATION OF ~g

It is still possible to calculate from the experimental isotherms the Gibbs energy change of transfer
of one CH2 group from the solution to the adsorbed layer. Let us assume that isotherms are
constructed with an homologous series of surfactants (Figure 6) with 10. 12 and 14 atoms of
carbon in the aliphatic chain onto a heterogeneous surface consisting of four different homoge-
neous domains. Any straight line parallel to the concentration axis intersects on each isotherm the
same homogeneous domain i. As a result. for a given value of the fractional surface coverage.
38

I)
+
2.0 • CIOH2I NH 3'CI-
• CI2 H2S NH 3'CI-
1.8 • CI,H29NH 3'CI- ."
a CI6H33NH3 'CI-
+ CI8H37NH3 'CI-
1
+
..
1.6

' ..
II,

I.'

Figure 8. Superimposition of the adsorplion isotherms of alkylammonium chlorides of different chain

lengths on biotite at pH 5.5 and 25°C [4].

InC.

10

-14

Figure 9. Relations between equilibrium concentrations and the number of atoms of carbon in the alkyl
chain at constant fractional surface coverage: the measurement of ~g.
 39

eqn. 11 can be rewritten:

(kT Log C* e,~a = - n ¥- + E'a (11 ')

If there is the same condensed state on the different homogeneous domains (the surface field does
not influence the state of the layer and ~g = ~g'),the slopes of the linear relations in the plot
(Log Ce VS n) must be equal, whatever the values of the fractional surface coverage, although the
origins of the lines will be different (Figure 9). The intersect at the origin increases with e because
EOa includes -cpoa,j which decreases if e increases, the different homogeneous domains being filled
from the most to the less energetic.
Such experimental results characterize without ambiguity the two-dimensional condensation on
heterogeneous surfaces and were found, for instance, with adsorption of alkylammonium chlorides
on biotite [5]. For very low surface coverage, eqn. 11 ° cannot be verified: this is indicative either
of the adsorption of surfactants as individual ions or of the formation of patches on the most
energetic sites with small aggregation numbers such that the thermodynamics of infinite adsorbed
layer cannot be used. The values for ~g usually found are near 1.15 kT, ifT > TK (liquid-crystal
state) and > 1.4 kT, if T < TK.

4.4. ON THE INFLUENCE OF THE NORMAL BOND ON THE NATURE OF AGGREGAnON

The adsorption of a surfactant with 10 atoms of carbon in the chain on different adsorbents A,
B,C, could respectively lead to the isotherms III, IV and V presented in Figure 6. The affinity of
the surfactant for the different solids may be classified following eqn. 10 used for heterogeneous
surfaces, n being constant:
- Strong afrmity for solid A - undersaturation corresponding to the end of the monolayer, < -3 kT.
- Medium affinity for solid B - undersaturation corresponding to the end of the monolayer,
-3 kT < ~g < -2 kT.
- Weak affinity for solid C - the isotherms often present a typical sigmoidal shape with a rising
part always located below but near the CMC, followed by a plateau region which is reached as
the CMC is approached.
In the case of strong or medium interactions between the ionic surfactants and the solid, there is
now no doubt that the two-dimensional condensation on heterogeneous surface theory is a fruitful
approach. However, from different recent works (Levitz and coworkers using nonionic
surfactants, Chandar et ai. [27], Partyka et ai. [28], Denoyel et ai. [29]) studying the anionic surfa-
ctants-alumina system) the following question arises: in the case of weak interactions, does the
growth of the adsorbed layer always proceed by the formation of patches and perhaps admicelles or
by the formation of micelles on the surface below the CMC ? Such an example of weak interaction
of surfactants for hydrophilic surfaces is presented: the nonionic surfactant-silica systems.

s. Systems with weak interactions : Adsorption of non ionic surfactants

Reversible adsorption of nonionc surfactants, such as alkyl (CnENp) or alkyl phenol polyethylene
glycol (CnPENp) occurs onto hydrophilic solids just below the CMC. The monomers adsorb
through hydrogen bonds between the polyoxyethylene chain and hydroxyl groups of the surface.
Then, the maximum of adsorption occurs often around the zero point of charge. In these systems,
40

.
r(lLmole.m- Z )
TIE a [Np.6]

x - _ - - - - T B E 8 [Np.t]

.-_---TXlOO [Np.9.S]

,...-_ _---._TX IOZ [Np.12.S]

C(wr'mole.l-1)

Figure 10. Experimental adsorption isotherms C8PENp surfactants from aqueous solution on silica.

the notion of a Krafft point is aimless. The three-dimensional aggregates are always in the liquid-
crystal state. As shown in Figure 10 extracted from the work of Levitz, each isotherm reaches a
plateau around the CMC which is typical of a weak. binding isotherm. The plateau stems from the
stabilization of the monomeric chemical potential slightly above the CMC. In the amount adsorbed
vs C e plot, this plateau decreases when the polar chain length increases. At this point, it is worth
noting that for a very short polar chain (CgPE6), a step isotherm can be observed, clearly remini-
scent of a two-dimensional condensation process, whereas for long chains the adsorption
isotherms appear to have a "Langmuir" shape but without a clear inflexion point_
The first important question to clear up about the structure of the adsorbed layer concerns the
possible occurrence of molecular associations. Several experimental arguments support the
existence of a very fragmented aggregative adsorption process [13] :
- when expressed in the plot (fractional surface coverage vs undersaturation or relative concen-
tration C/CMC), the isotherms of an homologous series (same polar head but slightly different
apolar chains) superimpose. This property, observed for the adsorption of CSPElO and C9PElO
on different types of polar solids (silica [30], quartz, kaolinite [31]), strongly suggests, as
described above, a local organization of adsorbed molecules very similar to that of the micelles in
solution ( the liquid-crystal state).
- above a fractional surface coverage of 0.1-0.3, the differential enthalpy of adsorption of
nonionic surfactants on hydrophilic silica gels shows an endothermic process, the amplitude of
which is very similar to the one observed for bulk micellisation [31,32].
- fluorescence spectroscopy all along the adsorption isotherm of a build-in chromophore or an
extrinsic probe (pyrene), gives clear evidence that adsorbed molecules are involved in finite
aggregates [9,10,30]. The average aggregation number in the adsorption layer below the CMC
and above a surface coverage of 0.2 turns out to be in the same range as the micellar size in
solution measured above the CMC. In fact, the results of Levitz and co-workers can be
summarised as follows:
a) the adsorbed phase of nonionic surfactants with short chains, like CgPE6, is always made of
 41

very large aggregates. In fact, their size is too large (the pyrene probe cannot "see" an
aggregate larger than two or three hundred molecules per aggregate) to be measured. As a
result, it is impossible to distinguish between a two-dimensional process and micelle
formation on the surface.
b) for long polar chains, there is a very interesting similitude in this case between the size of the
surface aggregates at the plateau of the isotherms and the size of the micelles in aqueous
medium above the CMC. For micelles, the aggregation number N)."Jc decreases if the number
of ethoxy groups Np increases, according to the relation [11]:

Log ~c = -1.1 Log Np + 7.264 (14)

e
c) an intermediate behaviour is displayed by C gPE9.5. From = 0.05 to 0.17, the number of
aggregates per unit area is constant but their size increases up to the size of micelles in
e
micellar solutions. It is a growth regime around some points of nucleation. From = 0.17 to
0.5, the size of the aggregates (N ::,~c # 160) does not increase anymore but their number
density increases. This is clearly a self-repeating regime. From 0.5 to 0.8, the number
density of aggregates is again stationary but the aggregation number increases : this is a
growing regime around the existing aggregates. In order to understand the transition from a
fragmented phase to a continuous one around e = 0.8, Levitz and co-worleers assumed that
the surface aggregates are oblate ellipsoids, like regular micelles. Increasing further the
e
amount of adsorbed molecules beyond = 0.8 implies at least some coalescence of the
surface aggregates (N)."Jc »200).
Finally, the adsorption of C12E6 on a flat Si02-water interface was investigated by neutron
reflexion [33]. From low to high surface coverage, the local and semi local structure of the
adsorbed layer appears to be a "fragmented" bilayer.
The point resulting from these investigations is that the adsorption process occurs mainly below
but near the CMC, i.e. in the range of weak normal interactions with hydrophilic solids. It involves
a surface aggregative process similar in some respect to that of the bulk micellisation taking place
above the CMC [13,14] and was formerly proposed by Klimenko and Kaganovskii [34].
With these results, the conclusions of the works presented by Chandar et al. [27) and
Somasundaran et al. [35) seem to be questionable.They used pyrene and dinaphthylpropane
fluorescence (to measure the microviscosity of the surface aggregates) probes to investigate the
structure of the adsorbed layer of sodium dodccyl sulphate at the alumina-water interface. The main
results of this very important worle conducted with ionic surfactants may be summarised as
follows:
- the adsorption takes place below but near the CMC, in the range of weak normal adsorbate-
adsorbent bonds;
- at low concentrations (Region I; C/CMC < 0.03; fractional surface coverage calculated from the
value of the plateau < 0.006), adsorption is likely to be due to an electrostatic attraction of
individual surfactant ions. Indeed, the aggregates are too small to accommodate a probe
molecule. This fact can be considered as a limitation ofFDS for the study of adsorption. It is the
reason why Levitz and co-worleer could not measure the size of the surface aggregates for low
surface coverages when using very hydrophilic surfactants;
- in region II (C/CMC < 0.1; fractional surface coverage < 0.1), the adsorbed layer is filled with
aggregates, the aggregation number of which increases from 50 to 170;
- in region III (C/CMC < 1; fractional surface coverage::; 1), adsorption corresponds to an
increase of the aggregation number up to 360;
42

- region IV corresponds to the plateau. The onset of the plateau coincides with the CMC as a result
of the relative surfactant monomer activity in the bulk beyond this value of concentration. The
maximun adsorption density value corresponds approximately to a closely packed monolayer of
25 A that means either that the layer (here necessarily in the hydrated-crystal state) is not
completely filled or that the aggregates are not hemimicelles as assumed by the authors.
They concluded that two-dimensional patches called hemimicelles are present on the surface in
region III because the environment formed by the adsorbed surfactants is viscous (100 cp, 12.5
times as big as in SDS micelles) but one of their results show clearly that in region III and IV the
fluorescence parameter of pyrene in the adsorbed layer is close to that obtained in a SDS micelle
(O.IM NaCl). This last result is in good agreement with the works of Denoyel et al. [29] on the
sodium octylbenzene sulphonate-alumina systems and of Partyka et al. [28] with the n-alkyl-
benzene sulphonate - alumina system. Using flow adsorption microcalorimetry and/or batch
adsorption microcalorimetry, these authors concluded that at low surface coverage two-dimensional
aggregates are formed on highly energetic but heterogeneous surfaces, corresponding to a sharp
decrease up to an apparent surface coverage # 0.07 (region II noted above), followed at higher
concentrations (regions III and IV) by the formation of micelle-like aggregates, because the
enthalpy of displacement becomes of the same order as the micellization enthalpy.
These results seem to explain why the aggregates revealed by Chandar and co-workers are, in
the domain of weak normal interactions, very fragmented and of such limited size. In this range, if
two-dimensional patches exist, corresponding to the formation of a hydrophobic surface, their
aggregation number might be high due to the solid surface formed by domains of weak energy,
i.e. of large surface area. A way to clear this point in this high range of concentrations would be to
be able to measure the evolution of the hydrophobicity (presence ofhemimicelles) or hydrophilicity
(presence of micelles) of the solid surface in range III and IV (by shaking the suspension with
cyclohexane).
In conclusion, it is now possible to answer the question: what are the conditions to obtain two-
dimensional patches or micelles on the surface and what are the corresponding driving forces? The
origins of the molecule association to form two-dimensional aggregates are :
- the importance of the normal bond qJ~ that forces the surfactant to adsorb on the solid with the
polar group of the amphiphile molecule fixed at the water-solid interface and the hydrocarbon
chain projected outward. The surfactant polar head affinity is greater for the surface than for
water. Thus, the surface becomes hydrophobic as the fractional surface coverage increases up to
the monolayer.
- the importance of lateral attractive bonds which provoke the two-dimensional condensation and
the formation of patches.
- the surface heterogeneity which controls the size of the aggregates. The aggregation number is
so great that it is always possible (except at very low surface coverage where the corresponding
very highly energetic domains are of limited size) to apply the thermodynamics of adsorbed
infinite phases.
The formation of micelles on the surface is only possible if a weak interaction between
monomers with a hydrophilic surface can generate a set of surface aggregates at eqUilibrium
concentrations below but close to the CMC. In these systems, the affinity of the polar group for the
surface is of the same order of magnitude as for water, such that the adsorbed molecules tend to
aggregate in three dimensional aggregates, very fragmented, called micelles: 1) due certainly to a
cooperative process involving strong lateral bonds, 2) but also in order to protect the aliphatic
chains from water leaving the polar portion of the molecules to associate with the water molecules
which surround them. The surface always stays hydrophilic and the thermodynamic model
 43

proposed must take into account more particularly the thermodynamics of small systems.

6. The three-dimensional condensation on substrate or "surface precipitation"

process

The flotation of sparingly-soluble salt-type minerals such as apatite, calcite, magnesite, baryte,
fluorite and sulphide minerals appears to be controlled by chemical interactions between anionic
collectors (fatty acid, sulphonate, xanthate collectors, chelating reagents [36]) with mineral cations,
and collector interaction appears to be controlled by solubility criteria. Under these conditions, the
thermodynamic model to be used must follow the well known thermodynamics of infinite three-
dimensional phases (the equilibrium concentration, being usually lower than the Krafft point, is
given by equations such as eqn. 3) and more precisely that relative to the three-dimensional
condensation on substrate: the surfactant being allowed to precipitate on the surface and/or in the
bulle In these systems, there are indications that collector uptake may go far beyond the monolayer
capacity [37,38] and the "abstraction" or the "retention" of surfactant by the surface should be
considered as surface precipitation rather than adsorption [39,40]. Special precautions must be
taken in order to interpret experimental results, for instance:
- the fractional surface coverage looses its physical meaning and it is possible therefore to use
different terms to talk on it, like "apparent" or "statistical surface coverage". An example of this
assertion can be found in a recent experimental work [38]: the adsorption isotherm of
amylxanthate on prcoxidised galena at 25°C, where an uneven distribution of the layer thickness
was found (see below §6.2).
- the only way to control the distribution of the molecules on the surface in the monolayer capacity
range, in orderto detect a real monolayer, would be to follow the corresponding evolution of the
hydrophilicity!hydrophobicity of the surface.
- the necessity to explain why in these systems the shape of the isotherm is dependent on the
solid/liquid ratio and on the total surface of the adsorbent and why the three-dimensional
condensation is not always detected on experimental isotherms with the infinite step
characteristic of this process [37,41 J.
For a better understanding, three different systems will be studied.

6.1. THE FLUORITE-SODIUM OLEATE SYSTEM

In this system for instance, the dissolution of the mineral puts fluoride ions in solution (ions that
can compete with the surfactant molecules), likewise, calcium ions can precipitate the surfactant if
the solubility of the calcium soap is exceeded, which can be the case when oleate is used as the
collector if (Ca2+) (On2 > Ksp. The values of the solubility product differ according to research
workers coarsely, for calcium oleate from 15.6 (42), 15.4 [43], 14.5 [44] to 12.4 (45), due to
impurities and pH. These discrepancies make it difficult to determine the mechanisms of
adsorption/abstraction in the lower part of the concentration range.
Sivamohan et at. [46] recently studied the adsorption of oleate onto natural fluorite at pH 9.45
(Figures II and 12). Diffuse reflectance associated with the FTIR spectroscopy, zeta potential and
turbidity have been used for the identification of oleate species adsorbed on the surface. In these
runs, the solid ratio was 7.5% (= 3.75g/50ml) and the specific surface area of fluorite was
1.14 m2/g. The total surface available for adsorption in these batch tests was 4.28 m2 and the total
amount of calcium ions in the bulk around 8x 10-6 mole or 1.9x 10-6 moVm 2 available for surface
44

Js II I
I
I cleo
I I
I I
l/ I

il
I I
I I

!
CD I®
I
® @
g,
li:;
I o>
I <.J
0 !
-------------- - - -- 2
v I
~
ra

-d
5 't:
iil

••-c- ~ HV 1

5
f+
~VI .
I

0 s ,-s
I I
10
-, 10
-, 10-
o
residual concentration(MX1)

Figure 11. Adsorption isochrone of oleate on fluorite at 25°C and pH 9.45.

L: I
I
I
I
I
I
c!.c
:;: I I I
E

!I~
I
I =>
! CD ® ~ @
l-
I!..
$- 70 ~oo
8.
.
....ra -60 ~
N
50

40 >'] .....
J 500

400
,....f.' II
f-
30 300

20 200

V""
.
10 100
'/ I
I

0
10's 10
., I .,
10
Ill!" .,
10
0
residual concentration (M.r1)

Figure 12. Zeta potential and turbidity as a function of residual concentration for pH 9.45 and
temperature 25°C.
 45

precipitation if the amount of calcium is artificially related to the surface.

In these runs, the pK used was 12.4, because that value was previously obtained by Jacquier
[45] with sodium oleate of the same origin. The onset of the precipitation of calcium oleate is
marked on the isochrone by an arrow noted C s (saturated concentration of calcium oleate). In
region I, below C s, the monotonic increase of surface coverage with residual concentration was
explained by the theory of the two-dimensional condensation on a heterogeneous surface, the
adsorbed layer being in a hydrated crystal state. The study of the corresponding diffuse reflectance
spectra suggested that either ionised carboxyl groups form chemical bonds with the calcium sites
on the surface after exchange with the hydroxyl groups (the oleate molecule is monocoordinated on
calcium sites) and/or the carboxyl groups adsorb on the surface through sodium counter ions. In
fact, the constant increase in turbidity against concentration also suggests the formation of bilayered
patches. There is no visible discontinuity characterising a change from a hydrophilic surface to a
hydrophobic one around the monolayer capacity.
In region II, above C s, surface precipitation up to the bilayer corresponds to a sharp increase of
the amount adsorbed and a sharp decrease in zeta potential. The diffuse reflectance spectra
(DRIFfS) shows clearly that the bilayer is filled with calcium oleate and sodium oleate in the
hydrated crystal state.
In region III surface coverage remains constant. Turbidity and zeta potential reach a plateau due
to the bilayer formation and the corresponding enhancement of the stability of the solid suspension
(the solid is hydrophilic). The nature of the adsorbed layer is the same as in region II.
At higher concentrations (region IV), surface coverage increases but zeta potential and turbidity
remain constant. That increase is an artifact due to spherical micelles formation at a value of7xlO- 4
mol/! (from recent NMR experiments [47] it can be claimed that the CMC value indicated in Figure
11 (1.67xlO-3 mole per liter as determined by the measurement of conductance) corresponds in fact
to the formation of oblates micelles while spherical micelles form at 7xlO-4 mol/!), these phases
being centrifugated during adsolption experiments and wrongly considered as adsorbed phases.
This formation is accompanied by a sharp decrease of the absorbance of the bands related to the
hydrocarbon chain, due to a change in the state of the adsorbed layer; the decrease in frequency of
the out-of-phase stretches of the CH2 groups from 2929-2926 to 2923 cm- 1 (Kellar et ai. [48]
suggest that the change from the hydrated to a liquid crystal state is accompanied by the decrease of
the frequency of this band from 2925 to 2919 cm -1) and the vanishing of all the bands, which are
characteristic in the carboxylate groups region of calcium oleate: the region of the spectra related to
these groups presents only a narrow band at 1561 cm- 1 which is characteristic of sodium oleate in
a solid state and in a disordered environment [49J.
The enrichment of the adsorbed phase in sodium oleate seems to be due to the release of calcium
ions from the external part of the bilayer and their adsorption on highly specific surface area
micelles. In this example, surface precipitation is limited to the bilayer because the total calcium
amount in the bulk is lower than the amount required to form a stoichiometric calcium oleate layer
on the external part of the bilayer (the capacity of this layer is 8.1xlO-6 mol/m2 oleate ions, so that
only half of the layer may be filled with calcium oleate.
With the same system, other research workers [42,44,50,51] found maximum isotherms. The
three-dimensional condensation on substrate is always related either to a sharp increase of the
amount abstracted [42] orto a vertical step [44J. In the work of Miller and co-workers the value of
the maximum, occurring at an eqUilibrium oleate concentration of about 3 xlO- 5 mol/!, corresponds
to 3.5 statistical layers of calcium oleate in the hydrated crystal state due to a high amount of
calcium ions of the system investigated (7 .lx 10-5 mol/m 2)(pKs used 15.6) .
An interesting feature of the isotherms are the results obtained by Rao et aI. [44,51] (pKs used
46

14.5). With natural fluorite, surface precipitation is preceded by adsorption of oleate to fonn a
monolayer, then a bilayer in a liquid crystal state (characterised by the value of the chain area at the
plateau and the low absorbance obtained with infrared experiments) at an equilibrium concentration
of about 5.7xlO· 6 mol/l (pH 10.0) and 4.9xlO· 6 mol/l (pH 11). Surface precipitation, is
characterised on the isothenns : I) by a vertical step, 2) by the presence on infrared spectra of two
bands, at 1537 and 1576 cm- I, in the range of the carboxylate groups, attributed to calcium oleate,
and 3) by a sharp increase of the sensitivity in measuring the absorbance of the bands related to the
hydrocarbon chains (related to a change in the state of the adsorbed layer from the liquid-crystal
state to the hydrated crystal state), and occurs around 7-7.5xlO- 6 mol/l. Surface precipitation is
also accompanied by a sharp decrease in zeta potential that can also be interpreted with the results
presented above as an increase in surface charge density due to the change of the adsorbed layer in
a hydrated-crystal state. The maximum occurs between I x 10-5 and 4xlo-5 mol/l for pH maintained
at 10.0 and around 2.5x I 0- 5 mol/l at pH 11.0. The maximum amount of calcium ions available for
precipitation in these runs was high, around 4.3xlO-5 mol/m 2 . In this case, the decrease of the
amount abstracted seems to be related at higher concentrations to the exchanges of the calcium ions
by the sodium counter ions, due to the increase of collector concentration.

30,-----------------------------,
SOLIDS (%)
o 005
x 0.10
'" 020

x x
x x

TI

-t:'o"
Vl
~ 10

o'"
6'"

10'5 10" 10'3

Equilibrium Oleate Cone (moles!l)

Figure 13. Adsorption isotherms of oleate onto synthetic fluorite at different solid/liquid ratio and at
pH 10.0.

With synthetic fluorite, the results obtained for low solid concentrations are close to those
described above. But in these runs, the total amount of calcium available for precipitation is lower
(around 2.65xlO-5 mol/m 2 ). With an increase in solid concentration, the preCipitation step was
found to be reduced and at higher solid concentrations the isothenn levelled off at the bilayer
without any precipitation step. In the last case, it was undoubtless demonstrated that the importance
of the maximum is a function of the solid/liquid ratio, i.e. the surface available for adsorption and
the amount of calcium ions in the bulk (Figure 13). In this condition, the maximum amount of
calcium available for surface precipitation was 6.64xlO- 6 mol/m 2, a value very close to that used
 47

by Sivamohan et al. [46) to obtain an isotherm limited to the bilayer. The beginning of the
exchange between calcium ions and sodium ions on the external part of the adsorbed layer is often
related to the disappearance of the bands related to calcium oleate and the decrease of the
absorbance of all the bands characteristic of the oleate molecule in the carboxylate group region
and in the hydrocarbon chain region. This phenomenon is related to a second change in the state of
the adsorbed layer from the hydrated crystal state to the liquid-crystal one.
The apparent lack of consistency between these results concerning the abstraction of oleate on
sparingly soluble minerals can be explained:
- by the total calcium ion concentration in the bulk controlling the stoichiometric composition of
the precipitated phase,
- by the real nature of the fluorite surface (more or less covered with calcium carbonate) which
controls the position of the isotherm on the X-axis. The isotherm obtained by Sivamohan and
co-workers is shifted towards higher oleate concentrations, i.e. higher Na ions concentration.
This phenomenon can be related to the surface charge of fluorite particles. The natural mineral
particles were found more (-35 mV for Sivamohan fluorite) or less (-15 mV for Rao fluorite)
electronegative in the range of pH chosen for adsorption experiments, whereas synthetic fluorite
particles utilised by Rao and co-workers was electropositive. Two consequences can be drawn :
1) in the first case, the negative electrical surface field is in a precursor stage opposed to
chemisorption (for the most electronegative fluorite, the monolayer capacity is reached for
lxlo-4 mol/l). Sodium and calcium ions may adsorb as counter ions.
2) in the second case, the positive elcctrical surface field favours adsorption. The isotherm is
located at low concentration (the monolayer capacity is reached around 5xlO- 6 mol/l) and
sodium ions cannot adsorb as counter ions. The band at 1555 cm- t can be attributed then
without any ambiguity to the monocoordinated surface calcium oleate.
In conclusion, an appropriate interpretation of the experimental isotherms imperatively needs the
knowledge of the amount of ions available for precipitation and the real nature of the surface.

6.2. THE GALENA-POTASSIUM AMYLXANTIIATE SYSTEM

This system (the surfactant presents a short chain and a high affinity for lead ions in solution,
pKs = 24) was recently investigated using chromatographic saturation front analysis, batch
adsorption tests and diffuse reflectance Fourier transform spectroscopy (DRIFTS) [38). The
following results were found:
- the galena (PKs = 28) was pretreated in order to obtain "clean galena", i.e. unoxidised galena
surface. Despite the precautions taken in order to avoid oxidation, galena was slightly oxidised
and the presence of traces oflead thiosulphate (pKs = 6.4) and elemental sulphur were detected.
The isotherm presents a Langmuir shape and a plateau for a value of the fractional surface
coverage around 0.13 (cross-sectional area of xanthate ion adsorbed on surface in the hydrated
crystal state was taken equal to 28.8 A2). The oxidation products were found to be responsible
for the xanthate abstraction probably under the lead xanthate form, but the affinity of xanthate
ions for clean galena was nil.
- clean galena was oxidised by dissolved oxygen in aqueous solution before surfactant adsorption
and the infrared spectrum showed weak absorption bands at 1122 and 983 cm- t , confirming the
presence of superficial lead thiosulphate. The corresponding isotherm also exhibits a Langmuir
shape with a plateau at a surface coverage of 0.5. The oxidation of galena by dissolved oxygen
was insufficient to provide the degree of surface oxidation necessary for an important
adsorption/abstraction. The infrared spectra clearly indicate that part of the surface was covered
48

with multilayers of stoichiometric lead xanthate (maximum 12.5% of the surface with three-
dimensional patches fonned with more than four layers), part was covered with a monolayer
(some infrared bands indicating a 1:1 coordination between the surface metal atom and the
xanthate radical) and part was empty [38].
- the galena was dry ground and a great superficial amount of oxidised species were found
(PbSzO) and PbCO). The adsorption/abstraction process of xanthate is limited at the fonnation
of lead xanthate [52].
- then, galena was wet ground, at different pH values, in a rod mill using stainless steel grinding
media. After grinding in acidic conditions, the surface of galena was covered with a weak
amount of PbS203. The surface concentration of lead thiosulphate increases with pH and in
alkaline grinding conditions, infrared experiments revealed the presence of high amounts of
basic lead carbonate. In these conditions the total amount of xanthate abstracted was always
found for higher concentrations, greater than five times the statistical fractional surface coverage.
A two-stage adsorption I abstraction process was discovered for amylxanthate. During the first
stage, a mixed layer is fonned made of 1-1 monocoordinated lead xanthate and dixanthogen. For
higher concentrations, it is composed of non stoichiometric lead xanthate, amyldixanthogen and
amylcarbonate disulphide. In the second stage, mainly dixanthogen is fonned when the stock of
lead ions available for surface preCipitation is consumed. Lead xanthate abstraction is always
accompanied with the vanishing of the bands related to superficial lead oxidised compounds. At
high xanthate concentrations, 1-2 coordinated lead xanthate can be fonned in addition. The
second adsorption stage process corresponds, in absence of frother, to complete flotation
[38,53].
This system is a very complicated one. For low concentrations, adsorption corresponds to
chemisorption of amylxanthate to form monocoordinated compounds and to superficial oxidation
of xanthate in dixanthogen, while for higher concentrations surface precipitation of lead xanthate is
prevailing, together with the oxidation of xanthate in dixanthogen and amylcarbonate disulphide.
Moreover, it was established that the total amount of xanthate adsorbed I abstracted is greatly
dependent on the amount of surface oxidation products, Le. on the amount of lead ions available
for surface precipitation, because oxidised compounds are more soluble than "clean" galena. The
first stage of the abstraction process is the surface precipitation oflead xanthate, followed by the in
situ oxidation (electrochemical step) of a part of the adsorbed molecules.

6.3. THE PYRITE - AMYLXANTHATE SYSTEM

It is well known that xanthate adsorbs on pyrite to fonn dixanthogen. The concept of surface
reactions proceeding by separate electrochemical steps is well established in many fields, in which
conducting materials are involved. Sulphides, being electronic conductors, can act as a source of
electrons and hence can support electrode reactions at the surface. A reaction such as the oxidation
of xanthate to dixanthogen would take place on the surface via two separate, simultaneous electrode
processes: the anodic oxidation of the adsorbed xanthate molecule coupled with the cathodic
reduction of adsorbed oxygen. In this way, the sulphide catalyses the formation of dixanthogen.
Recently, after grinding pyrite in different pH conditions, Cases et al. [54,55] confinned a
hypothesis expressed by Fuerstenau et al. [56] concerning the possibility of fonning dixanthogen,
according to a mechanism including the reduction of superficial ferric compounds to ferrous ones.
After wet grinding with stainless steel materials, pyrite is mainly oxidised into iron hydroxides or
oxyhydroxides, sulphur, and ferric sulphate. The adsorption/abstration of xanthate and the
fonnation of dixanthogen is clearly related to the vanishing (pH 12) or the decrease (PH 4 to 11) of
 49

all the bands attributed to ferric compounds in the region 1300-1000 cm- I , the ferrous species
being more soluble than the ferric ones. For concentrations higher than 5xlO-6 mol/l, multilayers
are formed; at the highest concentrations (5xlO- 3 - lxlO- 2 mol/l) at 25°C they reached around 25
(grinding pH 4.0 - adsorption in closed flasks [57]). 7 (grinding pH 5.1 - adsorption in closed
flasks [57]). 17 ( grinding pH 5.1 - adsorption in aerated conditions - influence of the oxygen
content [53]). 17.2 (grinding pH 12 - adsorption in aerated conditions [55]) statistically adsorbed
layers calculated from the value of the cross-sectional area of the xanthate ion. The increase of
amyldixanthogen surface concentration corresponds to a sharp decrease of zeta potential and
formation of a negative pyrite surface. The infrared experiments tend to prove that the superficial
amyldixanthogen thick layer is in a solid state because of the presence of the characteristic band of
the C-O-C stretching vibration at 1262 cm-! [52.58].
The species responsible for flotation, when iron is used as grinding material during wet
grinding. are different. The hydrophobicity of the mineral is due to the formation of ferric
amylxanthate for low additions of xanthate. For higher concentrations. the collector can abstract
through physically adsorbed xanthate ions [53]. The adsorbed layer is very likely limited to the
monolayer or the bilayer capacity.
In the two cases, adsorption/abstraction of collectors dissolves in a first step the whole or part of
the superficial oxidised species present on the surface. In these systems. where the surface is
higbly oxidised. a surface precipitation process could precede chemical interaction on the surface
that would occur by an electrochemical mechanism [59]. For these systems. a lack of information
prevents the use of a sophisticated thermodynamic approach.

a IC?lJ
. . . . . . . S:. lCZJJ'
e... s :

i:
o : 0
loge.::'~ -3 -2 -I 0
.:Ill

Figure 14. The different plots recommended for adsorption experiments. B represents the monolayer
capacity as determined by the classical method of the B point.

7. What kind of plot should be utilised to account for adsorption experiments

Usually adsorption experiments are conducted in order to answer a series of questions:

I) what are the fundamental laws that render an account for adsorption. i.e. laws extrapolable to
50

different adsorbents or adsorbates? The equilibrium laws always relate the fractional surface
coverage, q, to the equilibrium concentration or its logarithm (eqn. 1 or the classical
e
expression of the Langmuir isotherm). Then, the vs Ce or In C e plot appears as a natural
e
one. But the use of imposes an accurate determination either of the cross-sectional area of
the molecule in the condensed state and the specific surface area of the adsorbent or of a
marked plateau at the monolayer capacity. An error, for instance, in the specific surface area
will shift the break characteristic of the monolayer capacity from 1.
2) How to study, under good conditions, surface heterogeneity or microporosity [60] ? Figure
14 shows that to obtain a good perception of the filling of the isotherm in the range of the
monolayer, it is necessary to use a logarithmic abscissa. This filling is on a sub-vertical region
e
of the isotherm in the vs Cc plot. This plot will also be used to study the influence of the
length of the chain.
3) e
How to study the affinity of a surfactant for different adsorbents ? The vs 6.!l plot will be
preferred for all the reasons described in § 3 and 4. This plot will also be used in the study of
adsorption of an homologous series of surfactants onto a solid to verify the superimposition
of the different isotherms and then to appreciate the physical state of the adsorbed layer in
relation to the reference one.

Hydrophilic surface

~ T>U

.-----------i. ~~ T<Tk

1+2 3+4
2

Hydrophobic
surface

I I I I
0.0001 0.001 0.01 0.1 tC/CMC
------,-----,-----,-----,-----,-
-9.2 kT -6.9 kT -4.6 kT -2.3 kT 0 A/J
STRONG NORMAL WEAK NORMAL
BOND BOND

Figure 15. Schematic representation of different types of adsorption isotherms on heterogeneous and
hydrophilic surfaces.
 51

We do not recommend the use of Qa (expressed in mol/g). on the Y-axis. because the results do
not allow the comparison of the isotherms obtained on different adsorbents. nor the use of a
logarithmic Y-axis. Plotting results in the form of fractional surface coverage can mask the effects
encountered at very low surface coverage. i.e. the precursor stage of adsorption. Nevertheless. we
have always to keep in mind that in this region of the isotherm the determination of residual
concentration lacks of precision especially in the case of strong normal bonds. Moreover. for the
moment. any method is able to distinguish between localised molecules adsorbed as individual ions
or in aggregates in which the aggregation number is lower than 25 molecules. In fact. this use
greatly favours the low surface coverage domain with regard to the region corresponding to the
filling of the monolayer.

8. Conclusion

A figure will be presented for the sake of a conclusion. If the three-dimensional condensation on
substrate is possible. the amount of surfactant abstracted from the bulk depends on the amount
available in solution and necessary for precipitation (fluorite-oleate system for instance) or the
dissolution or some superficial phascs (galena and in some cases pyrite-xanthate systems). In
many cases, the thermodynamics of infinite three-dimensional phases will allow a good
interpretation of the abstraction/adsorption phenomena.
When the real adsorption phenomena are prevailing. different cases can be encountered and they
are described in Figure IS. Imagine a surface made of four homogeneous domains.
a) If the surfactant has a strong affinity for the adsorbent (high values of <p~0 whatever the value
of the Krafft point, the isotherm will present (or not for highly heterogeneous surfaces). the
steps characterising the two-dimensional condensation (00 > 4 leT) and corresponding to the
formation of patches on the surface with one layer. The size of these aggregates is a function
of the surface heterogeneity. For a givcn value of the undersaturation~!li'i # (<I>'h - <I> "A)!2. the
normal step appears on the isothcrm corresponding to the formation of the bilayer. Obviously
the nature of the solid surface will strongly influence the shape of the isotherm in the
monolayer region. In this process, the surface is hydrophilic for the surface coverage values
around zero and two, and hydrophobic in the monolayer capacity region.
b) If the affinity of the surfactant for the solid is weak. the system can be chosen such that. for
instance. the two first energetic domains will be filled with the formation of patches similar to
that described above. When the value of ~I!ir is reached two different possibilities can be
described:
bI) T < TK, in this case the only way to place extended molecule in the adsorbed layer is the
lamellar organization. Domains I and 2 are covered with a bilayer. On the empty domains
3 and 4. the patches that form in the adsorbed layer are bilayers. This is the region of the
application of the admicelle hypothesis [8] with an important restriction: the pseudophase
separation model cannot apply in this range of temperature and the classical laws of
adsorption are sufficient to explain the phenomenon. From the value of the surface
coverage corresponding to L\l!il the surface becomes and then always stays hydrophilic.
b:il T > TK, the domains 3 and 4 are fillcd with spherical or oblate aggregates. The surface
becomes hydrophilic and the suspension very stable if adsorption is realised with fine
particles.
For systems with weaker interactions, the formation of micelles on the surface is possible at the
onset of the isotherm.
52

In conclusion, the thennodynamic models discussed in the present work: predict some general
trends of the adsorption of surfactants on strongly or weakly interactive hydrophilic solid.

9. Acknowledgement

The authors wish to aknowledge the Ministere de la Recherche of France (phygis program) for its
fmancial support.

10. References

1. Cases, I.M., Poirier, J.E., and Canet, D.(1985) Adsorption a /'interface solide-solution des
tensio actifs ioniques " les systemes aforte liaison normale adsorbat-adsorbant et surface
hererogene, in I.M. Cases (ed.) Solid-Liquid Interactions in Porous Media, Technip
Publisher, Paris, pp. 335- 370.
2. Gaudin, A.M., and Ferstenau, D.W. (1955) Quartzflotation with cationic collectors, Trans.
AIME 202, 958-962.
3. Fuerstenau, D.W., and Raghavan, S. (1976) Some aspects of the thermodynamics of
flotation, in M.e. Fuerstenau (ed.) Flotation, AIME, pp. 21-65.
4. Cases, I.M., and Mutaftschiev, B. (1968) Adsorption et condensation des chlorhydrates
d'alkylamine a /'interface solide-liquide, Surface Science 2, 57-72.
5. Cases, I.M. (1979) Adsorption des tensio actifs a /'interface solide-liquide " Thermodyna-
mique et influence de l'heterogeneite des adsorbants, Bulletin de Mineralogie 102,684-707.
6. Cases, I.M., Canet, D., DoerIer, N., and Poirier, I.E. (1982) Adsorption des tensio actifs
ioniques a chaine alkyle a /'interface solides-solutions aqueuses, in 1. Rouquerol and K.S.W.
Sing (eds.) Adsorption at the Gas-Solid and Solid-Liquid Interface, Elsevier, Amsterdam,
pp.21-53.
7. Cases, I.M., Levitz, P., Poirier, I.M., and Van Damme, H. (1986) Adsorption of ionic and
nonionic surfactants on mineral solids from aqueous solutions, in P. Somasundaran (ed.)
Advances in Mineral Processing, SME, Littleton, pp. 171-188.
8. Harwell, I.H., Schechter, R., and Wade, W.H. (1985) Recent development in the theory of
surfactant adsorption on oxides, in I.M. Cases (ed.) Solid-Liquid Interactions in Porous
Media, Technip, Paris, pp. 371- 409.
9. Levitz, P., Azouz el Miri, Keravis D., and Van Damme, H. (1984) Adsorption of nonionic
surfactants at the solid-solution interface and micellization,' a comparative fluorescence decay
study, 1. Colloid Interf. Sci. 99, 484-492.
10. Levitz, P., Van Damme, H., Keravis D. (1984) Fluorescence decay study of the adsorption
of nonionic surfactants at the solid-liquid interface. 1. Structure of the adsorption layer on a
hydrophilic solid, 1. Phys. Chem . .8..8., 2228-2235.
11. Levitz, P., Van Damme, H., and Kcravis D., (1985) Fluorescence decay study of the
adsorption of nonionic surfactants at the solid-liquid interface, in I.M. Cases (ed.) Solid-
Liquid Interactions in Porous Media, Technip, Paris, pp. 473-484.
12. Levitz, P., and Van Damme, H. (1986) Fluorescence decay study of the adsorption of
nonionic surfactants at the solid-liquid interface. 2. Influence ofpolar chain length, 1. Phys.
Chem. 211, 1302-1310.
13. Levitz, P. (1991) Aggregative adsorption of non-ionic surfactants onto hydrophilic solidi
 53

water interface: Relation with bulk micellization, Submitted to Langmuir.

14. Bohmer, M.R., and Koopal, L.K. (1990) Association and adsorption of ionic flexible chain
surj'actants, Langmuir.6., 1478-1484.
15. Belle, I.P. (1979) Etude de quelques proprietes therrrwdynamiques et structurales des chaines
a
aliphatiques des milieux modeles de membranes l'aide d'une methode de simulation, These,
Universite de Bordeaux, 137 pp.
16. Poirier, I.E. (1984) Etude des mecanisme accompagnant l'adsorption des tensio-actifs
ioniques sur les sUrfaces solides, dans Ie cas de systeme a interactions faibles .' Application a
la recuperation par voie chimique du petro Ie contenu dans les gisements greseux, These,
INPL Nancy, 328 pp.
17. Brown, G.H. , Doane, I.W., Neff, V.D., and Ohio, K. (1971) A rewiew of the Struture and
Physical Properties of Liquid Crystals, CRC Press, Cleveland, 94 pp.
18. Belle, I.P., and Bothorel, P. (1977) Thermodynamical study of the phospholipid chain
structure in mono and bi/ayers, Nouveau 1. de Chimie 1. 265-267.
19. Mely, B., and Charvolin, 1. (1978) Configuration des chaines dans les phases lamellaires de
savons de potassium, Colloques nationaux du CNRS N° 938 - Physicochimie des Composes
Amphiphiles, pp. 41-44.
20. Cabane, B. (1981) Structure of the water/surfactant interface in micelles: an NMR study of
SDS micelles labelled with paramagnetic ions, 1. Physique 42, 847-859.
21. Marchal, I.P., Canet, D., Nery, H., Robin-Lherbier, B., and Cases, I.M. (1984) Nuclear
magnetic resonance study of alkylammonium chlorides. 3. Micelle characterization and
dynamic properties determined by protons and carbon 13 longitudinal relaxation, 1. Colloid.
Interf. Sci.ll, 349-359.
22. Denoyel, R., Rouquerol, F., and Rouquerol, 1. (1990) Thermodynamics of adsorption from
solution: experimental andformal assessment of the enthalpies of displacement, J. Colloid
Interf. Sci. l3.6., 375 - 384.
23. Poirier, I.E., and Cases, I.M. (1985) Sur l'origine et la nature de !'interaction normale
adsorbat-adsorbant dans les systemes a interactions faibles, in I.M. Cases (ed.)Solid-Liquid
Interactions in Porous Media, Technip, Paris, pp. 429-445.
24. Poirier, I.E., and Cases, I.M. (1991) Anionic surfactant adsorption onto silicate minerals:
the role of the cations, to be published in Colloids and Surfaces.
25. Phillips, M.C.,Williams, RM., and Chapman, D. (1969) On the nature of hydrocarbon chain
motions in lipid crystal, Chern. Phys. Lipids }, 234-244.
26. Cases, J.M., Cunin, P., Grillet, Y., Poinsignon, C., and Yvon, I. (1986) Methods of
analysing morphology of kaolinites: relations between crystallographic and morphological
properties, Clay Minerals 21,55-69.
27. Chandar, P., Somasundaran, P., and Turro, 1. (1987) Fluorescence probe study on the
structure of the adsorbed layer of dodecylsulphate at the alumina-water interface, I. Colloid
Interf. Sci., ill, 31-46.
28. Partyka, S., Rudzinski, W., and Brun, B. (1989) Calorimetric studies of adsorption of
anionic surfactant,l' onto alumina, Lanl,'ITluir~, 297-304.
29. Denoyel, R., Rouquerol, F., and Rouquerol, 1. (1989) Adsorption of anionc surfactants on
alumina; complementary of the information provided by batch and liquid flow micro-
calorimetry, Colloids and Surfaces 37, 295-307.
30. Levitz, P., (1985) Processus d'association de rrwlecules non ioniques: adsorption a /'interface
solide hydrophile-eau et micellisarion en phase aqueuse, These doctorat d'Etat., Orleans,
333 pp.
54

31. Denoyel, R. (1987) Etude thermodynamique de l'adsorption de molecules tensioactives a

/'interface oxyde mineral-solution aqueuse, These doctorat d'Etat, Marseille, 256 pp.
32. Lindheimer,M., Keh, E., Zaini, S., and Partyka, S. (1990) Interfacial aggregation of
nonionic surfactant onto silica gel: calorimetric evidence, 1. Colloid Interf. Sci.,Uli, 83-91.
33. Lee, E.M., Thomas, R.K., Cummins, P.G., Staples, E.J., Penfold, J., and Rennie, A.R.
(1989) Determination of the structure of a surfactant layer adsorbed at the silicate/water
interface by neutron reflexion, Chemical. Phys. Lett. 162, 196-202.
34. Klimenko, N., A., and Kaganovskii, A. M. (1974) Adsorption of non ionic surfactants in the
associated state from aqueous solutions on polar sorbent, Kolloidnyi Zh. 32, 151-154.
35. Somasundaran, P., Chandar, P., Turro, N.J., and Watennan, K. C. (1988) Micro-structure
of surfactants adsorbed layers at the solid-liquid interface : fluorescence and ESR
investigation, in K.S.E. Forssberg (ed.) Proceedings XVI Int. Min. Process. Congress,
Elsevier, Amsterdam, pp. 775-784.
36. Marabini, A., Cases, J. M., and Barbaro, M.(J989) Chelating reagents as collectors and their
adsorption mechanism, in K.S.V. Sastry and M.e. Fuerstenau (eds.) Challenges in
Mineral Processing, SME, Littleton, pp. 35-49.
37. Predali, I.J., and Cases, J.M. (1973) Thermodynamics of adsorption of collectors, in M.J.
Jones (ed.), Proc. X Int. Min. Process. Congress, IMM, London, pp. 473-492.
38. de Donato, P., Cases, I.M., Kongolo, M., Michot, L., and Burneau, A. (1990) Infrared
investigation of amylxanthate adsorption on galena: influence of oxidation, pH, and grinding,
Colloids and Surface 44, 207-228.
39. Hanna, H.S., and Somasundaran, P. (1976) Flotation of salt type minerals, in M.C.
Fuerstenau (ed.) Flotation - A.M. Gaudin Memorial Volume, SME, Littleton, pp.197-272.
40. Somasundaran, P., and Ananthapadmanabhan, K.P. (1986) Solution chemistry offlotation,
in P. Somasundaran (ed.) Advances in Mineral Processing, SME, Littleton, pp. 137-170.
41. Arnaud, M., Partyka, S., and Cases, J.M. (1989) Ethylxanthate adsorption onto galena and
sphalerite, Colloids and Surfaces, n, 235-244.
42. Hu, J.S., Misra, M., and Miller, J.D. (1986) Effect of temperature and oxygen on oleate
adsorption byfluorite, Inter. J. Min. Process . .1.8., 57-72.
43. Du Rietz, C. (1975) Chemisorption of collectors in flotation, in Proceedings XI Int. Min.
Proc. Congress, Istituto di arte mineraria e preparazione dei minerali, Universita di Cagliari
Pub., Cagliari, pp. 375-403.
44. Rao, K.H., Cases, J.M., and Forssberg, K.S.E. (1991) Mechanism of oleate interaction on
salt-type minerals. Part 1lI. Adsorption, electrokinetic and diffuse reflectance FT-IR studies of
naturalfluorite in the presence of sodium oleate, to be published J. Colloid Interface Sci.
45. Jacquier, P. (1982) Etudes physicochimiques d'un systeme phosphate-solutions aqueuses :
application a laflottation d'un minerai de phosphate sedimentaire agangue carbonatee, These
Docteur Ingenieur, Nancy, 292 pp.
46. Sivamohan, R., de Donato, P., and Cases, J .M. (1990) Adsorption of oleate species at the
fluorite-aqueous solution inteiface, Langmuir Q,637-644.
47. Mahieu, N., Canet, D., Cases, J.M., and Boubcl, J.C. (1991) Micellization of sodium oleate
in D20 as probed by proton longitudinal magnetic relaxation and self diffusion measurements,
1. Phys. Chern. (to be published)
48. Kellar, I.J, Young, C.A., Knutson, K., and Miller, I.D. (1991) Thermotropic phase
transition of adsorbed oleate species at afluorite suiface by in-situ FT-IR/IRS spectrosscopy,
to be published in 1. Colloid Interface Sci.
49. Cases, I.M. (1990) unpublished results.
 55

50. Miller, I.D., Su, I.S., and lin, R. (1989) Thermochemistry of oleate adsorption at the
fluorite/water interface, ColI. & Surf. 42, 71-84.
51. Rao, K.H., Cases, I.M., de Donato, P., and Forssberg, K.S.E. (1991) Mechanism of oleate
interaction on salt-type minerals. Part IV. Adsorption and precipitation reactions in relation to
the solidlliquid ratio in the fluorite-sodium oleate system, to be published in I. Colloid Interf.
Sci.
52. Cases, I.M., and de Donato, P. (1990) FTIR analysis of sulphide mineral surfaces before and
after collection: galena, Proceedings 2nd. Workshop Flotation of sulphide minerals, Lulea,
Sweden, 14 pp. (to be published in Inter. I. Min. Process.).
53. Cases, I.M., Kongolo, M., de Donato, P., Michot, L., and Erre, R. (1990) Interaction
between finely ground galena and pyrite with potassium amylxanthate in relation to flotation.
2. Influence of grinding media at natural pH. Inter. J. Min. Process. 30, 35-68.
54. Cases, I.M., de Donato, P., Kongolo, M., and Michot, L. (1991) Interaction between finely
ground pyrite with potassium amylxanthate in flotation : irifluence of pH. grinding. and
collector concentration, Proceedings 3rd. Intern. Symposium on Beneficiation and
Agglomeration - ISBA 91, Bhubaneswar, India, 11 pp.
55. Cases, I.M., Kongolo, M .. de Donato, P., Michol, L., and Erre, R. (1992) Interaction
between finely ground pyrite with potassium amylxanthate in flotation : 1. Influence of
alkaline grinding, Submitted to Inter. J. Min. Process.
56. Fuerstenau, M.C., Kuhn, M.e., and Eigillani, D.A. (1968) The role ofdixanthogen in the
pyriteflotation, Trans. AIME 241,148-156.
57. Cases, I.M. , de Donato, P., Kongolo, M., and Michot, L. (1989) An infrared investigation
of amylxanthate adsorption by pyrite after wet grinding at natural and acid pH, Coll. & Surf.
36, 323-338.
58. Milczarski, J. (1986) In situ spectroscopic study of xanthate adsorption on marcassite,
Coll. & Surf. n, 251-271.
59. Woods, R. (1976) Electrochemistry of sulphide flotation, in M.C. Fuerstenau (ed.),
Flotation, AIME, pp. 298-333.
60. Michot, L., Franr;ois, M., and Cases, I.M. (1990) Surface heterogeneity studied by a quasi-
equilibrium gas adsorption procedure, Lant,'lTlUir Q, 637-643.
1.3. ELECTROCHEMISTRY IN THE PLANT

S.R. RAO, G. LABONTE 1 and J.A. FINCH

Department of Mining and Metallurgical Engineering
McGill University
Montreal, Quebec H3A 2A7
Canada

ABSTRACf. Electrochemical mechanisms governing the mineral-reagent interaction and inter-

mineral interactions are reviewed. Two electrochemical factors of practical importance in flotation
are the pulp potential and galvanic interactions. Examples of industrial importance are discussed.
Two developments in sulphide flotation which are related to the electrochemistry of the system are
collectorless flotation and nitrogen flotation. The practical importance of each in the processing of
complex ores is described. Use of specific electrodes, measurement of pulp potential and its use to
monitor and possibly control flotation are discussed.

1. Introduction

The technology of the flotation of sulphide minerals has made some significant advances in the last
decade. One particular area generating practical interest is the recognition of electrochemical
reactions in sulphide flotation systems [1-3]. Basic electrochemical research and a better scientific
understanding of flotation have led to new developments in selective flotation of sulphide minerals
and in process control [4,5]. Many of these new developments are still in the early stages, but have
shown sufficient promise to predict that they will be integrated into sulphide flotation plant practice,
particularly for difficult to treat ores. Advances in the chemistry of sulphide flotation have been
extensively reviewed [6a,7,81. The intention here is to emphasise where electrochemical principles
are at play in the plant - how they may be identified, measured and manipulated.

2. Basic Considerations

2.1. ELECTROCHEMICAL INTERACTIONS

flotation is dependent on the formation of a hydrophobic surface. In the case of sulphide minerals,
this surface state usually arises through interaction with sulphydryl (or thiol) type collecting agents,

1 Present address: INCO Ltd., 1. Roy Gordon Research Laboratory, 2060 Flavelle Boulevard, Mississauga,
Ontario, Canada L5K lZ9

57
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 57-100.
© 1992 Kluwer Academic Publishers.
58

xanthates being the most common.

Two mechanisms of interaction between a sulphide and sulphydryl type reagents have been
identified. The first consists of a series of well-defined reactions leading to the formation of a
sparingly soluble surface chemical complex at the mineral surface. Such surface compounds could
be formed by ion exchange adsorption; for instance, exchange of xanthate for the suiphoxyl ions or
other oxidation products present at the sulphide mineral surface [7].
The second mechanism of interaction is electrochemical. Although this was recognised many
years ago it has been studied in detail only in the last 20 years. The end product may be the same as
that formed by chemical reaction, but the route is different. The electrochemical or charge transfer
mechanism consists of an oxidation (or de-electronation) step and a reduction (or electronation)
step. Example reactions will be described in subsequent sections.

2.2. THE UNIQUE NATURE OF SULPHIDES

In contrast to most nonsulphide minerals, sulphides are characterised by their chemical reactivity
with water and dissolved oxygen. This arises from the instability of sulphide sulphur. In sulphide,
sulphur occurs in its lowest oxidation state, -2. This can be oxidised to the higher oxidation states
of sulphur, 0, +2, +4 and +6 depending upon the activity of the oxidising agent (e.g. 02, Fe3+) in
the system. For example, oxidation can occur by the transfer of electrons from sulphur to oxygen,
a charge transfer mechanism. A series of charge transfer reactions of sulphide minerals in water are
known.

2.3. CHARGE TRANSFER REACTIONS IN WATER - WITHOUT SULPHYDRIL

REAGENTS

The oxidation reaction for any sulphide mineral in water may be represented by

(1)

where M is a bivalent ion and (S-) is an intermediate product in the oxidation. The metal ions
formed by oxidation of the sulphide may undergo hydrolysis and/or oxidation.
The following type of hydrolysis reactions have been suggested [9]:

M(H20)X2+ =M(OH) (H20)y+ + W + (x-y-l) H20 (2a)

M(OH) (H20>y+ =M(OH)z. ZH20 + W + (y-z-l) H20 (2b)

These hydrolysis species playa significant role in some of the flotation processes. For example,
some of them serve as activating species thus helping to promote the adsorption of collecting agent
[6b].
The sulphur intermediate (S-) may undergo oxidation through a series of reactions to form
sOi-. The first stage of this oxidation process is the formation of elemental sulphur

(3)

The sulphur formed may nucleate as a new phase or it may be present as a solid solution in the
sulphide.
 59

Oxidation of sulphur to sulphoxyl species like SxOl- occurs in stages. The formation of these
anionic species has been recognised in the adsorption of xanthate ions by exchar.ge adsorption [3].

2.4. CHARGE TRANSFER REACfIONS - WITH SULPHYDRIL REAGENTS

The common cathodic reaction is

(4)

Some typical anodic reactions are:

(i) X- -> Xadsorbed + e- (5a)

(ii) 2X- -> X2 + 2e- (where X2 = dixanthogen) (5b)

(iii) (5c)

Each charge transfer reaction corresponds to a definite electrochemical potential E related to the
Gibbs free energy change of the process,

~G= -zFE (6)

(where F is the Faraday constant and z the number of electrons transferred per unit mole.).
An anodic reaction can take place at a mineral surface, if a potential exists, at which both this
process and the cathodic reduction of oxygen proceed at finite and equal rates. Such a potential is
called a mixed potential. It should be noted that reactions need not proceed continuously at a finite
rate.

2.5. ELECTROCHEMICAL POTENTIALS AND THEIR IMPLICATIONS

2.5.1. Basic Considerations. Several notations for electrochemical potentials have been used.
The electrochemical potential (abbreviated as E), Oxidation Reduction Potential (ORP, redox, Eh),
rest potential, thermodynamic or reversible potential (E'), and standard potential (EO) all refer to
chemical reactions involving electron transfer having a definite associated electrical potential. There
are, however, subtle differences.
The electrochemical potential (E), and the Oxidation Reduction Potential (ORP, Redox) are
general terms. They do not imply any particular condition for the system being studied except that
a measurement was taken.
The rest potential, reversible potential (E,), standard potential (EO), and Eh have more precise
meanings. These notations imply that the system was at equilibrium when the measurement was
made. As such, the measured potential can be related to the thermodynamic properties of the
particular reaction occurring in the system. The state of the system (e.g. activities for ions,
pressure, P, and temperature, T) must be known.
For the standard state (unit activities, P = 1 atm, T = 25°C), the measured potential is the
standard potential of the reaction (EO). Furthermore, Eh implies that the potential (measured or
calculated) is referenced to the Standard Hydrogen Electrode.
60

TABLE 1 Notation
E general notation for an electrochemical potential, measured or calculated
E" thermodynamic (standard) potential of the reaction; referenced to the standard hydrogen
electrode unless otherwise specified. Tables provide values of EO for the most common
couples. By convention, these tables present the reactions as reductions; the higher the
potential (the more positive), the more likely the reduction will occur.
Er reversible potential for one couple. Under the standard conditions (T =25 °C, P = I atm,
unit activities for the ions), EO =Er.
Eh measured (or calculated) potential for simple or complex solutions. Always referenced to
standard hydrogen electrode (SHE).
Ep measured pulp potential for a slurry. Explicit mention of the sensing and reference
electrodes is required for the value to be interpretable.
Es solution potential for the solution part of the slurry. This may be a mixed potential if
the solution contains several ions. The sensing and reference electrodes must be stated.
Em mineral potential for the mineral particles in the slurry. This may be a galvanic potential
if there are two or more electrochemically active solids present (e.g. abraded iron media,
sulphide minerals).
ESCE measured potential with reference to a saturated calomel electrode, most commonly used
reference electrode. To convert the measured ESCE to Eh simply add the standard EOSCE
(0.247 V).
EAg/AgO measured potential with reference to a standard Ag/AgCl electrode. To convert the
measured value to Eh add the standard EOMIMO (0.222 V).

There are other types of potential which refer specifically to systems under dynamic conditions:
solution, mixed, galvanic, and pulp potentials. These names make reference to the non-equilibrium
(dynamic) situation when the potential was measured; one or more redox reaction pairs are
occurring at finite rates.
The eqUilibrium state is rarely achieved in flotation. The measured potential remains constant as
long as the reactions proceed at constant rates. The introduction of reduced and oxidised forms of
a reagent (even if the ratio of concentration is equal to that in the system), or an increase in the
solution mixing, unbalances the rates and a drift in the potential value is observed.
Mixed potential is the generic term for a potential measured in a complex system. Specific terms
are the solution potential (lOOking only at the ions in the system) and the galvanic potential (looking
only at the solid phases in contact).
The solid phases which can give rise to a galvanic potential are metals and semi-conductors. It
should be noted that the solids need not touch: the ions in the solution may transport the electrons
from one solid to the next by entering into oxidatiOn/reduction reactions at the solid surfaces. For
example, the Fe(III)/Fe(II) redox couple is reversible and can act as an "electron carrier" between
solids. The potential of the solution will approach that of the solids when the redox reactions are
reversible, but very often these potentials differ.
The pulp potential Ep has been used to describe the potential measured in a mineral slurry. This
potential is neither equal to the solution potential nor the galvanic potential of the minerals, but
takes a value in between these limits. For Ep to be interpretable, the sensing electrode must be
specified. For example, Ept specifies that the measurement was made with a platinum sensing
electrode. The notation used is summarised in Table 1.
In the present paper, potential values are expressed with reference to standard hydrogen
 61

electrode (SHE or S.H.E.) unless otherwise stated.

Detailed discussion on the fundamentals of electrochemistry may be found in the literature
[1,12]. Only the most important points will be covered here.
The oxidising or reducing conditions in a solution can be represented by E. For a system at
eqUilibrium, and being reversible, E is given by the Ncmst equation:

_ ° Rf [Reduced] (7)
E - E + zF In [Oxidised]

For a given couple involving z electrons, E will depend on:

i) the ratio of the activities (or concentrations) of the oxidised and reduced forms of the couple,
and
ii) the temperature, T. It should be noted that EO also depends on the temperature, since it is a
thermodynamic quantity (EO = - /'t.Go/zF).

u Oxidation Mixed
'6

--
0 potential
c
F:
c:
""
(I)

:::I
0 Potential
u
'6
.. ER2
0
:5
u Increasing mixing or
concentration

Reduction

Figure 1. Concept of the mixed potential for a complex solution.

E can be interpreted as the capability of the system to oxidise (or reduce) a given metal, a
mineral, or another ion. Solutions with a given E will oxidise metals, minerals, or ions which have
an Eo less than E. Conversely, the same solutions will reduce metals, minerals, or ions which have
an EO greater than E. Since EO is tabulated (with respect to the SHE), redox couples with higher EO
will oxidise the couples with lower EO. The electrochemical couples with higher EO are commonly
referred to as being more noble. In this context, noble does not mean electrochemically inert.
The main problem with thermodynamic quantities such as Eh and EO is that they give an
indication of the driving force for the reaction, but do not indicate if, or how fast, the reaction will
proceed. For example, thermodynamically, the oxidation of sulphides should produce sulphate
ions. In practice, sulphoxyl species or elemental sulphur is produced under flotation (and even
62

leaching) conditions.
When there are a number of couples present in equilibrium in the solution, there still exists one
electrochemical potential. For two couples Red l/OX 1 and RedzlOx2 :

(8)

As mentioned earlier, one of the couples will act as a reductant while the other one will act as an
oxidant (Figure I), depending on their respective Er. Because of this, the two couples cannot be
considered independently and the measured potential is a mixed potential with a value between the
E r of the two couples. Increasing the intensity of mixing, or the concentration of one of the redox
couples, causes E to shift towards the Er of the couple, which is favoured by the action taken.
The meaning of the potential, when there is a large proportion of sulphide particles present, is
uncertain. Since the flotation process is relatively rapid, sulphide minerals and the solution are
rarely at equilibrium. The measured potential is somewhere between the potential of the minerals
Em and that of the solution Es. The desired potential to be monitored is Em. This is best achieved
by using mineral sensing electrodes. However, a mineral electrode may degrade with time as its
exposed surface reacts with the dissolved components of the slurry. Further discussion on the
choice of sensing electrodes is given in section 3.

SOLUTION IN
S
. -.-
••••
:::: ....
]
CE C
E

SOLUTION FOR
ANALYSIS

1
GAS

Figure 2. Potentiostat measurement circuit. SCE, saturated calomel electrode (reference); CE, counter
electrode; WE, working electrode (mineral bed, electrical contact by platinum wire); P, potentiostat; M,
microcomputer processor [13,14].

2.5.2. Varying Electrode Potentials by External Power Source. In order to study the dynamics of
flotation systems it is necessary to follow the changes occurring when the potential of the mineral is
varying. This is performed by disturbing the state of the mineral surface. The most convenient
method is to impose a potential by an external power source. An electric current flows through the
mineral of a magnitude indicating how strongly the system reacts to the disturbance.
Measurement of current under the application of an external potential is called cyclic volt-
ammetry. The experimental curve showing the variation of current with applied potential is called a
voltammogram. When at certain potentials a large current flow is recorded, this indicates that a
 63

charge transfer reaction has occurred between the mincral and the solution. A possible set-up,
where the cell also serves for flotation, is shown in Figure 2. In this case, the mineral electrode is a
particulate bed electrode, (PBE); an alternative is a massive mineral electrode (MME).

2.5.3. Examples of Voltammograms. Voltammograms of two sulphide mineral systems are shown
in Figures 3-4. All show potential regions where large current flows are recorded. The reactions
corresponding to these potentials can be identified using thermodynamic data for the various
reactions that could occur in the particular system.

200

100
'"
~::I.
0
~
(/)
Z
W
0 -100
I-
Z
w
a:
a:
:::l -200
()

POTENTIAL, mV SHE

Figure 3. Voltammogram of pyrrhotite at pH 4 (from ref. [14]).

For example: The voltammogram of pyrrhotite at pH 4 in water (Figure 3) shows a potential

corresponding to current flow in the anodic region of 200 mV vs. SHE. This is compared with the
potential for the oxidation of pyrrhotite, leading to the formation of elcmental sulphur and sulphate,

FeS1.l3 + 7.52 H20 = Fe(OHh + 1.13 S04 2 - + 12.04 W + 9.78 e

Eh= 379 mV vs. SHE (9)

The experimental value is lower than this theoretical one, implying that the reaction is more
complex. Studies at different pH values led to the inference that the oxidation of pyrrhotite is a two
stage process [14]. The voltammogram of chalcopyrite (Figure 4) shows a charge transfer reaction
at 450 mV vs. SHE. This has been interpreted by the following redox reactions [IS]:
64

CuFeS2 = CuS + Fe2+ + S + 2e (10)

Eh = 0.232 + 0.0295 log (Fe 2+) = 55 m V (vs. SHE)

CuFeS2 = CuS + Fe3+ + S + 3e (11)

Eh = 0.412 + 0.0197 log (Fe3+) =294 mV

CuS = Cu 2+ + S + 2e (12)
Eh =0.590 + 0.0295 log (Cu 2+) =413m V

The potential most closely agreeing with the measured value is that of reaction 12.
Note that each of the chalcopyrite reactions leads to elemental sulphur. Since So is hydrophobic,
this implies that chalcopyrite could be floated without a collector if it has been superficially
oxidised. Rotation of chalcopyrite without collector, by some such mechanism, is now widely
accepted, and has been demonstrated in plant [16,17]; however, the exact form of the surface
species is still debated.

100r---------------------------------~

3
~o 0 ~--L--L--L--L~~~~--~~--L-~

-100LL__L__L~_ _~_L~~~~_ _L__L~~

-200 0 200 400 600 800
E vs SHE (mV)

Figure 4. Voltammogram of chalcopyrite.

The voltammogram for xanthate at pyrite shows a current flow at an anodic potential of about
100 mV (vs. SHE) [3]. This closely corresponds to the xanthate-dixanthogen potential (reaction
(5b». The voltammogram for galena treated with ethyl xanthate shows a current flow at 0 m V and
another at 120 mV (vs. SHE) [3]. The first charge transfer reaction is interpreted as the adsorption
of xanthate according to reaction (5a), the second as oxidation of xanthate to dixanthogen (reaction
(5b».
 65

3. Measurement Techniques

3 .1. INSTRUMENTAL ASPECTS

Cyclic voltammetry provides an elegant demonstration of the oxidation-reduction mechanism in

flotation. Measuring E in the flotation slurry should provide an indication of the readiness of the
sulphides to float.
As with any instrumental technique, the most critical component is the sensor. The sensor
consists of two parts: the sensing electrode and the reference electrode. Each will be considered
separately.

3.1.1. Sensing Electrodes. The "perfect" sensing electrode would monitor the (galvanic) potential
of the sulphide mineral particles in the flotation pulp. Since the particle-electrode contacts are at
random, this requires an electrode material which behaves similarly to the sulphides for the major
reagents used in flotation and an electrode shape which maximises the probability of contact
between the particles and the electrode.

Figure S. Noble metal electrodes (from left to right: black platinum wire, platinum (Pt) and gold (Au)
spirals, and finally Pt and Au foils).

Figure 5 shows noble metal electrodes used for a series of experiments. From left to right they
are: black platinum wire, platinum (Pt) and gold (Au) spirals, and finally Pt and Au foils. The
wires were 5 cm and 10 cm long for the black Pt wire and the spirals respectively. The exposed
66

surface of each foil was I cm 2. The black Pt electrode was obtained by platinising the plain Pt
electrode [18].
The main problem with the spirals is the weakness of the jOint to the epoxy casting. The problem
was solved on a laboratory scale by enclosing the electrode in a plastic tube with holes to permit
circulation of the slurry. An industrialised version of the electrode would most likely use a plastic
mesh to support the wire [18].
Mineral electrodes must be fabricated using mineral pieces from the ore. Contact between the
mineral and. for example. copper wire can be by a mercury drop. Epoxy resin is recommended to
seal the contact area between the electrode and the wire and also for a rigid support. The shape of
the mineral electrode is similar to that of noble metal foils.
For fundamental work. i.e. the study of mineral reactions. the mineral pieces have to be screened
either by visual inspection under the microscope for inclusions [19], or by rejecting electrodes
which differ greatly from the average rest potential value.
A comparison of mineral electrodes in water shows different potentials. (e.g. Table 2); when
immersed in the same mineral slurry. the electrodes yield much closer potentials.

TABLE 2. Potential for mineral electrodes in water at pH 7.0 and in mineral

slurries [18].
E (mV vs. SHE)
Mineral In water In Cu con. slurry
Chalcopyrite 195 443
Pyrite 278 415
Galena 144 429

Kastening et al. [20] showed that for single particle-electrode contact the ratio of the particle
contact area over the electrode area will influence the quality of the contact. For a large ratio (2 to
10). the contact approaches that of an ideal on-off switch. As the ratio decreases (1 to 0.5; i.e. the
electrode becomes larger. or the particle becomes smaller). the contact approaches that of a
switched R-C circuit. The quantity of charge exchanged between the particle and the electrode is the
same in both cases. However. the charge (electrons) will be exchanged over a longer period of time
for contacts with low particle area to electrode area ratios than with contacts with high ratios. As the
number of contacts per unit time increases. the equivalent R-C circuit for the former type of
contacts discharges to a lesser extent between contacts and eventually tends towards the ideal on-
off switch.
Figure 6 shows the typical response of sensing electrodes to an addition of solids. The response
is in line with this interpretation. The spirals have a fast response (on-off circuit) while the foils
have a slow response (switched A-C circuit). The changing potential with time observed for the
spirals is due to equilibration of the solids-solution system.
Verification ofthe performance of diverse noble metal electrodes in a flotation circuit has been
performed [21]. The same general observations were made with the exception that in the cleaning
section of the circuit. the differences between the electrodes were less acute than in the laboratory.

3.1.2. Reference Electrodes. The quality of the reference electrode is as important as the sensing
electrode. although previously reported verification procedures ascribed all measurement problems
to the sensing electrodes [22.23.24].
The physical distance between the reference electrode and the sensing electrode is important in at
 67

least two situations. First, when the measurement is made in extremely dilute systems, the small
number of ions in solution causes a large potential drop between the reference electrode and the
sensing electrode, even though the current involved during the measurement is very small (10- 10
amps). For measurements in dilute systems, the measuring components (voltmeter, analog to
digital convertor) must have the highest input impedance possible to minimise this current. Second,
during electrochemical investigations on the nature of the reactions occurring at the surface of the
metaVmineral (e.g. cyclic voltammetry), the senSing electrode is also a working electrode with
currents which may be up to 100 amps/cm 2. The potential drop due to the resistance of the solution
(IR drop) is minimised by bringing the reference electrode as close as possible to the sensing
electrode through a Luggin capillary [12].

Spirals

11----1
:.
Time

j Foils, minerals
! 1-----'
:.
..
Time

Figure 6. Typical responses of the sensing clectrodes to an addition of solids.

Figures 7 to 11 show different types of reference electrodes for industrial applications (mainly
for combination pH electrodes). Laboratory and single reference electrodes do not differ much
from these in terms of design.
One way to reduce the problem of electrolyte contamination is to buffer the reference electrode
from the system solution with an inert electrolyte, which requires two junctions to separate the
solutions (Figure 9). The double junction design is also used in laboratory investigations when it is
important to minimise the contamination of the sample investigated by KO.
The usual electrode construction is shown in Figure 7. Commonly the electrolyte is saturated
KO solution. Ions can freely diffuse from the electrolyte into the system and vice versa. After an
extended immersion, the electrolyte becomes depleted in KCI and/or loaded in foreign ions (Le.
becomes contaminated). If contamination of the electrolyte (yielding erroneous potentials) is
suspected, the KCl solution should be replaced with a fresh one. This is not possible with the gel
type electrolyte (the most recent design); a contaminated reference gel electrode must be discarded.
For dilute solutions it is important to diminish the streaming potential (due to the directional
diffusion of ions in a flowing system) developed between the pH selective membrane (or in our
68

case the noble metal or mineral electrode) and the reference. This leads to the particular example
shown in Figure 8, where the junction is around the pH membrane.

KCI
.Iurry 'U)--':::

L wooden or ceramic
liquid Junction

Figure 7. pH reference electrode.

--
Agi Agel element - f--- KCIIAgCI gel

L liquId)unction

Figure 8. Annular liquid junction gel filled reference electrode.

Sliver element KClgel Intermedlat. ammonium nltr.t. replac •• ble porou.

coated with Agel (NH .. NO~) gel ceramic jwtctlon
Junction

Figure 9. Double-junction diffusion type reference electrode.

 69

Another way to reduce the diffusion of ions is to use a "solid" electrolyte with multiple junctions
(Figure to).

I.'.,
u .. ,yolr

In.'....tlon
nowmountl"l

,1
• ft.
I.,.plcal

.' L dI.eh.fVI to
,tmolphl"
p, ... ure

Figure 10. Solid state permanently charged reference electrode .

••nlor
r.f.t,nee
Ih.rmocompln .. tor -0
'\

-f--' pox.,."""
I~

I - - "'11 AgCl ellment

woodin dow
".-
- .poxy birr'"

.. t . -
~I----- woodin plu. (4)
(KCI ,,'uut,d)

( )
4 - .... .....
Figure 11. Laboratory porous sleeve electrode.
70

Last (Figure 11), the laboratory porous sleeve which provides a high flow of electrolytes from
the electrode, reducing the risk of contamination by ions present in the solution, may be emulated
in industrial applications. This requires a relatively large reservoir of electrolyte. This reservoir
must be pressurised if the electrode is deeply immersed.
Most modem designs reduce or eliminate the risk of contamination of the electrolyte of the
reference electrode. Except for the double junction electrode (Figure 9), they do not help in
reducing the risk of having the junction clogged by deposits formed by reaction of the ions in the
solution and the electrolyte. AgjAgO reference electrodes are more prone to junction blocking than
SCEs due to the low solubility of most silver compounds. A clogged porous junction may give rise
to non- negligible junction potentials, thereby giving erratic readings.
A contaminated reference electrode may also introduce memory effects from solution to solution,
similar to a poisoned sensing electrode. These memory effects are more visible when performing
the measurements in weak (dilute) redox systems such as the process water in a concentrator, as
they have a very limited buffering capacity.
Details on pH and redox buffers for testing electrode pairs can be found in the ASTM standards
[25]. Also, the papers by Natarajan and Iwasaki [23,24] and Zhou and Chander [26,27] provide
details on testing procedures for memory effects for the electrode pair. A refined procedure for
testing the reference electrode independently from the sensing electrode can be found in [21].
Contrary to pH measurements, potential measurements cannot be automatically compensated for
the temperature of the solution or the slurry. This is due to the presence of mixed potentials in
complex systems and to ignorance of the main redox reactions present. The temperature of the
slurry in the plant will vary with the season (e.g. +2 to +300C for typical Canadian mills). While
the actual measurement cannot be compensated for temperature, the interpretative system (e.g. E-
pH diagrams) can be calculated at different temperatures.

3.2. ASPECTS OF PROBING AND SAMPLING

The methods of measurement fall into two basic categories: measurement in a sample of the pulp
(Le. off-line) and direct probing of the process pulp (Le. in-pulp), which can be discontinuous or
continuous (i.e. on-line).

3.2.1. Off-Line. In this approach, a sample is taken from the pulp and brought to the measuring
instruments. In this situation sample stirring may be optimised to achieve reproducible values. If
the measurements are made within a few minutes (e.g. < 5) of sampling, the measured value for
the temperature, for example, will be close to that of the process pulp. Unfortunately, this is not the
case for pH, E, and the dissolved oxygen content (DO). The oxygen demand of the pulp will cause
a consumption of the oxygen present in the sample, with the consequence that the measured pH
will be greater and the measured E and DO will be smaller than in the original process pulp. A
technique to measure the oxygen demand of the pulp was based on this observation [31]. Table 3
shows an example of the "error" on the pH introduced by this off-line technique, compared with
in-pulp.
It is possible to take measurements of the solution potential, Es, by filtering the sample [32]. It is
unlikely, however, that the solution thus presented to the probes is electrochemically the same as
that in the original slurry.

3.2.2. In-Pulp. Rather than bringing a sample of the process to the instruments, the instruments
are brought to the process. This requires portable instruments (pH-meters with automatic
 71

temperature compensation, E-meters, DO-meters) which are now readily available and at modest
cost.

TABLE 3. Effect of pulp sampling and sample transport on pH.

J:H
Sample In pulp Off-line (in lab.)
A 7.15 7.47
B 7.10 7.50
C 7.17 7.50
D 7.14 7.45
E 7.18 7.52
Notes:
1) About 5 minutes walk between the sampling point and the laboratory.
2) The probes were of the same type and the pH.meters were calibrated using the same
pH buffers.
3) The pH measurements were made using the automatic temperature compensation feature
of the pH·meters.

The advantage of this approach is that the measurements made are representative of the actual
conditions in the pulp. There are, however, some difficulties. First, it is not always easy to ensure
that the electrodes are actually immersed in the pulp and not just in the froth. The density of the
froth is often high enough to cause the electrodes to "float" in the froth layer. Attachment of
weights to the electrodes often helps. Also, the electrodes must be well sealed to take full
immersion in the pulp. Laboratory electrodes, contrary to their industrial counterparts, are often not
suitable for full immersion. The exit of the electrode cable and the integrated plastic top holder
should be sealed using silicone. A stringent test for immersability is full immersion in a 10 gil KO
solution. If a pH 7.0, or a E =0, is measured (indicating short circuiting of the electrode output),
then the electrode cannot be used in this fashion without additional sealing.
Second, there is always the possibility of poisoning of the sensing electrode, or the presence of
memory effects when going from one process point to another. The experience acquired during
some 60 plant surveys at both Les Mines Selbaie and Kidd Creek Mines indicates that this is not a
major problem if precautions are taken prior to the survey.
The sensing electrode must be thoroughly cleaned (e.g. mechanical polishing using 400 or 600
grit emery paper, followed by rinsing with process water) prior to each survey, if these are made
several days apart. For intervals less than 2-3 days, or prior to changing circuit (e.g. going from a
copper to a lead circuit) the electrodes may be soaked in dilute (5%) HO followed by a process
water rinse. Electrochemical cleaning [26,27] is as effective as mechanical polishing and is more
convenient and reproducible. Oeaning with dichromate solutions must be avoided as chromate ions
readily adsorb.
Measurements in the process water(s), must be performed prior to those in the process itself.
Ideally, the electrodes should be immersed in a pH buffer with a pH close to that of the process
water. Memory effects (originating from the sensing and/or the reference electrodes) are likely to be
observed otherwise, because process water is a very dilute redox system compared to process
pulps. It is strongly recommended to perform measurements in prepared solutions, such as those
suggested by the ASTM standard [25] and ~ a record of these measurements.
The third difficulty with in-pulp point potential measurements is that flotation circuits rarely
operate under steady state conditions. Since each survey may take several hours to complete, the
72

tacit assumption is that the circuit was operating at steady-state during that whole survey. It may be
unrealistic to expect to obtain sufficiently reliable data from plant surveys to determine relationships
between the metallurgical response of the ore and the pulp potential.

600 600
• ., • • \
\
\02(AO)
\
•• \02(AO)
\
H 2 O"",, • H 2 O\.
••
500 500
\ '\,
\,
\. \\
400 0 400
0
0
0

> >
0
0 0

! 300 Fr.lh wat.r

! 300 Fr8,h water o'C&
...
W
• wate,
•41. ...
W
• water
* rod mill * rod mil
• CuPbmlddl.
~4;. • CuPb mlddl. j.
200 • Zn 200 • Zn .~
•
Reul. water ReUI.wat.,
* ~:
o wat.r * :,,; o water
"'"*+
100 + rod min 100 + rod mil + * *
6. CuPb mlddl. 6. CuPbmlddl
o Zn o Zn

0 0
4 6 8 10 12 14 4 6 8 10 12 14
pH pH

Figure 12. (a) E/pH graph for October campaign. (b) E/pH graph for April campaign (ref. [32]).

600

400
I FLOTATION UNIT 4.
HFLOTATIONUNIT24 ,I
Q FLOTATION UNIT. IfH
F FLOTATIONUNCC 1 /ti eA
'1,~Q
E 2" CYC OIF Ir.nlilion
region l' A RMD

(;~
B ," BMD
-200 C ," CYC OIF
H· D 2" BMD
-400 o
o 2 4 10 12 14
pH

Figure 13. E/pH diagram for the grinding and flotation sections of a copper concentrator (RMD - rod
mill discharge, BMD - ball mill discharge, eyc OIF - cyclone overflow) [73].
 73

The easiest method to analyse the data obtained during the plant surveys is to plot the values on
E-pH (or Pourbaix) diagrams. If the flotation system has been the subject of extensive fundamental
investigations, it might be possible to intcrpret the E-pH values measured in relation to the known
(electrochemical) reactions in the system.
Unfortunately fundamental investigations have not been performed even for simple binary (e.g.
Cu-Zn) ores. The E-pH values obtained in the survey thus will not be interpretable. On the other
hand, this will indicate how the circuit operates. For example, what are the potentials at the start
and end of the roughing bank? Are the potentials measured in the laboratory similar to those
observed in the plant? Is there sufficient oxygen (air), or too much, going in the flotation cell to
promote selectivity?
Figures 12 and 13 show typical E-pH diagrams obtained in plants. In Figure 12, the solution
potential obtained by microfiltration of pulp samples is plotted against pH. It becomes apparent that
the points form a line parallel to the 02-H20 reduction line. In Figure 13, the pulp potential
(platinum electrode) is plotted. In this latter diagram, the observed line for the process is not
parallel to the oxygen reduction line or to the hydrogen reduction line. This suggests that redox
couples other than 02-H20 are active.
The experimental oxygen reduction lines (in oxygenated and deoxygenated distilled water) are
also shown on Figure 13. These experimental lines occur at lower potentials than the theoretical
line, either due to the formation of a Pt-O couple on the surface of the platinum electrode [33] or
due to a two step reduction of oxygen via the transitory H202 [34]. Had a gold electrode been used
to obtain these experimental lines, they would have occurred at much lower potentials.
In Figure 13, the apparent operating line for flotation (points D to I) is almost vertical. Such an
operating line appears to be common for copper rougher flotation circuits operated at natural pH, as
a similar operating line has been observed at Les Mines Selbaie and Kidd Creek Mines. This
operating line was interpreted as being a meta-stable H2S/HS- line following the description by
Peters [35]. It is somewhat troublesome to observe this H2S/HS- line and not a chalcopy-
rite/collector line.
On E-pH diagrams, all the lines presented are eqUilibrium lines unless they were calculated to
include kinetic effects (meta-stable lines) [35]. The standard diagrams for metals in water are given
by Pourbaix [36]. Some E-pH diagrams for minerals are beginning to appear.

3.2.3. On-Line. Such techniques are the foundation for the long-term objective of monitoring and
control ofthe electrochemical potential(s) in flotation. There are three main techniques to monitor
electrochemical potentials continuously in mineral slurries. The first is similar to industrial pH
measurements: sensing electrodes are immersed in the pulp at the desired location in the process.
The two other techniques have been developed around xanthate monitors and involve taking a
sample from the pulp and pumping it to the analytical instrument.
The dividing line between the three techniques is whether or not it is desired to monitor also the
collector concentration and whether or not the instrument is available. As regards instruments the
Elexan 80 (xanthate concentration and/or pulp potential monitoring and control) has been publicised
since 1985 [29,30]. The PIRANA (collector concentration and solution potential monitoring and
control) has also been discussed for some time [37]. To date, however, these instruments are not
generally available and have not been independently evaluated. Hence, it remains necessary to use
what is available - general purpose electrochemical potential monitor/controllers obtainable from the
same suppliers as pH monitor/controllers.
For monitoring reagent concentration, the accepted control philosophies for reagents, and in
particular collectors, aim at operating at starvation levels of reagent to maximise metallurgical
74

perfonnance and minimise reagent costs. If these control philosophies are correctly implemented,
there will not be much free collector to be monitored in the pulp. On the other hand, no control
scheme is perfectly tuned so that some collector residuals may be present.
Figure 14 shows schematically the Elcxan 80. The design of the pulp sampling system is similar

process stream

from
primary
sample,
(-1001/mln)

rn rn 000

t
mains

Figure 14. Schematic view of the Elexan 80 (Outokumpu).

to that of the Courier line of on-stream X- Ray fluorescent analysers. The initial design had long
lines between the primary sampler and the measuring cell. The revised design has very short lines
and in-pulp measurements have been considered.
The electrodes face the flowing pulp sample and one of the main (present) problems is the
abrasion of the electrodes. Mineral electrodes for potential measurement are usually made from
pieces of the mineral being floated. In the case of a zinc circuit (where it is not usually possible to
make a sphalerite electrode), the major iron sulphide gangue mineral may be used as the mineral
electrode (to control the depression of this undesired gangue mineral). This apparently has been
used with success to control the copper sulphate addition rate for optimum sphalerite/pyrrhotite
selectivity [38]
The xanthate concentration monitor is designed from a standard electro-analytical technique:
stripping voltammetry. In this technique, the current passed during the adsorption/desorption of the
collector onto the measuring electrode indicates the original concentration of the collector in the
solution. The collector must adsorb through electrochemical mechanisms to be monitored in this
fashion. A synthetic chalcocite electrode is used as the measuring electrode for the xanthate
concentration monitor [38]. The technique is reported to be insensitive to the presence of
electrochemically inert solids and calibration can be perfonned to take into account the presence of
electrochemically active solid particles [38].
It should be noted that potential monitoring and xanthate concentration monitoring cannot be
made within the same measuring cell due to the application of a potential for the laner measurement.
The PIRANA developed by BRGM (France) is built around a UV- spectrophotometer for
 75

collector concentration measurement [37]. Figure 15 shows the pulp sampling scheme and the
internal mechanism of a typical UV-xanthate monitor. Since UV spectrophotometry requires
samples free of any suspended solids, a solution sample is obtained from the pulp continuously by
micro-filtration (pore diameter < 1 JlIIl).

detector 1
Incld.nt 'adlatlon

signal processor

" 'ran.mi ••lon

0-10V
Illt,at.

4-20mA
control signal

..... air
pvc 50-400 kPa
probe
bod~

to drain

removable
lilt.,
cartridge

Figure 15. (Top) Operating principles of xanthate monitor. (Bottom) Schematic diagram of xanthate
monitor [37].

The main question in this case is whether the solution potential is related to the actual electro-
chemical conditions in the mineral slurry. It was shown that the selectivity of a copper/Zinc flotation
circuit could be modulated by controlling the potential at fixed values during pilot plant runs
[37,38]. However, some workers have expressed doubts on the long term viability of this
approach [39].
76

Results using black platinum electrodes may be relevant in this regard: the porous nature of such
electrodes provides an area exposed to the solution much greater than the one exposed for collision
with mineral particles and these electrodes consistently exhibit potentials which are at least 100 mV
more positive than noble metal and mineral electrodes and remain relatively insensitive to changes
in the system. They are thought to be monitOring the solution potential Es selectively [40]. In
contrast to a reported need to modulate the solution potential, Es, by 200 m V to have a significant
impact on Cu{Zn separation [37], in the case of the A 1 zone ore of Les Mines Selbaie, laboratory
and plant investigations showed that a modulation of the pulp potential, Ep, as small as 50 m V was
sufficient to change the response of sphalerite in the ore from depression to flotation during copper
roughing.
From the previous discussions it may be inferred that the Elexan 80 offers an approach which is
preferable over that of the PIRANA since the former measures Ep rather than Es. On the other
hand, as for off-line measurements of Ep, the time elapsed during the transport of the sample from
the process point to the analyser may be sufficient for complete consumption of the dissolved
oxygen present in the pulp thereby giving measured potentials which are lower than they are
actually in the pulp.
Regardless of the technique taken there will always be the question as to where the sample must
be withdrawn from, or where the electrodes must be immersed. This may be partially determined in
the laboratory by continuously monitoring the potentials during flotation tests. The final answer,
however, must be obtained in the plant.

RCu '/.

99.3

2.1
1.t

Figure 16, Response of a copper ore LO pH and Ep [42).

 77

As a hint for the field of applications opened by monitoring E on a continuous basis, Figure 16
shows an interesting response surface of copper recovery vs. Es (as monitored with the PIRANA)
and pH. There is a minimum in copper recovery at pH 9 and Es around -600 mY. Interestingly,
most copper flotation circuits are operated at pH 9 or so. It is evident from Figure 16 that Es (or
Ep) alone is not sufficient for process optimisation, neither is pH alone.

3.3. CONCLUDING REMARKS

The electrochemical potentials measured in flotation systems are mixed potentials. The practical
implication of this situation is that they cannot be readily related to fundamental (thermodynamic)
properties for the minerals and collectors, unless the kinetic data pertinent to the system are known.
Empirical relationships can be obtained between the pulp potential and the flotation response of the
ore. There may be difficulties in adapting these relationships to variation in process variables such
as ore type.
E-pH mapping ofthe flotation circuit may provide additional information on the operating regime
of the circuit and the predominant species involved in the process. If pulp heating is performed as
part of the process, E-pH-T mapping should be used. The eqUilibrium and meta-stable lines ofE-
pH diagrams required for interpreting the process E-pH (or E-pH-T) map can be calculated for any
temperature with the aid of thermodynamic analysis programs.
Gold spiral sensing electrodes appear to give the most reliable results for the measurement of the
pulp potential. They could be considercd as general purpose electrodes. The applicability of a given
noble metal electrode (material and shape) should be verified in the flotation circuit(s) against a
mineral electrode of the main sulphide being floated, if it is possible to obtain a mineral electrode.
The solution potential obtained by micro-filtration of the pulp or by using black platinum
electrodes is less sensitive to changes in the system than the pulp potential. A greater modulation of
the solution potential is required to affect the flotation response of the ore.
Finally, some of the practical details of obtaining pulp potential measurements have been
discussed for laboratory and plant applications. As for any instrumental technique, it will be
successful only if preventive maintenance of the sensor is implemented; and beware of commercial
gold electrodes as they may consist of only gold plating onto another metal (usually platinum).

4. Electrochemical Effects of Importance

4.1. PULP POTENTIAL AND FLOATABILITY

Since it is now well established that sulphide minerals interact with water and with flotation
collectors largely by charge transfer mechanisms and every such reaction corresponds to an
electrochemical potential, this suggests the possibility of controlling flotation selectivity by
manipulating the potential of the pulp. The pulp potential can be varied either by chemical methods
or by an external power source.
Chemical methods control the potential either through the activity of oxygen in the pulp, or by
adding suitable oxidising agents (to make the potential more anodic) and reducing agents (to make
the pulp potential more cathodic). The activity of oxygen in the pulp can be varied by the oxygen
content of the flotation gas - nitrogen enrichment to decrease the activity and oxygen enrichment to
raise it. The practice of adding modifying agents (depressants/activators) to control selective
flotation was largely established prior to any electrochemical interpretation of their role.
78

Direct control of the potential by an external power source has been suggested in recent years
based on laboratory testwork: it is given the name electrochemical flotation. In order to apply
potential to a mineral bed, a 3-electrode cell similar to the one in Figure 2 is used [43]. An example
of the investigations on flotation with varying applied potential follows. Aoditional examples may
be found in [44,45].

20
100 N
E
:::,
• chalcocite 16 0
o pyrlt. E
80 ::J..
.~
UI

~ 60
12 lii
0
>- c:
i
a...
0
>
0
u 8
0»
II: 40 0
UI
"0
c(
E
20 4 ~
1'0
a.
a.
c(
0 0
-400 -200 0 200 400
Potential mV (SHE)

Figure 17. Flotation recovery and etJ!yl xanthate adsorption as a function of potential for single mineral
chalcocite and pyrite beds at pH 9.2 [46].

4.1.1. Flotation o/Chalcocite (CU2S) and Pyrite (FeS2). Electrochemical flotation results for these
two minerals are shown in Figure 17 [46]. The figure also shows the results of xanthate adsorption
at these minerals. For chalcocite the flotation commences at -0.25 V and reaches 100% at -0.05 V.
A good correlation is observed between xanthate adsorption on chalcocite and the flotation recove-
ry. Similar observations are recorded for pyrite except for a decrease in xanthate concentration,
which is attributed to the formation of a new species called perxanthate. The higher potential for the
flotation of pyrite corresponds to the formation of dixanthogen as noted before.
The situation differs significantly when the two minerals are present together (Figure 18). In the
presence of chalcocite, flotation of pyrite is promoted at potentials between -0.15 V and +0.15 V,
where no flotation in the single component case occurred and is depressed above 0.15 V, where
flotation recovery reached 100% with the single mineral. The flotation of chalcocite from the
mineral mixture occurs in a similar potential range, but it is depressed to various degrees as
compared to chalcocite flotation as a single mineral.
Below -0.30 V, chalcocite undergoes reduction following the reaction.

(14)
 79

The HS- produced acts as a depressant lowering the flotation recovery of chalcocite. The flotation
of pyrite in the range -0.15 to +0.15 V was explained by the action of cupric ions generated from
the oxidation of chalcocite

(15)

Pyrite flotation from the binary mineral system was also observed when the mixture was pre-
conditioned at a potential of -0.35 V. This was explained by the action of HS- generated by the
reaction (14). The HS- ions are oxidised forming elemental sulphur which makes pyrite
hydrophobic.

100

..
• chalcocite
o pyrll.
-olngl. mlner81
80 minerai /"

~ ~!~~'::t .' . -_ .. -
'
>-
:. 60
----\1Iopen !
circUli t .. ____ •.....
j 40
---121 -0 35;

j
j ••• ~~~.
•••••

I:
!i
ji
20
il
-200 o 200 400
Potential mV (SHE)

Figure 18. Flotation recovery of chalcocite and pyrite at pH 9.2 as a function of potential for single- and
mixed-mineral beds [46].

4.1.2. Enhancing Flotation Recovery by Applied Potential. A recent study by Roos and cowOlxers
[47] using a modified Denver laboratory flotation cell, showed that applying a potential enhanced
flotation for a low concentration of collector (100 mg/l potassium ethyl xanthate). The flotation
recovery of covellite increased from 35% to 70% at an applied potential of 445 mY; that of
chalcopyrite rose from 30% to 45% at an applied potential of 345 m V. Further increases in applied
potential resulted in decreased recovery. This was explained by the decomposition of Cu-xanthate
species at the higher potentials.
It was observed [47] that there was a large difference in the potential between the flotation of
chalcopyrite and of covellite. The onset of the floatability of chalcopyrite occurred at -55 mY. This
is more cathodic than the potential for xanthate-dixanthogen equilibrium (+ 130 m V). This indicated
that oxidation to dixanthogen was not a significant factor for chalcopyrite flotation. For the flotation
of covellite the potential should be more anodic than 100 mY. The flotation of covellite increased
sharply until +450 mY. This indicated the influence of dixanthogen.
80

(A] (1) Exc ••• electron. r •• et with

dle.Dlved O. and H,O to
produce OH-

(2) Pr ••• nce d. OH'" le.d. to

formation of hydrophlnc

mild at.el
media

IBJ

with minerai

(2) Electron deficient F.··

r•• ct. to form lolubl.
oxide. and hydroxide.

Fe --+ Fe·· + 2e

Figure 19. Mechanism of galvanic interaction - a schematic representation [49].

TABLE 4. Rest potentials of sulphide minerals at pH 4 [5 o).

Mineral Rest p<>tential vs. SHE/volt
Pyrite 0.66
Marcasite 0.63
Chalcopyrite 0.56
Sphalerite 0.46
Covellite 0.42
Bornite 0.40
Galena 0.28
Argentite 0.28
Stibnite 0.12
Molybdenite 0.11

4.2. GAL V ANIC INTERACTIONS, GENERAL

As noted, flotation of a mineral from a mixture is often not consistent with the flotation of the same
mineral as a single species. This is due to electrochemical interaction between the two minerals.
These are called galvanic interactions [48). They occur when the minerals (or minerals and a metal
like iron present in the grinding medium) differ in electrochemical activity and form a galvanic cell.
 81

The mineral with the higher rest potential acts as as a cathode, while the mineral with the lower rest
potential is the anode (Figure 19).
Among all the sulphide minerals, pyrite has the highest rest potential (fable 4). The mineral may
be regarded as electrochemically the least active or most cathodic. Electrons flow from a less catho-
dic mineral to a more cathodic one. Pyrite thus serves as an electron acceptor. The less cathodic
mineral loses electrons and, for example, the sulphide is oxidised to sulphur. The formation of
elemental sulphur by oxidation of galena, sphalerite and chalcopyrite has been found to be enhan-
ced in the presence of pyrite at pH 2 [50]. This could reduce the selectivity in the flotation of these
sulphide minerals. The electrons drawn by pyrite (or the more cathodic mineral) ultimately transfer
to oxygen present in water, forming OH- ions. One way of hindering galvaniC interaction is to
increase the pH. The other way is to lower the activity of oxygen in water, thus depriving the
system of the ultimate electron acceptor.
An example of galvanic interaction is shown in Figure 20. This shows that the flotation of pyrite
as a single mineral is not affected by aeration, but when pyrite is in contact with pyrrhotite
decreasing flotation recovery is observed with aeration [51]. Electrons are drawn from pyrrhotite to
pyrite and transferred to oxygen. The hydroxide ions formed compete with xanthate and decrease
the hydrophobicity of pyrite.

100 r-----------------------------~

~90~~~~--~~--------__
>
II:
W
>
8w 80
II:
W • pyrite
~ 70 o Py -Po
>
a..

600-L-----~3~0~----6~0------~9LO------12~0~
AERA TION TIME, Min.

Figure 20. Flotation recovery of pyrite as a function of aeration time in the absence and in the presence
of pyrrhotite [51].

4.3. EFFECfS OF GALVANIC INTERACfIONS WITH GRINDING MEDIA [52-54]

The second main result of galvanic interactions is reflected in the effect of grinding media on tre
floatabilities of sulphide minerals. This has also been investigated by electrochemical measure-
ments.
During grinding, intimate contact is established among the minerals in the ore and between
minerals and the grinding media. In complex sulphide ores this leads to multi-electrode galvanic
cells. The galvanic interactions in such a system comprising two minerals and the grinding medium
are represented by an electrochemical model shown in Figure 21. The three solid phases act
according to their respective position on the rest potential scale. The following reactions occur:
82

1(2 02 + H20 + 2e =20H- (at the cathodic mineral) (16)

MeS =Me2+ + S + 2e (at the anodic mineral) (17)

Fe = Fe2+ - 2e (oxidation of the steel media) (18)

Me2+ + 20H- = Me(OHh (19)

Fe 2+ + 20H- = Fe(OHh (20)

The elemental sulphur formed can cause collectorless flotation and affect the selectivity of flotation.
The deposition of iron hydroxides on a mineral surface may have an activating or depressing effect
on flotation. The nature of the grinding medium can significantly affect flotation. (See also section
5.1).

Fe··

Figure 21. Model of galvanic interactions among two minerals and grinding media [54].

s. Electrochemical Principles in Plant Practice

5.1. EFFECT OF GRINDING MEDIA

As the flotation of a mineral is largely governed by pulp potential, the grinding media used often
have a maIked effect on flotation response and selectivity. The pulp potential varies depending
upon the chemical nature of the grinding medium. Where the grinding medium consumes oxygen it
lowers the pulp potential; a typical example of potentials around a grinding circuit is shown in
 83

Figure 22. The values will change if either grinding media type is changcd, or grinding circuit
configuration. For a given situation the values are often quite constant.
An example of the variation of grinding medium is shown in Table 5. The results suggest that
chalcopyrite was floated without collector under the more oxidising conditions associated with a
high chromium steel medium. This necessitated changing the proportion of collector added to the
rougher, scavenger and the first cleaner. Some examples of the effects of grinding media in specific
systems are discussed in the following sections. A detailed review of this subject has been recently
published [49].

TABLE 5. Composition of grinding media and collector distribution in a copper

circuit (A 1 zone ore of Les Mines Selbaie). Monthly collector distributions are
rounded to nearest 10%
% Collector to
Month Grinding Media Rougher Scavenger Cleaner
September 87 100% ca~t steel 20 40 40
April 87 100% cast steel 30 40 30
June 87 100% cast steel 20 40 40
August 87 33% high Cr 15 35 50
April 88 100% high Cr 20 20 60
August 88 100% high Cr 15 20 65
April 89 100% high Cr 0 60 40
August 89 100% high Cr 0 55 45

Feed

0.51 +130 9.2

Rod eall 2.04 +5 9.1
Mill Mill 3 91.3 +410
4 51.2 ·,85 9.3
5 57.7 +240 9.3
8 91.3 +270 9.3

Figure 22. Pulp potential, dissolved oxygen (DO) (per cent at saturation) and pH at selected sampling
points.
84

5.1.1. Galena. Most investigators [55,56] have observed that galena flotation using sulphydryl
collectors is faster following autogenous grinding or grinding with stainless steel than that
following grinding with mild steel. The promoting effect of stainless steel and autogenous media
has been related to a higher pulp potential. Flotation of galena inside a grinding circuit may be
difficult because of this.

5.1.2. Pyrite. This mineral floats best at higher pulp potentials due to the formation of dixanthogen
at the surface. Other factors attributed to the grinding medium have been found to influence flota-
tion. Galvanic interaction between the abraded iron and the mineral may cause formation of
elemental sulphur leading to collectorless flotation. The results reported by several investigators
[57-59] are not in mutual agreement.

5. 1. 3. Chalcopyrite. Autogenous or stainless steel grinding improves the flotation of chalcopyrite

with sulphydryl collectors. A correlation was identified between recovery, dissolved oxygen and
potentials: these are considered dominant factors in chalcopyrite flotation [60,61].

5.2. COLLECfORLESS ROTATION

One of the significant findings of electrochemical research on sulphide minerals is the recognition
that under certain conditions sulphide will be oxidised to elemental sulphur. Under certain, not well
defined conditions, the sulphur reacts with sulphide to form polysulphide species:

(x - 1) S + S2- -> Sx 2 - (21)

or metal deficient sulphides. This is most prominent at pH > 8, while elemental sulphur is
dominant at pH < 6 [62].
Any of these species cause the mineral surface to become hydrophobic to a certain degree. This
is the basis of collectorless flotation. (This should be distinguished from inherent floatability
exhibited by a few sulphide minerals like molybdenite, that is explained by their distinct surface
composition and structure [63]).
Collectorless flotation has been reported from time to time in industrial flotation [16,17], but it
has only recently attracted concerted attention, as the underlying electrochemistry of sulphide
flotation has become better understood.

5.2.1. Chalcopyrite. The cyclic voltammogram of chalcopyrite in water shows a strong anodic
wave (Figure 4). This is identified with the oxidation of sulphide to sulphur. This causes tre
chalcopyrite surface to become hydrophobic. One of the earliest studies on the collectorless
flotation of chalcopyrite was by Heyes and Trahar [64] who showed the influence of the oxidising
atmosphere. When ground in a glass mill chalcopyrite acquired hydrophobicity, but if ground in an
iron mill it did not. This was explained by the reducing conditions in the iron mill. Oxygen or an
oxidising agent is essential for the mineral to acquire hydrophobicity: for example, chalcopyrite
ground in an iron mill but with oxygen or an oxidising agent like NaOCI acquired hydrophobicity
[64].
Measurements of pulp potential under the conditions required for natural floatability show a clear
correlation between the potential and recovery of chalcopyrite (Figure 23). Flotation is observed
above about + 100 mV (SHE). This suggests that a certain degree of oxidation is essential. Since
the value of 0.1 V is very close to the potential of the charge transfer reaction of the oxidation of
 85

100
~--Q-B
90 __ potential •• t directly to It.
flotation valu. G!
%
~ ---potentlallnltlany reduced to /
:::i! 80
prevent flotation <equivalent I
I
~ 70 to grinding In an Iron min) I
W I
l- I
i'C 60 I
>
Do. /
I

0 50 I
U I
~
« /
I
:I: 40
U /
LL I
0 30 I
>
a: cp
I

W 0
>
0
20
I
I

U I
W
a: 10 ~
, I

G-------
o L-~ __~~~~~__~~__~~__~~~
-600 -400 -200 o 200 400 600
POTENTIAL mV (SHE)

Figure 23. General relation between applied potential and recovery of chalcopyrite [64].

100 r - - - - - . - - - - - - - - - - - , 2 0

80
15
t
>
a: 60
w
>
0
u ,,[J
w
a: 40 GRADE ...
"
...
b .8'
5
20
'8" 0.

'--____'--____'--____'--____'--__---'0
0
0.5 1.0 1,5 2.5
Na2S ADDED (kg/ton)

Figure 24. Collectorless flotation of chalcopyrite with sodium sulphide [65].

sulphide to sulphur, it is further evidence of induced floatability caused by the formation of

elemental sulphur on the mineral surface.
The collectorless flotation of chalcopyrite has been studied in detail with some ores [65,66).
Aoatability improves with increasing potential until a certain stage, when excessive oxidation has a
86

100
90
.....
?J<.
...., 80
...
>-
CI) 70
>
0
CJ 60
...::;,
CI)
Collectorless (mini-ceil)
0
50 ... Test #1
Q)
> 40 0 Test #2
~ 0 Test #3
IV
"3 30 {\, Test #4
Test #5
E
::;, 20 With collector

•
0
plant Cu-Pb rghr bank 13/01190
10 • plant Cu-Pb rghr bank 13/01/90

0
0 4 8 12 16 20
Cumulative Zn recovery (%)
Figure 25. Cu vs. Zn cumulative recovery with and without collector (M. Leroux, unpublished work).

100

90 galena-quartz

W o ceramic mill
I-
:;)
80 o ceramic mill with eulphlde
j 70
6 I t•• lmlll
I- pH 8
1/1
!!i 60 KEX 0
U.
PPG 400
~
50
~
>
II: 40 0
W
>
0
() 30
W
II:
20

10

0
-400 -200 0 200 400
PULP POTENTIAL mV (SHE)

Figure 26. Floatability-potential curves for galena at pH 8 after grinding in different environments [67].
 87

depressing effect. Treatment with Na2S was found to promote flotation in one study (Figure 24).
This may be because of removal of hydrophilic oxidation products from the mineral surface.
A continuous minicell test of collectorless chalcopyrite flotation in plant at Mattabi Mines showed
similar Cu/Pb selectivity but improved Cu{Zn selectivity compared with the plant performance with
collector (Figure 25).

5.2.2. Galena. Collectorless flotation of galena has been observed in a narrow range of potentials
+0.2 to +0.35 V [67]. It floats over this potential range when it is ground in ceramic or steel mills.
Some of the results reported on the collectorless flotation of galena are shown in Figure 26.
Flotation is more rapid after grinding in a ceramic mill than in a steel mill. When ground in the
presence of sulphide ion, the potential range for flotation is extended by 0.2 V to more reducing
conditions.
Collectorless flotation has also been demonstrated with several other sulphide minerals under
optimum conditions [14,66,68,69]. For example, collectorless flotation of pyrrhotite has been
found to be best in the potential region 250-450 mV (SHE). In this case the floatability may be
caused by the formation of the polysulphide species Fe(OH)aS2 although the formation of
elemental sulphur is not ruled out [14].

5.3. SULPHIDE FLOTATION USING NITROGEN

When an ore slurry is saturated with nitrogen it lowers the activity of oxygen in the pulp and
reduces the pulp potential. This may be expected, for example, to depress pyrite, and this has been
demonstrated by grinding the ore in a nitrogen atmosphere and conducting the flotation with
nitrogen [70]. However, if nitrogen was only introduced at the flotation stage this led to pyrite
flotation being promoted [71].

60r-------------------------------~

• Test #,(Air)
• Test #2 (N2)
54L-________ ________
~ ~L_ _ _ _ _ _ _ _ _~

80 85 90 95
Cumulative Zinc Recovery, %

Figure 27. Grade-recovery curves for reverse zinc flotation (T to 80 °e, then S02 addition) (I. Sandoval-
Caballero, unpublished work).
88

100 ,-------------------------------,

3 ~
b=====oooo==::::::::.-----
II> -:;/

'0 0
;; r-

-1 00 L..1.---:2"'0'-:0-'--0':---'--:2:-::0'-::0,.---J'--:4~0::-0--'--6::;-0-::'0:---'--;8:-:0~0

E vs SHE (mV)

100 ,-------------------------------,

3
'"
'0 0 l"==========::::;:-
"

-100 L..1.---2LO~0--'-~0---':-2~0~0~'--:4700~""-:6~0~0~~a~00

E vs SHE (mV)

100 ,-------------------------------,

.,3
'~ 0 k=--------==::::.----
)(

-,100 L..1.---2....O-O-'---0"-----'--2-0LO,.---J'-4....
0-0--'--6-0....0--'--a-0'-'0

E vs SHE (mV)

Figure 28. (a) Voltammogram of galena with xanthate in tap water; pH 8.6. (b) Voltammogram of
galena with xanthate in Kidd Creek recycle water untreated; pH 8.6. (c) Voltammogram of galena with
xanthate in Kidd Creek rccycle water treated with active carbon; pH 8.6.
 89

Although initially surprising, this is consistent with the results of galvanic interaction studies
described in section 4.2. Pyrite is more cathodic than sphalerite, hence it draws electrons which
transfer to 02 and the OH - ions formed make the pyrite less hydrophobic. In the absence of
oxygen, i.e. in presence of nitrogen, this galvanic interaction is hir.dered and pyrite stays
hydrophobic and floats readily.
Nitrogen flotation of pyrite has been conducted in continuous minicells in plant. The results are
consistent with laboratory test results [72]. Nitrogen also offers an alternative route for reverse
flotation of pyrite from base metal concentrates. This is illustrated by the results of reverse flotation
of pyrite from a sphalerite concentrate (Figure 27). The test was conducted at 80°C with addition
of S02, the conventional procedure for sphalerite reverse flotation. Significantly higher zinc grades
were obtained with nitrogen flotation compared with air. Similar results have been reported by
Johnson [73]. At least one Canadian plant is experimenting with by-product nitrogen from a
smelter.

5.4. EFFECfS OF WATER CHEMISTRY

Electrochemical studies using cyclic voltammetry have provided insight into the effects of water
chemistry on xanthate-mineral interaction and possible effects on flotation. This knowledge is of
importance, for example, in understanding the effecL~ on flotation of using recycle water which,
untreated, often carries dissolved heavy metal ionic and sulphoxyl (e.g. S2032.) species and
residual organics.

5.4.1. GalenLl-xanthate system. The flotation of galena in one study was depressed in the presence
of certain (unidentified) residual organics in recycle water [74]. The cyclic voltammograms of
galena with xanthate in tap water, in recycle water with residual organics and in the same recycle
water treated with active carbon (to adsorb the organics) are shown in Figure 28a, b and c.
Galena with xanthate in tap water shows an anodic wave at 250 m V. After the current rise
corresponding to a charge transfer reaction (probably xanthate chemisorption at galena) the current
reaches a steady value. The voltammogram of galena with xanthate in the recycle water shows a
continuously rising anodic wave. This suggests a continuous interaction of the mineral with the
organics in the water. This will interfere with galena-xanthate interaction and helps explain the
depressing effect on flotation recovery [74].

5.4.2. Pyrite-xanthate system. Several studies have shown that pyrite is activated by metal ions
like Cu 2+ and Pb2+. These could cause inadvertent activation of the mineral and lead to its flotation
even at alkaline pH, when it is expected to be depressed.
The cyclic voltammograms of pyrite in tap water, in a recycle water carrying Cu 2+, Pb2+ and
Fe3+ and in the same recycle water treated with ion exchange resin (to remove the cationic species)
are shown in Figure 29 a, b and c. With the mineral in tap water there is only one anodic wave.
With the mineral in recycle water the voltammogram is complicated, probably due to multiple
charge transfer reactions. The current flow at the lower potentials could be explained by charge
transfer reactions involving metal ions (Fe 3+ + e = Fe 2+ and Cu 2++ e = Cu+). This is supported by
the voltammogram obtained in the recycle water treated with ion exchange resin; it shows only one
anodic wave as in tap water. The results suggest that the pyrite surface is altered by the presence of
cationic species in water.
The cyclic voltammetric studies thus establish changes in the surface chemical reactions due to
the presence of organic and inorganic impurities in water. The results are consistent with
90

100

3
II>
'0
0
)( ~

-100
-200 0 200 400 600 800
E YS SHE (mV)

200

3 0
or0
)(

--200

-400
-200 0 200 400 600 800
E YS SHE (mV)

100

3 .....
II>
'0
0
::::::---
)(
..---:

-100 ~--~2~0~0~--~0--~~2~0~0~~4~0~0~~6~0~0~~8~0~0
E ysSHE (mV)

Figure 29. (a) Voltammogram of pyrite with xanthate in tap water; pH 8.6. (b) Voltammogram of pyrite
with xanthate in Brunswick external recycle water untreated; pH 8.6. (c) Voltammogram of pyrite with
xanthate in Brunswick external recycle water treated with cation exchange resins; pH 8.6.
 91

experience in flotation [74].

5.4.3. Pyrrhotite system. Cyclic voltammetric studies [75] on pyrrhotite have shown a number of
peaks which have been identified with various redox reactions. Figure 30 shows voltammograms
of pyrrhotite in water and the effect of Ca2+, S203 2- and S042- species. In distilled water the
following redox reactions are recognised:

al: initial oxidation of FeO.9S to Fe(OHMS2)S

a2: oxidation of (S2)S to Sand S042-
q: reduction of Fe 3+ to Fe2+
C2: reduction of So to S22-
C3: reduction of sl- to S2-/Fe 2+

In the presence of S2032- and S04 2- the anodic peaks move by up to 200 mY in the cathodic
direction for reactions al and a2. In thc presence ofthiosalts the potential range of the metastable
Fe(OH)x (S2)S is broadened. This could enhance the collectorless flotation of pyrrhotite which has
been observed in the range 0-200 mY [75].

600

N
'E 400
"
j.
200
~
III
Z
~ ol-+---+'#+---l---<f--!---+-+--''''':;4.,.;~---+----+-~

!z
~ -200
a:
::l
U
-400
11.2

-600 2.02

-600 -400 -200 0 200 400 600

POTENTIAL vs SHE (mV)

Figure 30. Yoltammograms for pyrrhotite electrode: effect of lime and thiosulphate additions, pH 9 [75].

In the presence of Ca2+ ions the sulphur reduction reaction c2 is enhanced relative to Fe(OHh
reduction (reaction c 1). Polysulphide/iron reduction (reaction C3) is detectable at -850 mY. These
changes suggest that Ca2+ ions in water alter the overall rate of oxidation of pyrrhotite. This was
explained by the adsorption of Ca(OHt ions especially at sulphide sites, suppressing further
oxidation. The following reaction was suggested
 92

..
100
~.
)::.:
80 ...(.;~;~.~ ....... +.................... ! ................... ~
?!
#-
Ii
u 60
0 ..; .................... ; ... ········· .. · .. ··f··· ..
:>
U
.2,..
:G 40
>
0
U
GI
II:
:>
U 20

0
-140 -40 60 160 260 360
Pulp Potential, mV (Au vs Ag/AgCI)

20 .,

16 .. j ................... (..................... ,! ...................;... ·········· .... ·.... i·

~.

.. ...... ..... ... (. . ... . ............ ;... '..................~... ·······1··

,..
000

:G 8 i o
.~
··· .. ····1(··· !.......................................... _.. . ........•.......................... -.;
>

•8
II:
rG
4 "'a.~
.~ ,,0,
....
',jI-

o ...,.................... 0 ........ .
-140 -40 360
Pulp Potential, mV (Au vs Ag/AgCI)

Figure 31. Plant performance - AI Zone of Les Mines Selbaie, November to December 1989 (hourly
means).
 93

(22)

Such a reaction would favour the adsorption of anionic reagents at sulphidic sites. In practice this
could lead to inadvertent flotation of pyrrhotite due to the appreciable concentration of Ca2+ ions
usually found in process water. Such flotation has in fact been observed [76]. It helps explain the
difficulty in pyrrhotite rejection in pentlandite flotation in the presence of Ca 2 + ions, for example,
when lime is used as the pH regulator. A complication, however, is that such low potentials as
examined in Figure 30 are not encountered in practice: there is a limit, therefore, to the
interpretative power of the cyclic voltammetry results.

5.5. CON1ROL STRATEGIES FOR SELECfIVE FLOTATION

Electrochemical potentials measured in the pulp can be considered indicators of the readiness of the
sulphide minerals to float.
Traditionally, flotation plants have used trial and error methods for selecting reagents and
appropriate concentrations to achieve the desired metallurgy. The knowledge of electrochemistry
gained in the last two decades points to the use of the pulp potential as a process indicator, perllaps
to the same extent that pH has been used to provide a basis for process control. Development of
such an approach requires laboratory and plant investigations to determine the response of the
process to this new measured process indicator. This is illustrated by the following case study.
A systematic study was conducted on selective flotation of chalcopyrite and sphalerite at Les
Mines Selbaie. By locating the Ep electrode near the conditioner, pulp potentials corresponding to
the fresh feed to the cireuit and the recycle from the cleaning circuit were obtained.
The plots in Figure 31 show the copper and zinc recoveries to the final copper concentrate
against the pulp potential measured in the copper rougher. Except for data scattering, most likely
due to variations in feed composition, there is a correlation betwen metal recoveries and the pulp
potential. The data demonstrate the possibility of using Ep to modulate selectivity.

TABLE 6. Effect of reagents on thc pulp potential (Copper rougher of Les

Mines Sclbaies)
A) Sodium cyanide dosage set-point
Ep, mV (Au vs Ag/AgCl
NaCN (glt) Mean Std
8 8.3 2.8
15 -5.6 3.8
25 -15.8 l.9
B) Variation of pH (using lime)
pH g" mV (Au vs A~A~Cl
Mean Std Mean Std
9.02 0.02 22.7 2.6
9.57 0.12 1.7 5.1
Note: Each senes of measurements were taken durmg a "steady-state" operatIOn of
30 min.

The data in Table 6 show the sensitivity of the measured Ep to the reagent dosage most often
used by the operators to control CuJZn selectivity. An increase in NaCN or pH set points leads to
94

more cathodic Ep. A possible control strategy would be to use pH to impart large changes in Ep
and fine tune the selectivity by controlling the dosage of cyanide.

6. Promise for Plant Applications

The results of extensive laboratory and limited plant investigations related to the electrochemical
interpretation of flotation systems makes the technology highly promising for improving plant
performance. Will the promise of greater selectivity through manipulation of the potential and/or
reduced reagent costs materialise in the future? There is no doubt that the answer is yes.
As to "when" and "how", there is less certainty. So far, the technology has roainly brought a
greater understanding of the chemical variables in flotation and a new approach to investigate them.
For example, Figure 16 showed an interesting response surface of copper recovery vs. Es (as
monitored with the PIRANA) and pH. There is a minimum in copper recovery at pH 9 and Es
around -600 m V. There are two regions of nearly equal high copper recoveries: i) pH 7 to 8 and
-200 to -400 mV and ii) pH 8.5 to 10 and 0 to -50 mY. It is evident that Es (or Ep) alone is not
sufficient for process optimisation, neither is pH alone. The maxima and the minimum in Figure 16
are extremely close in terms of pH and manipulating the process from one to the other using either
pH or Es (which may indirectly change the pH) alone would be difficult.
The "when" is probably the easiest of the two questions to answer. There are three independent
factors to be considered. First, for a given company, this will be when the economic situation is
difficult either due to depressed metal prices or more stringent environmental regulations, or due to
the treatment of more refractory/complex sulphide ores. Technological improvements will be
implemented only on economic merits.
Second, availability of engineers with a basic understanding of the fundamentals and the
methodology ofthe technology will be essential. There is nothing worse than a failed first-attempt
for cooling off even the most high-technology oriented management. Lack of skills in a new
technology is the most common cause of failure in its trial implementation.
Third, the means to manipulate the pulp potential in the plant must be proven for this
environment. Neither of the two commercially available industrial instruments have been tested
outside their Originating organisations. Hence, their claims for reliability and use in control are not
independently verifiable. At the moment, the technology is roughly at the same stage as that of on-
stream X-Ray Fluorescent analysis was in the early '60s. If history repeats itself, then it will take
about five years before the instruments start delivering their promised technical/economical
improvements at 80% or higher availability/reliability.
There are two possible "hows". The first one is linked to direct manipulation of the pulp
potential using large scale potentiostats. The second relies on reagents to achieve the same means.
These reagents can be presently-used modifiers and depressants or specific redox modifiers. The
technological and economic feasibility of each method will be dependent upon the solution of
several engineering problems.
The use of plant scale potentiostats will most likely require the following engineering solutions:
i) Designated electrochemical reactors. Flotation pulps are simply Particulate Electrodes (PE).
However, these PE cannot be compacted to a high pulp density as in the Packed Bed
Electrodes (PBE). Giving a value of 100% to the current efficiency of a Massive Electrode
(ME), the current efficiency of a PBE can be estimated at 80 to 90% while that of a PE can be
conservatively estimated at 10 to 20%. Hence, one would not use a conventional flotation cell
as an electrochemical reactor. The geometry of the current feeder, the counter and the
reference electrodes will be important factors. Books have already been devoted to the topic of
 95

electrochemical reactors but it is unclear if the designs presented are portable to flotation
plants.
ii) Industrial and safe potentiostats. A typical conditioning current in a PBE experiment is of the
order of 10 to 50 mA over at least 1 min. The mass of the PBE is typically 10 to 100 g.
Scaling these figures to a medium sise processing plant (e.g. 5000 tpd) implies current
capabilities between 3.5x103 A (1 mNg min) and 1.7x105 A (5 mNg min). The difference in
potential between the reference and the counter electrode could be as high as 100 V to sustain
the current required to maintain the applied potential to the working electrode (the flotation
pulp) at its set value (which is often less than 1 V). Such electrical power has to be safely
handled. Most likely. the know-how will have to be imported from electro-refining plants
(insulation. current distribution. intrinsically safe control devices. etc.). Since present mineral
processing plants have a steel structure onto which all equipment is attached. this means that
it will be difficult and expensive to modify present operating plants for direct potential
manipulation.
iii) Scale up factors. Most electrochemical investigations presently performed report only the total
cell current (see the cyclic voltammograms previously discussed). However. this cell current
will have to be converted to current density for scale up from the laboratory to the plant
environment. Contrary to the total cell current. the current density is independent of the mass
of the sample or its surface area. In very few cases do investigators report the specific surface
areas (m2/g) of the mineral samples used. which would permit current densities for given
sample masses to be calculated.
There may be other scale-up factors to be determined. particularly for the electrochemical
reactors. Given the engineering problems to be solved for the direct manipulation of the pulp
potential. it is the opinion of the authors that the chemical manipulation of the pulp potential is the
most promising method in the short term. Also. many plants presently manipulate reagent dosages
through automatic controllers. The only difference would be that for certain reagents. the set-point
will be given as a pulp potential set-point.
Still. there are also engineering problems to be resolved in the usc of chemical control of
potential. Here are a few of the most acute ones.
i) Coupling of pH and Ep modulation. Most pH modifiers (e.g. lime. sulphuric acid) modify the
pulp potential upon their introduction into the pulp. Similarly. most Ep modifiers (e.g. sodium
sulphide) also modify the pH of the pulp. If only one variable permits efficient manipulation
of the pulp potential (and consequently the metallurgical response) then the other one can be
left floating to whatever value it reaches. On the other hand. if both pH and Ep are critical.
then there is no choice. both must be controlled. There are several advanced process control
techniques which permit to decouple inter-related control variables. The best one may have to
be determined by trial and error.
ii) Assumed independence of the recovery vs potential curve upon the potential modifying
reagent. So far. probably 90% of the investigations have been performed using direct
manipulation of the pulp potential. In the other 10%. reagents (e.g. sodium sulphide. sodium
dithionate. hydrogen peroxide. potassium permanganate) were used to achieve the desired
modulation. Although some reports show no side-effect of the redox modifying reagent.
others do show significant shifting of the recovery vs potential curve [77]. In particular. the
on-set of flotation appears to be highly sensitive to the type of redox modifying agent. The
assumed independence will be confirmed or not in the near future as more investigations are
carried out.
iii) Use of nitrogen enriched or oxygen enriched flotation gas. The first concern is related to
96

occupational health and safety. Gas confinement and recycling may be needed. The second
concern is that some sections of the flotation circuit may require nitrogen-enriched flotation
air, while others may need oxygen enriched air. This implies higher capital and operating
costs since two compressors (in the case of flotation columns) or superchargers (conventional
cells) and two flotation gas distribution systems will be required.
Finally, there is the question of supplies of nitrogen or oxygen and their required purity. In this
aspect, integrated mine/mill/smeiter/refinery complexes have a net advantage over isolated mine/mill
operations. Modern smelters often require oxygen enrichment of the feed gases leaving an ample
supply of by-product nitrogen and probably some extra oxygen. Recently one Canadian plant
introduced such by-product nitrogen for testing in-plant. It is possible that simply recycling air will
essentially create a N2-atmosphere, as 02 is consumed by the sulphides in the pulp.

7. Acknowledgements

Funding from NSERC (Canada), FCAR and CRM (Quebec), Les Mines Selbaie and Kidd Creek
Mines is gratefully acknowledged.

8. References

1. Woods, R (1984), Electrochemistry of sulphide flotation, in M.H. Jones and J.T. Woodcock
(eds.) Principles of Mineral Flotation, The Australasian Institute of Mining and Metallurgy,
Victoria, Australia, pp. 91-115.
2. Woods, R, and Richardson, P.E. (1986) Theflotation of sulphide minerals - electrochemical
aspects, in P. Somasundaran (ed.) Advances in Mineral Processing, SME, pp. 154-170.
3. Woods, R (1976) Electrochemistry of sulphide flotation, in M.C. Fuerstenau.(ed.)
Flotation, AIME, New York, pp. 298-333.
4. Chander, S., and Gebhardt,1.E. (1989) Application of electrochemical techniques in sulphide
mineral flotation, in K.V.S. Sastry and M.e. Fuerstenau (eds.) Challenges in Mineral Pro-
cessing, SME, pp. 90-108.
5. Labonte, G., Leroux, M., Sandoval, I., Rao, S.R., and Finch, J.A. (1989) Some observa-
tions on copper-zinc separation from complex ores, in G.S. Dobby and S.R Rao (eds.) Pro-
cessing of Complex Ores, Pergamon Press, New York, pp. 193-202.
6. Fuerstenau, M.C., Miller, J.D., and Kuhn, M.C. (1985) Chemistry of Flotation, SME/
AIME, New York. (a) pp. 45-90; (b) pp. 115-118.
7. Leja, J. (1982) Surface Chemistry of Froth Flotation, Plenum Press, New York, Chapters 8
and 10.
8. Yarar, B., and Rao, S.R (1985) Froth flotation, in J.J. McKetta and W.A. Cunningham
(eds.) Encyclopedia of Chemical Processing and Design, Marcel Dekker, New York, Vol.
23, pp. 454-508.
9. Chander, S. (1988) Electrochemistry of sulphide mineral flotation, Minerals Metallurgical
Process.~, 104-114.
10. Johnson, N.W., Jowerr, A., and Heyes, G.W. (1982) Oxidation-reduction effects in galena
flotation: Observations on Pb-Zn-Fe sulphides separation, Trans. IMM 21, C32-C37.
11. Guy, P.I., and Trahar, W.I. (1982) The influence of grinding and flotation environments on
laboratory batch flotation of galena, Int. 1. Miner. Process . .12, 15-18.
 97

12. Bockris, J.O'M., and Reddy, A.K.N. (1970) Modern Electrochemistry, (2 vo!.), Plenum,
New York, 1432 pp.
13. Rao, S.R., and Finch, J.A. (1987) Electrochemical studies on the flotation of sulphide
minerals with special reference to pyrite- sphalerite. I. Cyclovoltammetry and pulp potential
measurements, Canadian Metallurgical Quarterly 2Q, 167-172.
14. Hodgson, M., and Agar, G.E. (1986) An electrochemical investigation into the natural
floatability of pyrrhotite, in P.E. Richardson, S. Srinivasan, R. Woods (eds.) Proceedings
of International Symposium on Electrochemistry in Mineral and Metal Processing, The
Electrochemical Society, Inc., pp. 185-201.
15. Kelsall, G.H., and Page, P.W. (1986) Aspects of chalcopyrite electrochemistry, ibid., pp.
303-332.
16. Boyce, J.H., Venter, W.J., and Adam, J. (1970) Beneficiation practice at Tsumeb concen-
trator, in D.O. Rausch and B.C. Mariacher (cds.) Lead & Zinc, AIME, New York, vo!. I,
pp. 542-570.
17. Lcpetic, V.M. (1974) Flotation of chalcopyrite without collector after dry, autogenous
grinding, CIM Bull. hl(746), 71-77.
18. Labonte G. (1987) Electrochemical Potentials in Flotation Systems: Measurement, Interpreta-
tion and Applications, M.Eng. Thesis, McGill University, 174 pp.
19. Shuey, R.T. (1975) Semiconducting Minerals, Elsevier, Amsterdam, p. 319.
20. Kastening, B., Busscher, N., and Asskamp, U. (1989) Electrochemical polarisation of
activated carbon and graphite powder suspensions. Part III. Charge transfer during single
collisions and contact resistance, J. Eleetroanal. Chern. 2.65., 77-101.
21. Finch, J.A., and Labonte G. (1989) Technical note-verification of electrodes for pulp
potential measurement, Minerals Eng.,2, 557-564.
22. Natarajan, K.A., and Iwasaki, I. (1973) Practical implications of Eh measurements in
sulphide flotation circuits, Trans. AIME~, 323-328.
23. Natarajan, K.A., and Iwasaki, I. (1970) Behavior of platinum electrodes as Redox potential
indicators in some systems of metallurgical interest, ibid. 247, 317-324.
24. Natarajan, K.A. and Iwasaki, I. (1973) Effect of poisoning of platinum electrodes on Eh
measurements, ibid. 254, 22-28.
25. ASTM (1986) Standard Practice for Oxidation-Reduction Potential of Water, ASTM Standard
D 1498-76 (reapproved 1981), ASTM Volume 11.01, pp. 261-267.
26. Zhou, R., and Chander, S. (1989) Effect of pretreatment on Eh response of a gold electrode,
in S. Chander and R.R. Klimpcl (eds.) Advances in Coal and Mineral Processing Using
Flotation, SME, pp. 57-64.
27. Zhou, R., and Chander, S. (1990) A comparison of gold, platinum and sulphide ion selective
electrodes as sensors for Eh measurement in sulphide solutions, SME, Prcprint no. 90-156.
28. Metrohm, Instruction for use, E 512 pH-meter, Metrohm Ag, Switzerland.
29. Heimala, S., Jounela, S., Rantapuska, S., and Saari, M. (1985) New potential controlled
flotation method developed by Outukumpu Oy, Proceedings XVth Int. Miner. Process.
Congress (Cannes), pp. 88-98.
30. Sipula, H., and Rantapuska, S. (1986) Flotation control based on measurement of the
electrode potential in the pulp, Control Seminar, Falconbridge, Sudbury, May 28-30.
3l. Spira, P., and Rosenblum, F. (1974) The oxygen demand offlotation pulps, Process. 6th
Canadian Mineral Processors Meeting, pp. 73-106.
32. Forssberg, K.S.E., Jonsson, H.R., and Palsson, B.I. (1985) Full scale test of process water
reuse in a complex sulphide ore circuit, in K.S.E. Forssberg (ed.) Flotation of Sulphide
98

Minerals, Elsevier, pp. 197-217.

33. Natarajan, K.A., and Iwasaki, I. (1974) Eh measurements in hydrometallurgical systems,
Minerals Sci. Eng. Q, 35-44.
34. Sato, M. (1960) Oxidation of sulphide ore bodies, I. Geochemical environments in terms of
Eh and pH, Econ. Geol.~, 928-961.
35. Peters, E. (1984) Sulphide mineral chemistry, Paper presented at the 23rd Annual Conference
of Metallurgists, CIM, Quebec (August).
36. Pourbaix, M. (1966) Atlas of Electrochemical Equilibria, Pergamon Press, Oxford, pp. 29-
50.
37. Barbery, G., and Cecile, J.L. (1986) Instrumentation for reagent control inflotation: present
status and recent developments, in P. Somasundaran (ed.) Advances in Mineral Processing,
SME, pp. 726-729.
38. Heimala, S., Jounela, S., and Saari, M. (1988) Flotation control with mineral electrodes, in
K.S.E. Forssberg (ed.), Proceedings XVIth Int. Miner. Process. Congress, Elsevier, pp.
1713-1722.
39. Ashton, B.E. (1984) On-stream analysis, in M.H. Jones and J.T. Woodcock (eds.),
Principles of Mineral Flotation, The Australasian Institute of Mining and Metallurgy, Victoria,
Australia, p. 290.
40. Labonte, G., and Finch, J.A. (1988) Measurement of electrochemical potentials in flotation
systems, CIM Bull. BJ.(Dccembcr), 78-83.
41. Rand, D.A.J., and Woods, R (1985) Eh measurements in sulphide mineral slurries, Int. J.
Miner. Process. U, 29-42.
42. BRGM Publication (1988), Le systeme PIRANA dans l'usine de flottation de Salsigne,
BRGM, France.
43. Walker, G.W., Stout III, J.V., and Richardson, P.E. (1984) Electrochemicalflotation of
sulphides. Reactions of chalcocite in aqueous solutions, Int. J. Miner. Process. .lb 55-72.
44. Richardson, P.E., Stout III, J.V. and Walker, G.W. (1984) Electrochemical flotation of
sulphides: chalcocite-ethyl xanthate interactions, ibid. 12,73-93.
45. Richardson, P.E., and Walker, G. W. (1985) The flotation of chalcocite, bornite, chalcopyrite
and pyrite in an electrochemical cell, Proceedings XV Intnl. Miner. Process. Congress, vol.
II, pp. 198-210.
46. Gebhardt, J.E., and Richardson, P.E. (1987) Differential flotation of a chalcocite-pyrite
particle bed by electrochemical control, Minerals and Metallurgical Process. 1, 140-145.
47. Roos, J.R, Celis, J.P. and Sudarsono, A.S. (1990) Electrochemical control of chalcocite and
covellite-xanthateflotation, Int. J. Miner. Process. 29, 17-30.
48. Majima, H. (1969) How oxidation affects selective flotation of complex sulphide ores,
Canadian Met. Quart. .8., 269-273.
49. Martin, c.J., McIvor, RE., Finch, J.A., and Rao, S.R. (1991) Review of the effect of
grinding media on the flotation of sulphide minerals, Minerals Eng. 1, 121-132.
50. Kocabag, D., and Smith, M.R (1985) The effect of grinding media in the flotation of
sulphide minerals, In A.D. Zunkcl, RS. Boorman, A.E. Morris and R.J. Wesley (eds.)
Complex Sulphides, The Metallurgical Society, pp. 55-82.
51. Nakazawa, H., and Iwasaki, I. (1985) Effect of pyrite-pyrrhotite contact on their
floatabilities, Minerals and Metallurgical Process. 2, 206-211.
52. Learmont, M.E. and Iwasaki, I. (1984) The effect of grinding media on galena flotation,
Minerals and Metallurgical Process. I, 136-143.
53. Adam, K., and Iwasaki, I. (1984) Pyrrhotite-grinding media interaction and its effect on
 99

floatability at different applied potentials, SME preprint no. 84-89.

54. Pozzo, R.L., Malicsi, A.S., and Iwasaki, I. (1990) Pyrite-pyrrhotite-grinding media contact
and its effect on flotation, Minerals and Metallurgical Process. 1, 16-21.
55. Rey, M., and Formanek, V. (1960) Some factors affecting selectivity in the differential
flotation of lead-zinc ores, Proceedings 5th Int. Miner. Process. Congress, IMM, London,
pp. 343-352.
56. Learmont, M.E. and Iwasaki, I. (1984) Effect of grinding media on galena flotation, Minerals
Metallurg. Process. 1, 136-143.
57. Finkelstein, N.P., Allison, S.A., Lovell, V.M., and Stewart, B.V. (1975) Natural and
induced hydrophobicity in sulphide mineral systems, in Advances in Interfacial Phenomena in
Particulate-Solution-Gas Systems, AIChE Symp. Series No. 150, vol. 71, pp. 165-175.
58. Cases, J.M., de Donato, P., Kongolo, M., and Michot, L. (1989) The influence of grinding
media on the adsorption of amyl xanthate on galena and pyrite, SME Prcprint 89-62.
59. Rao, S.R., Moon, K.S., and Leja, J. (1976) Effect of grinding media on the surface reactions
and flotation of heavy metal sulphides, in M.C. Fuerstenau (cd.) Flotation, New York,
vol. I, pp. 509-527.
60. Graham, R, and Heathcote, C.R (1982) The effect of the oxidation state of the pulp on the
selective flotation of chalcopyrite, Proceedings 12th C.M.M.J. Congr.(S. Afr. Inst. Miner.
Metal.), pp. 867-877.
61. Ross, V.E., and Van Deventer, J.S.1. (1985) The interactive effects of the sulphite ion, pH,
and dissolved oxygen on the flotation of chalcopyrite and galena, J. S. Afr. Inst. Min. Metal .
.£5.(1), 13-21.
62. Chen, K.Y., and Morris, J.C. (1972) Kinetics of oxidation of aqueous sulphide by oxygen,
Environmental Sci. Technol. Q, 529-537.
63. Gaudin, A.M., Miaw, H.L., and Spedden, H.R. (1957) Native floatability and crystal
structure, in I.H. Schulman (ed.) Proceedings 2nd Int. Congress of Surface Activity,
Butterworth, London, pp. 202-219.
64. Heyes, G. W., and Trahar, W.J. (1977), The natural floatability of chalcopyrite, Int. J.
Miner.Process.:1,317-344.
65. Yoon, R.H. (1981) Collectorlessflotation of chalcopyrite and sphalerite using sodium
sulphide, ibid . .8., 31-48.
66. Leroux, M., Rao, S.R, Finch, I.A., Gervais, V., Labonte, G., Ounpuu, M., Kim, I., and
MacMinn, N. (1989) Collectorless flotation in the processing of complex ores, in S. Chander
and R Klimpel (eds.) Advances in Coal and Mineral Processing Using Flotation, SME,
Littleton, Colorado, pp. 65-70.
67. Trahar, W.J. (1984) The influence of pulp potential in sulphide flotation, in M.H. Jones and
I.T. Woodcock (eds.) Principles of Mineral Flotation, The Australasian Institute of Mining
and Metallurgy, Victoria, Australia, pp. 117-135.
68. Hayes, R.A., Price, F.M., Ralston, J., and Smith, RW. (1987), Col/ectorlessjl.otation of
sulphide minerals, Mineral Processing Extractive Metallurgy Review 2, 1-20.
69. Kiljavainen, V.M., Laapas, H.R, and Heiskanen, K.G.H. (1989), Collectorlessflotation of
some nickel ores, in Today's Technology for the Mining and Metallurgical Industries, IMM,
London, pp. 281-286.
70. Rao, S.R. and Finch, I.A. (1987) Electrochemical studies on the flotation of sulphide
minerals with special reference to pyrite- sphalerite, Can. Met. Quart. 2.6., 173-175.
71. Martin, C.I., Rao, S.R., Finch, I.A., and Leroux, M. (1989) Complex sulphide ore
processing with pyrite flotation by nitrogen, Int. I. Miner. Process. 26,95-110.
100

72. Sandoval-Caballero, I., Leroux, M., Rao, S.R., and Finch, J.A. (1990) Testing nitrogen
flotation ofpyrite in mini-cells, Minerals Eng. .3., 369-373.
73. Johnson, N.W. (1988) Application of electrochemical concepts to four sulphide flotation
separations, Proceedings 2nd Intnl. Symposium on Electrochemistry in Mineral and Metal
Processing, The Electrochemical Society, Pennington, N.J. pp. 139-150.
74. Rao, S.R., Gehr, R., Finch, J.A., and Biss, R. (1989) Water treatment and flotation studies
on recycle water from mineral processing plants, Proceedings 12th Symposium on
Wastewater Treattnent, Montreal, Canada, pp. 99-119.
75. Hodgson, M., and Agar, G.E. (1989), Electrochemical investigations into the flotation
chemistry of pentlandite and pyrrhotite: process water and xanthate interactions, Can. Met.
Quart.~, 189-198.
76. Rao, S.R., and Finch, J.A. (1991) Adsorption of amyl xanthate at pyrrhotite in presence of
nitrogen and implications in flotation. Can. Met. Quart. 30 (in press).
77. Rohner, P., Di Pasquale, S. and Johnson, N. W. (1989). Pilot plant testing of metabisulphite
in the lead circuitfor Hilton ore, in G.S. Dobby and S.R. Rao (eds.) Processing of Complex
Ores, Pergamon Press, New York, pp. 183-192.
1.4. FLOCCULATION OF FINE PARTICLES

JOHN GREGORY
Department of Civil and Environmental Engineering
University College London
Gower Street
London WCI E 6BT
UK

ABSTRACf. The basic principles of colloid stability, destabilisation and flocculation are reviewed,
starting from the effects of simple salts. The role of adsorbed materials, especially polymers, on
stability is considered in some detail and some practical implications are discussed. Factors
influencing the rate of flocculation are outlined and the importance of particle size is emphasised.

1. Introduction

Many solid-liquid separation operations, including flotation, are greatly dependent on particle size
and in many cases separation is more effective for larger particles. Aggregation of fine particles is a
common method of increasing their size and improving the efficiency of solid-liquid separation.
The aggregation of particles is generally referred to as either coagulation or flocculation and
different distinctions are drawn between these terms, depending on the application - in this chapter
flocculation is used mainly as a generic term for the aggregation process and some cases where
different terminology is employed are mentioned.
Basically, flocculation can be treated in terms of two distinct stages (see Figure I). Firstly,
particles may need to be destabilised, so that any repulsion between them is reduced and then
collisions between particles must occur, giving aggregates or floes. Since dispersed particles
usually carry a surface charge, especially in aqueous systems, destabilisation often involves the
addition of materials which reduce the charge. With negatively-charged particles, for instance, salts
with specifically-adsorbing cations will be effective destabilising agents. Cationic polymers will
also be effective because of their tendency to adsorb on particles and so reduce or neutralize the
surface charge. Long-chain polymers may also act by bridging particles together, to give strong
flocs which are better able to withstand disruptive forces than aggregates formed simply as a result
of charge neutralisation.
Collisions between particles can be brought about simply by Brownian motion (diffusion), by
induced fluid motion or by differential sedimentation. In each case the collision rate can be derived
and, if the particles are fully destabilised, this gives the flocculation rate. In agitated suspensions,
flocs may be disrupted by shearing forces, giving an effective upper limit to floc size.
The basic principles of particle stability, the action of destabilising agents, and the factors
governing flocculation kinetics will be outlined in the following sections. For the most part, we

101
P. Mavros and K. A. Matis (eds.). Innovations in Flotation Technology, 101-124.
© 1992 Kluwer Academic Publishers.
102

® •
® •
•
•
• • •
•
•
Figure 1. Schematic illustration of the destabilisation of particles (a) and aggregation of destabilised
particles (b). The particles are shown as spheres and the destabilising agent as black dots.

will be concerned with rather idealised systcms. but thc concepts are applicable. at least qualita-
tively. to many practical problems.

2. Colloid Interactions

Particles to be flocculated may be initially very small and fall in the colloidal size range. which is
conventionally taken to cover sizes between I nm and I !ID1. although many particles of interest are
rather larger than 1 ~m. Such particles may interact with each other in a number of ways, which
can profoundly affect their flocculation behaviour. These interactions are especially important for
small particles, where they can easily outweigh other forces such as hydrodynamic drag and
gravity, which become dominant for larger particles. A number of important types of interaction
will be considered in the following paragraphs.

2.1. VAN DER WAALS INTERACTION

The universal attractive forces between atoms and molecules, known as van der Waals forces, are
known to operate between macroscopic objects also and play an important part in the interaction of
colloidal particles.
Essentially, there are two approaches to the calculation of van der Waals interaction between
macroscopic bodies. The first, due mainly to Hamaker [1], is based on the summation of all
interactions between molecules in the bodies and leads to quite simple expressions. However, the
assumption of pairwise additivity of intermolecular forces is open to doubt and another approach is
often used, based only on the bulk properties of the interacting media. This macroscopic approach
treats the interaction as arising from spontaneous electric and magnetic polarisations, giving a
 103

fluctuating electromagnetic field within the media and in the gap between them. In order to calculate
the force, the variation in electromagnetic wave energy with the separation distance has to be
determined. Lifshitz [2] derived an expression for the force between two semi-infinite media
separated by a plane-parallel gap and this treatment was later extended by Dzyaloshinskii et al. [3]
to include the case of two bodies separated by a third medium. Although there is no doubt about the
essential correctness of the macroscopic (or Lifshitz) approach, its application is limited to cases
where detailed knowledge of the dielectric responses of the interacting media over a wide frequency
range is available. Certain simplifications are possible, very clearly described by Hough and
White [4], but considerable computation is needed.
Because of these difficulties, the older, Hamaker approach is still widely used and, in many
practical cases, the results are not too different from more exact "macroscopic" computations. For
the case of parallel flat plates, separated by a distance d, the Hamaker result for the interaction
energy per unit area is:

(1)

where AI2 is the Hamaker constant for the interacting media (types 1 and 2). This depends on the
properties of the media, especially optical properties (see below). The negative sign, by conven-
tion, implies an attraction energy.
For spherical particles, radii al and a2, with minimum separation distance d (see Figure 2), the
simplest Hamaker expression is:

(2)

For the case of identical spheres, composed of medium 1, this reduces to :

(3)

where A II is the Hamaker constant for medium 1.

Eqns. 2 and 3 apply only to the case of very close approach of the spheres (d« a) and become
very inaccurate at separations greater than about 10% of the particle radius. Nevertheless, this
restriction is not serious in most cases, since the van der Waals interaction is only significant at
very close approach.
Approximate methods for calculating Hamaker constants are available [5], and the following
result for media 1 and 2 may be used:

A _ 2711wIWz(£I-l)(£2-1)
(4)
12 - 32(wl + Wz)(£1 + 2)(£2 + 2)

where 11 is Planck's constant divided by 2n, and £1 and £2 are the limiting dielectric constants of
media 1 and 2. In practice, these can be taken as the square of the refractive index for many
materials. The terms WI and Wz are characteristic frequencies for the media, which may be obtained
from optical dispersion data (see Hough and White [4]).
For similar media, eqn. 3 becomes:
104

(5)

For a range of materials, these simple expressions give good agreement with computations based
on the macroscopic approach.
The expressions given so far are for media interacting across vacuum. For problems of colloid
stability and flocculation, interaction through a liquid medium needs to be considered. It turns out
that the Hamaker expressions, eqns. 1 and 2, can still be used, provided that a modified Hamaker
constant is substituted. Thus, for the interaction of media 1 and 2 through medium 3, the
appropriate Hamaker constant, A 132, is:

(6)

Depending on the relative magnitudes of the different Hamaker constants, the presence of an
intervening medium can significantly reduce the interaction.
A useful approximation for Hamaker constants of different media is the geometric mean
assumption:

(7)

It can be shown from eqns. 4 and 5 that this assumption is only strictly valid when the two
materials have equal dispersion frequencies. With this and the corresponding expressions for A 13
and A23, eqn. 6 becomes:

(8)

For similar materials, I, interacting across medium 3:

(9)

So, for similar materials, the interaction should always be an attraction (positive Hamaker constant
and hence a negative interaction energy). However, for different materials, eqn. 8 allows the
possibility of negative Hamaker constants and hence van der Waals repulsion, which would arise
if the Hamaker constant for the intervening medium had a value intermediate between those of the
other two media. Although van der Waals repulsion is of very limited relevance to colloid
interactions, it is important in wetting phenomena [4] and in phase separation from polymer
solutions [6].
Typical Hamaker constants for aqueous dispersions are in the range 0.3 to lOxlO- 20 J. Dense
mineral particles have values towards the upper end of this range, whereas low density, especially
biological, materials have quite low values.

2.2. ELECfRICAL INTERACTION

Practically all particles in aqueous suspensions carry a surface charge, which can arise for a variety
of reasons, such as ionisation of surface groups and specific adsorption of ions. In an electrolyte
solution, the distribution of ions around a charged particle is determined by electrical interaction
with the surface and the randomising tendency of thermal motion. Oppositely-charged ions
 105

(counterions) will be attracted to the surface and can be either closely associated with the surface or
be distributed some way into the solution. The nature of the electrical double layer formed can
greatly influence the interaction between charged particles. At low ionic strength, the diffuse layer
of counterions around the particles extends to a considerable distance. Particles approaching each
other begin to experience repulsion when their diffuse layers overlap (see Figure 2) and so
repulsion can occur at quite large separations. By contrast, at higher electrolyte concentrations, the
diffuse layer is less extensive and particles need to approach quite close before repulsion is felt.

Figure 2. Interaction of charged particles in a) low and b) high ionic strength solutions, showing the
effect on the range of double layer repulsion.

In the context of particle interaction, an important quantity is the electric potential at the inner
boundary of the diffuse layer (i.e. close to the particle surface, but just outside a layer of closely
associated counterions - the Stern layer). This potential cannot be measured directly, but is believed
to be close to the experimentally-accessible zeta potential. The zeta potential can be obtained from a
number of electrokinetic techniques, in which there is relative motion between particle and solution
and a corresponding separation of charge (see Hunter [7] for a full account). For particles, the most
useful technique is electrophoresis, in which particles migrate under the influence of an applied
field. The electrophoretic mobility (velocity!field strength) is measured, from which the zeta
potential can be derived.
For two spherical particles, radii a 1 and a2, with zeta potentials 1;, 1 and 1;,2, the electrical
interaction energy, YE, as a function of the separation distance, d, can be written in the approximate
form:

(10)
106

where E is the pennittivity of the medium, z is the valency of the ions (assuming a symmetrical, z-z
electrolyte), and e is the elementary charge. The tenns Yl and Y2 are dimensionless functions of the
zeta potentials: Yl = tanh (ze~1/4kT) etc. The interaction decreases exponentially with separation
distance, with a decay length 1/1(, where x is the Debye-Huckel parameter. This is a function of
ionic strength and, for aqueous electrolytes at 25°, can be obtained from:

x = 2.3xl09 -V f CiZi 2 (m-l) (11)

where Ci is the molar concentration and Zi is the valence of ions of type i. The sum is taken over all
types of ion present in the solution.
The Debye-Huckel parameter is of crucial importance in detennining the range of electrical
interaction between particles. The corresponding length, I/x, can range from values near to I !IDl in
deionised water, to less than 1 nm in concentrated salt solutions. For 1-1 electrolytes the value is
about 30 nm in 10-4 M solutions and about 3 nm in 10-2 M solutions. (Note the dependence on the
square root of the salt concentration).
The pre-exponential tenn in eqn. 10 depends on the zeta potentials of the particles and is always
positive (repulsion) if the potentials are of the same sign and negative (attraction) for potentials of
opposite sign.
For identical particles eqn. 10 reduces to the simple result:

(12)

For small values of zeta potential, this simplifies still further to:

(13)

Increasing ionic strength has two effects on electrical double layer interaction. Firstly, the
corresponding increase in x causes a decrease in the range of repulsion, so that particles may
approach closer before they repel each other appreciably. Also, added salts cause a decrease in the
zeta potential of particles, which reduces the repulsion at a given separation distance. Both of these
effects are more pronounced with more highly charged ions, so salts of multi-valent ions are
expected to have a greater effect on colloid stability. In practice, highly charged counterions have
the greatest effect (see later).
The predicted exponential decay of the interaction with distance and a decay length of 1/x have
been confinned experimentally by direct measurements of force between mica sheets in various
electrolyte solutions [8], for distances greater than about 4 nm. At closer approach, hydration
forces become important.

2.3. HYDRATION EFFECTS

The nature of water close to a particle surface can be very different from bulk water, for a variety of
reasons. Since most particles carry a surface charge and hence ionic surface groups, some
hydration of these groups would be expected, by analogy with ions in solution. Some particles,
especially those of biological origin, have various types of hydrophilic material at their surface,
such as proteins and polysaccharides. These can have quite large amounts of "bound water", which
 107

plays a part in the interaction of such particles.

The approach of two particles with hydrated surfaces will generally be hindered by an extra
repulsive interaction. distinct from the electrical double layer repulsion. This hydration repulsion
arises essentially from the need for the surfaces to become dehydrated if true contact between
particles is to occur. This involves work and hence an increase in free energy of the system.
The most direct evidence for hydration effects comes from measurements of the force between
mica sheets separated by different salt solutions [8]. At low ionic strengths. the repulsion follows
the expected exponential form for double layer interaction. but at salt concentrations above about
1 mM an extra. monotonic. short-range force is apparent. owing to adsorbed hydrated cations.
This extra force increases with the degree of hydration of the counterions (Li+ = Na+ > K + > Cs+)
and decreases nearly exponentially over the range 1.5 to 4 nm. with a decay length of the order of
1 nm. Experiments on thin surfactant films have also provided evidence for hydration effects [9].
The range of these hydration forces is quite appreciable compared to the range of double layer
repulsion and they would be expected to have an effect on colloid stability. especially at high ionic
strength. The anomalous stability of latex particles at high salt concentration [10] and the inhibition
of bubble coalescence under similar conditions [11] may be examples of hydration repulsion.
With adsorbed layers of hydrophilic material. hydration can also lead to significant repulsion.
but this will be considered along with steric interaction in 2.5.

2.4. HYDROPHOBIC INTERACTION

When a surface has no polar or ionic groups or hydrogen-bonding sites. there is no affinity for
water and the surface is said to be hydrophobic. The nature of water in contact with such a surface
will be different from that of ordinary. bulk water. Ordinary water is significantly structured
because of hydrogen bonding between the molecules. The cooperative nature of this bonding [12]
means that quite large clusters of hydrogen-bonded water molecules can form. although these are
of a very transient nature. continually forming and breaking down in response to thermal energy
fluctuations. The presence of a hydrophobic surface could restrict the natural structuring tendency
of water. simply by imposing a barrier which prevents the growth of clusters in a given direction.
Water confined in a gap between two such surfaces would thus be unable to form clusters larger
than a certain size. For a narrow gap. this could be a serious limitation. resulting in an increased
free energy of the water in relation to bulk water. In other words there would be an attraction
between hydrophobic surfaces. as a consequence of water molecules migrating from the gap to the
bulk water. where there are unrestricted hydrogen-bonding opportunities and a lower free energy.
Attraction between hydrophobic surfaces has been measured directly [13] and can be of
surprisingly long range (up to about 80 nm) [14]. The systems used were mica sheets. with
adsorbed hydrocarbon and fluorocarbon surfactants to give hydrophobic surfaces. The attraction
was much stronger than the van der Waals force and of much greater range.
For particles dispersed in water. it is possible for their surfaces to have some degree of
hydrophobic character. For instance. negatively charged particles. with just enough adsorbed
cationic surfactant to neutralise their charge. would present hydrophobic "tails" of the surfactant to
the water. As well as eliminating electrical repulsion between the particles. an extra attraction could
be introduced as a result of the hydrophobic layer. By analogy with the measurements for mica.
such an attraction could be of quite long range and playa large part in promoting flocculation.
There has been a limited amount of work on the role of hydrophobic effects in flocculation.
although the position is far from clear.
108

2.5. STERIC INTERACTION

Adsorbed layers, especially of polymers, can playa very large part in colloid stability. In some
cases, small amounts of adsorbed polymer can promote flocculation by a "bridging" mechanism
(see 2.6). With larger adsorbed amounts, polymers can give greatly enhanced stability, by an effect
which is usually known as steric stabilisation. The most effective stabilizers are polymers which
have some affinity for the surface, but adsorb in such a way that segments of polymer chains
extend some distance into the aqueous phase. The simplest case is that of terminally-adsorbed
block copolymers, which have some segments which adsorb strongly on the particles and other,
hydrophilic segments which project into the aqueous phase. These polymers fonn adsorbed layers
like those shown schematically in Figure 3 and can give greatly enhanced stability. A well-known
example for aqueous dispersions is the range of non ionic surfactants, with hydrocarbon segments
providing the adsorbing part (by hydrophobic interaction) and hydrophilic "tails" of polyethylene
oxide, which can be of various lengths.

Figure 3. Effect of terminally-adsorbed polymer chains on particle interaction.

The stabilising action of such materials can be interpreted in fairly simple tenns. As particles
approach sufficiently close, the adsorbed layers come into contact and any closer approach would
involve some interpenetration of the hydrophilic chains. Since these chains are hydrated, overlap of
the layers would cause some dehydration and, hence, an increase in free energy and a repulsion
between particles. To a first approximation, the repulsion can be assumed to become infinite as
soon as the adsorbed layers begin to overlap, but zero at greater separations. This "hard sphere"
assumption is not strictly correct [15]. but is good enough for most practical purposes and leads to
some important conclusions.
It is reasonable to assume that the effective Hamaker constant for the adsorbed layer is rather
low, so that there is little van der Waals attraction between the layers. In that case, the most
important effect of the adsorbed polymer is to limit the attraction between particles by keeping them
a finite distance apart, where van der Waals attraction is lower. Contact between particles may then
lead to fairly weak aggregates, which can easily be broken by shear. The most important factor
detennining the degree of steric stabilisation is the thickness of the adsorbed layer relative to the
particle size. Since van der Waals attraction energy is proportional to particle size (eqn. 3), larger
 \09

particles will need thicker stabilising layers to confer the same degree of stability.
Sterically stabilised dispersions can be destabilised by changing the solvency of the medium for
the stabilising chains. With polyethylene oxide, adding certain salts (especially sulphates) and
increasing temperature can cause dehydration of the chains, hence decreasing the stabilising effect
and causing flocculation [16].
The case of block copolymers, illustrated in Figure 3 is not typical of most adsorbing polymers,
which are attached at many points along the chain, giving features known as trains, loops and
tails [17]. The effective thickness of the adsorbed layer is then more difficult to determine and can
be dominated by the tails, although these represent a very small proportion of the adsorbed amount.
Nevertheless, the principles outlined above are still generally valid; effective stabilisation is
achieved by fairly thick adsorbed layers, which are well solvated, so that overlap or
interpenetration of the layers cannot occur.
Steric stabilisation is a widespread phenomenon and in the older colloid literature materials acting
in this way were known as "protective colloids". The classic example is the stabilisation of gold
sols by gelatin, discovered by Faraday, but the effect has been exploited, unknowingly, since
ancient times, as in the preparation of stable dispersions of carbon black (ink) and of other
pigments.
In the aquatic environment, most particles have adsorbed layers of natural organic material, such
as humic substances [18], which can have a major effect on their colloidal behaviour.
Microorganisms produce extracellular polymers which may adsorb and have a great influence on
particle intcraction in biological systems. These natural polymers and organic materials are often
weak acids and are anionic at neutral pH values. It is thought that the rather low negative zeta
potentials found for most natural aquatic colloids are due to the adsorption of this anionic material.
The stability of such colloids is often higher than expected on the basis of zeta potential and ionic
strength, and it is likely that steric stabilisation plays an important part. Humic substances are
known to enhance the stability of inorganic colloids [19] and can cause greatly increased dosages
of flocculants in water treatment.

2.6. POLYMER BRIDGING

Long-chain polymers generally adsorb on particles in the manner indicated in Figure 4 and, with
large adsorbed amounts, can cause steric stabilisation as discussed above. With much less
adsorbed polymer, an individual chain can become attached to two or more particles, thus
"bridging" them together. In this way particles can form aggregates even though they may be
charged and repel each other. This effect, previously known as "sensitisation", is now widely
exploited in many industrial applications. In some texts the term "flocculation" is restricted to cases
where polymer bridging is the operative mechanism and "coagulation" refers to aggregation
brought about by reducing double layer repulsion. (This terminology is not employed here).
The essential requirements for polymer bridging are that there should be sufficient unoccupied
particle surface for attachment of pol ymer segments from chains attached to other particles and that
the polymer bridges should be of such an extent that they span the distance over which inter-
particle repulsion operates. Generally, the most effective bridging flocculation is found with linear
polymers of very high molecular weight (several million), so that extended loops and tails can
form, increasing the probability of attachment to other particles. Also, it is found that there is an
optimum dosage range for effective flocculation. At lower dosages, there is insufficient polymer to
form adequate bridging links between particles. With excess polymer, there is no longer enough
bare particle surface for attachment of segments and the particles become restabilised, which may
110

involve some steric repulsion. Figure 4 shows schematically the flocculation and restabilisation of
particles by adsorbed polymer. Bridging flocculation can give aggregates (flocs) which are much
stronger than those produced just by the addition of salts (Le. by reducing electrical repulsion).
Healy and La Mer [20] proposed that optimum bridging flocculation would occur when the
particle surfaces were half covered with adsorbed polymer, providing maximum opportunity for
favourable particle-particle contacts. However, there is some ambiguity in the concept of "surface
coverage" in the case of adsorbing polymers and the condition of half surface coverage is not easy
to establish. Some refinements to the La Mer model have been suggested [21,22].

(a) (b)

Figure 4. Schematic diagram showing a) flocculation and b) restabilisation by adsorbed polymer.

Many polymeric flocculants are ionic (polyelectrolytes) and the polymer charge introduces some
complications. In some cases, anionic polymers are found effective for the flocculation of
negatively-charged particles, despite the like sign of charge. The most common flocculants used in
practice are based on polyacrylamide, which may be hydrolysed so that amide groups are converted
to carboxylic acid groups. At neutral pH values these are ionised to anionic caboxylate groups. The
degree of hydrolysis, and hence the degree of anionic character, can be controlled during
manufacture and has been found to have important effects on the flocculation efficiency.
Michaels [23] showed that a certain degree of hydrolysis (about 30%) gave optimum flocculation
of silts and clays. The explanation depends on the fact that polyclectrolyte chains are expanded to
some extent, as a result of repulsion between charged segments. Below the optimum hydrolysis
level, the polymer chain is less extended and hence less able to form extended bridges between
particles. At higher charge densities, adsorption of polyelectrolytes on similarly-charged particles
becomes restricted by electrical repulsion and hence flocculation is less effective. There are ionic
strength effects also, although these have not been studied in detail. Higher ionic strengths eause
polyelectrolyte chains to be less extended, but should also make adsorption easier, and allow closer
approach of charged particles. There is also evidence of specific effects of certain ions. For
instance, the presence of calcium ions can promote the adsorption of hydrolysed polyacrylamide on
negatively charged particles, presumably by linking carboxylate groups with anionic sites on the
 III

particle surface.
When the particles and the polyelectrolyte are of opposite charge (as in the important practical
case of negative particles and cationic polymers), the possibility of de stabilisation simply by charge
reduction arises. In fact, in many cases, the action of cationic polymers can be explained in terms
of their strong adsorption on negative particles and the consequent reduction of double layer
repulsion, allowing aggregation to occur. The rather flat configuration adopted by polyelectrolytes
adsorbing on oppositely-charged surfaces [24] makes bridging contacts rather unlikely. For these
reasons, the most effective cationic flocculants are often those of high charge density, with
molecular weight playing only a minor role. There are important exceptions though, especially with
concentrated dispersions, where kinetic factors increase the importance of bridging interactions.
An important refinement to the simple charge-neutralisation picture is the so-called electrostatic
patch model [25,26], in which particle charge is not neutralised uniformly by adsorbed
polyelectrolyte. This condition can arisc whcn the chargc density ofthc polyelectrolyte is high and
that of the particle surface is low. In such cases each charged site on the particle surface cannot be
neutralised individually by an oppositely-charged polymer segment, simply for geometric reasons
(see Figure 5). Polymer chains, then, must adsorb to give "islands" or "patches" of charge,
surrounded by areas of opposite charge. Particles with polyelectrolyte adsorbed in this "patchwise"
manner can interact in such a way that oppositely-charged areas of different particles come into
contact, giving strong attraction. At low ionic strengths, this extra attraction can give measurable
increases in flocculation rate [27].

Figure 5. "Electrostatic patch" model.

2.7. COMBINED INTERACTION - COLLOID STABILITY

The earliest, and still the only truly quantitative, theory of colloid stability is that developed
independently by Deryagin and Landau [28] and Verwey and Overbeck [29] and now widely
known simply as DLVO theory. Essentially van der Waals attraction and electrical double layer
repulsion are assumed to be additive and combined to give the total energy of interaction between
particles as a function of separation distance. All other types of interaction, such as those discussed
112

in 2.3-2.6 above, are ignored. (These are now often referred to as non-DLVO forces).
When the zeta potential of the particles and the ionic strength are such that repulsion outweighs
the attraction, then there is a potential energy barrier, which tends to prevent contact of particles and
so prevent flocculation (see Figure 6). As the ionic strength is increased and/or the zeta potential
reduced (both effects usually result from an increased salt concentration), so the energy barrier
becomes lower and contact of the particles can occur more readily. Eventually, the barrier
disappears altogether and, in principle, particles can then adhere each time they collide. Such
particles are said to be fully destabilised and to undergo rapid flocculation. The latter term implies
nothing about the absolute rate of flocculation, but that the rate cannot be further increased by the
addition of more salt.

200r--r----------------------, 200
\
\ (a) \ (b)
\ \
\ \
\
\ V \ 150
150
\ VE

i I \. \\\
\
\
i\'\'" '. \
: \
E \
\

100
\
100 ~
\
\
\
>.
Ol
~ i: Energy.. \
\
\
\
Oi
c:
CD \
w
c: 50 : Barrier"" \
... "
\
\ , 50
c:
o! \ ,, 0

(T
'f.: i V... . . . . .~
?ii' "
.§ 0 i T " " " •.• ',------ __
....
" _-- .lB
....
0
c: i Primary ......~..::.:
...::l.;:";:;"''''''='~--1 E
iM;'rnimum \ Secondary
: Minimum

!~
-50 . V -50
A
_100~·-L __ ~ ____L __ _ _ _ _ __ L_ __ _ J
~
-100
o 2 468 10 0 2 4 6 8 10
Separation distance (nm) Separation Distance (nm)

Figure 6. Potential energy diagrams for thc approach of equal sphcrical particles, diameter 1 !Lm, in 1-1
electrolyte solution. The Hamaker constant is taken as 2 kT (about 8.3xlO· 21 J). a) Zeta potentials 30 mY,
salt concentration 50 mM: b) zeta potentials 20 mY, salt concentration 100 mM.

There are several features of the potential energy diagram, Figure 6a, which are important in a
discussion of colloid stability. If there is a barrier, then colliding particles must overcome this in
order to come into true contact. Potential energies are often expressed in kT units and, in Brownian
motion, particles will rarely have an energy of more than a few kT. Thus, a barrier height of20 kT
or more implies a very stable suspension, with only a minute fraction of colliding particles having
sufficient energy to surmount the barrier. Once the barrier is overcome, then the particles are held
by van der Waals attraction in a deep primary minimum, from which escape would be very
unlikely. In principle, the attraction is infinitely strong on contact of the particles (see eqn. 3), but
 113

short range effects, such as those caused by ion hydration (2.3), tend to keep particles from
coming into true contact and so the primary minimum is of finite depth.
Because ofthe different distance dependence of the van der Waals and electrical energies (eqns.
3 and 12), the former is always greater than the latter at sufficiently large separation distances. This
gives a secondary minimum in the potential energy curve, which can be responsible for the
formation of fairly weak aggregates. Since the interaction energies are directly proportional to
particle size, secondary minimum effects are more significant with larger particles (greater than
about 1 !ID1 in diameter). They are also more important at moderately high salt concentrations, so
that the range of electrical repulsion is reduced, allowing the particles to approach closer, where
van der Waals attraction is greater. The presence of an adsorbed polymer layer giving steric
repulsion (2.5) leads to a similar type of long-range attraction, but with more control over the
distance, and hence the depth, of the "secondary minimum".
In the presence of an energy barrier, only a certain fraction of collisions are effective and this
fraction is known as the collision efficiency, u. The reCiprocal of this value is known as the
stability ratio, W. By measuring the rate of flocculation (see later) as a function of salt
concentration, it is possible to find the critical concentration at which the onset of rapid flocculation
occurs. Simple theory [30] suggests that a plot of log W against log (salt concentration), should
show a linear decrease as far as the critical salt concentration, with log W being constant at higher
concentrations (see Figure 7a). Although such plots do not always show the expected linearity,
they are useful in determining the critical salt concentration at which rapid flocculation begins.

10gW 10gW

log c etc esc log c

Figure 7. Stability ratios plotted against salt concentration in the cases of: a) indifferent electrolytes and
b) specifically-adsorbing counterions. In the fonner case, the critical salt concentration is Cr, and in (b) there
are critical concentrations for flocculation (cfc) and restabilisation (esc).
114

The critical flocculation concentration of electrolytes can be calculated quite simply from the
appropriate expressions for the attraction and repulsion energies. Taking eqns. 3 and 12 as suitable
approximations for the van der Waals and double layer interactions respectively. these
contributions can be added to give the total interaction energy. VT. as a function of separation
distance:

(14)

When the critical salt concentration has been reached. the following conditions apply:

dVT/dd =O. and VT =0

Applying these conditions to eqn. 14 leads to the following result for the critical concentration:

c[ =3.82x 10-39 ( 12:6> (15)

where the constant is appropriate to aqueous dispersions at 25°C.

This approach is rather misleading. since the zeta potential of the particles (via the parameter y) is
itself dependent on salt concentration, whereas it is treated as constant in the differentiation of eqn.
14. In fact. for sufficiently high values of l;,. the term y becomes nearly constant (about 1). so that
the critical concentration should depend inversely on the sixth power of the valency z.
It has long been known that the destabilising effect of salts is strongly dependent on the valency
(especially the counterion valency) and the dependence on 1/z6 has become known as the Schulze-
Hardy rule. Although experimental data do show a strong effect of counterion charge, the sixth-
power law rarely applies. Usually. a somewhat less pronounced dependence is found and there are
good reasons for this. In practically all cases where particles are destabilised and undergo rapid
flocculation. the zeta potentials are quite low (typically less than 30 m V). It follows that the term y
can not be taken as constant, but depends strongly on the zeta potential, which leads to a much
weaker dependence of c[ on counterion charge. Overbeek (31) has discussed several modifications
to the simple theory and their effects on the Schulze-Hardy rule.
Although the destabilising effect of "indifferent" electrolytes (Le. those without specificaliy-
adsorbing ions) may be reasonably well explained by DLVO theory. in most practical cases we
need to consider specific effects of additives. Often. the action of flocculating salts can only be
explained by the adsorption of counterions and consequent reduction of zeta potential. In such
cases full destabilisation can be achieved at low ionic strength, provided that the zeta potential has
been reduced to some critical value. If the additive level is increased. the zeta potential can be
reversed and the particles become stable again (a process normally known as restabilisation).
Figure 7 shows the difference in behaviour of indifferent and specifically-adsorbing electrolytes on
the flocculation rate (as measured by the stability ratiO, W).
The range over which flocculation occurs depends on ionic strength. At low ionic strength,
flocculation may occur over only a limited range. since a small value of zeta potential (positive or
negative) is sufficient to stabilize the particles. As ionic strength is increased. flocculation can occur
over a broader range - less additive is required to destabilize the particles and more is needed to
give restabilisation. In principle. the critical levels should be given by DLVO theory (using a
modified version of eqn. 15). but there have been few investigations along these lines.
 115

3. Flocculation Kinetics

3.1. GENERAL

Most discussions of the rate of flocculation start from the classic work of Smoluchowski [32].
which laid the foundations of the subject. It is convenient to think in tenns of a dispersion of
initially identical particles (primary particles) which, after a period of aggregation, contains aggre-
gates of various sizes and different concentrations - nj particles of size i, n j particles of size j etc.
Here, nj etc. refer to the number concentrations of different aggregates and "size" implies the
number of primary particles comprising the aggregate; we can speak of "i-fold" and "j-fold"
aggregates. A fundamental assumption is that aggregation is a second order rate process. in which
the rate of collision is proportional to the product of concentrations of two colliding species. Thus.
the number of collisions occurring between i and j particles in unit time and unit volume, Jjj. is
given by:

(16)

where kjj is a second-order rate constant. which depends on a number of factors, such as particle
size and transport mechanism (see below).
It is then possible to write the following expression for the rate of change of concentration of k-
fold aggregates, where k = i + j:

(17)

The first tenn on the right hand side represents the rate of fonnation of k-fold aggregates by
collision of any pair of aggregates, i and j. such that i + j = k. Carrying out the summation by this
method would mean counting each collision twice and hence the factor 1/2 is included. The second
tenn accounts for the loss of k-fold aggregates by collision, and aggregation, with any other
aggregates. The tenns kij and kik are the appropriate rate constants.
It is important to note that eqn. 17 is for irreversible aggregation. since no allowance is made for
break-up of aggregates. Also, it is assumed that each collision results in the fonnation of a new
aggregate (Le. the collision efficiency and the stability ratio are both unity).
In principle. it would be possible to use eqn. 17 to derive the concentration of all aggregate types
at any time. but there are fonnidable difficulties, especially in assigning values to the rate
coefficients. These depend greatly on the nature of the particles and on the way in which collisions
are brought about. There are three important transport mechanisms in practice: I) Brownian
diffusion (giving perikinetic flocculation); 2) fluid motion (orthokinetic flocculation) and 3) diffe-
rential settling.

3.2. PERIKlNETIC FLOCCULATION

Particles in a dispersion are subject to random displacements due to the thennal energy of the
system and this effect can be seen microscopically as Brownian motion. For this reason, particles
will collide from time to time and it is possible, in simple cases, to calculate the rate of collision.
Smoluchowski calculated the number of particles colliding with a central. stationary particle in unit
116

time and then allowed for the fact that the central particle itself is one of many similar particles
undergoing Brownian motion. This approach allows the number of collisions between i and j
particles in unit time and unit volume. Jij. to be derived and hence the rate coefficient kif

(18)

where [.l is the viscosity of the medium and ai and aj are the radii of the particles (assumed
spherical).
For a uniform suspension of primary particles. radius al. the initial collision rate can easily be
calculated from eqn. 18. since only one type of collision (1-1) is involved. The rate of decrease of
the total particle concentration. 1lT. follows also directly from the collision rate. since each collision
reduces the total number by one (two primary particles lost, one doublet gained). The result is:

(19)

where kF (= 4kT/3[.l) is known as the flocculation rate constant and has a value of 6. 13xlO- 18
m3s- 1 for aqueous dispersions at 25°C.
The most noteworthy feature of eqn. 19 is that it does not include the particle size. which cancels
from eqn. 18 when aj = aj. (The reason for this is that increasing particle size has opposite effects
on the diffusion coefficient and collision radius. which exactly balance for equal. spherical
particles). The simple second-order rate expression of eqn. 19 can be integrated to give the total
particle concentration as a function oftime:

(20)

where no is the initial concentration of primary particles.

There is a characteristic flocculation time, tF. in which the number of primary particles is reduced
to half of the initial value (1lT =fl(/2) and this follows immediately from eqn. 20:

(21)

Another way of considering the flocculation time is as the average time in which a given particle
experiences one collision. From the value of kF quoted above. tF turns out to be about
1.6x 1017/no. So, for an initial concentration of 1016 particles per m 3, the flocculation time would
be about 16 s. For more dilute dispersions, the flocculation time would be correspondingly longer.
In a real flocculation process quite large aggregates are formed, representing a large reduction in the
number of particles. It follows from eqn. 20 that, for dilute dispersions, an extensive degree of
aggregation would require very long times (hours or days), so that perikinetic flocculation would
not be a practical operation.

3.3. ORTHOKINETICR..OCCULATION

It is a common observation that flocculation rates can be enormously increased by some form of
 117

agitation, such as stirring. This is because panicle motion is induced, which causes an enhanced
collision rate. It is very much simpler to quantify this effect when the fluid motion is laminar,
although this is not a realistic assumption for practical flocculation processes. Nevertheless, the
result for laminar flow can be adapted to other conditions and this approach is taken here.
In a uniform laminar shear field the fluid velocity varies uniformly in a direction normal to the
flow direction. In the Smoluchowski treatment of orthokinetic flocculation, the particles are
assumed to follow fluid streamlines and the collision frequency depends on the size of the particles
and on the velocity gradient, or shear rate, G. By considering a central sphere, radius aj, and
flowing particles of radius aj, it can be assumed that all of the latter panicles moving on streamlines
that bring their centres within a distance (aj + aj) of the centre of the "fixed" particle, will collide
with it (see Figure 8). The collision frequency can then be calculated quite simply, giving:

(22)

where Jjj again represents the number of i-j collisions occurring in unit volume and unit time, and
kjj is the appropriate rate constant.

..
z

u
w-------------I:~ I
a·
J

--------- -- --

G = du/dz

Figure 8. Model for orthokinetic collision of spheres in a uniform shear field.

As in the perikinetic case, the initial rate of decline of the total particle concentration, nT, for an
initially uniform suspension, can be derived:

-dfiT/dt =(16!3)n2Ga3 (23)

These expressions make clear why fluid motion is so effective in promoting collisions,
especially for larger particles. The dependence of the collision rate on the third power of panicle
size contrasts markedly with the perikinetic case, where size has little or no effect. By comparing
eqns. 19 and 23 it follows that the ratio of orthokinetic and perikinetic flocculation rate constants is
118

4Glla 3/kT. For a shear rate of only 10 s-l (corresponding to quite slow stirring) in aqueous
dispersions at room temperature. the ratio is unity for a particle radius of about 0.5 !lID. For higher
shear rates and, especially. for larger particles. the orthokinetic rate becomes much greater.
The form of eqn. 23 is such that a simple transformation gives the result in terms of the volume
fraction of particles. cp (= 4n:a3nJ3. for a concentration. n. of particles with radius a):

-dnr/dt = 4Gcpn/n: (24)

If cp is assumed to remain constant during flocculation (Le. the total volume of particles is
conserved). then eqn. 24 implies a simple first-order rate law. which can be integrated to give:

11T/ll() = exp(-4Gcpt/n:) (25)

The particle concentration should thus decrease exponentially with time and this type of
behaviour has been found experimentally in some cases [33]. However. the assumption of constant
volume fraction. while quite reasonable at first sight. has serious limitations. Although the true
particle volume (the sum of the volumes of the primary particles) should remain constant in a
closed system. the effective volume of an aggregate will not be simply the sum of the volumes of
the constituent particles. This assumption would only hold if the aggregating particles coalesced to
form spheres. In the case of "real" aggregates. the structure is more open and the collision radius
must be considerably greater than that for the equivalent sphere. We can conclude that. for real
flocculation processes. the rate would be rather greater than that predicted by eqn. 25.
When the fluid motion is not laminar. the simple Smoluchowski treatment cannot be expected to
apply. but there are alternative approaches. An early attempt to treat flocculation in turbulent flow
was that of Camp and Stein [34]. who proposed that the mean velocity gradient. G. could be
calculated from the power input. p. to the fluid (for instance. in a stirred tank):

(26)

This value can then be inserted. in place of G. in the previous expressions for orthokinetic
flocculation rate. such as eqn. 23. The result is surprisingly close to a more rigorous expression for
particle collisions in isotropic turbulence. but this apparent agreement is likely to be fortuitous (see
Spielman [35]). Nevertheless the simplicity of the Camp and Stein result makes it attractive for
design purposes.
It has been argued that the above approach can only apply to particles smaller than the
Kolmogoroff microscale of turbulence [36]. This is a characteristic length which depends on the
power input and the viscosity of the fluid. In turbulent flow there is a continuous range of eddy
sizes. from the "macroscale". depending essentially on the size of the vessel. down to very small.
energy-dissipating eddies. The latter have sizes below the Kolmogoroff microscale. n. which. for
the kind of turbulence in stirred tanks. is of the order of 100 [.lm. In practice, floes can grow
considerably larger than this and collisions of such flocs should be governed by inertial. rather than
viscous effects.

3.4. DIFFERENTIAL SETILING

Particles of different size or density will settle at different rates and this relative motion can cause
 119

particle collisions and hence flocculation. The collision frequency can be calculated very simply,
assuming that Stokes law applies and that a particle trajectory is linear until contact with another
particle. The result, for particles of equal density, is:

(27)

where g is the acceleration due to gravity, Qs is the density of the particles and Q is the density of
the fluid.
It is evident that differential settling will be more important when the particles are fairly large and
dense and, in such cases, it can be very significant in promoting flocculation. This is especially the
case in the later stages of a flocculation process, when the flocs have already grown quite large.

1 E-14 e------------------------------,

1 E-15

1 E-16

E
ctI
iii
1 E-17
---------
r:::
o ---------------- ----------
u

*
1 E-18
-----------------------------------------------------------------------------.
a: '\
Differential Settling ~ f
1 E-19 li
::

l
0.10 1.00 10.00
Particle diameter ( m)

Figure 9. Comparison of collision rate constants for different transport mechanisms (see text).

3.5. COMPARISON OF RATES

Since we have now covered the important collision mechanisms, it is useful to compare the
different rate constants as functions of particle size. For particles i and j, with radii aj and aj, the
collision rate constants are:

Perikinetic: kjj = (2kT/3!J.)(aj + aj)2/ajaj

Orthokinetic: kjj = (4!3)G(aj + aj}3
Differential settling: kjj = (2"tg/9!J.)(Qs-Q)(aj + a)3(aj - aj)
120

All of these results apply only to spheres. Even if the primary particles are spherical, collision rates
of aggregates will not be correctly given by these expressions, so that they will, strictly, only apply
to the very early stages of a flocculation process. Nevertheless, the variation of the rate constants
with particle size, as shown in Figure 9, still gives worthwhile information on the relative
significance of the different mechanisms. The rate constants, kij, are shown for the case of
collisions between two types of particle, one of which has a fixed diameter of 1 !lID. The diameter
of the second particle varies from 0.01 to 10 Ilm. The shear rate is taken as 20 s·1 and the density
of the particles is assumed to be 2 g/cm 3 . All other conditions apply to aqueous dispersions at
25°C.
As expected, the perikinetic rate is greatest when the second particle is very small. For particles
greater than about 0.5 Ilm diameter, the shear-induced rate is more significant and diffusion
becomes relatively unimportant for sizes above about 2 Ilm. Differential settling is of no conse-
quence until the second particle is larger than about 2 Ilm, but for larger particles this can become
comparable with the orthokinetic rate.

4. Flocculating Agents

4.1. INORGANIC SALTS

For charge-stabilised colloids, added salts can reduce both the effective surface potential and the
extent of the diffuse layer, giving a lower colloid stability (see 2.7). However, "indifferent"
electrolytes are not used in practice as flocculating agents, although their presence in water can
modify the behaviour of other additives. The reason for the ineffectiveness of simple salts is
probably that they do not generally allow the formation of strong aggregates, which can withstand
the shear forces encountered in most practical flocculation units. Double layer compression may
eliminate electrical repulsion between particles, but it is likely that short-range hydration forces are
still present (see 2.3) which prevent true contact. In many natural waters, adsorbed organic material
may also give steric repulsion (2.5) and simple salts can have little effect, except possibly at very
high concentrations.
Salts with specifically-adsorbing counterions can be very much more effective. By reducing the
particle charge to near zero, the concentration of "indifferent" counterions in the diffuse layer is
reduced, so that hydration effects may be much less important. There could be specific effects on
adsorbed organic material also, reducing any steric repulsion. In order to be effective, the
counterions must adsorb quite strongly, to neutralize the particle charge, and the strength of
adsorption can be greatly affected by chemical changes such as complex formation [37]. Chelated
ions may be very much more readily adsorbed than the uncomplexed ions and hence act as more
effective flocculants. The opposite effect may also be found, such as the case of the AIS04+
complex, which has less flocculating effect than the uncomplexed AI 3+, probably as a result of the
reduced charge. Hydrolysis of ions can also have a dramatic effect on their adsorption and
flocculation behaviour, which is especially relevant to iron and aluminium salts, very commonly
used as fIocculants. Before discussing these salts in more detail. it should be stressed that. despite
apparent complexities of behaviour. specifically-adsorbing counterions act predominantly by
neutralising particle charge and reducing repulsion between particles. In excess amount they may
cause charge reversal and restabilisation. "Chemical" effects. such as complexation and hydrolysis,
may greatly change the adsorption affinity. but the basic charge-neutralisation model remains valid.
By far the most commonly used inorganic fIocculants are iron and aluminium compounds,
 121

especially in the water and effluent treatment areas. Although it was originally thought that their
effectiveness could be explained in terms of the highly-charged Fe 3+ and AJ3+ ions and the
Schulze-Hardy rule, this is now known to be greatly over-simplified. Because of hydrolysis, the
simple ions do not exist in significant amounts in solutions around neutral pH and the hydrolysis
products are responsible for any destabilising effects on particles.
Taking aluminium as an example, the following sequence of hydrolysis steps can be written for
increasing pH, which represent the progressive replacement of water molecules in the hydration
shell of A13+ by hydroxyl groups:

A13+ > Al(OH)2+ > Al(OHh+ > Al(OH)3(S) > Al(OHk

Hydrolysis involves a loss of positive charge, until the uncharged hydroxide is formed. This is
practically insoluble in water and so forms a precipitate, which is initially amorphous, but may
slowly form crystalline gibbsite. This precipitate is of great practical importance in many
flocculation processes. At still higher pH values the precipitate dissolves to give the aluminate ion.
The simple hydrolysis scheme above is by no means complete and it is known that several
polymeric hydrolysis products can form including the dimer AI2(OH)44+ and "AIl3" polymers such
as Aln04(OHh/+, the structure of which is now well understood [38]. These polymeric
products could be thought of as intermediates in the precipitation process, but they can be quite
long-lived. The polymerisation and precipitation processes are markedly enhanced in the presence
of sulphate ions, which is significant, since aluminium is very often applied as the sulphate salt
("alum").
The hydrolysis products are strongly adsorbed by many particles and, because of their positive
charge, can cause charge neutralisation and reversal [39). This can explain some observed
flocculation behaviour, but is not the whole story. In many practical applications, especially when
the particle concentration is quite low (river waters for instance), the precipitation of aluminium
hydroxide plays a very important role. When alum is added to a water of around neutral pH, the
hydroxide forms fairly rapidly, initially as a colloidal precipitate, with subsequent growth to form
quite large flocs. During this precipitation and growth, many of the particles originally present in
the water become coated with hydrolysed species or colloidal hydroxide particles and enmeshed in
the growing flocs. The original particles may be said to be "swept" out of suspension by the
hydroxide precipitate and, for this reason, the process is known as "sweep flocculation".
It has long been known that, for dilute suspensions, the optimum flocculation conditions are
those which give the most rapid and complete precipitation of aluminium hydroxide and that the
required AI concentration is largely independent of particle concentration [40]. The optimum pH is
usual\y in the range 6.5 - 7.5, i.e. near the region of minimum solubility of the hydroxide. These
findings are entirely consistent with the "sweep flocculation" concept. The main advantage over a
simple charge-neutralisation process is that the rate of flocculation can be markedly increased by the
hydroxide precipitate. In orthokinetic flocculation, there is a strong effect of solids concentration,
as shown by eqn. 25. For dilute suspensions the collision rate can be very low, so that a long
stirring period would be needed to produce large flocs. The introduction of a new solid phase, in
the form of a hydroxide precipitate, may increase the effective particle concentration many-fold and
transform the kinetics.
Because of the finite times required for hydrolysis, adsorption, precipitation and flocculation, the
exact sequence of events after adding alum to a suspension is very difficult to describe in detail. It
is likely that some of soluble hydrolysis products are adsorbed on particles rather than forming
precipitates. Further polymerisation and surface precipitation may occur after adsorption or
122

colloidal precipitates may deposit on particles. Attempts to model such processes in some detail
have been made (e.g. Dentel and Gossett [39]), but it is doubtful whether these efforts will be able
to provide truly predictive models, bearing in mind the complexity of the systems. At higher
particle concentrations, it is likely that adsorption and charge neutralisation become more important
than "sweep flocculation", largely for kinetic reasons.
From the above discussion, it is not surprising that the mixing process can have a large effect on
flocculation with hydrolysing salts [41]. The traditional method of application is to have a fairly
brief "rapid mix" period at quite high shear, followed by a much longer period of slow stirring,
during which orthokinetic flocculation takes place. During the rapid mix period, the hydrolysis,
adsorption and precipitation processes are probably completed, together with some flocculation. In
water treatment, this stage is often tenned "coagulation" and the subsequent, slow-stirring phase as
"flocculation", but this tenninology is not generally followed in other areas.
The action of ferric salts is broadly similar to that of aluminium salts, with hydrolysis and
hydroxide precipitation playing major parts. The hydrolysis reactions of Fe(III) have been
thoroughly reviewed by Flynn [42]. The range of pH over which effective flocculation can be
achieved is somewhat broader than for aluminium salts. Also, the role of sulphate in promoting
polymerisation and precipitation is much less important.

4.2. POLYMERIC FLOCCULANTS

Very many flocculants used in practice are organic polymers or polyelectrolytes. These can be
based on natural products, such as starches, alginates, chitosan or extracts from various types of
seed [43]. However, in practice, most polymeric flocculants are of synthetic origin, with
polyacrylamide and its derivatives being the most widely used.
Aqueous polymerisation of acrylamide can give products of varying molecular weight, up to
very high values (about 20 million). Although polyacrylamide is nominally nonionic, some
hydrolysis of amide groups usually occurs, giving carboxylate groups and hence a degree of
anionic character (except at low pH). The degree of hydrolysis can be controlled to give
polyacrylamides with different charge density. It is also possible to prepare cationic
polyelectrolytes based on polyacrylamide, by copolymerisation of acrylamide with a suitable
cationic monomer, very often dimethylaminoethyl acrylate or methacrylate. These are quatemised
after polymerisation to give strongly cationic products, whose charge density depends on the
proportion of cationic monomer used. There are many other types of polymeric flocculant available
and a comprehensive survey has been given by Halverson and Panzer [44].
As discussed earlier (2.6), bridging of particles by adsorbed polymer can give very effective
flocculation and many polymers undoubtedly act in this way. It follows that adsorption is a
necessary step and that some favourable interaction must exist between particle and polymer to
promote adsorption. Among the most important interactions are hydrophobic bonding, which
promotes the adsorption of non-polar segments on hydrophobic surfaces and hydrogen bonding,
which enables suitable polymer segments (such as amide groups in polyacrylamide) to attach to
hydroxyl groups on oxide surfaces. In many cases it is found that a certain concentration of a
divalent metal ion, such as calcium or magnesium, is needed to promote the adsorption of anionic
polymers on negative surfaces.
In the case of opposite charges, as with cationic polymers and negative particles, there is no
difficulty over adsorption, because of the strong ionic interaction. However, in such cases, the
adsorbed polymer may adopt a rather "flat" configuration [24], which would reduce the opportu-
nity of bridging contacts. In fact, as mentioned earlier, charge neutralisation or "electrostatic patch"
 123

effects can account for many of the observed effects.

5. References

1. Hamaker, H.C. (1937) The London-van der Waals attraction between spherical particles,
Physica~. 1058-1072.
2. Lifshitz, E.M. (1956) Theory of molecular attractiveforces, Soviet Physics JETP 2.73-83.
3. Dzyaloshinskii. I.E., Lifshitz, E.M., and Pitaevskii, L.P. (1960) Van der Waals forces in
liquid films , Soviet Physics JETP .ill, 161.
4. Hough, D.B., and White, L.R. (1980) The calculation of Hamaker constants from Lifshitz
theory with applications to wetting phenomena, Adv. Colloid Interf. Sci . .H. 3-41.
5. Gregory, J. (1969) The calculation of Hamaker constants, Adv. Colloid Interf. Sci. 2.,
396-417.
6. Van Oss, C.J., Omenyi, S.N., and Neumann, A.W. (1979) Negative Hamaker coefficients.
/I. Phase separation of polymer solutions, Colloid Polymer Sci. ill, 737-744.
7. Hunter, R.J. (1981) Zeta Potential in Colloid Science, Academic Press, London.
8. Pashley, R.M., and Israelachvili, J.N. (1984) DLVO and hydration forces between mica
surfaces in Mg2+, Ca 2+, Sr2+ and Ba2+ chloride solutions. J. Colloid Interf. Sci. 97,446.
9. C1unic. J.S .. Goodman, J.F.. and Tate. J.R. (1968) Adsorption of inorganic iors in black
foamfilms, Trans. Faraday Soc. 61, 1965-1970.
10. Healy. T.W., Homola, A., and James. R.O. (1968) Coagulation of amphoteric latex colloids:
reversibility and specific ion effects. Faraday Disc. Chern. Soc.~, 156-163.
11. Lessard, R.R.. and Zieminski. S.A. (1971). Bubble coalescence and gas transfer in aqueous
electrolyte solutions. Ind. Eng. Chern. Fund. 10,260-269.
12. Symons, M.C.R. (1989) Liquid water - the story unfolds, Chern. Brit., 25.. 491-494.
13. Israelachvili. J.N., and Pashley, R.M. (1984) Measurement of the hydrophobic interaction
between two hydrophobic sUrfaces in aqueous electrolyte solutions, J. Colloid Interf. Sci. 2.8..
500.
14. Claesson, P.M., and Christenson, H.K. (1988) Very long range attraction between unchar-
ged hydrocarbon andfluorocarbon sUrfaces in water. 1. Phys. Chern. 92, 1650-1655.
15. Napper. D.H. (1983) Polymeric Stabilisation of Colloidal Dispersions, Academic Press, New
York.
16. Napper. D.H. (1970) Flocculation ofsterically stabilised dispersions, J. Colloid Interf. Sci.
~, 106-114.
17. Cosgrove, T. (1990) Volume fraction profiles of adsorbed polymers. J. Chern. Soc. Faraday
Trans. 86, 1323-1332.
18. Tipping, E. (1988) Colloids in the aquatic environment, Chemistry and Industry, No. 15,
485-490.
19. Jekel, M.R. (1986) The stabilisation of dispersed mineral particles by adsorption of humic
substances, Water Research 20,1543-1554.
20. Healy, T.W., and La Mer, V.K. (1964) Energetics of flocculation and redispersion by
polymers, J. Colloid Sci. 19, 323-332.
21. Hogg, R. (1984) Collision efficiency factors for polymer flocculation, J. Colloid Interf. Sci.
lQ2, 232-236.
22. Moudgil, B.M., Shah, B.D., and Soto, H.S. (1987), Collision efficiency factors in polymer
flocculation offine particles. J. Colloid Interf. Sci.ll.2., 466-473.
124

23. Michaels. A.S. (1954) Aggregation of suspensions by polyelectrolytes. Ind. Eng. Chern. 46.
1485-1490.
24. Lyklema. I .• and Fleer. G.l. (1987) Electrical contribution to the effect of macromolecules on
colloid stability. Colloids Surfaces 2.5..357-368.
25. Kasper. D.R. (1971) Theoretical and experimental investigations of the flocculation of
charged particles in aqueous solutions by polyelectrolytes of opposite charge. PhD Thesis.
California Institute of Technology.
26. Gregory. 1. (1973) Rates offlocculation of latex particles by cationic polymers. J. Colloid
Interf. Sci. 42. 448-456.
27. Gregory. 1. (1976) The effect of cationic polymers on the colloidal stability of latex particles.
J. Colloid Interf. Sci. ~. 35-44.
28. Deryagin. B.V .• and Landau. L.D. (1941) Theory of the stability of strongly charged
lyophobic sols and of the adhesion of strongly charged particles in solutions of electrolytes.
Acta Physicochim. URSS H. 733-762.
29. Verwey. E.J.W., and Overbeck. 1.Th.G. (1948) Theory of the Stability of Lyophobic
Colloids. Elsevier. Amsterdam.
30. Reerink. H. and Overbeck. J.Th.G. (1954) The rate of coagulation as a measure of the
stability of silver iodide sols, Disc. Faraday Soc. ~. 74.
31. Overbeek. J.Th.G. (1980) The rule of Schulze and Hardy, Pure and Appl. Chern 52.
1151-1161.
32. Smoluchowski. M. (1917), Versuch einer mathematischen Theorie der Koagulationskinetic
kolloider wsungen. Z. Physik. Chern. 22,129-168.
33. Higashitani. K .• Miyafusa, S .. Matsuda. T .. and Matsuno. Y. (1980) Axial change of total
particle concentration in Poiseuilleflow, J. Colloid Interf.-Sci. 77.21-28.
34. Camp. T.R.. and Stein, P.e. (1943) Velocity gradients and internal work in fluid motion. J.
Boston Soc. Civ. Eng. .3Q. 219-238.
35. Spielman. L.A. (1978) Hydrodynamic aspects offlocculation, in KJ Ives (ed.)The Scientific
Basis of Flocculation. Sijthoff and Noordhoff, Alphen aan den Rijn. pp. 63-88.
36. Cleasby. J.L. (1984) Is velocity gradient a validflocculation parameter? 1. Env. Eng. llQ.
875-897.
37. Matijevic. E. (1973) Colloid stability and complex chemistry. J. Colloid Interf. Sci. 43.
217-245.
38. Bottero. J.Y .• Axelos. M .• Tchoubar. D .• Cases. I.M .• Fripiat. U .• and Fiessinger. F.
(1987) Mechanism of formation of aluminum trihydroxide from keggin AIJ3 polymers.
1. Colloid Interf. Sci. ill, 47-57.
39. Dentel, S.K., and Gossett. 1.M. (1988). Mechanisms of coagulation with aluminum salts.
J. Am. Water Works Assn. 80. 187-198.
40. Packham. R.F. (1965) Some studies of the coagulation of dispersed clays with hydrolysing
salts. J. Colloid Sci. ZQ, 81-92.
41. Amirtharajah. A.. and Trusler. S.L. (1986) Destabilisation of particles by turbulent rapid
mixing. J. Env. Eng. ill, 1085-1108.
42. Flynn. e.M. (1984) Hydrolysis of inorganic iron(lI/) salts, Chern. Rev. 84.31-41.
43. Jahn' S.A.A. (1988). Using Moringa seeds as coagulants in developing countries. J. Am.
Water Works Assn. W. 43-50.
44. Halverson. F .• and Panzer. H.P. (1980) Flocculating agents. in Kirk-Othmer: Encyclopedia
of Chemical Technology (3rd Edition). Wiley. New York. vol. 10, pp. 489-523.
1.5. APPLIED MINERALOGY AND MATERIALS BALANCING
PROCEDURES. EV ALUA TIONS OF FLOTATION CONCENTRATORS

W.PETRUK
CANMET,
555 Booth St.,
Ottawa, Onto KIA DCl
Canada

ABSTRACf. Quantitative mineralogical analysis combined with materials balancing of the mineral
data provide a reliable method for evaluating the behaviour of a concentrator with respect to a
specific ore. The quantitative mineralogical analysis involves determining size distributions of
minerals in unbroken ore pieces to predict the grind for liberating the minerals and determining
quantities, sizes, and degrees of liberations of the minerals in mill products from a concentrating
unit. The materials balancing calculation determines recoveries of each type of mineral-bearing
particle at each point in a flowsheet. Factors that affect recoveries by flotation of Zn, Cu, Pb, Ag,
Au and Sn from volcanogenic base metal orcs have been evaluated by this technique. Recoveries
are discussed for sphalerite from the Brunswick No. 12 deposit in New Brunswick, Canada; for
chalcopyrite, galena, silver and gold from several Canadian deposits; and for Sn from the Kidd
Creek deposit in Ontario, Canada.

1. Introduction

Flotation concentrators need to be assessed both routinely and periodically to obtain maximum
recoveries. Routine assessments are performed to monitor operations, and periodic assessments to
evaluate performance and to identify processing problems. The assessment process involves
collecting and analysing representative samples and interpreting the data. Techniques which can be
used to analyse samples and interpret data are chemical analyses, qualitative mineralogical
investigations, quantitative mineralogical analyses, and materials balancing. The chemical analyses
include XRF analysis and assays. XRF analyses are generally done every ten minutes for process
control and for monitoring concentrator performance. Chemical assays are performed as a
reference for the XRF analysis and on sample suites for concentrator evaluations. Assay data for
sample suites can be processed by materials balance to determine the elcmental recoveries
throughout a circuit.
When problems are indicated by the chemical assays and materials balancing procedure, a
qualitative mineralogical investigation might identify the cause, but a quantitative mineralogical
analysis is needed for detailed interpretations [1]. Quantitative mineralogical data include mineral
quantities, size distributions of free and unliberated mineral grains, percentage of mineral that is
liberated, percentage of mineral unliberated but in recoverable types of particles, percentage of

125
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 125-148.
© 1992 Kluwer Academic Publishers.
126

mineral unliberated but in particles that can be liberated by regrinding, and percentage of mineral
unliberated and unrecoverable. Material balancing is applied to the quantitative mineralogical data to
determine the behaviour of free and unliberated mineral grains during processing. This paper
reviews applications of quantitative mineralogical analysis to flotation and discusses mineralogical
characteristics which have a bearing on recovery of Zn, Cu, Pb, Ag, Au and Sn from volcanogenic
ores.

2. Quantitative Mineralogical Analyses

2.1. MINERAL QUANTmES

Data on mineral quantities provide background information for designing a flowsheet and for
evaluating a circuit, and are used directly for materials balancing. Mineral quantities can be
determined by point counting, image analysis studies of polished sections of ores, x-ray
diffractometer analysis and calculations from chemical assays. The x-ray diffractometer analysis
technique is the least accurate and calculations from chemical assays the most accurate.
Unfortunately, in many instances, the calculations from chemical assays cannot be performed,
because some elements occur in several minerals and their distributions among minerals cannot be
defined. In such cases the mineral quantities need to be determined by either point counting or
image analysis of polished sections. Unfortunately the amount of mineral displayed on the polished
surface is not always representative of the sample. It is recommended that the samples be analysed
chemically for at least one element prior to point counting or image analysis so that the analysed
mineral quantities can be verified. If the error is greater than 15% of the amount present a second
polished section should be prepared and analysed. Another technique of obtaining realistic mineral
quantities involves analYSing a suite of samples and calculating a materials balance. The materials
balance program adjusts the mineral quantity data to best fit (Table 1).

2.2. SIZE DISTRIBUTIONS OF MINERALS

Size distributions are performed on minerals in unbroken ores to predict a grind for liberating the
minerals and on ground materials to determine the sizes of unliberated grains. The size
distributions of mineral grains in unbroken ores can be determined by analysing polished sections
by two techniques, depending upon the mode of occurrence of the mineral. If the mineral occurs as
discrete grains the surface area of each grain in the polished section is measured and the area %
covered by the grains in each size range is calculated. If the mineral occurs in interconnected
masses or veins, its size distribution is determined by chord analysis. The proportion of chord
length within each size range is assumed to be equal to the quantity of mineral in each size range.
The proportion of mineral that would be liberated in each sieved fraction of a ground ore can be
calculated from size distribution data for a mineral in unbroken ore by using appropriate liberation
models [2-5]. The calculated liberation values are approximate and in some instances are not even
close to measured liberations. The poor correlation occurs largely because liberation is dependent
upon the grinding method and on bonding between minerals. Weakly bonded mineral grains break
preferentially along grain boundaries. whereas strongly bonded'mineral grains break randomly.
Most theoretical liberation models assume random breakage and a uniform bonding between
minerals. Unfortunately the bond strength between minerals is different for each mineral pair and is
 127

stronger in recrystallised (metamorphosed) ores than in unmetamorphosed ores.

TABLE 1. Comparison between Analysed and Adjusted Pyrite Contents in Samples

from a Circuit
Flotation Cell Analysed Adjusted Error
(wt %) (wt %) (%)
Feed 9.5 10.2 -6.9
Rougher conc. 20.6 21.2 -3.0
Rougher tail 11.7 10.8 +8.4
Scavo conc. 38.8 39.0 -0.5
Scavo tail 9.4 10.2 -7.8
C1. 1 Conc. 16.7 12.2 +36.9
C1. 1 Tail 41.8 37.8 +10.6
C1. 2 Conc. 8.5 9.3 -8.8
C1. 2 Tail 34.3 36.5 -6.0

Since weakly bonded grains tend to separate readily during grinding (preferential breakage),
they tend to produce liberated grains that arc nearly the same size as the grains in unbroken ore. On
the other hand, well bonded grains break randomly and liberation is not achieved, until the grains
are broken into smaller sizes than the grain sizes in unbroken orcs. It has been found that polished
sections of weakly bonded grains display incipient fractures and pits along grain boundaries,
whereas polished sections of well-bonded grains do not. The incipient fractures and pits are used
in image analysis studies to detect bond strength between grains and to determine grind sizes for
separating weakly bonded grains [6,7].
The author uses the Finch-Petruk model for predicting liberation by random breakage and an
empirical liberation model [8] for preferential breakage. The theoretical model [3] predicts the
liberation in each sieved fraction and the empirical model predicts the "minimum" and "optimum"
grinds for liberating the minerals in an ore. The term "minimum" grind, as used in the empirical
model, is defined as the grind at which mineral liberation begins to increase substantially with
decreasing particle size (grind). The condition for "minimum" grind is generally met for an ore
with preferential breakage, when the grind produces a particle size distribution that is equal to the
size distribution of the mineral in the unbroken ore. The term "optimum" grind is defined as the
grind size at which the incremental increase in mineral liberation becomes small as the particle size
decreases. The condition for optimum grind is generally met for an ore with preferential breakage
when the Kso grind size is equal to K30 of the size distribution for the mineral in unbroken ore[9].
An example of an application of the empirical model is given by an evaluation of chalcopyrite in
the Trout Lake ore [1,9]. The size distribution for the chalcopyrite is given in Figure 1. It is
interpreted that the "minimum" grind for chalcopyrite liberation is Kso =75 IllTl and the "optimum"
grind is Kso = 30 1llTl. The company grinds the flotation feed to Kso = 68 IllTl and obtains a high
recovery[IO]. It has been determined that most of the chalcopyrite in this ore is coarse grained and
is liberated by preferential breakage. A small proportion of the chalcopyrite in the ore is, however,
very fine grained and is liberated by random breakage.

2.3. MINERAL LIBERATIONS

The procedure for determining mineral liberations involves analysing polished sections of the
128

ground material and measuring the proportion of mineral that is liberated, partly liberated and
unIiberated.
In some instances a grain which appears liberated in polished sections may be unliberated since,
in the third dimension, another mineral could be atlached to it. This factor is minimised and can be
disregarded when relative instead of absolute values are used and the fractions analysed have a
wide size range. The author measures mineral liberations by subdividing the mineral as occurring
in particles which contain 0-10, 10-20, 20-30, 30-40, 40-50, 50-60, 60-70, 70-80, 80-90,
90-99.9 and +99.9 area % of the mineral. Free mineral grains are in the +99.9 area % category.
Dala for chalcopyrite in a Copper rougher flotation feed is shown in Figure 2. The bottom curve in
the figure gives the cumulative size distribution for totally liberated chalcopyrite (+99.9 vol %
particles); the second curve is the size distribution for chalcopyrite in +90 vol % particles (90-99.9
plus +99.9 vol % particles); the third curve is size distribution for chalcopyrite +80 vol % particles
(80-90 plus 90-99.9 plus +99.9 vol % particles), etc. The right ends of the curves give the total
amounts of chalcopyrite that are in each type of particle, i.e 40 % in +99.9 vol % particles, 61 %
in +90 vol % particles, 68 % in +80 vol % particles, etc. The top curve in Figure 2 is the size
distribution of chalcopyrite in all particles.

SIZE DISTRIBUTIONS

120

c 100
u
m
u
I 80

60

40
w
t OPTIMUM GIRIN[)
20
" oL-__L -_ _ ~ _ _~_ _~_ _- L_ _- L_ _~_ _~

13.2 26.5 37.5 73 106 150 212 300

micrometres

-.- cp In ore --+- Flotation feed --+- Cu cone

Figure 1. Size distributions of chalcopyrite in unbroken ore and sieve analysis of flotation feed and Cu
concentrate. Predicted minimum and optimum grinds, using the preferential liberation model, are indicated.

3. Evaluations of Mineral Behaviour During Processing

3.1. OPTIMUM RECOVERY

It is assumed that the maximum percentage of mineral that can be recovered from an ore is equal to
the percentage of mineral that is in types of particles which are recovered in the concentrate. It has
been detennined [11] that concentrates recover both liberated and partly liberated grains, depending
upon concentrate grade.
In particular zinc concentrates grading 50 to 54 wt% Zn (out of about 60 wt % Zn in sphalerite)
 129

LIBERATION OF CHALCOPYRITE
ROUGHER FLOTATION FEED: Trout Lake ore
Amount cp
100
91 %

80 82 %
P 68 %
E
R 60 61 %
C 40 %
E 40
N
T
20

0
0 3.3 4.7 6.7 9.4 13.3 18.8 26.5 37.5 53
SIZE (in micrometres)

-90 vol % cp-l- -80 vol % cp--*- -70 vol % cp-B- -60 vol % cp
----- -40 vol % cp-4-- -20 vol % cp-8-- -10 vol % cp ---£- Total cp

Figure 2. Line graph showing size distributions, quantities and relative proportions of liberated, partly
liberated and un liberated chalcopyrite in feed to a rougher flotation cell (cp=chalcopyrite).

tend to recover particles containing more than 70 vol % sphalcrite and chalcopyrite concentrates
grading 20 to 24 wt% Cu (out of 34 wt% Cu in chalcopyrite) tend to recover particles containing
more than 50 vol% chalcopyrite. It was proposed [Ill that the term "liberated" be used to describe
the type of particle that is recovered in a particular type of concentrate.
Recoverable particles are also size dependent. For example flotation cells recover "liberated"
sphalerite between about 5 and 100 11m in diameter and "liberated" chalcopyrite between about 3
and 75 11m. An ore must, therefore, be ground to produce the "liberated" grains within a specific
size range. Consequently optimum recovery is percentage of mineral that, in a properly ground
product, is in "liberated" grains with respect to the required concentrate grade and in the
appropriate size range for the concentrating process.
The types of particles recovered in concentrates can be determined by analysing a suite of
samples, including feed, concentrate and tailings from a circuit and calculating, by materials
balance, the recovery of each type of particle in the concentrate. Results for chalcopyrite recovery
in the Cu concentrate from Trout Lake ore are given in Figure 3. They show that 98 to 100 % of
the chalcopyrite is recovered in 3.3 to 106 11m particles containing more than 99.9 vol % (free)
chalcopyrite and in 10 to 53 ~ particles containing 60 to 99.9 vol % chalcopyrite. The recoveries
drop drastically for particles containing less than 60 vol % chalcopyrite. Hence, for Trout Lake
ore, the term "liberated" chalcopyrite applies to particles containing more than 60 vol %
chalcopyrite (Figure 3b). It is noteworthy that the minimum size for high recovery is 3.3 /1ffi for
free chalcopyrite particles and increases to 10 11m for particles containing 60 to 70 vol %
chalcopyrite. It is also noteworthy that the recoveries of chalcopyrite in minus 10 /1ffi particles
 120"------------------------------------------, 120,1-------------------------------------------,
""o

R R +
e e
c c
o o
e e
r (a) r (b)
y y
'II. 'II.
20 20
01L-__i -__ ~ ___ L_ _ ~ _ _~_ _ _ _L __ _~_ __ L_ _- L__~ oI >----c=~ "'-'1 'Sf
0-3.3 4.7 6.6 9.4 13.2 16.7 26.5 37.5 53 75 106 0-3.3 4.7 8.6 9.4 13.2 18.7 28.6 37.5 63 76 106
Grain Sizes (micrometres) Grain Sizes (micrometres)
• 80 .. cp --+- 80-90" cp • 90 'II cp -+- 80-90" cp ---*- 70-80 .. cp 80-70 .. cp
---*- 70-80" cp 80-70 .. cp 50-60 .. cp ---+- 40-50" cp ---A- 10-20 .. cp 0-10 .. cp

120rl---------------------------------------------,

R
e
o
v
e "
r "
y
'II. (c)
20

OLI__- L__ ~ ____L __ _L __ _ ~ _ __ L_ _ ~ _ _ _ _L __ _i __ _ ~

Figure 3. Recoveries from flotation feed in Cu concentrate with respect to different sizes of
0-3.3 4.7 6.6 9.4 13.2 18.7 26.5 37.5 53 75 106
chalcopyrite grains and particle types: (a) chalcopyrite grains in particles containing +90, 80-90,
Grain Sizes (micrometres)
70-'80, and 60-70 vol % chalcopyrite, (b) chalcopyrite grains in particles of all types from 0-10 to
- - . &0 .. cp -+- .150 .. cp ---*- 10-60 .. cp -3- 0-10 .. cp +90 vol%. (e) summary diagram.
 131

containing 1 to 60 vol % chalcopyrite is the same as for minus 10 IJ.Il1 particles containing 60 to 90
vol % chalcopyrite. This indicates poor selcctivity for particles smaller than 10 IJ.Il1 in diameter.

3.2 . MATERIALS BALANCE OF LIBERATED GRAINS

The liberation for each mineral should be sub-divided into the following categories:
1) proportion of mineral occurring as "liberated" grains,
2) proportion that is unliberated but could be liberated by regrind (mineral occurs as large grains
in particles which contain enough mineral to be liberated, for example particles containing 30
to 70 area % of the mineral), and
3) grains which cannot be liberated by regrind (small grains normally in particles with less than
30 area % of mineral).
The amount of mineral in each category is processed by materials balance and the recovery of
mineral in each category of particle is calculated. The recovery of chalcopyrite in different types of
particles in the Trout Lake Cu circuit is given in Table 2 as an example. It is apparent from Table 2
that the circuit is operating efficicntly since 97.5% of the + 70 vol % chalcopyrite is recovered. On
the other hand 14% of the chalcopyrite in the ore is in 70-30 vol % particles and only 44.6% of this
chalcopyrite is recovered. Regrinding the scavenger concentrate and first cleaner tail might increase
chalcopyrite liberation and in turn chalcopyri te recovery.

TABLE 2 Recovenes 0 fChal copynte . ( cp;).III d·ffi

I erent Partlc
·1e Categones.
Sample Recoveries (%)
Free cp +70%c~ 70-30%cp -30% cp_
Feed 100.0 100.0 100.0 100.0
Rougher conc. 108.0 106.8 73.4 64.0
Rougher tail. 3.1 6.8 66.0 103.7
Scavo conc. 2.7 4.3 10.6 23.9
Scavo tail 0.4 2.5 55.4 79.8
Cl.l tail 8.4 9.3 28.8 43.8
Cl.2 tail 5.4 4.7 6.6 5.4
Cl.I conc. 105.0 102.3 5l.2 25.6
Cl.2 conc. 99.6 97.5 44.6 20.2
Circuit conc. 99.6 97.5 44.6 20.2
Circuit tail. 0.4 2.5 55.4 79.8
Prop. cp· 59.7 8l.5 14.0 4.5
* Prop. cp = proportIon of chalcopynte In each category of partJcle

3.3. ROLE OF IMAGE ANALYSIS-MATERIALS BALANCE EVALUATION

The image analysis-materials balance evaluation is an expensive procedure, but is cost-effective

when performed at appropriate times. It should be perfonned:
1) on products from a pilot plant test to help in designing the flow sheet,
2) on a suite of products from a concentrator to detennine whether major circuit modifications
would improve recoveries, and
3) on a suite of products from any operation when major recovery problems are encountered and
132

chemical assays and mineralogical investigations have not provided an adequate answer.

4. Mineralogical Evaluations of Volcanogenic Base Metal Ores

Volcanogenic base metal ores contain about 1.5 to 15 wt % Zn, 0.2 to 5 wt % Cu, 0.01 to 4 wt %
Pb, 10 to 500 ppm Ag, 0.2 to 7 ppm Au, 0.02 to 0.3 wt % Sn, and small amounts of other
elements. The Zn, Cu and Pb are recovered by flotation in Zn, Cu and Pb concentrates,
respectively, and the Ag and Au are recovered as by-products in the same concentrates. Most of the
Sn is rejected, but some can be recovered from the tails in a separate Sn concentrate. High Zn, Cu
and Pb recoveries are obtained when the ore is relatively coarse grained, but many volcanogenic
depOsits are fine grained; hence, recoveries are frequently low. To improve recoveries it is
necessary to fine-tune the concentrator with respect to the mineralogical characteristics of the
specific ore. Mineralogical characteristics which affect mineral recoveries from Canadian
volcanogenic base metal ores have been studied at CANMET by microscopy and image analysis of
mill products since about 1975 and the findings are summarised in the following parts of this
paper.

4.1. CHARACTERISTICS OF VOLCANOGENIC DEPOSITS

Volcanogenic dcposits generally consist of pyritc-rich massive sulphide bodies which contain
sphalerite, chalcopyrite and galena masses and are intcrlayered with volcanic rocks. Pyrite is the
major mineral and is commonly massive. The massive pyrite is gcncrally composed of relatively
small anhedral grains. Relics of framboidal pyrite and large recrystallised pyrite grains are present
in most deposits. Numerous interstices are present between the anhedral pyrite grains. The
interstices are usually filled with associated minerals such as sphalerite, galena. chalcopyrite.
quartz. carbonates and. to a minor extent. tetrahedrite-freibergite. Most of the interstitial fillings
form discrete grains. but where the pyrite has bcen weakly fractured. the interstitial minerals occur
as hairline veinlets along the fractures. In metamorphosed ores some of the pyrite is partly
recrystalIised to relatively large euhedral pyrite crystals. Some of the euhedral pyrite crystals
contain encapsulated inclusions of other minerals. particularly galena and sphalerite. Variable
proportions of the chalcopyrite. sphalerite and galena occur as masses. dependent upon the type of
volcanogenic ore. The main sulphide minerals in volcanogenic deposits are pyrite. sphalerite.
galena and chalcopyrite; minor minerals include pyrrhotite. magnetite. arsenopyrite, barite.
monazite, and tetrahedrite-freibergite; and trace minerals include cassiterite. stannite, pabstite,
bournonite. meneghenite. wittichenite. pyrargyrite. stephanite. acanthite. silver. electrum. gold-
silver-mercury alloys. dyscrasite. naumannite. hessite. volynskite, miargyrite, freieslebenite.
clausthalite. andorite. owheeite, polybasite, eosalite, kobellite. native antimony. native bismuth and
marcasite.

4.2. FACTORS WHICH AFFECT ZINC RECOVERY

4.2.1. Sphalerite Liberation. The sphalerite in volcanogenic base metal ores occurs as masses, as
interstitial fillings in massive pyrite and as inclusions in large recrystallised pyrite grains. Since
sphalerite is the primary ore mineral in these ores, the grinding circuit is usually designed to grind
the ore to a size that will liberate the sphalerite and prepare it for flotation [12]. During primary
grinding. which is about 80% minus 75 ~m (200 mesh) for most volcanogenic ores, much of the
 133

massive sphalerite is liberated but some remains attached to grains that were inclusions in the
massive sphalerite.
Some of the sphalerite, which occurred in interstices between pyrite, is partly liberated by
breaking moderately bonded pyrite grains along grain boundaries and releasing the enclosed
interstitial fillings [6,13]. In the primary grinding circuit discharge, however, much of the
interstitial sphalerite remains attached to pyrite.
Regrinding is required to liberate the relatively large sphalerite grains which remain attached to
pyrite. The degree of regrinding depends upon the quantity and grain size of the unliberated
sphalerite, and on the grain size of the pyrite. Intense regrinding may produce slimed sphalerite
because sphalerite is much softer than pyrite and breaks into smaller grains than the sieve grain size
for the reground ore [8]. For example, in a study of ore from Brunswick Mining and Smelting, the
grain size of liberated sphalerite was 80% minus 39 ~, at a primary grind of 80% minus 54 !lm;
in the regrind stream, the free sphalerite was 80% minus 29 !lm, at a regrind size of 80% minus
37 ~. The small proportion of the sphalerite that occurs as inclusions in recrystallised pyrite is
seldom liberated because intense grinding is required to break the host pyrite crystals across grain
boundaries.

4.2.2. Sphalerite Properties Which Affect Flotation. Characteristics which affect flotation of
liberated sphalerite are the maximum and minimum sizes of the particles, activation by copper, and
possibly, grain surface activities. Characteristics which affect the flotation of unliberated sphalerite
are the quantities of sphalerite in the particles (degree of liberation), sizes of particles containing
unliberated sphalerite, sizes of sphalerite inclusions, activities of secondary copper sulphides, and
possibly, grain surface activities. A suitable equipment for determining the percentages of liberated
and unliberated minerals, sizes of the liberated mineral, percentages ofunliberated mineral in host
particles, percentage of secondary copper sulphides and other related features is an automatic
microbeam-assisted image analysis system [14-17].

TABLE 3 Recoveries of "Liberated" Sphalerite from Zn Circuit Feeds

Size range (~) A Kidd Creek C Kidd Creek Selco Mattagami Average
3.3-4.7 82.2 94.5 94.0 82.7 88.3
4.7-6.6 97.2 91.2 94.1 88.2 92.7
6.6-9.4 98.7 94.1 97.8 91.1 95.4
9.4-13.2 97.0 97.3 90.4 94.1 94.7
13.2-18.8 95.4 98.7 94.0 95.9 96.0
18.8-26.5 98.7 99.1 94.0 99.5 97.8
26.5-37.5 100.0 99.1 97.3 99.8 99.0
37.5-53 100.0 98.1 98.9 99.9 99.2
53-74 100.0 100.0 97.4 100.0 99.3
74-106 100.0 94.5 100.0 98.6
106-150 100.0 100.0 100.0
Total sample 95.9 95.9 93.1 94.1 94.7
%"Lib." 86.4 86.2 88.0 86.9

4.2.3. Recovery of Different Sizes of "Liberated" Sphalerite. Analyses were conducted to

determine recoveries of different sizes of "liberated" sphalerite grains in commercial Zn
134

concentrates. Suites of samples composed of feed, concentrates and tails were obtained from
several concentrators and were analysed by image analysis. Recoveries of "liberated" sphalerite
("liberated" sphalerite in this test was defined as sphalerite in particles containing more than 90
vol% sphalerite) were determined for grains from each size range of the Tyler series from 3.3 to
150 !lm. The results (Table 3) showed that more than 93% of the "liberated" sphalerite was
recovered in Zn concentrates from Zn circuit feeds.
The average results for the four concentrators show that recoveries of "liberated" sphalerite in
grains 26.5 to 106 !lm are 99 to 100%. The recoveries decrease with decreasing size to 95% for
9.4 to 13.21lffi grains, but are still 88% for grains 3.3 to 4.7 Ilffi in diameter (Table 3). Results for
bench tests show that recoveries decreased from 98 - 100% at 18.8 to 741lffi to 95% for grains 9.4
to 13.2 Ilffi in diameter. The recoveries continue to decrease to 70% for grains 3.3 to 4.7 Ilffi in
diameter and are around 60% for grains 1 to 3.3 Ilffi in diameter (Table 4). The higher recovery for
the minus 10 !lm sphalerite in concentrators than in bench tests suggests that commercial size
roughers and cleaners recover more fine grained "liberated" sphalerite grains. The maximum size
of sphalerite that tends to float has been determined by Gaudin [18] as being around 150 Ilffi, and
the writer has confirmed this grain size in many instances.

T A BLE 4 . Recovery 0 f"·be :;pl a ente f rom Faro Sample: Benc h Test.
LI rated"Shl·
Size range (Ilffi) Recovery of "liberated" sphalerite (%)
1-3.3 60
3.3-4.7 70
4.7-6.6 84
6.6-9.4 90
9.4-13.2 95
13.2-18.8 91
18.8-26.5 98
26.5-37.5 99.3
37.5-53 98
53-74 100.0
Total sample 92.7
% "Lib." in conc. 97.7

4.2.4. Sphalerite Activation by Secondary Copper Sulfides. Minor amounts (0.01 to 0.1 wt %) of
secondary copper sulphides such as covellite, chalcocite, digenite and native copper may occur in
the ore. When more than 0.05 wt % is present they activate sphalerite to such an extent that
separate Cu, Pb and Zn concentrates cannot be produced without corrective measures. The
activation can be suppressed by using enough sodium cyanide to react with all of the Cu 2+ ions in
solution. In some instances other reagent~ are used to depress activated sphalerite.

4.2.5. Liberated Sphalerite in Copper Concentrates. The Zn content of Cu concentrates from base
metal orcs varies from about 2 to 8 wt%. When it is higher than 3 wt%, there is a loss of revenue
and it may create a problem in copper smelting. The Cu concentrate from the Trout Lake deposit at
Rin Ron, Manitoba [10] contains around 5 wt% Zn which accounts for 10 to 12% of the
sphalerite in the ore. Image analysis studies have shown that about 80% ofthe sphalerite in the Cu
 135

concentrate is "liberated" and has the same grain size as the "liberated" sphalerite in the Zn
concentrate. Hence, about 10% of the "liberated" sphalerite in the ore is activated by the reagents in
the Cu rougher circuit. The reason for sphalerite activation is unknown, but tests at CANMET cr.
Negeri, personal communication) with sphalerite depressants have reduced the sphalerite content of
the Trout Lake Cu concentrate without much loss in Cu recovery. Some of the sphalerite in the
Trout Lake ore has the "chalcopyrite disease" [19], but the proportion of the "liberated" sphalerite
with the chalcopyrite disease in the Cu concentrate is small. A study on Brunswick Mining and
Smelting and on Caribou ores has shown that only about 10% of the chalcopyrite disease that
occurred in sphalerite in unground ore remained in sphalerite grains when the ore is ground [20].
The low proportion of chalcopyrite disease in ground products with respect to the amount in
unground ore suggests that, during grinding, a large proportion of the sphalerite breaks along
cleavage planes. Since many of the chalcopyrite globules in sphalerite are aligned along the
cleavage planes, they are liberated from the sphalerite during grinding. On the other hand, it is
possible that the sphalerite which hosts chalcopyrite disease may still contain sub-micrometre
chalcopyrite grains. Such chalcopyrite grains, if present, would activate sphalerite.

4.2.6. Recovery 0/ Sphalerite/rom the Ore 0/ BMS. A study of the characteristics of sphalerite in
the ore of Brunswick Mining and Smelting (BMS) and of the behaviour of the mineral in the
concentrator was undertaken in 1976 to determine whether it is possible to improve Zn recoveries
by modifying the concentrator [21]. Zinc distributions were as follows: Zn concentrate 74.5%, Pb
concentrate 7.2%, bulk concentrate 3.3%, Cu concentrate 0.1 % and concentrator tails 14.9%. The
study was performed by: (1) collecting samples of the feed, concentrate and tails from each bank
of flotation cells, (2) determining, with an image analyzer, the proportion of sphalerite that
occurred as "liberated" (+90 vol%) and unliberated grains of each size range in each sample, and
(3) performing a materials balance [22] to calculate the recovery of "liberated" and unliberated
sphalerite of each size at each point in the flowsheet. It was found that the zinc concentrate
recovered 96% of the +90 vol% ("liberated") sphalerite from the zinc circuit feed; the main loss
was as sphalerite slimes (-3.3 ~m) [21]. Hence, it was concluded that the zinc concentrator was
operating efficiently. On the other hand, 5% of the sphalerite in the ore was lost to tails as
unliberated sphalerite grains that range from 15 to 75 ~ in diameter, with the mean size being
26 ~. It was evaluated that a regrind to 80% minus 400 mesh would liberate enough sphalerite to
recover an additional 3% Zn and this is equivalent to about 6,300,000 kg of Zn per year. A series
of grinding tests was conducted at Lakefield Research during the period of 1980 to 1982 [23]. The
grinding tests confirmed the mineralogical evaluation and the regrind circuit in the concentrator was
modified from an open to closed circuit regrind. Initial Zn recoveries after the modification were
around 82%, but in several years, decreased to around 78%.
The concentrator was resampled in March 1987 and the samples were re-studied [24].
Fortunately, the performance of the Zn circuit was poor during the sampling campaign and
emphasised the problems which had developed. Zinc distributions were as follows: Zn concentrate
75.3%, Pb concentrate 2.9%, bulk concentrate 5.3%, Cu concentrate 0.4% and concentrator tails
16.1 %. The recovery of "liberated" (+95 vol%) sphalerite in the Zn concentrate from Zn circuit
feed was 88% rather than 95%. The loss of "liberated" sphalerite to the tails was 8.5% of the
sphalerite in the ore, with 4.8% in "liberated" grains smaller than 9.2 ~m and 3.7% in "liberated"
grains larger than 9.2 ~. The unlibcrated sphalerite (7.6% of the sphalerite in ore) in the tails was
in grains which are too small to be liberated by further regrinding. The re-study showed that the
ore had been ground to the proper size to achieve liberation, but that the flotation cells lost large
amounts of the slimes (minus 9.2 ~m "liberated" particles). It reconfirmed that a coarser grind
136

would not achieve the desired liberation. Another study [6] has shown that the regrind of 80%
minus 37 !ID1 is the optimum grind for BMS ore and that finer grinding would increase liberation
only by a marginal amount. The company has undertaken research on the flotation conditions to
find a technique for recovering fine grained "liberated" sphalerite.

4.3. FACfORS WHICH AFFECf COPPER RECOVERIES

4.3.1. Mode of Occurrence of Chalcopyrite. The chalcopyrite content in some volcanogenic

deposits is high and is nearly absent in others. Furthennore chalcopyrite-rich volcanogenic
deposits frequently contain a chalcopyrite zone, a Cu-bearing feeder zone and a chalcopyrite-
sphalerite zone, as well as massive pyrite. The chalcopyrite in the chalcopyrite zone occurs as
relatively large grains and masses; in the feeder zone as disseminated grains in rock, and in the
chalcopyrite-sphalerite zone as relatively coarse-grained intergrowths with sphalerite and pyrite.
The chalcopyrite in massive pyrite occurs as interstitial fillings between pyrite grains, as
chalcopyrite veinlets in fractured pyrite, and as minute inclusions in pyrite. A small portion of the
chalcopyrite also occurs as chalcopyrite disease in sphalerite.

4.3.2. Chalcopyrite Liberation. The chalcopyrite masses and the discrete chalcopyrite grains in the
various zones are approximately the same sizes as the sphalerite masses and discrete sphalerite
grains. Hence, the grind required to liberate the chalcopyrite is generally the same as the grind for
liberating the massive and coarse grained sphalerite in the same ore. In contrast, whenever a
volcanogenic deposit has a low copper content, the chalcopyrite liberation in the primary grinding
circuit is low because the mineral is fine grained. Most of the chalcopyrite in such deposits occurs
in massive pyrite in interstices between pyrite grains, and will be partly liberated and partly
exposed when the ore is ground and reground to a size that separates the pyrite grains (established
from pyrite grain sizes). Consequently, the chalcopyrite liberation is higher than would be expected
from a liberation model (both random and preferential breakage) which uses chalcopyrite grain
sizes to predict liberation. Nevertheless, much of the fine grained chalcopyrite remains attached to
the pyrite even in reground ore, and relatively low grade Cu concentrates (20 to 24 wt% Cu) must
be produced to obtain acceptable Cu recoveries.

4.3.3. Chalcopyrite Properties Which Affect Flotation. Characteristics which affect flotation of
liberated chalcopyrite are maximum and minimum sizes of the particles, and activation of the
chalcopyrite. Since chalcopyrite is readily activated, the main factor to be considered is particle
size. It has been shown in the first section of this paper that liberated chalcopyrite grains 3.3 to
106 !ID1 in diameter float readily.
Characteristics which affect flotation of unliberated chalcopyrite are particle sizes and degree of
liberation of the chalcopyrite. A test on concentrating chalcopyrite at different flotation times has
shown that liberated chalcopyrite grains were fast floating, and the time required to float
chalcopyrite increased as the amount of chalcopyrite in the particles decreased (or conversely, as
the amount of gangue in each particle increased). It has been shown in the first sector of this paper
that partly liberated chalcopyrite grains 10 to 53 11m in diameter tend to be recovered in Cu
concentrates.

4.3.4. Recovery of Chalcopyrite from Low-Copper Volcanogenic Ores. The deposit of

Brunswick Mining and Smelting (BMS) is an example of a low-copper volcanogenic ore. The
deposit contains about 0.3 wt% Cu, mainly in the pyrite-rich zones as interstitial fillings between
 137

pyrite. A small proportion of the chalcopyrite also occurs in a chalcopyrite zone as large grains. A
test on the ore established that particles containing more than 50 vol % chalcopyrite tend to be
recovered in Cu concentrates grading 22 to 24 wt % Cu. Hence + 50 vol % chalcopyrite particles
are considered as "liberated" [6]. It was detennined that 43% of the chalcopyrite was "liberated" at
a grind of 80 % minus 75 Jlm (200 Mesh), and 59 % was "liberated" at a regrind to 80 % minus 37
Jlffi (400 mesh). The liberation did not increase to any extent with further regrinding. An evaluation
of concentrator perfonnance showed a Cu recovery of 67 %; 61 % being "liberated" chalcopyrite
and 6 % unliberated chalcopyrite. Most of the liberated chalcopyrite grains were recovered in the
Cu concentrate in 2 to 30 Jlm particles (a large proportion being between 4.7 and 13.2 Jlffi), and
some liberated chalcopyrite grains 8 to 60 Jlm in diameter were lost to the tails. Most of the
unliberated chalcopyrite associated with sphalerite was recovered in the Zn concentrate, and
unliberated chalcopyrite associated with pyrite was recovered in the Cu concentrate and tails.

4.3.5. Recoveries ojChalcopyritejrom High-Copper Volcanogenic Ores. The Trout Lake deposit
is an example of a high-copper volcanogenic ore. Thc deposit contains about 2 wt % Cu, 3 wt %
Zn and 0.05 wt % Pb. Most of the chalcopyrite occurs in the chalcopyrite, chalcopyrite-sphalerite
and feeder zones and is relatively coarse grained. At a grind of 80 % minus 56 Jlffi around 92 % of
the chalcopyrite is "liberated" (for Trout Lake ore "liberated" chalcopyrite was detennined to be in
particles which contain more than 60 vol % chalcopyrite). The size distribution of the "liberated"
chalcopyrite (+ 60 vol % chalcopyrite particles) in the ore is 80% minus 26 Jlm. Around 92 % of
the Cu in the ore is recovered in the Cu concentrate grading about 28 wt % Cu. About 4 % of the
Cu is in the Zn concentrate and 4 % is in the tails. Recovery of "liberated" chalcopyrite is 98 %
with the free chalcopyrite in the concentrate being 80 % minus 24 Jlffi. Most of the chalcopyrite lost
to the Zn concentrate is attached to sphalerite, but some occurs as large "liberated" chalcopyrite
grains. Some of the chalcopyritc in the tails is attached to pyrite, but some is in "liberated" grains
of all sizes.

4.4. FACTORS WHICH AFFECT LEAD RECOVERIES

The galena in volcanogenic ores is usually finer grained than the major economic minerals
(sphalerite and chalcopyrite) which are used to establish the grind size for the ore. Consequently,
galena liberation in ground products is frequently low. To obtain acceptable Pb recoveries, low
grade Pb concentrates (about 20 to 65 wt% Pb) are produced. Stoichiometric galena contains 86
wt% Pb. A summary of the grades and recoveries for the ores of Brunswick Mining and Smelting
in New Brunswick [21], Heath Steele Mines in New Brunswick [25], Trout Lake ore in Manitoba
[10], and the Faro deposit in the Yukon [26J is given in Table 5.
The galena in volcanogenic ores occurs as intergrowths with sphalerite, chalcopyrite and pyrite;
as interstitial fillings between pyrite grains; as masses; as veinlets in pyrite; as minute inclusions in
pyrite; and as rounded inclusions in large recrystallised pyrite grains. The intergrowth galena is
well bonded to the sphalerite, chalcopyrite and pyrite [27]. The rounded galena inclusions in
recrystallised pyrite occur as small grains encapsulated in the large pyrite crystals.

4.4.1. Galena Liberation During Grinding. During grinding, the parts of the ore that contain
galena intergrowths break randomly rather than along grain boundaries, because the intergrowth
galena is well bonded to the other minerals, particularly sphalerite. In contrast the pyritic portions
of the ore break along pyrite grain boundaries and other zones of weakness [6] and thereby release
138

the interstitial mineral grains including galena. The very minute galena grains (1 to 5 !IDl) in pyrite,
however, arc not liberated during primary grinding nor during regrinding because they are so
small. The rounded galena inclusions enclosed in large reerystallised pyrite crystals are not
liberated during primary grinding because the pyrite crystals are difficult to break.

TABLE 5. Assays and Distributions of Lead in Ores

Ore Assay (Pb, wt%) Distributions (Pb,%)
Pb Cu Zn Pb Cu Zn
Feed conc. conc. conc. Tails Feed conc. conc. conc. Tails
Faro 3.6 62.0 2.4 0.6 100 80.1 - 5.7 14.3
BMS 4.3 35.0 15.4 2.4 1.2 100 64.9 1.5 8.0 21.1
Heath 1.6 28.9 5.0 1.1 0.7 100 49.8 8.8 4.4 37.0
Steele
Trout 0.04 . 0.07 0.38 0.02 100 . 13.1 45.5 41.4
Lake

During primary rod mill grinding, most of the galena masses arc liberated and are present as
relatively large free grains and as slimes (minus 4 !IDl grains). In most grinding circuits, the large
free galena grains are directed to the cyclones where they pass into the cyclone underflow and
thence to the primary ball mill. The liberated galena grains are thus reground in the primary ball
mill into smaller grains and galena slimes. The small galena grains and slimes pass into the cyclone
overflow as flotation feed. It appears reasonable to float the large free galena grains from the rod
mill discharge before the grains arc ground into slimes in the ball mill. Most grinding circuits,
however, do not have a Pb circuit for recovering the liberated galena from the rod mill discharge.
The benefits of such circuits arc probably too small and associated problems too large to warrant
installation and operation.
A laboratory scale liberation study was carried out on galena from the ore of Brunswick Mining
and Smelting [6]. The size analysis of galena in unbroken ore was determined by image analysis
(Table 6). It was calculated, using the galena size analysis and the preferential breakage liberation
model [8], that 57% of the galena would be liberated at a grind of 80% minus 37 !IDl; and that 71 %
of the galena would be liberated at a grind of 80% minus 22 !IDl.

TABLE 6 Size Analysis· of Galena in Unbroken BMS Ore

Size range (!IDl) wt%**
0-4.7 12.9
4.7-9.4 11.9
9.4-18.7 18.2
18.7-37.5 22.2
37.5-75.0 23.1
75-150 11.7

.Total
detenmned by Image analYSIS
100.0

·"measured area%; area% =vol% =wt%

A study of Pb concentrates from BMS ore established that particles which contain 50 vol% or
more galena tend to be recovered. Hence, the +50 vol% galena particles are referred to as
 139

"liberated" galena. Grinding tests conducted on the same ore showed that about 28% of the galena
in the rod mill discharge was "liberated", and an additional 44% of the galena was "liberated" in the
equivalent of the company's flotation feed. No increase in "liberation" was obtained in the
equivalent of the company's regrind mill. These tests indicate that about 72% of the galena in the
ore is "liberated" in the primary grinding circuit, and that there is little or no increase of "liberation"
in the regrind mill. The results agree with the performance of the galena in the concentrator; 65 to
70% of the galena is recovered in the company's Pb concentrate with or without regrinding
[21,24].

4.4.2. Composition of Galena in Volcanogenic Ores. The galena in volcanogenic ores contains
trace amounts of Ag, Bi and Se in solid solution. The Ag content in galena varies from a few ppm
to 1 wt%, with an average Ag content around 0.1 wt %. When large quantities of galena are
present in the ore, the solid solution Ag in galena accounts for a substantial proportion of the total
Ag. For example. the galena in the ore of Brunswick Mining and Smelting contains about 30% of
the Ag in the ore [21] and the galena in the Faro deposit contains about 65% of the total Ag.
The Bi distribution in galena is more erratic than the Ag distribution. The quantity varies from
below the microprobe detection limit (250 ppm) to 2.4 wt% Bi. A linear relationship between Bi
and Ag contents was noted for the Heath Steele galena [25]. and Bi-rich galena grains from the
ores of Brunswick Mining and Smelting have high Ag contents (>0.6 wt% Ag). The same
relationship was not observed for the galena from the Trout Lake deposit in Manitoba.
The galena in the Trout Lake ore contains trace amounts of Se. There is no upper limit for the Se
content in the galena since the mineral forms a solid solution with c1austhalite (PbSe) and grains
with every composition between galena and c1austhalite have been found in the ore [19]. Trace
elements, such as selenium, occur as part of the galena structure; hence, they follow galena in the
processing circuit.

4.5. FACTORS WHICH AFFECT SILVER RECOVERY

4.5 .1. Mode of Occurrence of Silver. The average Ag teneur in volcanogenic deposits varies from
around 10 to 500 ppm, but most deposits contain 50 to 100 ppm. The silver in unmetamorphosed
deposits occurs as a constituent of tetrahedrite-freibcrgite, and as solid solutions in galena,
sphalerite, chalcopyrite, and pyrite. In metamorphosed deposits, the element also occurs as a
constituent of tetrahedrite-freibergite, and as solid solutions in galena, sphalerite, chalcopyrite,
pyrite and marcasite. A small proportion of the Ag in metamorphosed deposits is, however,
present as minute grains of silver, electrum, Au-Ag-Hg alloy. pyrargyrite, stephanite, polybasite,
miargyrite, freislebenite, andorite, owheeite, dyscrasite, naumannite, hessite. volynskite, kobellite
with up to 8 wt % silver, cosalite with up to 2 wt % silver, and probably other Ag sulphides,
sulphantimonides, selenides and tellurides. Acanthite may also be present, but generally occurs in
zones of surface oxidation and along faults.
The tetrahedrite-freibergite commonly occurs as irregular grains associated and intergrown with
sphalerite and galena, and as veinlets associated with galena. The silver content in the tetrahedrite-
freibergite varies from about 1 to 35 wt %; the average is generally around 15 wt % Ag.
The silver content in solid solution in galena, chalcopyrite, sphalerite and pyrite varies from
grain to grain, and the average amount in each mineral varies from deposit to deposit. The silver
content in individual galena grains ranges from several to 6000 ppm [21,28] in individual
chalcopyrite grains from several to 1600 ppm, in individual sphalerite grains from several to 310
ppm, and in individual pyrite grains from 0.1 to 340 ppm [28]. The average Ag content in galena
140

ranges from 600 ppm in the Trout Lake deposit to 1500 ppm for the galena in the Heath Steele
deposit. The average Ag content in sphalerite varies from 15 ppm for some deposits to 100 ppm
for other deposits. The average Ag content in pyrite is around 10 ppm [19,21, Chryssoulis,
personal comm.].
The pyrargyrite, stephanite, and other silver sulphides and sulphantimonides, selenides, and
tellurides occur as small grains scattered throughout the deposit. The native silver generally occurs
as minute grains in sphalerite and as veinlets and minute inclusions in pyrite [29].

4.5.2. Metallurgy. Silver recovery from base metal ores involves recovering Ag-bearing minerals
in several concentrates. Most of the silver in tetrahedrite, pyrargyrite and chalcopyrite is recovered
in the Cu concentrate; silver in solid solution in galena is recovered in the Pb concentrate. The
silver in solid solution in sphalerite is recovered in the Zn concentrate. The minor silver-bearing
sulphantimonides are generally rejected to the tailings. It is noteworthy that some tetrahedrite is
recovered in the Pb concentrate, a minor amount is in the Zn concentrate, and some is in the tails.
The silver in the Cu and Pb concentrates is recovered in the metallurgical process (smelting), but
generally no payment is received for the silver in the Zn concentrate. For most volcanogenic ores,
about 10 % of the silver is lost in the Zn concentrate and 20 to 30% in the tails. About 20 to 50%
of the silver in the Zn concentrate is soluble in cyanide solution:;, and may occur as a coating of
Ag2S on sphalerite. The materials-minerals balance for Ag among the minerals in the Brunswick
12 deposit is given in Table 7 as an example.

4.6. FACTORS WHICH AFFECT GOLD RECOVERY

Most volcanogenic base metal deposits contain trace to minor amounts of gold (0.1 to 7 ppm). The
gold content in some deposits is too low to warrant recovery; in other depoSits, the metal is a
valuable by-product [30,31] and in still others [32], the gold content is high enough to be the
primary ore mineral.

4.6.1. Mode of Occurrence of Gold. Gold occurs in all zones of mineralised volcanogenic
massive sulphide ores. The gold teneur in the chalcopyrite-bearing feeder zones and in the
chalcopyrite zones tends to be higher than in the pyrite-sphalerite and the pyrite-sphalerite-galena
zones. Hence, a positive correlation between the gold content in the ore and the CU/CU + Zn ratio is
common.
Three modes of occurrence have been observed for gold in volcanogenic ores:
1) gold associated with chalcopyrite and chlorite schist,
2) gold associated with pyrite and arsenopyrite, and
3) "invisible" gold
Gold associated with chalcopyrite and chlorite schist occurs in the chalcopyrite-bearing feeder
zones and in the chalcopyrite zones [30,31]. The gold grains range from 5 to several hundred
micrometers in size. The mineral occurs as veins and veinlets along chalcopyrite-chlorite bounda-
ries and appears to have concentrated into masses at the noses of microfolds. Discrete gold grains
also occur in massive chalcopyrite.
Gold in the massive pyrite is associated with pyrite and arsenopyrite. The gold occurs as
I) minute grains and v~inlets along pyrite grain boundaries and in fractures in pyrite,
2) as minute grains associated with galena, freibergite and other late minerals in fractures in
pyrite,
 141

3) as "invisible" gold in pyrite and arsenopyrite, and

4) as gold inclusions in recrystallised pyrite grains.
The gold grains in pyrite fractures and along pyrite grain boundaries in massive pyrite range from
less than 1 to 50 !lID in diameter, but most are 1 to 10 !lID.
The gold associated with other minerals such as galena and tetrahedrite also occurs in veinlets in
pyrite and occasionally in fractured sphalerite and chalcopyrite. Late vein forming minerals occur in
the fractures and the gold is intimately intergrown with them. In some instances, earlier minerals
such as interstitial sphalerite and chalcopyrite are fractured and contain the gold-bearing veinlets.
The "invisible" gold occurs in pyrite and in arsenopyrite. The amount of "invisible" gold in the
arsenopyrite in the Trout Lake ore is about 50 times the amount of gold in pyrite. Since pyrite in
volcanogenic ores is several orders of magnitude more abundant than arsenopyrite, most of the
"invisible" gold in volcanogenic ores is contained in the pyrite. The amount of "invisible" gold in
pyrite in the Trout Lake ore varies from less than 1 ppm to 6 ppm, and averages 0.7 ppm. The
amount of "invisible" gold in the arsenopyrite in the same ore ranges from 5 to 130 ppm, and
averages 30 ppm [30). The "invisible" gold in pyrite accounts for about 6 to 9 % of the total gold
in the Trout Lake ore, and in arsenopyrite for about 1 to 3 % of the gold in the ore.

TABLE 7. Distn . 'butlOns

. 1 ver d ue to 1 erent M'mera s.
an d assays 0 fS'1
Host mineral Feed OJ Pb Zn Bulk SecZn Tails
% conc% conc% conc% conc% conc% %
tet. (dist) 52.2 16.3 17.3 2.0 1.8 0.9 13.9
galena (dist) 34.7 0.5 20.7 3.3 1.5 0.4 8.3
sphalerite (dist) 4.1 - 0.2 3.2 0.1 0.2 0.4
acanthite (dist) 5.6 - - 2.8 - - 2.8
pyrite (dist) 3.4 - 0.2 0.1 - - 3.1
total Ag (dist) 100.0 16.8 38.4 11.4 3.4 1.5 28.5

Assays
tet. (Ag,ppm) 61 4591 284 14 249 92 19
galena (Ag,ppm) 41 149 339 23 200 48 11
sphalerite (Ag,ppm) 5 - 3 22 14 20 1
acanthite (Ag,ppm) 7 - - 20 - - 4
pyrite (Ag,ppm) 4 - 3 1 - - 4
total (Ag,ppm) 118 4740 629 80 463 160 39
tel - tetrahedrite-frelberglte.

4.6.2. Composition of the Gold. Gold in volcanogenic ores does not have a specific composition,
but is alloyed with various amounts of silver, and in some instances, mercury. The gold in the
Trout Lake deposit has every composition within the gold-bearing field of the Au-Ag-Hg ternary
system, and has been described as a Au-Ag-Hg alloy [30). Most of the gold analysed from
volcanogenic ores (Caribou, Point Leamington) contains some silver and is electrum [33,34). It is
assumed that the gold composition is dependent upon the composition of the mineralising solutions
as well as on the depositional conditions.

4.6.3. Origin of the Gold. The occurrence of minute gold grains along fractures in pyrite, of gold
142

in late veinlets in pyrite, of gold associated with late minerals that occur as late veinlets in pyrite,
and of gold along pyrite grain boundaries suggests that gold in the pyrite-sphalerite and pyrite-
sphalerite-galena zones is a late mineral. The presence of "invisible" gold in pyrite and arsenopyrite
suggests that most of the gold in the pyrite-sphalerite rich parts of the volcanogenic orebodies may
have been deposited as "invisible" gold. The "invisible" gold can be released from pyrite by solid
state diffusion to fractures and grain boundaries, and by recrystallisation of the pyrite [35]. The
pyrite would still contain "invisible" gold if the release is incomplete. It is assumed that
recrystallised pyrite grains do not contain "invisible" gold. Gold in volcanogenic deposits is,
therefore, a late mineral in the paragenetic sequence, and is associated with other late minerals in
veinlet~.

4.6.4. Examples of Gold in Volcanogenic Deposits. Volcanogenic deposits occur throughout the
world and contain trace to significant amounts of gold. Examples of the occurrences of gold in a
few deposits are given below.
The volcanogenic deposits in the New Brunswick mining area of Canada contain minor amounts
of gold. The Brunswick No. 12 deposit, which consists mainly of pyrite-sphalerite and pyrite-
sphalerite-galena zones, contains less than 1 ppm gold, largely as "invisible" gold. Only a rare
occurrence of electrum was found in polished sections of samples from the footwaIl chalcopyrite
zone [36]. Approximately 80 % of the gold in the ore is rejected to the pyrite-rich flotation tailings.
The Caribou deposit in New Brunswick contains about 1.5 ppm Au. Minute grains of gold and
electrum have been observed along fractures and grain boundaries in pyrite [33].
Gold in the volcanogenic deposits in the Flin Ron mining area grades from about 1.5 to 3 ppm.
The metal is present in all zones of the deposits. In feeder zones, it occurs as veinlets and masses
associated with chalcopyrite and chlorite. In chalcopyrite zones, in pyrite-sphalerite zones and in
pyrite-sphalerite-galena zones the gold occurs as fracture fillings in pyrite, as gold grains along
pyrite grain boundaries, as gold associated with late minerals (galena and tetrahedrite) in pyrite
fractures, and as "invisible" gold in pyrite and arsenopyrite.

4.6.5. Recovery of Gold from Volcanogenic Deposits. Studies of mill products indicate that gold
is poorly bonded to pyrite and strongly bonded to chalcopyrite. Hence, during grinding most of the
minute gold grains that occur in fractures and along pyrite grain boundaries will tend to break free
from the pyrite. Only a few gold grains will remain attached to the pyrite, and only the minute gold
grains which occur along rehealed hairline fractures will be unliberated. In contrast, much of the
gold associated with chalcopyrite wilJ be attached to or enclosed in chalcopyrite. The free gold
grains wiII range from about I IAffi in diameter to the largest grains (nuggets) in the ore.
Since gold is a by-product from volcanogenic ores, most mineral beneficiation concentrators do
not have gold circuits. The classical concentrator consists of a series of Cu, Pb and Zn circuits.
The Cu circuit is the first in the series. It usually has a regrind circuit, and produces a relatively fine
grained copper concentrate. The Cu circuit tails are fed to a Pb circuit which produces a lead
concentrate. The lead circuit tails are fed to a zinc circuit that also has a regrind circuit and produces
a zinc concentrate and a final concentrator tails. Gold and its variants float extremely well in the
copper circuit, and most of the liberated and exposed gold grains, as well as gold attached to
chalcopyrite, are recovered in the Cu concentrate. Any liberated and exposed gold that was not
recovered in the copper concentrate float~ we\1 in both the Pb and Zn circuits, which tend to act as
gold scavengers. Hence, all or nearly all liberated and exposed gold grains are recovered in the
Cu, Pb and Zn concentrates. Only very small gold grains attached to large pyrite grains, inclusions
 143

of gold in pyrite, and "invisible" gold in pyrite and arsenopyrite are lost to tailings.
For Trout Lake ores, about 60% of the gold is recovered in the Cu concentrate, 10% in the Zn
concentrate, and 30% is lost to tails. The gold recovered in the Cu concentrate occurs as liberated
grains 1 to 100 f-lm in size, and as unliberated grains attached to chalcopyrite and pyrite. The gold
recovered in the zinc concentrate is present as minute unliberated grains attached to pyrite and
sphalerite. The gold lost to tailings occurs as unliberated gold grains smaller than 1.5 f-lm in
diameter and as "invisible" gold in pyrite and arsenopyrite. Some of the minute unliberated gold
grains in the tailings are attached to large pyrite grains and some are enclosed in pyrite. About 30%
of the gold in the tailings is exposed and dissolves by cyanidation. An additional 35% of the gold
is enclosed in pyrite, but can be exposed by grinding the ore to minus 18 !1m. Undoubtedly the
latter form of gold occurs along partly healed microfractures in pyrite. The remaining 35% is
"invisible" gold. Hence, normal flotation operations collect all the gold that can be recovered by
flotation. Only unliberated gold grains smaller than 1.5 f-lffi in diameter and "invisible" gold are lost
to tailings.

4.7. FACTORS WHICH AFFECT TIN RECOVERY

Volcanogenic base metal deposits contain about 0.05 to 0.3 wt% Sn. The tin is present mostly as
cassiterite, but a small proportion occurs as stannite, pabstite and as solid solutions in pyrite
[21,28]. The Kidd Creek deposit near Timmins, Ontario, which contains 0.1 to 0.2 wt% Sn, was
studied in detail as an example of thc behaviour of cassiterite [37]. About 82% of the Sn in the ore
is rejected to the tails, mostly as cassiterite. The company had recovered the cassiterite from the
tails by gravitational methods, but the Sn circuit was discontinued due to low recoveries.

4.7.1. Mineralogical Characteristics of Mill Products. The Kidd Creek concentrator consists of
Cu, Pb and Zn circuits in series and produces Cu, Pb and Zn concentrates. The Cu circuit tails is
fed to the Pb circuit, the Pb circuit tails is fed to the Zn circuit, and the Zn circuit tails is the
concentrator tails. The feed, concentrate and tails samples were analysed chemically and the results
were processed by the MATBAL materials balance program [38]. The weight of each product and
the recoveries of the elements in the products were calculated. The results showed that 13.9% of
the Sn is in the Zn concentrate and 81.3% is in the concentratortails.

TABLE 8 Liberations of Cassiterite

Sample Libemtion %, Material Balance
Head 55
Cuconc. 46
Pb conc. 4
Zn conc. 25
Cu tails 55
Pb tails 56
Zn tails 61
Sn conc. 77

The samples were analysed with the microprobe-SEM-IPS image analysis system [16], and
mineral distributions were calculated with the MATBAL materials balance program. The results
144

show that 12.8% of cassiterite (Sn02) in the ore was in the Zn concentrate and 82.7% in the
concentrator tails.

4.7.2. Cassiterite Liberation. The size distribution of the particles in the head sample shows that
the ore had been ground to 55% minus 44 !Lm. The percentage of cassiterite present as free grains
(cassiterite grains without inclusions of other minerals) was determined by image analysis for the
mill products and for a Sn concentrate which had been produced by the company in their gravity
Sn circuit (Table 8). Size distributions show that most of the free cassiterite grains in the feed range
from about 1 to 18 JLffi in diameter, whereas most of those in the gravity Sn concentrate range from
9 to <18 !-tm in diameter (Table 9). The largest free cassiterite grains in the products (Sn
concentrate and tails) are about 40 !Lm in diameter although cassiterite grains up to 110 !Lm in
diameter have been observed in the unbroken ore [39].

TABLE 9 Size Distributions of Free Cassiterite.

~

Size range Head Cu conc. Pb conc. Znconc. Tails Sn cone.

JLffi %wt %wt %wt %wt %wt %wt
0-4.6 24.6 27.4 0 4.9 25.5 2.1
4.6-9.2 34.3 36.3 0 12.6 35.3 15.5
9.2-18.4 29.5 36.3 100 39.3 29.4 62.0
+18.4 11.6 0 0 43.2 9.8 20.4
wt% ofcass. 54.6 1.0 0.1 3.2 50.3

The unliberated cassiterite grains were classified as occurring in particles which contain 70-100,
30-70, and 0-30 vol% cassiterite. The proportions of cassiterite in each class of particle and the
distribution of cassiterite among the different particles were combined with the distributions of free
cassiterite. The results show that the cassiterite in the head sample occurs largely as free grains and
as inclusions in particles that contain less than 30 vol % cassiterite (Table 10).

TABLE 10. Distributions of Cassiterite.

Class of particle % Distribution
(vol. % cass.) Head Cu conc. Pb conc. Znconc. Tails
Free (100%) 54.63 1.02 0.09 3.08 50.44
70-100% 7.23 0.16 1.57 0.84 4.66
30-70% 10.84 0.17 0.27 1.61 8.79
0-30% 27.30 0.92 0.38 7.08 18.92
Total 100.0 2.27 2.31 12.61 82.81

The percentages of unliberated cassiterite grains associated with sulphides and with silicate
minerals were determined. It was found that, apart from the zinc concentrate where the cassiterite
occurs as inclusions in sphalerite, about 50% of the unliberated cassiterite is combined with
sulphides and 50% with silicates.

4.7.3. Discussion of Cassiterite from the Kidd Creek Deposit. The mineralogical study showed
that most of the tin occurs as cassiterite, and that most of the cassiterite in unbroken ore occurs as
inclusions in sphalerite. The cassiterite grains range from 1 to 110 !Lm in diameter in unbroken ore,
 145

but more than 85% of the cassiterite in crushed ore is smaller than 18 !-lm in diameter. The
predicted cassiterite liberation, using the cassiterite size distribution and the liberation models of
Finch and Petruk [3] (random breakage) and Petruk [8] (preferential breakage) is that 50 to 60% of
the cassiterite would be liberated at a grind of 100% minus 18 !-lm. The observed cassiterite
liberation in the feed, which had been ground to only 55% minus 44 !-lm, was 55%. The observed
high cassiterite liberation at the much coarser grind than predicted suggests that selective breakage
occurred during grinding. Presumably the sphalerite broke along cleavage planes and released the
small cassiterite inclusions. Most of the liberated cassiterite was preferentially rejected from the Cu,
Pb and Zn circuits to the concentrator tails.

TABLE II Distributions* of Sised Cassiterite Grains

Casso Class Size (IJ.ID) Head Cu conc. Pb conc. Znconc. Zn tails
Free 0-4.6 13.39 0.28 0.0 0.15 12.96
4.6-9.2 18.75 0.37 0.0 0.39 17.99
9.2-18.4 16.18 0.37 0.09 1.21 14.51
+18.4 6.31 0.0 0.0 1.33 4.98
Total 54.63 1.02 0.09 3.08 50.44
70-100 0-4.6 0.31 0.04 0.03 0 0.27
4.6-9.2 0.41 0.05 0.17 0 0.19
9.2-18.4 6.19 0.10 1.32 0.62 4.15
+18.4 0.32 0.0 0.05 0.12 0.15
Total 7.23 0.16 1.57 0.84 4.66
0-30 0-4.6 5.67 0.27 0.02 0.39 4.99
4.6-9.2 7.28 0.18 0.03 1.24 5.83
9.2-18.4 8.47 0.10 0.10 3.18 5.09
+18.4 5.88 0.37 0.23 2.27 3.01
Total 27.30 0.92 0.38 7.08 18.92
* in percent

The unliberated cassiterite grains in the Kidd Creck feed range from 1 to 37 IJ.ID; most occur in
particles that contain less than 30 vol% cassiterite. It may be possiblc to liberate some of the
cassiterite by regrinding, but since about 50% of the unliberated cassiterite is associated with
silicates (about 22% of the cassiterite in ore), regrinding the tails to 100% minus 18 !-lm would be
required to increase the liberation significantly. Recovering the cassiterite locked in pyrite would
involve floating the pyrite and regrinding the pyrite concentrate to liberate the cassiterite. The
cassiterite in pyrite tends to occur along grain boundaries; hence, it would be liberated when the
pyrite has been ground sufficiently to separate the pyrite grains (size distribution of fine grained
pyrite in ore).
In view of the low Sn content in the ore and the high cassiterite liberation in the tailings, it was
proposed that recovery of cassiteritc from the tailings be carried out without regrinding. The
gravity Sn concentrate produced by the company contained grains which were largely 9 to 37 IJ.ID
in diameter. Since about 60% of the cassiterite in the ore is in grains smaller than 9 micrometres,
this amount is unrecoverable by gravitational techniques. Hence the low Sn recovery obtained by
the company (less than 10%) is consistent with the characteristics of the cassiterite. It may be
possible to obtain better recoveries by cassiterite flotation using aminophosphonic collectors [40].
146

The sulphides, however, would have to be floated prior to cassiterite flotation. Cristovici and
McCarthy [40] have successfully floated more than 90% of the sulphides from the Kidd Creek tails
without regrinding, but did not evaluate cassiterite flotation.

5. References

1. Petruk, W. (1991) Chemical and mineralogical analysis for performance assessment, in D.

Malhotra, R.R. Klimpel and A.L. Mular (eds.) Proceedings Evaluation and Optimisation of
Metallurgical Performance, Ann Arbor, Mi.: SME, pp. 181-19l.
2. King, R.P. (1979) A model for the quantitative estimation of mineral liberation by grinding,
Int. J. Miner. Process. Q,207-220.
3. Finch, I.A., and Petruk, W, (1984) Testing a solution to King liberation model, Int. l.
Miner. Process.1b 305- 311.
4. Barbery, G., and Leroux, D. (1986) Prediction of particle composition distribution after
fragmentation of heterogeneous materials: Influence of texture, in R.D. Hagni (ed.) Process
Mineralogy VI, Warrendale, PA. : TMS, pp. 415-427.
5. Klimpel, R.R. (1984) Some practical approaches to analysing liberation from a binary
system, in W. Petruk (cd.) Process Mineralogy III, AIME/SME, New York, pp. 65-81.
6. Petruk, W. (1988) Ore characteristics that qffect breakage during grinding, in D.l.T. Carson
and A.H. Vassiliou (eds.) Process Mineralogy VIII, Warrendale, PA: TMS, pp. 181-193.
7. Petruk, W. (1989) Image analysis of minerals, in W. Petruk (cd.) Short Course on Image
Analysis, Can. Mineralogist, pp. 6-18.
8. Petruk, W. (1986) Predicting and measuring mineral liberations in ores and mill products and
effect of mineral textures and grinding methods on mineral liberation, in R.D. Hagni (ed.)
Process Mineralogy VI, Warrendale, PA: TMS, pp. 393-403.
9. Petruk, W. (1991) An image analysis and materials balancing procedure for evaluating ores
and mill products to obtain optimum recoveries, in Proceedings 23rd Ann. Meeting, Canadian
Mineral Processors, CIM, Ottawa, Canada, paper 19.
10. Healy, R.E., and Petruk, W. (1989) A mineral evaluation of the behaviour of metallic
minerals in the Trout Lake Concentrator circuit, Department of Energy, Mines and Resources,
CANMET, Mineral Sciences Laboratory, Investigation Report, MRP/MSL 89-87 (IR).
11. Petruk, W. (1990) Measurements of mineral liberation in connection with mineral bene-
ficiation, in W. Petruk, R.D. Hagni, S. Pignolet-Brandom and D.M. Hausen (eds.) Process
Mineralogy IX, Warrendale, PA: TMS, pp. 31-36.
12. Petruk, W. (1990) Mineralogical characterisitlcs that affect recoveries of sphalerite from
volcanogenic base metal ores, in T.S. Mackey and R.D. Prengaman (eds.) Lead-Zinc '90,
Warrendale, PA: TMS, pp. 655-664.
13. Petruk, W., and Smith, M.M. (1988) Ore characteristics that affect breakage during grinding,
Department of Energy, Mines and Resources, CANMET, Report 88-5E.
14. Grant, G., Miller, P, Reid, A.F., and Zuiderwyk, M.A. (1981) Feature extraction and
population distribution from complex are particles by QEM*SEM image analysis,
Proceedings 3rd Eur. Symp. on Stereology, LjUbljana, pp. 233-238.
15. Jones, M.P. (1983) The characterisation of ores and mineral products by automatici image
analysis of mineralogical features, in l.P.R. de Villiers and P.A. Cawthorn (eds.) ICAM 81,
Geol. Soc. South Africa, Spec. Pub. 7, pp. 449-464.
16. Petruk, W. (1988b) Capabilities of the Microprobe Kontron image analysis system:
 147

application to mineral beneficiation, Scanning Microscopy 2(3), 1247-1256.

17. Pignolet-Brandom, S., and Reid, K.l. (1988) Mineral characterisation by QEM*SEM, in
D.J.T. Carson and A.H. Vassiliou (eds.) Process Mineralogy VIII, Warrendale, PA: TMS,
pp. 332-346.
18. Gaudin, A.M. (1939) Principles of Mineral Dressing, McGraw Hill Book, Co., New York.
19. Healy, R.E., and Petruk, W. (1988) The mineralogical characteristics that affect the metal
recoveries from the Cu. Zn. Pb and Ag ores from Manitoba. Part 1. An investigation of the
ore mineralogy of the Trout Lake deposit, Department of Energy, Mines and Resources,
CANMET, Mineral Sciences Laboratories, Investigation Report, 88-61 (IR).
20. Hughson. M.H. (1979) An image analysis study of exsolution chalcopyrite in sphalerite from
the East Sulphide Orebody of the Caribou deposit. New Brunswick, Department of Energy,
Mines and Resources, CANMET. Mineral Sciences Laboratories. Investigation Report
MRP/MSL 79-5(IR).
21. Petruk. W., and Schnarr, 1.R. (1981) An evaluation of the recovery offree and unliberated
mineral grains. metals and trace elements in the concentrator of Brunswick Mining and
Smelting Corp. Ltd.. CIM Bull 74(833), 132-159.
22. Laguitton, D. (1980) Material balance of mineral processing flowsheets. Fortran program
Matbal2. Users Manual, Department of Energy, Mines and Resources. CANMET. Mineral
Sciences Laboratories, Investigation Report MRP/MSL 80-33(IR).
23. Hendrik, D.W. and Ounpuu, M. (1985) Modification to the Zinc flotation circuit of
Brunswick Mining and Smelting Corporation Limited, in A.D. Zunkell, R.S. Boorman, A.E.
Morris and R.J. Wesley. (eds.) Complex Sulphides. Warrendale. PA: TMS, pp. 271-292.
24. Coleman, M.M, Gilders, R.. May. R.A., Melluish. J.• Petruk, Wand Surgis, L. (1990) An
investigation of the behaviour of free and unliberated mineral particles in the Brunswick
Mining and Smelting Corporation Limited concentrator, in W. Petruk, R.D. Hagni, S.
Pignolet-Brandom and D.M. Hausen (cds.) Process MineraLogy IX. Warrendale, PA: TMS.
pp.71-79.
25. Chen, T.T. and Petruk, W. (1980) Mineralogy and characteristics that affect recoveries of
metals and trace elements from the ore at Heath Steele Mines. New Brunswick. CIM Bull.
73(823),167-179.
26. Petruk W .. and Pinard, R.G. (1990) Image analysis evaluation of mineral behaviour during
beneficiation: bench test of a base metal ore from the Faro deposit. Yukon, in W. Petruk.
R.D. Hagni, S. Pignolet-Brandom and D.M. Hausen (eds.) Process Mineralogy IX,
Warrendale, PA: TMS. pp. 37-48.
27. Petruk, W., Chen, T.T., and Wilson. J .M. (1990) Mineralogical characterisitics that affect
recoveries of galena from base metal ores, in T.S. Mackey and R.D. Prengaman (eds.) Lead-
Zinc '90, Warrendale, PA: TMS. pp. 665-672.
28. Cabri, L.J .• Campbell, J.L., Laflamme, J.H.G., Leigh, R.G., Maxwell, J.A. and Scott,
J.D. (1985) Proton-microprobe analysis of trace elements in sulphides from massive-sulphide
deposits, Can. Mineralogist n, 133-148.
29. Petruk, W. (1965) Mineralogical investigation of zinc and copper ore from the North and
South zones of the Texas Gulf sulphur deposit. Kidd Township, Ontario, Department of
Mines and Resources, CANMET, Mineral Sciences Division, Investigation Report IR-65-89.
30. Healy, R.E., and Petruk, (1990) Petrology of Au-Ag-Hg alloy and "invisible" gold in the
Trout Lake massive sulphide deposit. Flin Flon. Manitoba. Can. Mineralogist.2.8., 189- 206.
31. Sinclair, LG.L. (1990) Some practical applications of ore microscopy at Les Mines SeLbaie.
Quebec, in W. Petruk, R.D. Hagni, S. Pignolet-Brandom and D.M. Hausen (eds.) Process
148

Mineralogy IX , Warrendale, PA : TMS, pp. 81-92.

32. Kerr, D.J., and Mason, R. (1990) Are-appraisal 0/ the geology and ore deposits 0/ the
Horne mine complex at Rouyn-Noranda, Quebec, CIM Bull . .83.(936), 34 (abs).
33. Jambor, J.L. (1981) Mineralogy 0/ the Caribou massive sulphide deposit, Bathurst area,
New Brunswick, Energy Mines and Resources, CANMET Report 8l-8E.
34. Melluish, J.M. and Stowe, KC. (1990) Evaluation o/high and low grade ores o/the Point
Leamington massive sulphide deposit, in W. Pctruk, R.D. Hagni, S. Pignolet-Brandom and
D.M. Hausen (eds.) Process Mineralogy IX, Warrendale, PA: TMS, pp. 61-70.
35. Boyle, R. W. (1979) The geochemistry 0/ gold and its deposits, Geol. Survey of Canada,
Bull. 280.
36. LeBlanc, S. (1989) Massive sulphide-associated gold: Brunswick No. 12 mine, Bathurst,
New Brunswick, B.Sc. Thesis, Department of Geology, University of New Brunswick.
37. Petruk, W. (1990) Mineralogical characteristics that affect recoveries 0/ cassiterite/rom the
volcanogenic base metal ore 0/ Kidd Creek Mines in Ontario, in T.S. Mackey and R.D.
Prengaman (eds.) Lead-Zinc '90, Warrendale, PA: TMS, pp. 673-679.
38. Laguitton, D. (1985) The SPOC manual, CANMET Report SP 95-1/3.2E, CANMET,
Department of Energy, Mines and Resources, Ottawa, Canada.
39. Petruk, W., and Owens, D. (1965) Mineralogical investigation 0/ a zinc-copper-silver-lead
ore/rom the North Zone o/the Texas Gulf Sulphur deposit in Kidd Township near Timmins,
Ontario, Department of Mines and Technical Surveys, Mines Branch, Investigation Report
IR- 65-105.
40. Cristovici, M.A., and McCarthy, D.R. (1987) Investigation to remove sulphides/rom Kidd
Creek Mill tailings prior to cassiterite flotation. Progress report, Department of Energy, Mines
and Resources, CANMET, Mineral Science Laboratories, Investigation Report, MSL
87-71(IR).
1.6. LIBERATION BY SIZE REDUCTION. CONSEQUENCES AND
IMPROVEMENTS ON FLOTATION KINETICS

M.R. MACHADO LEITE

Mining Department
University of Porto
Rua dos Bragas
4099 Porto Codex
Portugal

"This model will be a simplification

and an idealisation and, consequently,
a falsification. It is to be hoped that
the features retained for discussion
are those of great importance
in the present state of knowledge"
A.M. Turing, Chemical basis of morphogenesis,
1952, Phil. Trans.

ABSTRACT. The first part of this lecture aims at emphasising the important role played by particle
size on the general formulation of a separation process by means of its influence on the liberation
status of a given ore. Liberation, seen as a static state of a particulate system, is described by
means of a marginal probability density of grade, the computation of which compels us to define
and use a grade distribution density conditional on size. Liberation is then considered as a dynamic
state evolving along comminution time. Both cases of non-discriminating (transgranular) and
discriminating (perigranular and transgranular preferential) comminution arc referred. For the non-
discriminating case, a limit probability argument is used to justify the introduction of the Euler Beta
Law as a well suited tool for describing the liberation state of a comminuted ore. since it is able to
predict the grade distribution of a given iso-size fraction. Considering this previous general model
of liberation by size reduction as a good support. flotation kinetics is reviewed in order to provide
evidence for the kinetic rate constants (seen as non-wettability parameters) as dependent on the
grade distribution conditional on size. Simulation of traditional experiments. such as washability
curves and release analysis. supported by the modelled liberation phenomenon and a good
condensation of the kinetical parameters. is shown to be feasible. Finally. some numerical
experiments of an elementary flotation kinetic model are performed in order to enhance the
influence of size and degree of liberation on flowsheet performance.

1. Introduction

A particle population or a particulate system is a mass of comminuted particles ready for delivery to

149
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 149-170.
© 1992 Kluwer Academic Publishers.
150

a flotation separation process. The word population stresses the uncontrolled variability of the
individual particle properties, i.e. their statistical character, while the word system underlines the
interactive (competitive and/or cooperative) character of the particles behaviour within a mineral
processing unit. Any study of this non-linear interactive behaviour obviously implies the
knowledge of the statistical distributions of the most relevant particle properties in a particulate
system.
First of all, we shall notice that the notion of a distribution of particle properties, when trying to
obtain experimental data relevant to a satisfactory description of the particle behaviour, implies such
hard work that it very often leads to particle-by-particle analysis and observation. This type of
laboratory work, however crucially important, is prohibitively expensive without access to
automated means. Image analysis has been pointed out as vital for controlling and tuning grinding
and separation circuits. However, for design purposes, data must come from unground geological
specimens and we do not hope to have, in the near future, a computer system able to predict
property distributions of comminuted particles solely based on image analysis.
Thus, we have been thinking for years that a straightforward theoretical approach, starting from
first principles and leading to the development of computational models of comminution as a
process of generating new particle properties, which will be able to predict the resulting property
distributions is a prerequisite to the building of the interpretation algorithms to be incorporated in
the image analysis systems.
Our view about this kind of problem is well described by Turing's quotation on the first page of
this paper: in fact, we are inclined to think that the fundamental requirement to be made is that
models will remain under our conceptual control at all times; to comply with this, models must
always keep separately the different physical phenomena involved. From a theoretical point of
view, models are conceptual, physical and mathematical forms which describe reality, built under a
deliberate intention of simplicity. A model, in its idealism, turns itself away from reality in order to
enhance some of its features and aspects. Taking advantage of this, we can use models of
elementary operations and equipments as gnoseological windows upon a complex reality, i.e., as
conceptual templates, whose shortcomings in fitting experimental data act as detectors of eventual
misadjustments of our theoretical notions.

2. Particle property distributions. The importance of particle size on the

liberation state of a given ore

Among the various particle properties, two stand out as the most important in Mineral Processing
Technology: size (s) and grade or mineral content (g). Even though both these properties of
individual particles in a population are random variates, they are effectively related to each other.
Since the main target of mineral processing is the production of a concentrate, first of all a
separation process, seen as a way to increase the mineral content in the concentrate, may be
described as an operation depending on particle grades alone. However, it is well known that
individual particle grade is a function of its degree of liberation, itself strongly dependent on the
individual particle size. The relationship between grade and size is quite obvious in the fact that
large particles have grades that tend to cluster around the average grade of the ore, while fine
particles are expected to have more extreme grades, respectively, both on the gangue range and on
the concentrate range. On the other hand, in the case of separation by flotation, the interactive
behaviour of particles due to their properties makes for size being important in two different ways:
its influence both on the physics of the separation, e.g. controlling the collision mechanism
 lSI

between particles and bubbles, and on the individual particle grade, on account of its control upon
the degree of liberation.
In this case, a joint probability density of size and grade f(s,g) may be defined as a more suitable
tool to use for separation modelling. Our main task will be, however, to build up such a
description. In order to attain this goal, we should review some important features:
- the joint probability density of size and grade f(s,g) allows us to define

f(s,g) ds dg (1)

as the mass or volume fraction of particles from the population that have sizes and grades within
the intervals ]s-ds/2, s+ds/2[ and ]g-dg!2, g+dg!2[, respectively;
- the marginal size distribution density des) is then:

1
des) = f f(s,g) dg (2)
o
so that d(s).ds is the mass or volume fraction of particles from the population that, independently
of their grade, have sizes within the interval ]s-ds/2, s+ds/2[
- the marginal grade distribution density h(g) is then:

h(g) = f f(s,g) ds (3)

o
so that h(g).dg is the mass or volume fraction of particles from the population that,
independently oftheir size, have grades within the interval]g-dg/2, g+dg!2[.
- and, as another probability density, f(s,g) may be broken up into the product:

f(s,g) = e(gls) . des) (4)

of the marginal size distribution density and the Grade Distribution Conditional on Size, c(gls),
i.e. the grade distribution of iso-sized particles of size s.
The marginal size density des) may be experimentally obtained by means of sieving the material
through a standard Tyler-ASTM series. The weight fraction of material retained on each sieve of
aperture si is then:

Si-l
di = f d(s).ds =M(Si_l) - M(Si)
Si

where M(Si) represents the size distribution function (or cumulative distribution function of size).
Unfortunately, a similar laboratory procedure cannot be implemented yet for the experimental
determination of the grade distribution functions, because these would force expensive and tedious
work on particle by particle analysis, even if we thought of the image analysis tools. As a matter of
fact, at the present state of knowledge in the field, the only thing that image analysis is able to do
accurately is surface grade determinations of planar sections of polished rock specimens or
polished preparations of ground material. However, for liberation analysis purposes, volume
152

grades are always needed in order to describe the phenomenon.

3. Liberation as a static state

Liberation, as a static state of a particulate system, is adequately described by the marginal grade
distribution density, h(g). Eqns. 3 and 4 allow us to write:

h(g) = f f(s,g).ds = f c(gls).d(s).ds (5)

o 0

In fact, as far as a separation process (such as flotation) is concerned, given a cut grade go and a
Tromp curve ~(g-go) for describing the equipment efficiency, the exact knowledge of the function
h(g) will be enough for predicting the properties of the final concentrate product:

Qconc(g) =~(g-go) h(g)

where Qconc(g) is not yet the normalised density distribution of the concentrate. To obtain the
function conc(g), it should be necessary to calculate the normalising factor Q, which is the total
amount of material recovered in the concentrate product:

Q = JQconc(g).dg

and, finally,

conc(g) = Qco~C(g)

4. Liberation as a dynamic state

In the early sections, grade and size distributions have been thought of as a static process, i.e.,
des), h(g) and f(s,g) have been assumed as referring to some well defined state of a particulate
system, the actual properties of which result from the previous life of particles during processing,
especially comminution. However, a more powerful concept in mineral processing research is to
consider liberation as a dynamic state evolving along comminution time. In the next section we will
attempt to develop it through a step by step analysis.

4.1. COMMINUTION AS A HEREDITARY PROCESS

In order to follow the evolution of particle properties, the transformations undergone by them must
be followed along the comminution process.
As soon as a particle breaks, it ceases to exist as an individual one and is replaced by several
new different daughter particles. A means of defining, at any moment and with no ambiguity, the
descendants of a given particle must be found. For achieving this goal we will use Bastenaire's
formal theory as a tracing tool [11. Let us choose a geometrical point B, which we shall calI a
 153

Bastenaire point, at random in the total volume V of the solid matter of the particulate system, at
one of its possible states of evolution at time to, with probability uniformly distributed in the
volume V. The probability of point B falling in a given particle Pi of volume v i, size Si and grade
gi, independently of size or grade (which we call the event Ei), is:

Prob { Ei } = Prob { B E Vi } = V
v·
(6)

The probability of B falling in one of the particles of size Si < s and grade gi < g (cumulative
probability), is:

Prob { B E Vi : si < S, gi < g } = L~

i: Sj<s gj< g
= Fto(s,g) (7)

i.e. it coincides with the previous empirically defined joint cumulative distribution of size and grade
at time to, FtO(s,g).
Bastenaire's theory allows us to define genuine random variates, s and g, the jOint probability
distribution of which coincides with the empirical (volumetric or mass) joint distribution function
of size and grade.
Now, let us consider a later stage of evolution of the system, at time t1: particle Pi may have
been broken into a certain number of daughter particles and let us call Pki the particle to which
point B now belongs (Vki, Ski, and gki being, respectively, its volume, size and grade). Given the
event Ei, the probability of B is uniformly distributed throughout the volume Vi =}:Vki of the
daughter particles. The probability of Ski being less than some given value, sand gki being less
than some given value, g, is then computed by:

Prob { B E vki: Ski < S , gki < g I Ei } = 'L.J

" Vk'v/ = FttltO(S,gIEi) (8)
k: Ski<s gki< g

FtlltO(s,gIEi) represents the joint cumulative distribution of size and grade at time t1 of particles
issued from the breakage of Pi on the instant to.
From this eqn. 8 we may compute the a priori probability of (Ski < s , gki < g) independently of
the event Ei, as the expected value for the conditional probability on the event Ei, taken upon the
whole conditional events, is:

Prob { B E Vki : Ski < S , gki < g } =

= E [ Prob { B E Vki: Ski < S , gki < g I Ei } . Prob {Eil]

Combining eqns. 6,7 and 8, the last equation may be computed:

Prob ( B E Vki : Ski < s, gki < g ) = I. [ I. v ~. i ] ~ =

i k: Ski<S gki<g 1
154

vki
=I.i I. V = FlI (S,g) (9)
k: ski<s 8ki< 8

If we consider both Ski and gki as continuous variables, we may write eqn. 9 as follows:

or, in the differential form, using the density distributions,

(10)

This remarkable result [2] shows that:

1) the size and grade of the particle in which the Bastenaire point B falls are truly random
variates, the probability of which coincides, at any particular instant in time, with the
empirically defined joint size and grade distribution;
2) the size and grade joint cumulative distribution Fu (s,g) (or its density distribution, fu (s,g) )
at any given stage tl of the comminution process depends on the size and grade joint
distribution at some previous stage to, which demonstrates that the comminution process has
the distinctive Markovian hereditary property. Eqn. 10 is then called the forward comminu-
tion prediction equation, which obeys a Chapman-Kolmogorov-Smoluchowski type equation;
2) the size and grade joint distribution density fn (s,g) also depends on the size and grade joint
distribution, conditional on the stage to, of the daughter particles issued from the original
particle Pi of grade gi and size si - ftlltO(s,glsi,gi) - that is called the joint grade-and-size
transitionfunction. This transition function generally depends on:
a) the particular original instant to;
b) the time1apse (tl-to);
c) the original size Si and grade gj of the parent particle Pi;
d) on the size s and grade g of the daughter particle;
e) may eventually depend on other unaccounted relevant properties of the population and/or
the comminution machine, especially on fl(s,g) itself.

4.2. THE TRANSITION FUNCfION

The joint size and grade transition function fIIIIO(S,glso,go) has a very simple and clear physical
meaning: let us assume that the initial particulate system is made-up of iso-size, iso-grade particles,
the common size and grade of which are so, go; in this particular case, the joint size and grade
density fl oCsi,g0 takes the particular form of a double Dirac Delta function:

Eqn. 10 then becoming

meaning that the joint size and grade transition function fU1 10 (s,glso,go) represents the joint size
and grade density of the comminution product after a time lapse (tl-tO) when the feed is iso-size
 155

and iso-grade.
Let us now investigate the fonn of the transition function for an infinitesimal small comminution
time tl-to =,:\.t. After this small time interval, the probability of a particle, of initial size Sj and
grade gj, having been broken is given by:

(11)

where KtO(sj,gj), called the destruction rate function, (also called selection function), depends on
the particle's size and grade and on the particular time considered and has dimensions of volume
fraction per unit time, i.e. inverse time.
On the other hand, if the particle actually is broken, there is a definite probability of Bastenaire
point B belonging to a daughter particle of size s and grade g, this probability being dependent not
only on s and g, but also on the average way in which Pj breaks up. This probability is called the
formation function (also referred to as the breakage function)

Its units are volume fraction per unit size and grade, i.e., inverse size, since grade is
dimensionless.
Since a particle cannot give rise to particles of size larger than itself and also given that s and g
are essentially positive variates, we shall compute the transition function as follows:

0 for s>Sj

{ 1- KtO(sj,gj)':\'t for S=Sj and g=gj

f tO ,tO+llt(s,glsj,gj) = (meaning there was no breakage)

KtO(sj,gj),:\.tBtO(s,glsj,gj) for s<Sj

(meaning there was breakage)

or, in more compact fonn:

f tO ,tO+llt(s,glsj,gj) = [l-Kw(sj,gj)':\'t 1 <'i(Sj-s) <'i(gj-g) +

+ KtD(sj,gj),:\.t BtO(s,glsj,gj) (12)

4.3. GENERAL EQUATION OF COMMINUTION

Inserting eqn. 12 into eqn. 10, we will obtain the joint size and grade distribution at time t+':\'t:

ft,t+llt(s,g) = f dgj f ft,t+M(S,glsj,gj) ft(sj,gj) dSj =

o
1
= f dgj [ 1 - Kt(sj,gj) ':\'t 1<'i(Sj-s) <'i(gj-g) ft(sj,gj) dSj +
o
156
1
+ 6.t f dgj f Kt (Sj,gj) Bt(s,glsj,gj) ft(sj,gj) dSi =
o
= ft(s,g) - ft(s,g) Kt (s,g) 6.t + 6.t f dgj f Kt (Sj,gj) Bt(s,gl Sj,gj) fr(sj,gj) (13)
o
the physical meaning of which is quite simple indeed:
- the first term stands for the original volume of particles (parent particles) of size s and grade g;
- the second one stands for the volume of those particles which disappeared by comminution
during the time interval 6.t; and
- the third term stands for the volume of particles which acquired size s and grade g by
comminution of larger particles of size Sj>s. Bringing fr(s,g) into the first member of the
equation, dividing by 6.t and passing to the limit 6.t = 0 we will obtain the differential equation

1 ~

dfr~~g) = _ Kr(s,g) fr(s,g) + f dgj f Kr(sj,gj) Bt(s,glsj,gj) ft(sj,gj) (14)

o
Eqn. 14 is built under the assumption of mass conservation and set up as a Liouville equation in
(s,g) phase space [3], so that it is referred to as the instantaneous equation of batch comminution
[2], which is similar to an equation originally derived by Herbst [4], in a discrete form, from
population balance considerations.

4.4. LIBERATION AS A DYNAMIC STATE

The concept of liberation as a dynamic state results directly from the assumption of comminution as
a liberation production process itself, instead of a simple size reduction or surface increase
phenomenon, and is adequately described by the destruction rate function, K(so,go) of the parent
particles, and by the size and grade formation function, B(s,glso,go) of daughter particles issued
from the parent ones [3].
As we saw in §2, the size and grade formation B may be broken up into a product of tre
marginal size formation function, A, and the grade formation function, conditional on daughter
particle size, C, both on parent particle size and grade, so that:

B(s,glso,go) = C(gls,so,go) A(slso,go) (15)

We should note that, since comminution is a first order Markovian process, knowledge of daughter
particle size renders parent size irrelevant, Le.:

aC(gls,so,go) = 0
aso

Le.

C(gls,so,gO) => C(gls,go) (16)

 157

S. Non-discriminating transgranular comminution

Comminution is said to be non-discriminating transgranular, if there is no relation whatsoever

between the mechanism of fracture and the properties of the in situ grains of the rock texture. We
do realize, of course, that this concept is so idealised that there is not any real comminution process
that remotely resembles it; it is, however, a good starting point for a discussion of liberation in a
general and conceptual approach.
For non-discriminating transgranular comminution, certain simplified forms apply for the
descriptor functions K and A. Being so, since both destruction rate and size formation must, in this
case, be independent of parent particle grade, this leads us to:

JKj~,g) = 0

i.e.,
K(s,g) => K(s) (17)

and
JA(slsO,gO) =0
agO

i.e.
A(slso,go) => A(slso) (18)

The assumption of the simplified forms eqns. 17 and 18 makes for the general description of
comminution as a liberation-producing process, in the non-discriminating case, being separated
into:
a) the standard description of comminution as a size-reduction process, by means of the conven-
tional destruction rate or selection, K(s), and size formation or breakage, A(slso), functions;
b) the description of grade distribution within an iso-size fraction of the particle population,
C(gls,so); which allows us to rewrite eqn. 14 as :

1
dft(s,g)
-d-t- = - Kt(s) ft(s,g) + J dgj J Kt(Sj) Ct(gls,gi) At(slsi) ft(Si,gi) (19)
o
Standard size-reduction theory is nowadays a well established field, regarding both parameter
condensation techniques and parameter fitting methodology [5,6] in order to obtain Kt(sj) and
At(slsj); finding an adequate analytical form for Ct(gls,gj) has been for years the central topic of
many researchers.
As a first step, Madureira [7] pointed out that a limit probability argument would lead to a Beta
form function: in fact, particle formation by comminution is a complex random process arising
from multitudes of microscopically different events and unpredictable factors which suggest that an
error propagating mechanism is at work and gives rise to macroscopically different products;
Rocha e Silva [8] pointed out that error propagation in comminution results from a non-linear
bifurcation mechanism. We now make the further assumption that no single factor, or small group
of factors, fundamentally controls the process, which is consistent with the non-discriminating
158

argument accepted. Consequently, the multiplicity and the equipotency of all the factors as a whole
allow us to invoke the limit distribution theorem for the volumes VI and V2 of two mineral species,
SI and S2, in a binary texture, which are genuine gamma variates [3,7]. A theorem by Johnson [9]
states that, given two independent gamma variates, the new variate g =Vt!(VI+V2), which is the
volumetric grade of species SI, has a Beta distribution fonn, that allows us to assume:

Ct(gls,gO) = j3(g;p,q) = Cpq gp-I (1 - g)q-I (20)

where g is constrained to the range ]0, 1[. Cpq is a constant and the parameters p and q, whieh
control the Beta function's shape, must be functions of parent particles grade, gO, and, of course,
of daughter particle size, s, in order to describe the desired dependence of these shapes on particle
size:

p = fl (gO,s) and q = f2(gO,S) (21)

5.1. EULER'S BETA FUNCTION

We are now going to investigate how the shape of the Beta density changes with changing p and q,
i.e. with changing particle size, always keeping in mind that the Beta distribution is a descriptor of
statistical properties of the whole population and becomes meaningless for individual particles.

5.1.1. The (p,q) Diagram. The (p,q) Diagram of Figure 1 is divided into four regions and the
small diagram inside each region shows the general shape of the plot of C(gls,go) against g.

1":)1
I. DOD - - - - - - - - - - - - - - - ~ - - - - - - - - - - - - - - -

1
[I .'.~j[.
D.DO::J+---------I1c:---------
0.008 1.000

Figure 1. Beta (p,q) diagram.

 159

In the region p,q> I, both mineral (value) and gangue are unliberated, i.e. all panicles contain
mineral and gangue locked together; in fact, at both ends of the g interval there is no substantial
area; the curve has its maximum at:

M_ P- I (22)
g -p+q-2

so that most particle grades cluster around this value which is near the mean grade. g*, of the ore,
the form of which is:

(23)

We should note that this clustering is more intense for higher values of (p+q); (p+q) is thus seen
to be a measure of panicle size, because, as we know, larger panicles tend to have grade near the
average grade of the ore. In the region p<l ,q> I, there is an infinite maximum of C(gls,go) at
g = 0: it means that there is a significant amount of pure, or nearly pure gangue panicles. Since in
the extreme g= I the frequency is null, in this region some gangue is liberated but mineral is not.
We note that p < q so that g* < 0.5.

Figure 2. Iso-grade lines in the (p,q) diagram.

Conversely in the region p > I , q < I, with g* > 0.5 following the above idea, a significant
amount of panicles are pure or nearly pure mineral (value), so that it is the region where some
mineral is liberated but gangue is not. Finally, in the region p.q < I, there are two infinite maxima
of C(gls,go) at g = 0 and g = 1. meaning that both mineral and gangue are liberated.
From this we infer that:
- sm: q = f1 (sm) = I is the particle size at which mineral starts to liberate;
- sg: p = f2(sg) = I is the panicle size at which gangue starts to liberate;
- and finally. the lines p = 1 and q = 1 in the (p,q) diagram represent the liberation barriers of
160

gangue and mineral.

5.1.2. Iso-Grade Lines. Comminution being transgranular and non-discriminating as we have

assumed, the average grade g* = p/(p+q) of the iso-size fraction described by C(gls,go) must be
inilependent of the size s and equal to the average grade g** of the ore:

g*=~ =g** (24)

p + q

Thus, iso-grade lines in the (p,q) diagram are straight lines through the origin:

1 *
q=p~g
(25)

so that iso-grade fractions obtained by non-discriminating comminution of an ore of average grade

g* must match with one of the straight lines of eqn. 25, the slope of which fits the average grade of
the ore. These results may be seen in Figure 2. Inspection of the way in which the shape of
C(gls,go) varies along an iso-grade line (see also Figure 1) clearly shows that particle size must
increase with the distance from the origin. On the other hand, it is also shown that when g* < 0.5,
gangue becomes liberated at sizes coarser than those of the mineral; conversely, mineral liberates at
coarser sizes than gangue when g* > 0.5. This is a well known fact since Gaudin's [10]
pioneering work: in transgranular, non-discriminating comminution, liberation of the major
constituent occurs at sizes larger than the in situ grain size of the minor constituent.
These interesting properties of the Euler's Beta density fully justify its choice as a descriptor of
the liberation process since, among the currently used probability distributions, it is the only one
possessing such a variety of shapes, which arc indispensable to describe liberation, and that has the
range ]0,1 [ for the domain of variation of the independent variable

5.1.3. Iso-Variance Lines. Since variance is a measure of the dispersion of the particle grade
values around the average grade of the ore, it can be used for an effective global evaluation of the
state of liberation of a given particulate system. Thus, it may be interesting to observe how the
variance of grade g (given daughter particle size s, and parent particle grade gO) does depend on
parameters p and q, and ultimately, on size s (as we have been pointing out in the last paragraphs).
According to Johnson [9], the variance of Beta distribution is computed as follows:

02 *- p.q (26)
gls,g - (p+q)2.(p+q+ I)

If, instead of the variance itself, we take the ratio between the actual variance - eqn. 26 - and the
maximum value for the variance, compatible with the given average grade g*, we shall have the
value 1;2 as a first assessment of the liberation state of a size fraction, s:

(27)

The circumstance which makes for the maximum of variance of grade occurs when both mineral
and gangue are fully liberated, that is, when the distribution is represented by two Dirac delta
functions centered at grades 1 and 0, respectively:
 161

C(gls,gO) = (I-g*) beg) + g* beg-I) (28)

thus:
02gls ,g* max = g*2.(l_g*) + (l_g*)2.g* = g*.(I-g*) = (P~~)2 (29)

Reworking eqn. 27 we shall obtain:

I
1;2 = p+q+l (30)
that is:
I - 1;2
q=1;2_p (31)

which means that iso-variance lines are parallel straight lines of slope -45 degrees.

Figure 3. Iso-size lines in the (p,q) diagram.

5.1.4. Iso-Size Lines. Since we have agreed that 1;2 is a measure of the liberation state of the iso-
size fraction, s, and depends only on (p+q), we infer that iso-variance lines must also be iso-size
lines, so that:

p + q = <I>(s) (32)

and inserting eqn. 32 into eqn. 25 we get:

p = g* .<1> (s)
q = (l - g*) .<1> (s) (33)
162

and, finally:

q = <I> (s) - P (33')

Now that we know their form, the problem is how to graduate the iso-size lines, that is, how to
rule the (p,q) diagram in terms of particle size, s, which is the conditioning variate for C(gls,go).
Establishing the precise analytical form of <I>(s) from first principles has been a problem worked on
by some authors [7,11] for years; direct experimental work at this stage is, of course, barred both
by the sheer dimension of the work involved and by the impossibility of obtaining real non-
discriminating transgranular comminution. We have succeeded in solving the problem by means of
a Monte Carlo technique which has the additional advantage of having a clear physical meaning
and, thus, being easily conceptually controlled at all stages. This technique has been developed by
Cavalheiro [12,13] and is built up as follows:
1. a three-dimensional synthetic ore texture is generated in computer memory; most of the
textures tried were made out of Dirichlet-Voronoi polyhedra homogeneously and isotropically
grown around Poisson-distributed original seeds, which accurately mimic granitoid textures
[14-16];
2. the texture is sampled according to the assumption of non-discriminating transgranular
comminution: randomly placed particles of all sizes and shapes are taken from it;
3. each particle is examined for volume and surface grade determinations in each mineral
species;
4. plane sections of the 3-dimensional textures are also sampled and treated as 2-dimensional
entities by the same techniques;
5. in this important step, the Beta distribution form eqn. 20 is fitted to the C(gls,go) data
obtained by simulation in the above steps 2, 3 and 4, the goodness of fit of which is then
judged by an adequate statistical criterion. It is important to note that the limit probability
conjecture was vindicated in every tested case; as expected, this result is valid for bulk
(volume) as well as for planar grades, while it came as a total surprise that the Beta
distribution fined the particle surface grade data almost as well as it did the bulk and planar
grade data.
6. plots of the (p+q) = <I>(s) adjusted values vs. size, s, are then examined to find a plausible
fitting function that conveniently approximates <I>(s). For the case of Poisson-Voronoi
textures and for non-discriminating comminution, Cavalheiro [12] has found:

<I> (s) = K s exp(K s) (34)

to be a very satisfactory fitting function. Again, this type of fitting function applies to surface
data almost as perfectly as to bulk and planar data;
7. calculations of the empirical K values for a large number oftextures showed that it varies with
the in situ grain size, as expected; since dimensional analysis suggests that the product K.s
must be dimensionless, Cavalheiro wrote:

K=C/z (35)

where z is the in situ (textural) average grain size and C is a constant depending only on
texture type, the nature of which allowed the author to suggest the name of topological
constant for it.
 163

5.2. HOW BETA DIsmmUTION AND TOPOLOGICAL LAW DO WORK ON

LmERATION PREDICTION

For a good understanding of how this model does work on liberation prediction, let us imagine a
sample of a ground ore in a given state of comminution. The available information about that
sample is the average grade, g, in the valuable mineral and the overall discrete size composition, ci,
extended over a wide range of n size intervals. Prediction of the liberation state of this sample is
adequately described by the marginal grade distribution, as we have already pointed out in previous
pages, or in this particular case by the discrete grade composition, ti, extended over a wide range of
m grade intervals.
Since it is accepted that the ore texture is of the granitoid type, characterised by the in situ
mineral particle grain size, z, the model presented in the last paragraphs will be applied.

5.2.1. Liberation predicting algorithm. The target defined in the last paragraph can be achieved as
follows:
a) Restrictions:
- the ore texture must match a granitoid type;
- the comminution must be non-discriminating transgranular so that all the discrete iso-size
fractions will have the same mean grade (equal to the initial ore average grade) at each
comminution stage;
b) Input data:
- vector matrix representing the discrete sieve series, Ii, of n elements,
)In,ln_l[ ... [li+I' li[ ... [\I,lmax[;
- vector matrix representing the grade discretisation of m clements,
)0. gIl ... [gj' gj+JI ... [gm. 1[;

5/z

Figure 4. Liberation model.

164

- vector matrix referring the size composition of the sample, d(li), which will be of course a
real ground product or, alternatively, a simulated product derived from a comminution
algorithm;
- average grade of the whole sample, g;
- average in situ mineral grain size, z.
Figure 4 provides in graphic form the overall background for following the liberation algorithm,
since it collects all the relevant functions used in the model. First of all, for each size interval i we
must compute the mean size Imi; then, the ratio between the mean size of each size class and the in
situ texture grain size, lmi/z, is introduced in the plot on the left hand side of Figure 4, in which
Cavalheiro's topological law (eqns. 34 and 35) is plotted, to compute <I> (lmJz), the origin intercept
of the iso-size line lmi/z, accordingly with eqn. 33'.
From eqns. 33 we can determine both values of p and q parameters that match the condition of
ore average grade equal to g, which will be represented by the coordinates of the interception
between the iso-size line lmilz with the respective iso-grade line, g. Since we have just determined
the p and q values, the next step will be the integration of Beta function for each j grade interval
[gj' gj+l[, to obtain the grade histogram conditional on size:

gj+l
f ~(g;p,q) dg = c(gmjllmV (36)
gj

that can be computed from Kummer's procedure [17].

At this point, the joint histogram of size and grade, f(lmj,gmj), where lmi and gmj are the size
and grade mean values in the respective size and grade classes, will be computed by the product,
according to eqn. 4 :

(37)

The marginal grade histogram, h(gj), as the decisive target, is easily obtained through eqns. 3, by
integrating eqn. 37 for all size classes, that is:

h(gj) = L f(lmj,gmj) (38)

i

which fully describes the liberation state of the given sample.

If, instead of starting from a ground sample in a given state of comminution, we were able to
perform a simulation with the algorithm that models the comminution both on size and grade, the
joint histogram of size and grade, f(lmi,gmj) would be obtained directly from its own results.

5.3. WASHABILITY CURVES OF A SEPARATION PROCESS

Let us imagine that the sample of ore, referred to in the last paragraph, is submitted to a separation
process, such as flotation, for example. Technically a1l the work will develop under the assumption
of the existence of a technical cut grade which refers to the minimum grade allowed in the
concentrate product. On a conceptual basis, conSidering liberation alone, this means that all
particles in which the valuable mineral is liberated above the purity level corresponding to that cut
grade, PM, (volume grade or, in the case of flotation, surface grade), should be carried into the
 165

concentrate product.
The volume fraction of particles, LM, that matches the criterion is:

1
LM = L L c(gmjllmi) (39)
i gmj=PM

which is the fonnal representation of the well known washability curves. The average grade of the
whole amount of particles recovered in the concentrate product can be computed as follows:

1
L Lg.c(gmj>lmi)
i gmj=PM
(40)
1
L L c(gmjllmi)
i gmj=PM

Since gM may be thought of as the average fraction of valuable mineral in the concentrate product
and go as the overall volume fraction of valuable mineral in the feed ore, it follows that the volume
fraction of valuable mineral recovered in the concentrate - the recovery - is:

(41)

This recovery, called the ultimate recovery, derives its name from the liberation status of the ore
alone, as a geometrical factor.

5.4. FLOTATION KINETICS BUILT UNDER LIBERATION ANALYSIS

Rotation kinetics, as the study ofthe variation of froth overflow product properties with time and
the quantitative identification of all rate variables, is described by many authors [18-20] as a first
order process, according to its analogy with chemical reactions. Since grade and size act as
important controlling variables, flotation perfonnance will depend on the phenomenological
classes, mi,j, i.e. the mass of particles in each size and grade intervals, obtained from the original
joint grade and size distribution. Thus, first order kinetics will be described by:

dmi.i<t) / dt = - Ki,j miit) (42)

where mi,j(t) is the mass of particles with identical flotation rate constant Ki,j, i.e. the frequency
distribution of the flotation rate constants Ki,j, and t is the time. The rate constant Ki,j can itself be
broken up, at least, into the product of the probability of collision of particles to bubbles, i.e. at
least, a function of size and the probability of adhesion to bubbles after collision that depends on
the particle floatability after collection and is closely related with the particle grade:

Ki.j = PCj(s) PajCg) (43)

As we saw in §S.2, given the size composition of the sample, its average grade and the in situ grain
size, the developed algorithm for liberation prediction can be used to compute the mass mi,j of each
166

phenomenological class, for Ihe case of Voronoi textures and non-discriminating comminution.
Using eqn. 43 and a good parameter condensation technique, a simulation algoriIhm for batch
flotation can be derived by integrating eqn. 42 on time, t:

met) =~LmjiO) . exp(-Kj,j-t) (44)

1 J

The reader should notice Ihat, instead of using Ihe traditional linear combination of different rate
constants, whose frequency distribution is back calculated on a mathematical basis alone, Ihe
liberation prediction algorithm can be used to compute Ihese flotation rate constant frequencies.
This procedure allows, indeed, for a suitable description of the flotation kinetics on a
phenomenological basis.
Although Ihis paper doesn't aim mainly at studying the flotation kinetics, we can go through
parameter condensation a little more in order to improve our approach. Thus, the probabilities of
collision and of adhesion can be drawn up as functions of size and grade, respectively, in Ihe
simplest way, as follows:

(collision factor) (45)

(floatability factor) (45')

where Ql and Q2 are global condensed parameters. It can be seen Ihat PCj decreases wiIh
decreasing particle size, while Paj increases wiIh increasing particle grade. In such a model:
Ihe collision factor is related to particle size and also to bubble diameter Ihrough Ihe parameter
Ql; although it acts as a simple hydrodynamical factor, its influence on the final recovery is
not negligible.
the floatability factor, alIhough directly dependent on the particle grade gmj and, consequently,
on its liberation state, is also affected by the condensed parameter Q2 Ihat establishes a
positive relationship between floatability and grade due to collector activity.

5.5. APPROACH TO THE CONCEPT OF SEPARATION EFFICIENCY

Separation efficiency, Tl, of a flotation process must be split up into three factors of different
meaning, as we have just described

Tl = TJG· Tlp . rtr (46)

I. a geometrical factor, TlG, directly related with the liberation status of the sample, which is
expressed by Ihe ultimate recovery; at this level the separation is performed by unerringly
sorting particles aecording to Iheir grade;
2. a physical factor, Tlp, which expresses the difference between the ultimate recovery and Ihat
produced by sorting particles according to their floatability (which depends boIh on grade and
on the collector effectiveness, pH and Eh of the pulp, and so on); that difference is a result of
Ihe more or less fuzzy relationship between floatability and grade, which is quantitatively
expressed by Q2 in our model;
3. a technical factor, rtr, which expresses the fact that a real separation is never perfect; in our
case, Ihis imperfectness is originated by Ihe interference of the effect of particle size in the
 167

collision mechanism, through parameter QI; in this particular way, size is a particle property
loosely related to grade.

6. Numerical Experimentation with the Batch Flotation Simulator

This section aims at presenting some results of numerical experimentation, showing how a
liberation algorithm can be embedded in a flotation simulator.
The algorithm, developed in Pascal, has as input values the ore average grade, g and the in situ
texture grain size, z, which are used to compute Euler's Beta distribution c(gls) that can be seen in
Figure 5: in the coarse size range neither gangue nor valuable mineral are liberated, while in the
fine size range both are well liberated.

EULER's BETA DISTRIBUTION

/0"",

Figure 5. Euler's Beta distribution.

The following step of the algorithm generates several size distributions according to the
operator's request. Each size distribution is described by its maximum size. Afterwards, the
operator is asked about the value of parameters QI and Q2, which will be used for drawing the
collision and adhesion mechanisms of particles to bubbles performance. Will the promise of greater
selectivity through manipulation of the potential, and/or reduced reagent costs, materialise in the
future? There is no doubt that the answer is yes.
In order to simulate diverse experimental conditions, the program was exploited with 5 different
sets of parameters: 4 feed products simulating different degrees of grinding and 2 collectors of
different selectivities:
168

AvgFccd In situ Maximum Collector

Number Grade Grain Size Feed Size Selectivi ty
1 0.1 1 1.00 High
2 0.1 1 0.75 High
3 0.1 1 0.50 High
4 0.1 1 0.25 High
5 0.1 1 0.25 Low

From experiments number 1 through 4 data were collected to plot Figure 6, which shows the
influence of the degree of liberation in the flotation kinetics behaviour: fine sized material (max feed
size =0.25) makes for better recovery than coarse sized products (max feed size = 1.(0).

SIMULATION OUTPUT
Different Liberation States

~ 0.5
~O.4~~~--------------------~

0.3 -IdIf------------- --------------1 c:---c-:-----:-,

0.2

0.1 +---~--~----~--~------__I
o 10 20 30 40 50 60
Flotation Residence Time

Figure 6. Influence of the liberation state.

SIMULATION OUTPUT
-ll1~
~ 0.9 ....-
=::11
====:;j. .--+--
Recov/low S.1ect

'iO.B ~ +-- ~~ :Ov/HiQ"'I Select

~ 0.7 r~- ----

=e/LOW Select I
0.6 it Grade/High Select
0.5

~ 004 +t'If-----------------------------1 r~- ..-d-Cr-~-d~-_-O~-O-.I'

j 0.3 -.------------.- ,-,-"- IGrain Size = 1 I

0.2 +O----I~O----20----3~O----4~O----5~O-----j60 !Max Size .::0: 0.25 I

Flotation Residence Time

Figure 7. Influence of collector selectivity.

 169

It is very important to note that the product with maximum feed size equal to the in situ texture
grain size exhibits a deficient liberation status. This is, indeed, another important statement that
agrees with Gaudin's work [10]: in non-discriminating comminution, the liberation of a given
constituent occurs at sizes smaller than the in si tu texture grain size.
In Figure 7 the data from experiments 4 and 5 have been plotted; it can be seen that the collector
selectivity affects both the recovery and the average concentrate grade: with low collector selectivity
the recovery increases slightly while concentrate grade reduces sharply.
Finally, in Figure 8 we can visualize the flotation rate constants frequency distributions for each
size class. This plot shows that the highest flotation rates conform with the intermediate size
classes, which agrees with industrial practice: in fact, coarse sizes have a low flotation rate on
account of their insufficient liberation while fine sizes do not have a good flotation performance
because of the decrease in their collision efficiency with bubbles.

SIMULATION OUTPUT
Average Flotation Rate Constants

S:':::I======-~'.-.~.---1"-J..,,-.~=--::~-.~--.-..-.--'-1:1 1
.1::~ Selectivity I

~ 0.07 ' \ Low Selectivltt

"r,,- ~~
~ 0.06
o
u 0.05
l'
0.04
J7 '\ - :
I
~

.e 0.03 \ --
j \ ~ Grade = 0.1
~ 0.02 ---------
c;:
~O.Ol
j \ ICroin Size:::: 1

'" 0 ---,---,
~ \.i 'T"
0.001 0.01 0.1 10
Size (mm)

Figure 8. Flotation rate constants frequency distribution.

The last statement leads to a technological conclusion of great importance. In a traditional

flotation circuit, the tailings of a rougher cell, although being a complex product since its low
floatability derives from different physical conditions (low efficiency in collision for fine particles
and insufficient liberation of coarse particles), are normally fed to the same scavenger cell, without
concern for the size factor. We think that this problem should be taken into account while designing
flotation and regrinding circuits.

7. Extensions to the multi-species case and to discriminating comminution

The binary composition model may be extended to the case of n coexisting mineral species,
according to Madureira[3], by simple replacement of the Beta distribution by the Dirichlet-Filon-
Isserlis distribution.
The main problem here seems to be the extension to the diSCriminating mechanism of
comminution. Although we may imagine some fairly plausible mechanisms, such as peri granular
(that is, preferential breakage along grain boundaries) and discriminating trans granular (preferential
breakage within certain mineral species) and some research work is being done, so far we have:
170

- neither a clear criterion nor standard experimental strategies to decide which, if any, particular
mechanism is at wolk in a specific comminution process;
- not even a way of telling which comminution descriptor function K(Si), C(gls,gD, or A(slsi) (see
eqn. 19) is affected by each of these discriminating mechanisms and how.

8. References

1. Bastenaire, F. (1965) Etude Theorique du Mode de Generation des Distributions

Granulometriques, Revue de L'Industric Minerale, Janvier, 39-50.
2. Madureira, C.M.N., and Leite, M.R.M. (1982) Metodos Cibernericos em Mineralurgia.
Fundamentos Estatisticos do Estudo Fenomenologico dos Processos de Tratamento de
Sistemas Particulados, Departamento de Minas da Universidade do Porto.
3. Madureira, C.M.N. (1988) Size, Grade and Liberation: A Stochastic Approach to the
Fundamental Problem of Mineral Processing, in Proceedings XVI International Mineral Pro-
cessing Congress (Stockholm), Elsevier, Amsterdam, pp. 1823-1834.
4. Herbst, J.A., Rajamani, K. (1985) Development of a Multicomponent-Multisize Mineral
Liberation Model, 114th A.I.M.E. Annual Meeting, N. Y., Feb.
5. Leite, M.R.M. (1984) Moagem Nao-Linear em Moinho de Barras: Metodo Global para a
Detenninaqao de Parametros Cineticos, PhD Thcsis, Departamento dc Minas da Univcrsidade
do Porto.
6. Leite, M.R.M. (1989) Kinetic Models for The Simulation of Crushing Circuits, Minerals
Engineering 3.(112),165-180.
7. Madureira, C.M.N. (1978) Teor, Calibre, Libertaqao: um Tratamento Estocastico do
Problema Fundamental da Preparaqao de Minerios, 20 Congresso da Ordem dos Engenciros,
Porto.
8. Rocha e Silva, J.A.C. (1989) Calibre: Conceito e Medida, PhD Thcsis, Departamento de
Minas da Universidade do Porto.
9. Johnson, N.L. (1960) An Approximation to the Multinomial Distribution: some Properties
and Applications, Biometrika Q 93-102.
10. Gaudin, A.M. (1939) Principles of Mineral Dressing, McGraw-Hill Book, New York.
11. Barbery, G.H., and Gateau, G. (1979) Liberation Analysis with the Help of Image Analysis,
in J. Laskowski (ed.) Proceeding XIII Int. Min. Proc. Congress, pp. 365-394.
12. Cavalheiro, A.A.T. (1984) Libertaqao de Minerios por Cominuiqao: Modelo Matematico, PhD
Thesis, Departamento de Minas da Universidade do Porto, Boletim de Minas, 22(3).
l3. Cavalheiro, A.A.T. (1990) Prediction of Coal Amenability to Dry Gravity Separation, 21st
Annual Meeting of Fine Particles Society, San Diego, California.
14. Pielou (1979) Mathematical Ecology, Charles Griffith & Co. Ltd.
15. Cox, D.R. and Isham, V. (1980) Point Processes, Chapman & Hall.
16. Serra, I. (1982) Image Analysis and Mathematical Morphology, Academic Press.
17. Abramowitz, M. and Stegun, LA. (1972) Handbook of Mathematical Functions, Dover.
18. Imaizumi, T., and Inoue, T. (1963) Kinetic Considerations of Froth Flotation, in Proceedings
6th Int. Min. Proc. Congress (Cannes), pp. 581-593.
19. Kapur, P.e., and Mchrotra, S.P. (1973) Phenomenological Modelfor Flotation Kinetics,
Trans. IMM, C229-C234.
20. Lynch, A.J., Johnson N.W., Manlapig E.V., and Thornc, C.G. (1981) Mineral and Coal
Flotation Circuits, Elsevier, Amsterdam.
1.7. ELEMENTS OF PHYSICALLY -BASED MODELLING OF THE
FLOTATION PROCESS

H.J. SCHULZE
Research Institute of Mineral Processing
0-9200 Freiberg, Saxony
Germany

ABSTRACT. The general aim of physical modelling of flotation is to precalculate floatability of a

component as a function of flotation time. on the basis of physical. physico-chemical and
hydrodynamic parameters. e.g. interparticle and hydrodynamic forces. Drastic simplifications are
necessary in modelling. in order one to be able to describe the overall process clearly.

1. Introduction

Rotation is a macroprocess composed of a very large number of individual microprocesses taking

place simultaneously and successively in time and in space. Rehbinder. the great Russian colloid
chemist. therefore called flotation "the encyclopaedia of colloid science".
Drastic simplifications are necessary in modelling. so that it is possible to describe the overall
process clearly. The reduction to the essential microprocesses (elementary steps) is one such
possible way of simplification.
The most important microprocesses are [1] :
- the approach of a particle to a bubble in the flow field;
- the formation of a thin liquid film. its rupture at the critical thickness. herit. connected with the
formation of three-phase-contact (Lp.c.). and
the stabilization of the particle / bubble aggregate against external stress forces.
The simplest form of a kinetic equation describing floatability is the variation of the particle
number per unit time. in a given volume of apparatus. It depends on : the number of particle/
bubble collisions per time. z. on the particle and bubble numbers. np and nb. and on the individual
probabilities of interaction:

(1)

where
- P e is the probability of collision in the flow field in the surroundings of a bubble.
- Pais the probability of adhesion.
- Ptpc is the probability ofLp.C. formation after film rupture, and
- Pstab is the probability of particle / bubble stability against external stress forces.

171
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 171-180.
© 1992 Kluwer Academic Publishers.
172

2. Probability of Particle Collision

Not all particles. which move within the bubble projection path (Rb). can collide with the bubble.
but only those within a streaming tube of limiting collision radius. Re. Therefore. the general
collision probability is defined by the ratio of Rc and Rb :

(2)

In recent years a comprehensive description has been given in this field [2-5].

particle trajectory

Rp
/---\
sliding
vp1 - --
\
/
~~/
/
/
\
/ I
1\
/

fluid stream
lines

Figure 1. Schematic representation of particle interaction with a gas bubble by colliding and sliding.
 173

dB Imm

0.2
0.5
1.0 '''_ o ANFRUNSdB= 1.0mm
2.0 ~_
3.0
+• ANFRUNS dB =0.6mm
SCHULZE, GOTTSCHALK
1:dS = 2-Smm, dp=148jJm
2:d B= 2-4mm,d p =240jJm
Re8-30
3:dB=2-Smm,d p =148jJ1'i'I
ReB'" 300

Figure 2. Probability of particle collision: comparison between theory and experimental data (see
ref. [7]).

3. Probability of Particle Adhesion

When a particle approaches a bubble to within a small distance, two types of interaction may occur
in the vicinity of the latter, as has been found by flow tube experiments during the past few years
[6,7] :
- impact (collision) processes, in which the bubble surface is strongly deformed, an extended thin
liquid film is formed and the particle is repelled, unless attachment takes place during the first
contact.
- sliding along the surface with a weak surface deformation ("point contact").
Whether the particle can be attached or not depends crucially on the duration of contact (collision
time, '" or sliding time, lsi), in relation to the film drainage time until its rupture (induction time, tJ.
Therefore, the time required for attachment, "'on (collision or sliding), must be greater than, or
equal, to the induction time. Hence, P a consists in principle of two parts, the adhesion probability
174

by collision, Pac, and the adhesion probability by sliding, Pasl [8]. Which of these dominates
depends on many factors and has not been investigated in detail. It may be stated, however, that
collisions occur with large, heavy particles moving normally to the bubble surface, while sliding
occurs with fine and light particles and at small normal velocities.
The most important value for precalculating the adhesion probability is the critical thickness of
film rupture. It is well known that the properties of thin liquid films depend on interparticle forces
[9,10]. The best fit relation between hcrit and the experimentally easily measurable parameters,
advancing contact angle, 6 a and surface tension, 0 is [II] :

hcrit = 16.6 [0 (1 - cos6J] 0.22 (3)

This power law may be found physically from the theory of thin film stability [12] and the
theory of contact angles of wetting films [13].
It was found that the adhesion probability by collision is exponentially distributed [14,15].
Hence the following equation holds:

tc - tmin
Pac = 1 - exp [- --t1-.- ] (4)

.-.-~~~----~-------,----~nml
__I-----r- 40nm P~

lOnm
q05f-----II--\-----I'1r--

------I
i

q01L-_ _,!,;-_ _~_ _~---l........,;;;:;'.

ISO 200
Rplpm

Figure 3. Probability of particle adhesion by collision. Pae : spheres; p' ae : broken particles. RB = 0.5
mm, VB = 10 em/s, Rmin = 2.5 !-lm, (J = 70 mN/m, Qp = 2.5 g/em3 • Parameter: herit .
 175

All the required values may be calculated [16,17]. The most important results are summarised
(Figure 3) :
- Pac is a function of particle size; for smooth, spherical particles it decreases drastically with
increasing particle size, whereas for rough (broken) particles it is exactly the opposite, it
increases with particle size.
- The influence of the degree ofhydrophobization of the particle surface, which is characterized by
herit, is not very important for smooth particles, but in the case of broken particles Pac increases
drastically with the increasing of hent.
The probability of adhesion by sliding, Pasl, can be derived from the equation of particle motion
in the flow field of the bubble. The complete solution seems, however, to be so complicated for
the moment, that again some simplifications must be made [18-20] :
- particles move quasi-stationarily, inertia-free on an almost circular path across the bubble
surface;
- as the sliding path, L, is much greater than the film thickness, hp, furthermore dL/dt » dhp/dt is
valid;
- the tangential component of fluid flow at a free bubble (CB = 4) can be approximated by the
potential flow;
- for the completely rigid bubble (CB = 1) intermediate flow is suitable as an approximation [23].
From the force balance (gravity, centrifugal, flow, resistance to lifting [21]) a complete set of
equations of the particle motion may be derived. The solution can be found by numerical
integration only. It gives the thickness of film in terms of the particle position. Attachment is
possible, when the critical thickness is reached during sliding in the range of touching angle, {jJ and
{jJ = 90°. The touching angle, at which hcrit is marginally reached is denoted as the critical angle,
{jJcrit. Particles touching the bubble at {jJ:5: {jJcrit are able to adhere (Figure 4).
Therefore, the attachment probability by sliding may be defined as

P asl =sin2 {jJcrit (5)

Pasl depends strongly on the character of flow (potential or intermediate flow), the degree of
mobility of bubble surface (rigid or free movable) and on the critical film thickness as well as the
particle size. The smaller the particle the greater is P asl. But the main influenee on P asl arises from
the real bubble rising velocity, VB : the smaller the bubble rising velocity (bubble size) is, the
higher is P asl (Figure 5).

4. Probability of Aggregate Stability

If the attachment capillary force, F ca, between particle and bubble is stronger than the sum of all
external stress forces, then this aggregate remains stable on its long way from the place where it is
formed to the froth layer.
It is possible to calculate the probability of aggregate stability from the foree balance of spherical
particles in the liquid/gas interface [22] :

F - Fd t
Pstab = 1 - exp [_ ca'F e ] (6)
det

Here, Fearn is the maximum capillary force and Fdet the sum of all external detachment forces.
176

trajectory

circle with herit

r--- -
L

Figure 4. Almost circular movement of particles by sliding; two possible particle trajectories near the
bubble surface; (1) during sliding, the critical thickness, herib of rupture of the thin liquid film, between
the particle and the bubble surface, is reached; (2) herit is not reached.

The additional acceleration, ac, which detennines the detachment forces depends on the intensity
of the turbulent vortices. hence finally on the energy dissipation. E, in a given apparatus volume.
We assume that particle - bubble aggregates. the dimension of which corresponds to those of the
turbulent vortices. are moved mainly by the centrifugal acceleration in the vortex;

(7)
ac= (RB + Rp)1/3

For a bubble flowing around laminarily. e.g. for conditions realised in column flotation for
example. ac depends mainly on the tangential vclocity of the particle across the bubble surface. in
the case of an unretarded bubble (CB = 4) and in a potential flow approximately

(8)

where vps is the particle settling velocity.

 177

to 1,040nm
Pasj 40nm
t J
Q5 lan~ tQ4qs 40nm "m~1
q4 --

0.3 q3 BOnm
40nm
Q2 -
Q2 5an~
nm
IOnm
0,1 0.1
Semis
} "."fOanl$

so ISO 200
-Rpl)m

Figure 5. Probability of particle adhesion by sliding. (Left) intermediate flow, Cs = 1; (right) poten-
tial flow, cB = 4. Rs = 0.5 mm, VB = 10 cm/s, Rmin = 2.5 11m, (J = 70 mN/m, Qp = 2.5 g/cm 3 . Para-
meter: herit.

Using the example of two modelling systems - flotation of quartz and flotation of potash salt - it
may be shown in which way floatability and elementary steps correlate as a function of variables.
Figure 7 shows that the best quartz floatability, in the pH range from 10 to 11, coincides with the
optimum of all variables of the investigated elementary steps. It seems also that the instability of the
thin liquid film - as indicated by the induction time ti - does not play any significant role in
flotation, being quite the same in both reagent regimes (Figure 8).

5. References

1. Schulze, H.J. (1984) Physico-chemical Elementary Processes in Flotation, Elsevier,

Amsterdam.
2. Weber, M.E., and Paddock, D. (1983) Interceptional and gravitational collision efficiencies,
J. Colloid Interf. Sci. 94, 328-335.
3. Plate, H., and Schulze, H.J. (1991) Modelling of the overall flotation process based on
physico-chemical microprocesses, Proceedings XVII Int. Miner. Process. Congress.
4. Schuch, G .. and Laffler, F. (1978) Uber die Abscheidewahrscheinlichkeit von Feststoff-
partikeln an Trapfen, Verfahrenstechnik, Mainz 12,302-306.
5. Duchin, S.S., Rulev, N.N., (1986) Kaaguljacija i Dinamika Tonkich Plenak, Naukova
Dimka, Kiev.
178

P.
1,0
q8
~\ ~ t-........ - ........" , a-goo

\ \ '-, ~
slab q6
\I \ '" '\ ~,

1\ " ,
Q4 \ -

i \ \

+\-
\ i
I I
I
I
0,2 I
I !
i
I \ \ i
I
I

0,1
I +---J:-\ !\
i ,
q08
1 I I'
,
0,06 t-

I
----

i
Opt.
I
a-60 0
I, \ I
I I
I

0,0
a-5 0 a - 50

-- f. 1 Wlkg
L =
I
a_90° -.-l
~ .
I
i

I
- - - f. - 100 Ikg W
I
I I I
25 50 75 100 125 150 175 200
Rp l)Jm

J<'igure 6. Probability of stability of particie/bubble aggregates against external stress forces. RB =

0.5 mm, 0 = 70 mN/m, Qp = 2.5 g/cm 3 .

6. Schulze, H.J. (1981) Experimentelle Untersuchung der hydrodynamischen Wechselwirkung

von Partikeln mit einer Gasblase, Aufbereitungstechnik, Wiesbaden, 22, 253-264.
7. Schulze, H.l. (1989) Hydrodynamics of bubble-mineral particle collision, Min. Process.
Extract. Metall. Rev.,2, 43-76.
8. Schulze, H.J. (1992) Flotation as a heterocoagulation process, in B. Dobias (ed.) Coagulation
and Flocculation, Marcel Dekker (in press).
9. Scheludko, A. (1967) Thin liquidfilms, Adv. Colloid Interf. Sci. 1, 391-437.
10. Schulze, H.J. (1975) Einige Untersuchungen aber das Zerrei{3en danner Flassigkeitsjilme auf
Feststoffoberfliichen, ColI. Polymer Sci. Darmstadt 253,730-732.
11. Schulze, H.l., and Bizer, 1.0. (1987) Stability of thin liquid films on Langmuir-Blodgett
layers on silica surfaces, ColI. & Surf. 21, 209-224.
12. Radoev, B., and Scheludko, A., (1983) Critical thickness of thin liquid films, 1. Colloid
Interf. Sci. 95, 254-265.
13. Frumkin, A.N.D. (1938) 0 lavlieniach Smatschivanija i prilipanija puzirkov, Shumal Fiz.
Khimija 12,337-348.
 179

70- 5 moll I DCA

700
70
t- 75
E
'- :c:
50 E 60
"
lQ

25
50

2 2 4 6 8 pH 12
30
200 after 8,3 s

E 760 20
c w'"
Q)
" 720
}; 80 70
40 + 10- 2molll
KCI
L-~2~L4~6~~8~1~0~'2
60 120

50 100
~ 40
.... • "" 80
Q)
~ 30 60
20 40

10 20

2 4 6 8 70 12 246810/2
300

A.r." 2 roo

0\
Sm,

100

2 4 6 8 /0 /2 2 4 6 8 10 /2

Figure 7. Correlation between flotation recovery (r) and critical film thickness (heri'), film drainage
(induction time, tJ, expansion velocity of Lp.c. (vTPc), surface tension (0), dynamic receding (8 r) and
advancing angle (8,), attachment forces (F), maximum particle size floatable (dpm ) and centrifugal
acceleration (a e). Material: spherical quartz particles (aqueous dodecylamine hydrochloride solution,
lxlO- 5 M).
180

100
60 r--
SO
."
~
60 ..... 40 r-
-
0

~ 40 I,,"
20
20
2

qs
-
10 r-- r--
~ Q6
.....
t 0.4
r-- .e 5
... 0,2 .....
2 '-' 2

Figure 8. Effect of reagent regime on flotation recovery (r), maximum particle size floatable (dpm ),
centrifugal acceleration (ae) and induction time (ti). Material: broken potash (KCl) particles in saturated salt
solution (1 : Rofamin T 152 glt, octanodiol 14 glt, amysed 232 glt; 2 : Rofamin D 63.5 glt, Octanodiol
19 glt, Petroleum 29 glt).

14. Geidel, Th. (1985) Hajtwahrscheinlichkeit von Mineralkornen an LUftblasen, Autbereitungs-

technik, Wiesbaden, 26, 287-294.
15. Rasernann, W. (1988) On attachment probability of bubble/particle contact, Int. J. Miner.
Process. 24, 247-267.
16. Schulze, H.J., and Radoev, B., (1989) Investigations of the collision process between
particles and bubbles, Int. J. Miner. Process. n,
263-278.
17. Evans, L.F. (1954) Bubble-mineral attachment in flotation, Ind. Eng. Chern. 46, 2420-2424.
18. Sutherland, K.L. (1948) Physical chemistry offlotation, J. Phys. Chern. 52, 394-425.
19. Dobby, G.S., and Finch, J.A. (1986) A model of particle sliding time for flotation size
bubbles, J. Colloid Interf. Sci. 109, 493-498.
20. Mileva, E. (1990) Solid particle in the boundary layer of rising bubble, Coll. Polymer Sci.
Darmstadt 2..6B., 375-383.
21. Saffrnan, P.G. (1965) The lift on a small sphere in a slow shear flow, J. Fluid Mech.
22, 385-400.
22. Schulze, H.J. (1989) Determination of adhesive strength of particles within the liquid/gas
interface, J. Colloid Interf. Sci. 128, 57-65.
23. Yoon, R.H., and Luttrell, G .H. (1989) The effect of bubble size on fine particle flotation,
Min. Process. & Extract. Metall. Rev. 5.,101-122.
2. Chemical Technology Aspects
2.1 . FLOTATION KINETICS

C. EK
University of Liege
Rue Armand Slevart, 2
B-4000 Liege 1
Belgium

ABSTRACf. Rotation kinetics is briefly introduced and its history is reviewed. Main theoretical
approaches are discussed, and the kinetic models are presented in detail. The application of
flotation kinetics to modeling and simulation of the circuits is shortly surveyed, then some
industrial results are used to show how the models fit to experimental results.

1. Introduction

The kinetic study of the flotation process includes the determination of all the factors that influence
the rate of concentrate production [1-3J This rate can be defined according to various meanings, but
in the operations carried out on ores, it is generally measured as absolute recovery (unity fraction)
divided by the time (kinetic constant units: s'! or min'!).
When the physico-chemical phenomena occuring in a flotation pulp are examined, it is possible
to define a number of stages taking place from the cell feed to the froth overflow.
The flotation cell has to perform several functions:
1) homogeneous suspension of the pulp;
2) air introduction as small dispersed bubbles;
3) turbulent agitation to bring bubbles and minerals into contact;
4) transfer of the pulp from cell to cell;
5) froth formation in a quiet zone,
Beside construction characteristics, such as height/surface ratio, impeller diameter/cell length
ratio, etc. [41, it results from points 1 to 5 that two important cell factors have to be controlled:
a) the peripheral speed of the impeller (or rotation speed);
b) the air flow, in absolute or relative units.
As far as the feed pulp is concerned, it can be admitted that the transfer probability of a mineral
particle from the pulp to the froth is a product of probabilities, However, it is today not yet possible
to ascribe quantitative values to most of these probabilities and the present results are not usable to
control an industrial flotation process,
In view of these shortcomings, it has been proposed to consider the flotation process as a
chemical reaction, but the collision between atoms and molecules gives way here to the collision
between mineral particles and air bubbles. This approach has given acceptable results leading
eventually to the computer control of flotation plants [5,61.

183
p, Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 183-210.
© 1992 Kluwer Academic Publishers.
184

2. Microkinetic rate steps

2.1. GENERAL S1EPS

The general steps that are taking place when a pulp is floated in a cell can tentatively be distributed
as follows:
1) probability of adhesion of collector to detennined mineral particles;
2) probability of these particles to enter the turbulent zone of the cell;
3) probability of collision between mineral particles and air bubbles in this zone;
4) probability of aggregation of particles and bubbles;
5) probability of transfer of the aggregate to the froth;
6) probability of hold-up of the aggregate in the froth.
Some infonnation on these different steps [7,8] will briefly be given in the following sections.

2.2. PROBABILITY OF COLLECTOR ADHESION

This very important step of the flotation process has not yet found a complete theory to explain thc
various research results that have been published a long time ago and are periodically reviewed in
the specialised literature.
Whatever the real mechanism can be, it is generally supposed, in the kinetic studies, that the
mineral to be floated is properly conditioned before flotation begins. It is thus considered that this
step is achieved when the pulp is introduced into the cell.
This assumption is really a simplification, namely in the industrial operations where numerous
ores are treated without separate conditioning and where the first collector addition is made in the
first flotation cell. In this case, it is implicitly agreed that collector adhesion is very rapid. The
examination of real operations shows, however, that the flotation results in the first cells of a bank
can be erratic due to the conditioning going on for a certain time.

2.3. PROBABILITY OF CROSSING THE AGIT A1ED ZONE

This probability is linked to the hydrodynamic characteristics of the cells.

Referring to the theory of chemical reactors, it has been confinned that a continuous flotation cell
is pcrfonning as a continuous stirred tank reactor (CSTR) [9,10].

2.4. PROBABILITY OF COLLISION AND AGGREGATION

These two probabilities arc dependcnt on complex physico-chcmical and hydrodynamic phenomena
that are difficult to isolate for specific study [11,12]
Collision phenomena have been partially explained by meteorological theories, for example the
capture of solid particles by spheres or thc dispersion of fogs.
Due to experimental problems when in rcal pulps, the thcory of particle-bubble adhesion is only
slowly improving; it requires a better knowledge of the physico-chemistry of the surfaces and the
related film layers. A very good Russian research is often taken as an example, but it covers only a
small range of particle sizes [13 J.
It is also to be noted that particle-bubble adhesion can be due to the nucleation and growth of a
gas bubble on the solid surface, what is generally called "gas precipitation": these phenomena
complicate the theoretical analysis.
 185

2.5. PROBABILITY OF TRANSFER TO THE FROTH AND HOLD-UP

When the aggregate is fonned, its density is such that it comes up to the pulp surface. During this
transfer, it is submitted to various forces tending to its destruction.
In addition, it has been observed that another probability should be introduced: transfer of the
solid back to the pulp, due to the bubble rupture and the drainage of the particle in interfacial fIlms,
or to an eroding action of the agitated pulp at the bottom of the froth [14].

2.6. CONCLUSION

In this brief review, it was intended to show that the comprehensive knowledge of the various
physico-chemical phenomena governing the transfer from the pulp to the froth of well conditioned
particles is still incomplete and unsatisfactory for practical purposes.
Many experimental results are needed to base a general fundamental theory, but there appears to
be complex relationships rendering this task very troublesome. A good approach seems to be the
design of experimental systems where some probabilities are maintained equal to unity while one or
two characteristics only are measured [15].

3. Historical survey

3.1. INITlAL OBSERVATIONS

It is generally agreed that the first paper concerned with flotation kinetics was published by Garcia-
Zuniga in Chile, in 1935 [16]. This scientist had already observed that recovery is an exponential
function of time.
Soon after, Beloglazov [17] and Schulunann [18,19] develop the conditions to be applied in the
laboratory for continuous studies at steady-state.
In 1948, Sutherland [20] perfonns a theoretical study of the particle-bubble adhesion and derives
a rather complicated fonnula that he applies to glass sphere flotation. This fonnula contains many
tenns that are very difficult to measure and its utili ty is low at present.
In fact, the interest for flotation kinetics has been revived by a paper of Arbiter in 1950 [21],
when he proposes a second-order equation to represent the results of Zuniga, Beloglazov and
Sutherland, as well as his own results obtained in laboratory batch tests and other data produced by
industrial cells.
The approach used by Arbiter is the chemical engineering one, where it is admitted that the
mineral quantity transferred to the froth is a direct function of its concentration when most
parameters are maintained constant (type of mineral, pulp density, pH, reagent addition, aeration,
etc.):

- dC/dt =k(C)D (1)

k is the rate constant and C is the mineral concentration. The exponent n is considered as the order
of a chemical reaction and its value is set at 2 by Arbiter. The original equation proposed by Arbiter
is written:

dR/dt =k(R* - R)2 (2)

186

where R is the recovery at time t and R* is the ultimate recovery for an infinite time. Integrating
with R = 0 when t = O. it becomes:

(R*f kt
R=I+R*kt (3)

or

tIR = (R*~2 k + tIR* (4)

When drawing tIR in terms of t. the slope permits the calculation of R * and of the ordinate. k. A
shortcoming of this function is its lack of sensibility when t is high. as R hardly varies and the
figure represents t in terms of t! Arbiter put himself on guard against this fact and advised to check
the first points of the graphs. measured when t is low.
In the discussion of this paper. Morris [22] refutes the second-order equation and proposes. with
graphical proof only. the first-order equation. finally written:

R*
lnR* _ R=kt (5)

The graphical proofs of Morris are very weak. as he checked only one series of Arbiter's results
and the second proof concerns a single test [23]. where collector was regularly added during froth
removal.
In 1952. Morris [24] publishes the results of discontinuous tests. including particle size analysis
of the concentrates. He uses the first -order equation. but to obtain linear graphs. he has to admit
that a fraction C* of the mineral to be floated does not float in the test conditions:

- dC/dt = k(C - CO) (6)

and
In(C o - CO)
kt = In(C - C*) (7)

The value of C* is obtained by trial and error and fixed by the linear aspect of the graph!
Hukki [25] criticizes this approach and proposes other display methods. and namely. when R* is
taken equal to I:

first-order: t = In(l - R)jk (8)

second-order: l/t = (kjR) - k (9)

and the graphical displays proving the models are:

t= f(ln(l - R» (10)
I/t = f(l/R) (11)

All those discussions are specious and absolutely useless. for a linear display is mainly dependent
on the scales of the graphs: it is obvious that with a long time scale and a short scale for R or C.
one must always obtain a straight line! However. Hukki is still in favour of the second-order
equation.
 187

In 1956, Debruyn and Modi [26] show, with continuous tests in steady-state, that for pure
quartz, the first-order equation is acceptable for sizes under 65 mesh and with low pulp density.
Shortly after that, Horst and Morris [27] move to a 1.5-order equation:

dRldt = k(R* - Rl5 (12)

and, by integration:

1 kt I
(R * _ R)O.5 = '2 + (R*)O.5 (13)

The experimental batch tests, including a back-feed of the solution entrained in the froth, were
carried out on a pyrrhotite-pyrite concentrate, and it is difficult to take into account the role of the
former conditioning and flotation! Anyway, here again, the graph does not prove that the 1.5-order
equation is better than the first-order onc.
From 1956, Huber-Panu has studied [28] a theoretical kinetic equation that he has checked by
batch tests [29,30]. This equation is very complex and can only be simplified at a rough guess.
However, the final equation still contains 2 factors to be determined from test results, and
nevertheless the agreement between actual and calculated results is not satisfactory.
So, in the beginning of the sixties, theoretical ideas are confused because the experimental
methods are unable to give unquestionable results. It seems essential to forget how to get linear
graphs by any means and better to try to rely upon physical phenomena [31].

3.2. SOME INTERPRETATIONS

3.2. 1. Meaning of C*. The term C' introduced by Morris [24] to obtain linear kinetic relationships
is shown by Bushell [32] to be the concentration of the floatable mineral remaining in the pwp after
an infinite time; the proof is given by laboratory tests and industrial results. The value of C* is
obtained graphically as the asymptote to the curve (C,t).
The disadvantage of this method is mainly due to the lack of precision of the graph when time is
not very long. Furthermore, results are often better represented by two straight lines than one, as
shown by Bushell's own drawings. Finally, when chemical conditioning of the mineral is constant
in tests where only physical factors are changing, one may not qualify as "non floatable" the
mineral fraction that has not floated due to a lack of time.
Anyway, many scientists have used Bushell's formula when selecting first-order kinetic
equations.

3.2.2. Research work at Leeds University. The research work carried out in the Department of
Mineral Technology has helped to clarify many ideas.
The laboratory method includes a continuous flotation at steady-state: continuous operations are
preconditioned pulp feeding in the cell, froth overflowing and tailings extraction.
At steady-state, simwtaneous collection of froth and sink permits the computing of the kinetic
constant. Using the first -order equation, it is possible to write:

- dC/dt = R =kC (14)

where C is the mineral concentration in the sink (or tailings), when the flotation cell behaviour
188

corresponds to a CSTR. The results of parallel tests with various C, drawn on a diagram (R,C),
have to give a straight line, the slope of which is a function of k: this is effectively the case, and
this line passes through the point (0,0) when the C· correction is applied (33).
It has also been shown that the kinetic constant for the froth drainage is proportional to the feed
concentration (34) and that, in a two-component system, the amount of free gangue transferred to
the froth is proportional to the amount of free gangue in the pulp [35].
Most of this work has been performed in coal flotation tests, but the conclusions, supported by a
rigorous experimental technique, are obviously valid for other ores.

3.2.3. Model based on mass trans/er. Among the valuable papers written for the book "Froth
Flotation", Arbiter and Harris [1] have surveyed the former work on flotation kinetics and have
shown that mass transfer has been incompletely addressed so far.
Consequently, these authors propose a more elaborate model based on the mass flows through
the cell and on the mass transfer from pulp to froth. They write differential equations describing
two cases:
a) flotation with continuous ore feeding and continuous concentrate and tailings removal, in the
initial conditions (semi-continuous flotation) and at steady-state (continuous flotation);
b) flotation with continuous concentrate removal, water addition and tailings remaining in the cell
(most usual laboratory test).
The first-order kinetic equation is used throughout.
In a supplementary paper, Harris and Rimmer [14] show how the use of a computer allows the
study, in fixed conditions, of the influence of the variations of some factors of the two-phase
differential equation. They are able to reproduce the behaviour of experimental systems described
earlier and to explain some deviations observed in more simple models.

3.2.4. Model with mineral activity coefficient. Jancarek has explained the results published earlier
by a model including the activity of the mineral. and not its concentration (36). Using the
differential equation

- dC/dt= kC rn (m = I) (15)

this author obtains. after simplification:

1 at
log (1 - R) = (1 + bt)P (16)

where a and b depend on the experimental conditions, and b and p are written:

(17)

2 -m
p=~ (18)

The value of p is read on a graph giving log (t/Iog 1/(1 - R» in terms of log t, and it is shown that p
is generally equal to 1. and so

m=3/2
 189

However, here one can observe that when t is high, R hardly varies and the graph represents log t
in terms of log t!
In fact, close examination of the graphs shows that in many cases, the first points fall into a
different straight line than the last ones, and the overall straight line is only a graphical compromise
without real value.
Russian authors [37] have used Jancarek's formula, but it can be agreed now that it can be
neglected in a discussion on flotation kinetics.

3.2.5. Model with variable kinetic constant. In 1963, Imaizumi and Inoue [38] introduced a new
idea on flotation kinetics, stating that the integral ofthe first-order equation, generally written:

should be

C = Co fF(k) e- kl dk (19)
o
where F(k) is the distribution function ofk, or "k spectrum". As

CICo = 1 - R,

the function F(k) can be graphically determined on the graphs (log (1 - R), t). The method consists
of a decomposition of the curves into tangents, beginning by the right lowest part of the curves.
The same remark can here be made as for the C* determination according to Bushell: the value is
not precise when the flotation time is short [32).
The Japanese authors consider the kinetic constant k as a floatability criterion for the minerals,
the main factor influencing the k variations being the particle size [38].
As far as industrial flotation is concerned, Imaizumi and Inoue represent the results obtained in a
n-cell bank by the formula:

1 - R = fEet) e- kl dt (20)
o
where

tn - J
E(t) = e-l!l' (21)
(n - 1)! t*n

is the distribution function of the retention time of the pulp (t * is the mean retention time in each
cell). One also can write:

(22)
190

or

R = I - (I + k )D (23)

Eqn. 23 shows that R can be represented in tenns of log k. which gives a curve like the Tromp
washability curve. By drawing such curves for various flotation circuits. it is possible to design the
circuit giving the best selectivity. It is namely shown that the roughing circuit with n infinite gives
the maximum selectivity; it can be observed that this circuit corresponds to batch flotation. so k
appears to be a real floatability criterion.
The japanese work has been the basis of Loveday's work [39,40). who researched a fonnula to
represent the distributions. For the retention time distribution. he admitted that the flotation cells are
acting like CSTR. and for the kinetic constant distribution. he chose a "gamma" function of the
following type:

f(k) = c ka e-bk (24)

that gives a rather good correlation with experimental results. It is also possible to compute an
average kinetic constant. written:

a+ I
kav=b+t (25)

a (a > -I) and b (b > 0) are characteristics of the initial distribution of the k constants. The
mathematical treatments pennitting to choose the distribution functions have thrown some light on
the study. but it remains to determine the influence of aeration. rotation speed. reagent
concentration. etc. on the k constant distribution.
Inoue and Imaizumi have given another paper [41) on the same subject. and they admitted that
the kinetic constant distribution follows a modified "gamma" function. which they discussed
thoroughly. However. as in their previous paper. the lack of detailed experimental results does not
pennit testing of other kinetic models.

4. Present approach of flotation kinetics

4.1. THEORETICAL STUDIES

It is not possible to discuss in details all the papers published on the theoretical approach of
flotation kinetics. and this review will be limited to three papers only. respectively due to Haynman
[42). Huber-Panu and coworkers [43) and Kapur and Mehrotra [44). The choice here is
questionable. but it is intended to simply pinpoint how flotation kinetics has been tackled during the
last decade. or so.

4.2. MODELOFHAYNMAN

This fundamental model is based on a fonner work of Mika and Fuerstenau [45). in which flotation
is considered as a multistage process taking place in the pulp and in the froth. The flotation of
homogeneous particles by means of narrow range bubbles is described with the help of physico-
 191

mathematical equations involving all the pulp and froth variables.

Haynman acknowledges that Mika and Fuerstenau have introduced a mathematical simulation
that takes into account the multiple process stages, as they worked out the concept of
"subprocesses" and showed this concept to be necessary to select a sufficient number of variables
when coping with the whole actual process.
This author has simplified his description of the process by restricting the subject to the flotation
of homogeneous particles by bubbles of a narrow size range. The factors chosen are supposed to
characterize the state of a small pulp volume and are selected as follows for the qualitative
description of the pulp:
a) concentration of "free" particles, depending on time and coordinates, and expressed as Cp(x,t);
b) concentration of air bubbles n(x,t,m p), where mp is the degree of mineralisation of the
bubbles, expressed as mp(x,t), mean number of particles attached to an air bubble;
c) movement of the pulp.
In view of describing the pulp state, the following differential equations are developed:
a) concentration of free particles in the pulp;
b) concentration of bubbles in the pulp;
c) concentration of particles attached to air bubbles.
In order to describe the froth, the author has choosen the following variables:
a) specific interface of the froth volume p(x,t);
b) mineralisation of the gas-liquid interface in the froth m«x,y);
c) concentration of free particles in the froth c«x,y);
d) water content w(x,t) of a unit volume of the froth layer.
The froth state is accordingly described by four differential equations:
a) specific surface of the froth layer;
b) water content of the froth;
c) degree of mineralisation of the solid-liquid interface in the froth;
d) concentration of free particles in the froth.
The resolution of these equations permits, for instance, the description of the macroscopic
recovery:

( f (cp(x,t)+n(x,t) mp(x,y)} dx + f (cr(x,t)+p(x,t) mr(x,t)} dx )

e(t) = 1 __.:t..vJLP_ _ _ _ _ _ _ _ _---.lVCLf_ _ _ _ _ _ _ __
fCp(x,O)dx
vp

It is obvious that solving this equation, like others in the same paper, requires overcoming some
mathematical difficulties and the availability of a large amount of information related to the various
subprocesses in the pulp and froth volumes.
After a number of simplifications, Haynman has finally arrived at a very simple equation:

de/dt = K(l - e) (26)

4.3. MODEL OF HUBER-PANU

Huber-Panu and his coauthors have drawn upon years of study of flotation kinetics (as ex amplified
by numerous papers) to publish a very interesting paper in 1976 [43].
These authors consider the factors which influence the flotation process to be divided into three
192

main groups:
a) factors detennined by the feed ore, the flotation reagents and the process stages preceding the
flotation operation;
b) factors detennined by the design of the flotation cell (mixing and aeration characteristics);
c) factors detennined by the pulp transport charactcristics through a continuous flotation bank.
To establish a general kinetic equation, the authors take into account a number of parameters that
can be simply described as follows:
a) the size distribution of the valuable mineral particles, generally considered to be made up of
several size fractions, each having its specific recovery;
b) the floatability parameter, which is the value of the mass fraction of the valuable mineral
particles which is floatable in a size fraction;
c) the floatability distribution for valuable mineral particles in the same size range;
d) the flow characteristics for continuous flotation:
- specific volumetric flow rate of pulp
- ratio of input and output volumetric flow rate of pulp in a cell of the bank
- deviation between the mean valuable mineral concentration in the pulp of a cell and the
same output concentration.
In order to develop a usable model for the flotation process, the following assumptions are
made:
a) the kinetics of flotation of floatable particles can be assumed to be proportional to the number
of air bubbles and to the concentnltion of the floatable particles;
b) in the valuable mineral particles, there are some which are non-floatable in the cell conditions;
c) in an industrial flotation bank, steady-state conditions prevail;
d) when the flotation cells are identical, mixing and aeration conditions are identical;
e) non-ideal mixing behaviour is similar in all the cells;
f) the volume flowrate ratio is the same in all the cells.
These conditions result in a general model equation for flotation recovery, that can be simplified
in particular cases well described in the literature and already discussed in part in this text, taking
into account the cell operating behaviour, the size distribution of the feed particles and the
floatability distribution of these particles:
a) for the batch flotation of a monodisperse material with a rectangular distribution of
floatabilities:

R = R*(l --lcf-)
I ·kt
(27)

b) for the continuous flotation of mineral particles characterised by a gamma distribution of

floatabilities:

(28)

c) for the batch flotation of a monodisperse material where all the particles have the same
floatability:

(29)
 193

d) for continuous flotation in a bank of n cells where mixing is perfect and the concentrate
volume negligible, assuming that t* is the mean retention time in a cell:

R = R*(! - (l - kt*Y") (30)

e) when the floatable particles of the same size have the same floatability, and the relation
between size and floatability is linear, the equation for batch flotation is

(31)

The authors also give some details on the experimental systems used to test the model equations, as
well as the results of comparative examination of the fitness of the eqns. 27 to 31 to the
experimental values.

4.4. MODEL OF KAPUR AND MEHROTRA

These authors have recently published an excellent paper [44) summing up a lengthy work on
flotation kinetics, as examplified by numerous articles in the specialised literature. This paper
contains also a considerable literature survey.
The approach used here is the classification of the kinetic models in a P-S matrix, where each
type of modcllies at the intersection of phase P (single or multiphase system) and feed species S (a
single class of species, or a discretely distributed, or continuously distributed, mass of particles).
A Single phase-single species model is supposed to follow the well-known first-order kinetic
model, possibly modified to take into account the direct entrainment of solids from pulp to froth or
the zero-time correction.
A single phase discretely distributed species model, where the feed is divided into some classes
of homogeneous species, and where the flotation does not proceed to completion, is represented by
the common first-order equation:

(32)

When each size fraction is divided into two classes with fast and slow floating properties, one
obtains:

(33)

where z is the slow floating fraction and kj > k2.

A single-phase-continuously distributed species model can take many forms according to the
type of distributed rate constant: the best known distributions are the gamma function and the
rectangular distribution.
When coping with multiphase systems, the complexity and sophistication of the single phase
systems are increasing; the rate parameters must be estimated from experimental data and the most
simple model is probably the most advisable. In their paper, Kapur and Mehrotra discuss the
different aspects of the kinetics of these systems, and suggest that significant progress is to be
expected later in that regard.
The authors also examine the structure of the flotation rate constants, lumped and apparent, as
194

well as the effects of feed and process variables.

Finally, they introduce the kinetic aspects in a study of the simulation of flotation circuits, taking
into account the feed specifications, the flotation model, the split factor and the circuit
configuration.

4.5. CONCLUSION

From this short survey of some papers on flotation kinetics, it appears that the theoretically
constructed models, whichever their scientific interest, cannot presently offer a satisfactory answer
to practical kinetic problems, as for example the introduction of this parameter into circuit
simulation.
On the other hand, the empirical models are probably acceptable at the present time, when their
use is restricted to a macroscopic description of the flotation process.
It will be shown later how the actual results, from laboratory studies as well as from too scarce
industrial tests, can be represented by these kinetic models.

S. Kinetic models

For several years, Klimpel, Aplan and their coworkers have produced many papers related to
flotation kinetics and its use in various aspects of the process.
The interest of these studies seems to lie in the large survey they have made of the flotation
models, and the testing of these models in empirical conditions on different ores.
It is not in the scope of this paper to make a comprehensive review of all the papers they have
published; however, it is possible to get a general overview of their work through the study of one
of these papers, published in 1985 [46]: the title is self-explanatory: 'Model discrimination in the
flotation of a porphyry copper ore'.
In spite of the fact that some duplication may occur, it seems to be advisable to quote largely
from this paper to get a full picture of the available kinetic models.

5.1. MODEL 1: CLASSICAL FIRST-ORDER MODEL

This simple model has already been formerly discussed, and has appeared lastly as eqns. 29 and
32: it describes the flotation of a monodisperse feed containing particles having a constant
floatability. It should also be noted that if a probability function is substituted for the exponential
rate term, the equation becomes analogous to the limiting case of the probability model proposed by
Kelsall [47].
As will be reported later, some authors propose to incorporate a correction term for the
identification of zero time.

5.2. MODEL 2: FIRST-ORDER MODEL WITH RECTANGULAR DISTRIBUTION OF

FLOATABILITIES

This model has been formerly discussed as eqn. 27: monodisperse feed with rectangular
distribution of floatabilities. It has also been derived by considering the mechanical rate of mass
removal from the froth [48]. It is believed that the rectangular distribution of f10atabilities gives this
model added flexibility and, therefore, it should be a better form of the first-order one.
 195

5.3. MODEL 3: ARST-ORDER TWO-STAGE MODEL

By considering the flotation system as being composed of discrete pulp and froth volumes, Meyer
and Klimpel [48] proposed a physical model incorporating rate terms describing the mass transfer
of a component from the pulp to the froth and from the froth to the concentrate. This concept can be
represented as:

where A = concentration of component in the pulp, B = concentration of component in the froth

and C =recovered component.
By assuming that the rate of drainage from the froth is minimal, the k3 value is much greater than
the k-3 value and this latter value can be dropped out. The mathematical form of the model is
written:

(34)

This equation is a three-parameter model describing a first-order two-stage process, and it is similar
to the model derived by Harris and Rimmer [14].
It can be observed that eqn. 34, when k3 is large, can be written:

i.e. the equation of model 1.

5.4. MODEL 4: ARST-ORDERREVERSIBLE MODEL

The reversible model describes the transfer of a component from the pulp to the froth (kt) and the
subsequent drainage of a portion of this component from the froth (k-4). The mathematical form of
this model is written:

(35)

It is obvious that this equation is reduced to eqn. 29 when k4 is negligible.

5.5. MODELS: RJLLY MIXED REACTOR MODEL

Imaizumi and Inoue [38] have shown that flotation results can be represented by an expression
analogous to the equation describing the time concentration for a series of fully mixed reactors.
Rearranging this equation into recovery terms and taking the number of reactors as one, the
following model equation is obtained:
196

R=R*(I- ---:-t- (36)

+ kS

This model describes the first-order time-recovery equation of a flotation component from a
monodisperse feed with an exponential distribution of floatabilities: this feature gives an added
flexibility over the classical first-order model.

5.6. MODEL 6: GAS/SOLID ADSORPTION MODEL

Meyer and Klimpel [48] described a general gas/solid adsorption kinetic function as follows:

(37)

It must be noted that this model is of a very different form than any of the previously described
models: it does not incorporate an ultimate recovery parameter. Furthermore, it can be shown that
m is generally greater than I, and this fact results in R increasing to values greater than I for long
times, and decreasing thereafter.

5.7. MODEL 7: IMPROVED GAS/SOLID ADSORPTION MODEL

Since model 6 (eqn. 37) was of the same form as the Langmuir adsorption isotherm, when
parameter m = 1, it was improved to get the following equation:

(38)

This model is similar to the model proposed by Bull [49], and it can be easily checked by a simple
mathematical transformation.
It is also observed that this model is similar to model 5 when k7 is changed to Ilks.

5.8. MODEL 8: SECOND-ORDER MODEL

Already proposed by Arbiter [21] in 1951, this model is a two-parameter expression describing the
flotation of a monodisperse feed with particles having a constant floatability:

_ *2 kgt (39)
R- R 1 + R*k8 t

It must be noted here that the fit of calculated time-recovery profile and the optimal R * value are
found to be identical to that of models 5 and 7, but this form is not as good statistically as those
determined by the first-order forms. Furthermore, the influence of the square of R * can be too
heavy when R* is becoming large.
 197

5.9. MODEL 9: SECOND-ORDER WITH RECTANGULAR DISTRIBUTION OF

FLOATABILITIES

This model has been proposed by considering the second-order general mass balance of a flotation
component having a rectangular distribution of floatabilities for each size:

(40)

5.lD. MODEL 10: THREE-PARAMETER, FAST AND SLOW FLOATING COMPONENTS

Kelsall [47], recognizing that the particles to be floated have a distribution of floatabilities, has
proposed this model suggesting the treatment of fast (kr) and slow (ks) floating components:

(41)

not very different from eqn. 33. It should be noted that the ultimate recovery term is absent.
Generally speaking, this type of model does fit the experimental results better than other models,
due to the supplementary parameter. However, the authors observe, as for other models with more
than two parameters, that the confidence intervals are systematically large, so that differences
between tests become completely masked. Therefore, they do not encourage the use of such
models.

5.11. MODEL 11 : FOUR-PARAMETER, FAST AND SLOW FLOATING COMPONENTS

This model brings a slight improvement over model 10, by the introduction of the ultimate
recovery:

(42)

5.12. MODEL 12: THREE-PARAMETER MODEL WITH PARTICLE FLOATABILITIES

PROPORTIONAL TO SIZE DISTRIBUTION

This model has been proposed by Huber-Panu et al. [43] and previously discussed here above
(eqn. 31):

e-k121t _ e-k12ut
R - R*(l ) (43)
- - k12u - kl21

S.13. MODEL 13: THREE-PARAMETER GAMMA DISTRInUTION

This model, already largely discussed, is here written:

(44)

as shown in eqn. 28.

198

The comparison of the results of the experimental tests carried out on a porphyry copper ore has
shown that:
a) the fit of the model to the results is:
- excellent for models 2 (best overall), 5 (except when R* tends to 1),7 (id. 5), 8 (id. 5), 11,
12 and 13;
- good for models 6 (except when R*« 0.8) and 9 (id. 5);
- fair for models 1,3,4 and 10;
b) confidence limits are:
- excellent for models 2, 5, and 7;
- good for models 6, 8 and 12;
- fair for models 1,3,9, 10 and 13;
- poor for models 4 and II.

5.14. MODELS WITH ZERO-TIME ADJUSTEMENT

In a large series of papers, Agar and coworkers from INCO Research Laboratories have used
flotation kinetics to a great extent (at the industrial point of view) in order to better explain and
control various flotation operations.
Agar has definitely adopted the first-order kinetic equation according to model 1 of Klimpel, but
he considered [50) that this equation had to be modified to include a zero-time factor:

R = R *(l - e-k(t+t*») (45)

The time correction factor is added because of the difficulty in physically assigning time zero [51).
This is normally set when the air flow starts, but it takes a finite time for a stable froth to
accumulate and to overflow, so a negative correction time should be required: this is often observed
with the least floatable component. On the other hand, the most hydrophobic solids may have some
air attached during the conditioning period, which makes them float faster than they normally
would: this can cause a positive time correction. When total time is of the order of several hundreds
seconds, it has been observed that the time correction is comprised between 0 and lD s, and this
correction can be neglected in many tests.

6. Application of flotation kinetics

Many scientists and engineers in the field have been in the past and are now using flotation kinetics
to study in depth and with more details the influence of many parameters on this process.
Here again, it is not in the scope of the paper to cover the whole range of applications, but
simply indicate some trends and results obtained so far.

6.1. INFLUENCE OF PARTICLE SIZE

From scratch, this is probably the most studied factor and it has been already discussed in 1942 by
Gaudin and coworkers [19). These scientists have tried to relate flotation rate with particle size and
coefficient of mineralization: they obtained good results, but with an amount of collector largely
higher than usual.
In 1952, Morris [24) concluded from his test results that maximum flotation rate was obtained in
 199

an intelTIlediate size range, and that the rate was decreasing on either side of this range; on the other
hand, it was difficult to detelTIline whether the flotation rate constant k varied directly with particle
size or with its logarithm on the fine side of the maximum.
In a study of the effects of particle size and feed rate on the flotation rate distribution in a
continuous cell [52], Kapur and Mehrotra have shown that the apparent rate distribution is
independent of the slurry flow rate; the size distributions reduce to a self-similar fOITIl when the
apparent rate constant is rescaled by a median rate constant. The scale parameter quantifies the role
of particle size in the structure of the rate constant and exhibits a characteristic maximum at an
intelTIlediate size.
In a paper reporting on some factors influencing kinetics in sulphide flotation, Klimpel and
Hansen [53] remark that there is a tendency for both the overall R* and k values to increase with a
finer grind. Loss of valuable metal is greatest in the very fine and coarse sizes, an observation
already made by many authors. It is also noted that there is some discrepancy between the values of
R* and k for a mixture and a composite of these individual values for different size fractions.

6.2. INFLUENCE OF AIR FLOW RATE

In a series of papers [54-57], Laplante and coauthors have studied, in a specially designed batch
cell, the influence of air flow rate and frother concentration on bubble size distribution and flotation
rate; it is observed that the flotation rate constant increases to a maximum and then decreases as air
flow rate is increased. When investigating the flotation of pure galena, it was found that the overall
rate constant increased with increasing air flow rate and with decreasing froth thickness; the froth
transport constant (that concerns the transfer over the cell lip) increased similarly. When floating
galena and slightly conditioned chalcopyrite, the selectivity of the transfer from the slurry to the
froth was found to be the highest at intelTIlediate air flow rates, and lowest at very low or very high
ones.
The last paper [57] is a very thorough investigation of the effects of air flow rate and froth
thickness on the major transfer mechanisms in batch and continuous flotations; the test feeds used
were a deslimed sphalerite concentrate and pure silica, alone or mixed.

6.3. INFLUENCE OF BUBBLE SIZE

Ahmed and Jameson [58] have carried out experiments in a cell in which bubbles of known size
could be generated independently of the turbulence levels, which could be controlled by varying the
impeller speed. The mean bubble size ranged from 75 to 655 Ilm. Polystyrene latex particles,
quartz and zircon, giving a wide range of densities, were floated as fine particles, less than 50 !lID
diameter. It was found that the flotation rate was very strongly affected by the bubble size, there
being an increase of up to one hundred-fold when the bubble size was reduced from 655 to 75 !lID.
For light particles, high agitation increases the flotation rate, and for particles of high density, it
seems that optimum results are achieved with fine bubbles at impeller speed just sufficient to
provide full suspension of the particles.
Considering the combined influence of bubble size and turbulence, Spears and Jordan [59] have
shown that when floating fine galena (-40 Ilm), particle size and turbulcnce are major factors.
Within the size range studied (75-142 !lID), bubble size was only a minor factor. When particle size
increases, the efficiency of collision and the number of collisions increase, leading to an increase in
flotation rate. Turbulence increases the velocity of bubbles and particles relative to the fluid. This in
tum leads to an increase in the number of collisions, and consequently the probabilities of collision
200

and flotation increase. Turbulence and bubble size are inversely related through dissipation energy.
A hundred-fold increase in dissipation energy causes only a two-fold decrease in bubble size.
Thus, the effect of turbulence on the probability of collision was much greater than the effect of
bubble size in the size range studied.

6.4. REAGENT ADDmON

This factor has been largely investigated, probably because it is much easier to control than
physical factors like air flow rate, feed rate, bubble size or turbulence.
The group of Dow Chemical Company, headed by Klimpel, has been very active in that field. In
a long series of papers [46,53,60-62]. these scientists have covered a large number of conditions,
and it can be suggested that the important contribution due to these people is essentially based on
the application of kinetic models selected after very practical studies on real ores.
It is not possible to enter here into many details, but it must be stressed that this group has tried
to select reagents. and namely frothers (that are often forgotten in this type of investigation), using
he R* and k values simultaneously ltO conclude from the test work. To quote one of their
conclusions: "The change in the parameters with reagent changes can help explaining. from a
physical point of view. why different conditions of cell operation give different results. The
separation of equilibrium recovery from the rate of achieving such recovery has important
implications when relating batch laboratory results to continuous plant-scale operation".
At INCO, Agar and his group have used flotation kinetic results extensively to study various
aspects of the flotation process [50]. Using the modified first-order rate eqn. 46, they claim that the
kinetic parameters R· and k provide some insight into the mechanism by which separation is altered
by reagent additions.
The kinetic analysis permits identification of the optimum separation conditions. provides a
sound technical procedure to calculate the optimum separation time and, finally, identifies the
optimum separation without having to do experiments at every conceivable time.
Because the optimum separation time is dependent on the reagent additions. the use of the rate
parameters for evaluating the influence of reagents is a more powerful technique than the
conventional batch test procedure.
In another paper [63], a kinetic study was used to select a satisfactory rock depressant in the
flotation of a copper-nickel ore. Locked cycle tests carried out at the bench level were used to
confirm the results simulated from batch data. There was good agreement between the predicted
and the experimental locked cycle test results.
Marin and Molina have studied the collecting properties of xanthoformiate type reagents for the
flotation of copper sulphide ores [64]. Results of kinetic tests were compared to those obtained by
using diethylxanthoformiate. They determined the R* and k values from Model I equation for
sulphide and gangue flotation and could calculate the optimum separation time and separation
efficiency. For instance, they could show that methyl-isopropylxanthoformiate is a faster collector
than diethylxanthoformiate. while obtaining the same selectivity.
The use of kinetic analysis and separation efficiency factor has been recently demonstrated for
the comparison of frother performance by Hosten and Tezcan [65]. The tested frothers were a
polypropylene glycol. pine oil and MIBC; the rate equation chosen was Model 2. From the results,
it appears that the performance of the frothers tested is rate controlled. However, it must be
stressed that many authors suggest that laboratory tests are not very convenient for frother
evaluation.
 201

6.5. PREDICTION OF GRADE-RECOVERY CURVES

Ball and coauthors [66] have extended a two-phase, distributed parameter model to obtain grade-
recovery curves and relate these to the initial feed distribution. They present a simple expression for
the concentrate grade in terms of the zeroth and first moments of the concentrate distribution; an
example is given for a binary ore containing two pure minerals, each having non distributed rate
constants.
For extending the model to the distributed rate constant case, a cubic polynomial was used to
simulate the feed distribution. It can be shown that grade-recovery curves are dependent on the
initial feed distribution. In spite of the interest of such a study, it appears that there was no follow-
up later on.

7. Flotation kinetics and modeling

It is obvious that flotation circuit modeling is based to a great extent on flotation kinetics, as has
been shown years ago by pioneers in the field [67-69].
Today, there are a number of examples of groups working on this problem, and results arc
becoming most interesting and usable, even when the solution is only partial.
Here again, the INCO group, headed by Agar, has provided practical papers showing how to
use kinetic results. One of these papers [70] reports a technique by which the design of a flotation
circuit may be optimised: three interrelated criteria are defined and the application of these criteria to
the selection of the optimum residence time for each stage of a flotation circuit is demonstrated. The
criteria are:
a) add no material to the concentrate that is lower in grade than the feed to that separation stage;
b) maximize the difference of recovery between the desired mineral and the gangue;
c) maximize the separation efficiency.
Three independent systems were examined experimentally and it was shown that application of
any of the criteria yielded similar results.
It was demonstrated how the application of rate measurements (criterion 2) could lead to the
selection of a set of optimised times for several stages in a flotation circuit and that these could in
tum be used to simulate the results of a locked cycle test. The predictions were checked against the
experimental results from a locked cycle test, and there was substantial agreement between the
predicted and observed results.
In a more recent paper, Agar [5] has shown how to use this technique for an ore containing
mainly pentlandite, chalcopyrite, pyrrhotite and rock, and feeding a circuit comprising rougher,
cleaner and cleaning scavenger. The parameters are the respective rate constants (according to
model I with zero-time correction) and the split factors (determined as shown in reference [71]). It
is concluded that a satisfactory agreement is obtained between the predicted continuous results, the
locked cycle test results and the continuous plant results. Once again, the zero-time correction is
negligible.
Kalapudas [6] has studied a method to predict flotation time and cell volume for full-scale plant
on the basis of laboratory batch data. The scale-up factor is defined as a ratio of continuous full-
scale and laboratory batch retention times. The value of the scale-up factor was determined
experimentally for different sulphide ores: the sampling in industrial flotation circuits was
performed and simultaneously the feed pulps of these circuits were floated in the laboratory. The
experimentally determined scale-up factors can be used for computing the required retention times
202

and cell volumes for new circuits when the lab test have been carried out. The criteria used to select
retention times of single flotation stages are similar to those of Agar.
Barbery and coworkers have studied the scale-up of flotation circuits:
a) they assess a methodology for the calibration of parameters describing the kinetic behaviour of
minerals in flotation, using a conventional laboratory cell; the kinetic model adopted
throughout is the model with two classes of fast and slow floating components (eqn. 34); the
methodology is applied to the case of a continuous pilot plant testing a copper-zinc are from
Chessy. The parameters measured in the laboratory and in the pilot plant are compared; scale-
up factors are given for: equilibrium recovery, first-order rate constants for hydrophobic
minerals and entrainment coefficients for hydrophilic minerals, in rougher, scavenger and
cleaner sections of chalcopyrite and sphalerite flotation; it is concluded that the main
parameters for scale-up are the rate constants (for a volume scale-up ratio of 30, the decrease
in rate constants is 4.15) [72];
b) they have performed the same type of sampling on a rougher-scavenger bank, fitted with 1 m3
cells, in the copper circuit in Mattagami; they also have developed a programme (called
CALFLOT) to perform the complete design of the entire circuit and to simulate the scale-up
when the flotation bank is well known [73].
Vedrine and coworkers [74] have used, in the same kind of approach, a kinetic model according
to Model 7 (eqn. 38), summed up for four different size fractions; the application was pan of the
design of a general simulator for the flotation of the ponuguese Moinho complex sulphide are.
The simulation and optimisation of flotation circuits have also been discussed in details by Kapur
and Mehrotra [44].

8. Industrial results

8.1. INTRODUCTION

It is easily acknowledged that a great number of laboratory batch flotation results are available, and
although many scientists have not published detailed results (so it is impossible to check the fit with
the 13 different Klimpel models), they permit statistical evaluation of the practical application of
these models.
At the University of Liege, like in other places, a programme has been designed to evaluate the
Klimpel models, restricting these to the following: 1 - 2 - 5 to 10 - 13. All in all, it can be claimed
that some of these models are never giving the best fit and can be dropped out of the list: for
instance, model 7 is always slightly worse than model 5, model 8 is always worse than model 7
and modelS is comparable to model 6; models 9 - 10 and 13 sometimes give the best fit, but the
overall best models are I and 2.
However, it seems to be most imponant to know how flotation banks behave in practice, and it
will be reponed hereunder on available results from industrial sources.

8.2. SOME INDUS1RIAL RESULTS

Arbiter [21] reproduced the results of Garcia-Zuniga [16]; when testing the models, it appears that
the best fit is obtained by model 1 for four out of five ores, and by model 6 for the oxide copper.
In the same paper, two sets of results out of three, for continuous flotation, are best fitted by
modell, except the molybdenite (model 9).
 203

In 1962, Lewis and Morris published an excellent analytical work on the flotation banks in the
London mill, where sulphide minerals are floated from a complex sulphide ore [75]. The results
have been the subject of many subsequent comments, but here again, the different models can be
tested for the flotation banks giving a sufficient number of values:

Best fit : model

Bulk flotation copper 13
zinc 13
sulphur 13
Copper rougher copper (very bad fit) 6 corr. : 1 *
zinc (very bad fit) 6 corr. : 1 *
insol. 2
Copper cleaner copper 6
insol. 10
Copper-talc copper (very bad fit) 1
insol. (bad fit) I
Zinc rouJ!her zinc (very bad fit) 6
Zinc scavenger zinc (bad fit) I
Zinc cleaner sulohur (bad fit) I
* corr. : graphically corrected for zero-time
What can be concluded from this study is that flotation of 13 species in 7 different operations
kinetically proceed according to 5 models, with some apparent relationship between model and
flotation stage.
Ek has proposed an evaluation of published results to date in a review paper [76], and other
papers containing detailed results from industrial flotation bank sampling are listed in the references
[10,49,74,77,78].
The fact that the two unique results concerning pneumatic flotation cells [9,79] are both in
agreement with model I must also be noted.
Comparisons between pilot plant and industrial plant results are proposed by several authors,
like Arbiter and coworkers [4] and Davis [80]; they show that there is no direct relationship
between the models fitting these results in both cases.

8.3. EXAMPLE OF FLOTATION KINETICS CALCULATION

A pneumatic flotation cell, having a length of 62 ft and treating a chalcopyrite ore, has been
thoroughly sampled; the results are shown in Table I. In this table, the column "Weight %" is
obtained from the copper balance in the different stages in the cell, and the column "Cu recovery
%" is obtained from a general balance of copper in the whole cell.
These results were studied on a computer, testing 9 out of the 13 models proposed by Dowling
et al. [46]; Table 2 presents the results of the calculations and the error associated with each model.
It appears that Modell fits these results better than the other ones; it has also the advantage of
being the simplest one.
The experimental data, with the best-fit line, are plotted in Figure I. It must be stressed that the
units of the kinetic constant are ft'l, but it is easy to convert it into min-I, when the total flotation
time is known.
 204

TAB LE 1. Expenmental flotatlOn results f rom a pneumatIc cell treating chalcopyrite.

Pulp Concentrate from the interval
Cu% Cu% Weight % Cu recovery %
Feed 0.808 100.0 100.0
28.4 1.619 56.909
After 11 ft 0.354
23.0 0.724 20.610
After 21 ft 0.186
15.2 0.382 7.186
After 31 ft 0.127
10.8 0.209 2.793
After41 ft 0.104
7.6 0.078 0.734
After 51 ft 0.098
6.7 0.191 1.584
Tailings 0.085 96.797 10.184

Recovery
1.0 r----'---------------,

40 50 60
Time

Figure. Fitting of Model I to the experimental data of Table I; flotation time is expressed as ft.

9. Conclusion

In this paper. it has been shown that flotation kinetics is presently approached by the application of
various models to kinetic results obtained in the laboratory. in pilot plants or in industrial
operations. Successful use of the kinetic parameters is demonstrated by the modeling and
simulation of flotation circuits.
As far as the microscopic approach of this problem is concerned. progress is continuing at a
good pace. but much effort is still needed to obtain a complete satisfactory picture of the various
processes involved.
 205

TABLE 2. Results of fitting models to the experimental data.

Model No. Enur

1 R=R*[I- exp(-kjt)], R* = 0.8973, k j = 0.0925 0.5584 %

2 R=R*[I-{1- eX~~~2t) l], R* = 0.9928, k2 = 0.1964 1.7891 %

5 1 3.5335 %
R=R*[I- - - t1, R* = 0.9999, ks = 7.5215
1 +-
ks

6 k6t 2.7169 %
R = (l+kt;t)ffi' kt; = 0.1325, m = 0.9872

7 k7t 3.5335 %
R = R* 1 + k7 t ,R* = 0.9999, k7 = 0.1330

8 R=R*2 1 + ~~k8t ,R* = 0.9999, kg = 0.1330 3.5335 %

9 R = R* [1 - In(l~ k9t)l, R* = 0.9999, k9 = 0.4619 4.8017 %

10 R = (1-<p)(l-exp(-klOft)) + <p (1 - exp(-klOst)), 0.6722 %

<p = 0.1142, klOf = 0.0944, klOs = 0.0019

13
R=R*[1 -~], R*= 0.9368, ~= 27.5178, p= 2.9287 1.7960 %

10. References

1. Arbiter, N., and Harris, C.C. (1962) Flotation kinetics, in D.W. Fuerstenau (ed.) Froth
Flotation, AlME, New York, pp. 215-246.
2. Ek, C. (1968) Contribution a !'etude cinttique de laflottation, Collection des Publications de
la Faculte des Sciences Appliquees de l'Universite de Liege Q. 3-61.
206

3. Weiss. N.L. (1985) SME Mineral Processing Handbook. SME-AIME. New York.
4. Arbiter. N.• Harris. C.C.• and Yap. RF. (1968) A hydrodynamic approach toflotation scale-
up. Proceedings VIIIth Intern. Miner. Process. Congress (Leningrad). paper D-19.
5. Agar. G.E. (1985) The optimization offlotation circuit design from laboratory rate data.
Proceedings XVth Intern. Miner. Process. Congress (Cannes). 2. pp. l00-111.
6. Kalapudas. R (1985) A study of scaling-up of laboratory batch flotation data to industrial size
flotation. ibid. £.. pp. 112-121.
7. Jameson. G.J .• Nam. S .• and Moo-Young. M. (1977) Physical factors affecting recovery
rates in flotation. Miner. Sci. Engng. 2.. 103-118.
8. Schulze. H.J. (1984) Physico-chemical Elementary Processes in Flotation. Elsevier.
Amsterdam; see also Schulze. in this volume.
9. Niemi. A. (1966) On the dynamics of a pneumatic flotation cell. Acta Polytechn. Scan..
Chern. incl. Metall.Series 49.
10. Niemi. A. (1966) A study of dynamic and control properties of industrial flotation processes.
ibid. 18..
II. Rao. S.R. (1974) Surfaceforces inflotation, Miner. Sci. Engng. Q, 45-53.
12. Laskowski. J. (1974) Particle-bubble attachment in flotation. ibid. Q.223-235.
13. Derjaguin, B.V .. and Dukhin, S.S. (1961) Theory of flotation of small and medium-size
particles, Bull. IMM 651, 221-246.
14. Harris, C.C., and Rimmer. H.W. (1966) Study of a two-phase model of the flotation
process, Trans. IMM Series C, 153-162
15. Tomlinson, H.S., and Fleming, M.G. (1963) Etudes cinhiques de flottation, Proceedings
VIth Intern. Miner. Process. Congress (Cannes), 677-691.
16. Garcia-Zuniga, H. (1935) The efficiency obtained by flotation is an exponential function of
time, Boletin Minero de la Sociedad National de Minero 47, Santiago, 83-86.
17. Beloglazov, K.F. (1939) Kinetics offlotation process, Tsvetnye Metally 2. 70-76.
18. Schuhmann, R (1942) Flotation kinetics I - Methods for steady-state study of flotation
problems, J. Phys. Chern. :16., 891-902.
19. Gaudin, A.M., Schumann, R, and Schlechten, A.W. (1942) Flotation kinetics 1/ - The effect
of size on the behaviour of galena particles, ibid. 46.902-910.
20. Sutherland. K.L. (1948) Physical chemistry of flotation: Part 1/: Kinetics of the flotation
process. ibid. 51,394-424.
21. Arbiter. N. (1951) Flotation rates andflotation efficiency, Min. Engng . .3... 791-796.
22. Morris, T.M. (1951) Discussion of ref. [21], ibid. l, 991-992.
23. Ludt, RW., and Dewitt, C.C. (1949) The flotation of copper silicate from silica, Trans.
AIME (Mining) 184.49-51.
24. Morris, T.M. (1952) Measurement and evaluation of the rate offlotation as a function of
particle size. Min. Engng. ;1, 794-798.
25. Hukki, RT. (1953) Discussion of ref. [24], ibid.,1, 1122-1124.
26. Debruyn, P.L., and Modi. H.J. (1956) Particle size and flotation rate of quartz, ibid . .8., 415-
419.
27. Horst, W.R., and Morris, T.M. (1956) Can flotation rates be improved ?, Engng. Min. J.
157 (October), 81-83.
28. Huber-Panu, H. (1956) Beitriige zur Kinetik der Flotation, Revue de Metallurgie (Bucarest)
1.113-120.
29. Huber-Panu. H., and Georgescu, B. (1959) Sur les equations cinetiques des indices
techniques de laflottation, ibid. 2, 189-211.
 207

30. Huber-Panu, H., and Georgescu, B. (1961) Untersuchungen aber den Einfluss der Belaftung
auf die Flotation-kinetik, ibid . .6, 253-267.
31. Ek, e. (1967) Interpretation de la cinetique de flottation d'un minerai sulfure de cuivre,
Industrie Chimique BeIge, 36e Congres de Chimie Industriel1e, Bruxelles, 2., 1-5.
32. Bushell, e.H.G. (1962) Kinetics offlotation, Trans. AIME (Mining) ill, 266-278.
33. Jowett, A., and Safvi, S.M.M. (1960) Refinements in methods of determining flotation rates,
ibid. 217, 351-357.
34. Harris, e.C., Jowett, A., and Ghosh, S.K. (1963) Analysis of data from continuous flotation
tests, ibid. 227, 444-447.
35. Jowett, A. (1966) Gangue mineral contamination offroth, Brit. Chern. Engng.ll, 330-333.
36. Jancarek, J. (1964) Ein Beitrag zur Theorie der Flotationsgeschwindigkeit, Bergakademie
4-5,257-263.
37. Kakovskii, LA., Grebnev, A.N., and Silina, E.!. (1961) The relationship between the
floatability of mineral particles of various sizes, their structure and the consumption of
collectors, Tsvetnye Metally .31.(August), 7-17.
38. Imaizumi, T., and Inoue, T. (1963) Consideratons cinetiques sur laflottation a la mousse,
Proceedings VIth Int. Miner. Process. Congress (Cannes), pp. 699-715.
39. Loveday, B.K. (1966) An investigation into the kinetics offrothflotation, Ph. D. Thesis,
University of Natal.
40. Loveday, B.K. (1966) Analysis offrothflotation kinetics, Trans. IMM.6.2, 219-225.
41. Inoue, T., and Imaizumi, T. (1968) Some aspects of the flotation kinetics, Proceedings VIIIth
Intern. Miner. Process. Congress (Leningrad), paper S-15.
42. Haynman, V.J. (1975) Fundamental model offlotation kinetics, Proceedings XIth Intern.
Miner. Process. Congress (Cagliari), pp.537-559.
43. Huber-Panu, H., Ene-Danalache, E., and Cojocariu, D.G. (1976) Mathematical models of
batch and continuous flotation, in M.C. Fuerstenau (ed.) Flotation, AIME, New York, pp.
675-724.
44. Kapur, P.C., and Mehrotra, S.P. (1989) Modelling of flotation kinetics and design of
optimum flotation circuits, in K. V.S. Sastry and M.e. Fuerstenau (eds.) Challenges in
Minerai Processing, AIME, New York, pp. 300-322.
45. Mika, T.S., and Fuerstenau, D.W. (1968) A microscopic model of the flotation process,
Proceedings VIIIth Intern. Miner. Process. Congress (Leningrad), paper S-4.
46. Dowling, E.C., Klimpel, R.R., and Aplan, F.F. (1985) Model discrimination in the flotation
of a porphyry copper ore, Miner. Metall. Process. 2, 87-101.
47. Kelsall, D.F. (1961) Application of probability assessment offlotation systems, Trans. IMM
70, 191-204.
48. Meyer, W.C., and Klimpel, R.R. (1984) Rate limitations in froth flotation, Trans AIME
(Mining) 274, 1852-1858.
49. Bull, W.R. (1966) The rates offlotation ofmineral particles in sulphide ores, Austr. IMM
Process. 220 (December), 69-78.
50. Agar, G.E., and Barrett, J.J. (1983) The use offlotation rate data to evaluate reagents, CIM
Bull. 76(851), 157-162.
51. Agar, G.E. (1987) Simulation in mineral processing, in B. Yarar and Z.M. Dogan (eds.)
Mineral Processing Design, NATO ASI Series E, no. 122. pp. 268-287.
52. Mehrotra, S.P., and Kapur, P.e. (1975) The effects of particle size and feed rate on the
flotation rate distribution in a continuous cell, Int. J. Miner. Process. 2, 15-28.
53. Klimpel, R., and Hansen, R. (1981) Some factors influencing kinetics in sUlfide flotation,
208

SME-AIME, Preprint 81-14.

54. Laplante, A.R, Toguri, I.M., and Smith, H.W. (1983) The effect of air flow rate on the
kinetics offlotation. Part 1: the transfer of material from the slurry to the froth, Int.J. Miner.
Process. 11, 203-219.
55. Laplante, A.R, Smith, H.W., and Toguri, I.M. (1983) The effect of air flow rate on the
kinetics offlotation. Part 2: the transfer of material from the froth over the cell lip, ibid. 11,
221-234.
56. Laplante, A.R, Smith, H.W., and Toguri, I.M. (1984) The effect of air flow rate on the
kinetics offlotation. Part 3: selectivity, ibid . .u, 285-295.
57. Kaya, M., and Laplante, A.R (1985) investigation of batch and continuous flotation kinetics
in a modified Denver laboratory cell, Proceedings 24th CIM Conf., paper 18.4.
58. Ahmed, N., and Jameson, G.J. (1985) The effect of bubble size on the rate of flotation of
fine particles, Int.J. Miner. Process. 14, 195-215.
59. Spears, D.R, and Jordan, e.E. (1989) The effect of turbulence on the flotation rate of galena
when using fine bubbles, in S. Chander and R.R. Klimpel (eds.) Advances in Coal and
Mineral Processing using Flotation, SME, Littleton, pp. 77-84.
60. Klimpel, RR., Hansen, R.D., Mayer, W.C., and Nimerick, K.H. (1979) Laboratory
characterisation of the influence of reagent changes on coalflotation, SME-AlME, Preprint
79-11.
61. Klimpel, RR (1980) Selection of chemical reagents for flotation, SME-AlME, Preprint 80-
34.
62. Klimpel, R.R (1988) Methods of improving the economic performance of industrial scale
mineral flotation circuits, SME-AlME, Preprint 88-171.
63. Agar, G.E., Styles, G.H., Lyons, B.G., and Kipkie, W.B. (1987) Selection of rock
depressants based on laboratory kinetic studies, CIM Bull. 8Q(907), 45-51.
64. Marin, G., and Molina, E. (1988) Characterization of collectors through flotation rate data, in
S.H. Castro Hores and J. Alvarez Moisan (eds.) Froth Flotation, Elsevier, Amsterdam, pp.
329-340.
65. Hosten, e., and Tezcan, A. (1990) The influence offrother type on the flotation kinetics of a
massive copper sulphide ore, Minerals Engng. 1, 637-640.
66. Ball, B., Kapur, P.e., and Fuerstenau, D.W. (1970) Prediction of grade-recovery curves
from aflotation kinetic model, Trans. AIME (Mining) 247, 263-269.
67. Morris, T.M., and Edwards, RM. (1964) Analog computers find application in control of
flotation circuits, Min. Engng.l§.(8), 67-69.
68. Faulkner, B.P. (1966) Computer control improves metallurgy at Tennessee Copper's flotation
plant, ibid.N(11), 53-57.
69. Cooper, H.R (1966) Feedback process control of mineral flotation. Part I: Development of a
model for froth flotation, Trans. AIME (Mining) 2.3.5., 439-446.
70. Agar, G.E., Stratton-Crawley, R., and Bruce, T.I. (1980) Optimisation the design of
flotation circuits, CIM Bull. 11(824),173-181.
71. Agar, G.E., and Kipkie, W.B. (1978) Predicting locked cycle flotation test results from batch
data, ibid. 71(799), 119-125.
72. Bourassa, M., Barbery, G., Broussaud, A., and Conil, P. (1988) Flotation kinetics scale-up:
comparison laboratory batch test to pilot plant processing, in E. Forssberg (ed.) Proceedings
XVlth Int. Miner. Process. Congress, Elsevier, Amsterdam, pp. 579-588.
73. Barbery, G., Bourassa, M., and Dionne, L. (1988) Mise a l'echelle des circuits deflottation,
Ind. Miner.- Mines et Carrieres - Les Techniques 1Q, 8-14.
 209

74. Vedrine, H., Broussaud, A., Conil, P., and De Matos, C.S. (1990) Modelisation de la
cinetique deflottation d'un minerai sulfure polymerallique. Ind. Miner.- Mines et Carrieres -
Les Techniques 12..79-87.
75. Lewis. F.M .• and Morris, T.M. (1962) Analysis of operating flotation plants. Part 1:
Operating datafrom a sulfide flotation plant, the Londnn mill, in D.W.Fuerstenau (ed.) Froth
Flotation, AIME. New York. pp. 455-481.
76. Ek, c. (1970) Cinetique de flottation - Etudes de laboratoire - Essais industriels, Rudy.
1.a(3-4), 113-121.
77. Forssberg. K.S.E., Frykfors, C.O., and Palsson, B.J. (1982) Study of the kinetics of
jlotation in a bulkjlotation circuitfor galena and sphalerite, Proceedings XIVth Int. Miner.
Process. Congress (Toronto). vol. IV, paper 16.
78. Frew, J.A. (1982) Variation offlotation rate coefficients in zinc cleaning circuits, Int. J.
Miner. Process. 2,173-189.
79. Phelps Dodge, Ajo mill (1966), personal communication.
80. Davis, W.J.N. (1964) The development of a mathematical model of the lead flotation circuit at
the Zinc Corporation Ltd., Aus. IMM Proceedings 212,61-89.
2\0

Banks of flotation cells

2.2. MIXING AND HYDRODYNAMICS IN FLOTATION CELLS

P.MAVROS
Department of Chemistry (Box 114),
Aristotle University
GR-540 06 Thessaloniki
Greece

ABSTRACf. The flotation process is a complex interaction between the physicochemical characte-
ristics of the species involved and the hydrodynamic and operating conditions in flotation cells.
Efficient cell design requires knowledge about the power to be provided and the time to be spent in
the cell by each one of the phases. A review of new work in terms of mixing is presented. with
particular attention to the effects of turbulence on flotation. The experimental and theoretical work
on residence time ditributions in flotation cells is also presented.

1. Introduction
"Rotation cells are invented. not designed"
K.J. [yes

Rotation is a separation process used mainly for removing solid particles or particulate matter from
aqueous dispersions. It is widely used in mineral processing for beneficiating ores and recovering
valuable minerals; it is estimated that in the future it will treat more than 95% of the total ore
reserves [IJ. It is also used in water and wastewater treatment. either for recovering valuable
material (e.g. scarce metal ions etc.) or for removing unwanted species (e.g. toxic metal ions. ink.
pigments etc.).
It is probably very old: Herodotus refers to a similar process for cleaning gold grains [2aJ.
using tar-covered feathers. and it appears in an early Persian manuscript for cleaning lapis lazuli
[2bJ.
The process is rather complex. since it depends upon a multitude of parameters:
- geometrical (vessel configuration. particle size. rotor-stator design etc.);
- operational (gas and pulp flowrates. gas sparger design. metallurgical requirements in terms of
grade and recovery etc.);
- chemical (hydrophobicity - natural or induced - ofthe desired particles as well as gangue. effect
of added surfactants on the particle surface as well as the gas-liquid-solid dispersion (bubble
diameter. froth stability etc.) • surface oxidation etc.).
In the following sections. the various cells - mechanically agitated flotation machines and
flotation columns - will be briefly examined; the latest work concerning mixing requirements will
be reviewed. The hydrodynamic characteristics of the cells will also be examined. in particular the
residence time distribution of the various phases involved.

2Jl
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 211-234.
© 1992 Kluwer Academic Publishers.
212

2. Flotation Cell Characteristics

The process may be regarded as consisting of two stages: in the first, the contents of the pulp are
well mixed with appropriate reagents, so that the wanted species becomes hydrophobic, collides
with gas bubbles and hopefully adheres to one of them. The probability of a successful attachment
is given by the following equation:

(1)

where Pc is the probability of collision, PAis the probability of this collision resulting in the
adhesion of the particle onto the bubble surface and Po is the probability of detachment from the
bubble. For particles smaller than 100 ~, Po is approximately 0 [3].
The probability of collision decreases with the size of the particles: the finer the particle, the less
probable it becomes for it to encounter a bubble. Thercfore, smaller air bubbles are required to
maintain performance while processing fine particle systems: it has been calculated that a decrease
of bubble size by half results in an eight-fold increase in particle attachment [3].
In the second stage, the bubbles with their load disengage at the top of the vessel, where a froth
rich in particles is formed and has to be removed. Less frequently it is the unwanted species which
is made hydrophobic and floats to the top, while the valuable material is removed from the bottom
of the vessel. Some of the gangue particles are often entrained, too, with the concentrate, usually
fine particles with a diameter less than 30 !lm [4]. The addition of a wash-water stream on top of
the froth has a beneficial effect [5,6], since it entrains them back into the pulp.
Mineral processing plants use mechanically-agitated cells extensively nowadays; flotation
columns are also becoming increasingly popular. New flotation machine designs are appearing,
too, like the centrifugal pneumatic flotation machine [7.8]. In water and wastewater treatment. on
the other hand. dissolved-air agitatorless cells have been mainly used [9].

2.1. MECHANICALL Y-AGITATED FLOTATION CELLS

The first flotation cells to be used in mineral plants were mechanically-agitated. dispersed-air
vessels of varying design. Harris [10] and Young [11] have reviewed the various designs; the
mechanical cell characteristics will be discussed hcre only briefly.

(a) (b)

Figure 1. Flotation equipment including rotors and stators. (a) Sala; (b) Outokumpu.
 213

It is obvious that a flotation cell must comprise two distinct regions:

- the first one is a region of intense turbulence, where vigorous mixing is required to bring all the
species into close contact;
- in the second region, conditions of quiescence must prevail, allowing the bubbles to disengage
undisturbed, in order to minimise the probability of particle detachment.
The mechanically agitated cells usually comprise a mixing region, with the agitator system
located near or at the very bottom of the cell. A stator often shrouds the rotor; two such designs are
illustrated in Figure 1. Air is introduced, usually through the agitator shaft, either by the agitation-
induced air suction or by an external blower. Once out of the stator region, the bubbles are
progressively loaded with particles which then disengage through the upper, quiescent region of
the cell.
Cells are usually installed in banks; the tailings of one cell are fed into the next one while the
concentrate is removed from the bank side. The latest trends in cell design are mainly towards
large-volume cells: cells of 60-85 m 3 have been reported.
Alternative cells have also been examined, e.g. the agitator-less Davcra cell [12], the air-sparged
hydrocyclone [13].

2.2. FLOTATION COLUMNS

The need to process fines and the decreased floatability of such systems in agitated cells led to other
designs being investigated; the most successful new flotation cell design has been the flotation
column [5].

Colltctioll. ZOllA!

Figure 2. Flotation column.

It is most probable that they were invented in Canada, back in the '60s [14], although there is a
report that flotation columns have operated in China since 1961 [15]. Their main features may be
seen in Figure 2: they consist of a tall, tubular vessel, with a gas-sparging system at the bottom and
a concentric froth-collecting lip at the top. The diameter of the column may be up to 2.5 m, while
its height may go up to 10 or even 12 m. The pulp is fed in at approximately two thirds of the total
height and flows downwards, against a swann of riSing bubbles. The mixing of the phases is
214

accomplished by the countercurrent flow of the gas and pulp.

The region below the feed point is called the collection zone, where in fact the colliding of
bubbles and particles occurs, whereas the zone above it consists of the froth.
The gas is introduced at the bottom of the column. Several gas spargers have been used; they are
either multiple-orifice plates or tubes (with orifice diameters about 0.5 mm) or porous spargers
made of glass, stainless steel, rubber [14], elastic membranes [17] or filter cloth [18]. The rubber
ones seem to be the most promising ones, since all others tend to be clogged by small particles and
need frequent replacement.
The U.S. Bureau of Mines has presented a different kind of gas sparger [19], which makes use
of the dissolved-air technique: water and air are fed in a special packed-bed device, where air is
dissolved under high pressure in water. This is then fed at the bottom of the column through
nozzles fitted onto shafts; this arrangement permits the removal of the shaft for repair/replacement
of the nozzles, without shutting down the column, which is of obvious benefit to the operator.
The pressure release means that air "precipitates" out of the water into very fine bubbles; their
size lies in the range from 40 to 70 /illI [20]. This is extremely beneficial for fine particle flotation
[9,21], since bubbles tend to nucleate on or very close to the particles and this increases the overall
flotation probability, according to eqn. (1).
Although the columns were rather late starters, more companies are nowadays retrofiting the old
plants, replacing banks of mechanical cells by columns [22], thus achieving raised grades [23] and
even power savings. The U.S. Dept. of Energy [24 considers them "to have an inherent advantage
over standard mechanical pneumatic flotation machines in cleaning coal" [25].

~AIRBUBBLE
~ PATH
€11
is) 0

.
~
t3iS)
REFLUX WATER
PATH

x = gangue
..
LOW DENSrTY RODS • = valuable specie
(5o.Gv. < Pulp) o = air bubble
H = high density rod
L == low density rod

Figure 3. Features of the Wemco-Leeds flotation cell.

Alternative designs for the columns have also been proposed. Among them (see Finch and
Dobby [5] for a review of all column designs) : the Leeds column (Figure 3) with horizontal rods,
claimed to be very efficient in the reduction of ash in coal cleaning [26,27]; the use of internal
paeking [25,28]; provision of agitation for enhanced mixing [29]; separation of the column in
stages, as in distillation ones [30]. The Jameson eell [31] is a special column, in that it has no
collection zone: the pulp is fed through a tall downcomer, where it is mixed with air and is released
 215

into a froth zone; this column is claimed to give better results than both mechanical cells and
conventional columns.

3. Mixing in Flotation Cells

When entering the cell, air and pulp are mixed so that, on the one hand they are finely dispersed
and, on the other hand they are brought into contact so thal the bubble-particler aggregates may be
formed. The mixing of the phases is achicved either by agitation, in the case of mechanical cells, or
as a result of the countercurrent flow of pulp and bubbles.
The agitation provided in mechanical cells results in the development of a velocity field; the
resulting shear stress is responsible for the dispersion of the fluid and solid into fine particles,
which may then encounter each other. So far, shear stress is beneficial to the process. However,
once the aggregate is formed, it may again disintegrate, if the shear stress is too high, and this is to
be avoided. The shear stress, t, may be calculated from the power consumption per unit volume
(ffY) and the dynamic viscosity !.l of the medium:

t= ...J(PfY).!.l (2)

or from the shear rate, y, which is the velocity gradient, i.e. the rate of change of velocity with
distance in the cell :

(3)

Velocity measurements in the cell may thus yield the local values of shear stress and rate, whereas
power measurements permit the calculation of the cell average shear stress. Bowen [32] has given
an example of mixing vessel design problems related to shear rates. It should be pointed out that
this cell average is not sufficient for correlating mixing under various vessel geometries and
operating conditions.

TABLE 1. Dimensionless groups of the main parameters involved in mixing.

Dimensionless Group Range [27] Physical meaning
Power number, Np P g-l N- J D-) 0.5 - 5 Ratio of actual to
theoretical power
consumption
Flow number, NQ Q.N- 1 D-3 IxlO- 2 - 7xlO- 2
Froude number, NFr N L Dg- 0.1 - 5 Ratio of inertia to
I!ravitv
Reynolds number, NRe gD 2 NI-l- 1 5xl0 4 - 2xlO6 Ratio of inertia to
vi scous forces
Weber number, NWe N:lD:lo-1 - Ratio of inertia to
surface forces

Power consumption is one of the main variables of the process. Other important parameters
involved in the study of mixing conditions in flotation cells and mixing vessels, in general, are : the
volume V of the cell, the agitation speed N, the agitator diameter D, the air flowrate Qa, the density
216

soo
curve 1 curve 2 curve 3 curve 4 5 curve 6

*
curve

100

W~ $ -*-
-*- ~ ~
0.
Z
~ I crlb I I
c:r:I::cr::J ~
WID = 1/5 WID = 1/5 WID = 1/8 WID = 1/8 WID:' 1/8 WID = 1/8
10

5
--------------~=========1
~~==~~~==~======~4~5
------------------------6

Figure 4. Power number as a function of Reynolds number and turbine agitator style.

of the fluid mixture Q. the gas-liquid surface tension GOL etc. Experimentally determined values of
these parameters are usually correlated using dimensionless groups of parameters. e.g. the
Reynolds number. the Froude number. Table 1 provides a list of the main dimensionless numbers.
as well as the physical meaning of some of them. These groups have been used to correlate data
obtained while studying the mixing process in single. two- and three-phases agitated chemical
reactors.
The power consumption in agitated reactors is usually plotted against the Reynolds number. as
in Figure 4. Correlations vary with the impeller style. but at high NRe. it remains practically
constant; this is also the case for the flotation cell, which usually operates in the range 10 5 < NRe
< 106 [33].
In flotation cells, however, there are other parameters, e.g. the rotor-stator configuration, the
type of baffling etc. and their effect is difficult to quantify so that a single correlation covering all
may be derived. Degner and Treweek [34] have reported that data obtained in large (8.5 and
14.2 m3) Wemco cells correlate according to

and Qa - In(N) D2.4

which is somewhat different from the correlations indicated in Table I. In another study, Schubert
[35] presented data for specific power consumption for two different-size flotation cells, where
(PN). NQ. Nand Qa correlate reasonably well, although no explicit correlation was given.
Fallenius tabulated data for power consumption for several sizes of different cell manufacturers
[36]; if these are plotted against ND2 (Figure 5), then a plot comparable to that of Figure 4 is
obtained indicating that, in general, a correlation similar to the one used in chemical reactors may be
expected for flotation cells. too.
One should note, however, that: (a) there are several features of the flotation cell, which are
difficult to quantify, e.g. the configuration of the rotor-stator, the design of the agitator etc. and (b)
 217

the mechanical flotation cell configuration is far from standard, as compared to the agitated
chemical reactor, whose design obeys specific rules (Figure 6).

10

..!...

1lE ~ 1:1

•a
Outokumpu
Wemco

~
::l Denver
...C
<l> 0 Minemet
~
& II Mekhanobr

.1
.1 10

Figure 5. Power consumption of flotation cells (data from Fallenius [36]).

3.1. FLOW PATIERNS IN MECHANICALLY AGITATED CELLS

The fact that agitators in mechanical flotation cells are located near or at the botom of the cell means
also that the flow pattern generated will be very different from those obtained in stirred chemical
reactors. Figure 7 illustrates the patterns generated by a Rushton turbine; it is obvious that it
depends upon the clearance of the agitator from the cell bottom: if positioned at its usual height (as
shown in Figure 6), the flow pattern is distinctly different (see also Figure 8), depending again
whether the flow generated by the agitator is mainly axial or radial.

,
- -

- &-
H

...c:J
L...J

Figure 6. "Standard" configuration of stirred chemical reactor agitated with a Rushton turbine. d = D/2,
b = D/S, I = D/4, np = 6, H = D, C = H/6 - H/3, Bw = O.1D, nIl = 4.
218

I- -
/' ,
"'-, \ I ---
, I .. - .. ' \ I I '
, , , .. -~\ \ I
,
I 'I
" "
",
\ I
, '"
,
\ \ I
I ' , ,,
I ,
,,
I
I
,I,
", I' ' I ~
'. '
,',
III I I

,:',l~~:~.::I-I::~~,.
I I II" ,

.. ,.. .... -."

I I
\\ J" '\', II
," _ .. ' \ \ ..

.. .....
I I , ,
\. ... J , ...... _ ' " '''...
":"::,,! ""- ..,

Figure 7. Axial and radial flow patterns produced by agitators in chemical reactors.

Row patterns in flotation cells must be completely different from the ones in chemical reactors.
The fact that a quiescent zone is required at the top of the cell, so that a stable froth may be
established, means that the agitator must be located almost at the bottom of the cell. The toroidal jet
coming out of the rotor (jstator) causes at the same time the suspension of the solid particles, which
might otherwise tend to settle at the bottom.

10~--------~--------~

7
6
5

3
2

Figure 8. Particle settling due to poor agitation.

Once out of this region, the liquid may circulate in the cell, eventually either coming back to the
agitator region or flowing out of the cell. The study of flow patterns is linked to the study of
residence time distributions (see §4) and is important for the design of the cell. Determination of
the flow pattern in a vessel may be done
- either directly, using an optical method (by keeping track of a small coloured particle going
 219

round the vessel)

- or by indirect methods, using tracers, following their dispersion thr;)Ugh pH, temperature or
conductivity measurements at pre-detennined points and matching the results to theoretical
models [37,38].

3.2. SUSPENSION OF SOLID PARTICLES

One of the actions the agitator is required to perfonn is the suspension of the solid particles, so that
they may encounter the air bubbles. It is obvious that, if power is not sufficient, particles will settle
at the bottom of the vessel, either under the impellcr or near the cell walls, as shown in Figure 8.

Finger Imp Double - Finger Imp. Tri.ngte-8I.dt Imp. Blade Imp. Double - Blade Imp.
(EFR) (OFR) (OrBR) (fBR) (DBRI

600 2.50
2.25
500 2.00
1.75
~
'".f§.
400 1.50
g
I:
~
~
til
1.25
Z 300 1.00 ~
0.75

200 0.50
DFR EFR DrBR EBR DBR
Impeller Type
Figure 9. Specific power input (fN) and agitation speed N for I-sec suspension of solids for various
impellers (data from ref. [35]).

Suspension is considered complete, when all particles are in motion and no particle remains at
the bottom of the cell for more than a given short time, e.g. a few seconds. Zwietering [39] has
proposed to correlate the power input with the J-sec criterion, N(ls), Le. the agitator speed
corresponding to a maximum time of one second spent by the particles at the bottom of the cell.
Schubert [35] presented experiments with various solid particles and found that N(ls) correlated
well with Np as :

N(ls) =K(Np Q)-a (4)

220

with (K,a) varying: (1.3xlO\ 0.47) for ballotinis (0.3-0.5 mm) and (9xlOZ, 0.47) for both
ballotinis (0.1-0.16 mm) and sylvinite ore (-125 mm). As the slope of the lines is the same in both
cases, it seems probable that with more experiments the values of K could possibly be correlated
to the solid particle properties (size, density, floatability etc.). He also experimented with the
agitator style; Figure 9 shows that the number and position, as well as the form of the blades is
important, both in power consumption and suspension of solids.
Other workers have correlated the power input with NJS, which is similar to N(ls). Starting
from turbulence models, theoretical correlations have been developed for NJS :

NJS = f(g, L1Q, QL, T, Np, D, dp) (5)

where T is the cell diameter (or its equivalent in square ones) and dp the mean particle diameter.
Nienow (40) has reviewed the work done on particle suspension and the corresponding N JS and
PJS correlations for stirred chemical reactors mainly.
Finally, little work has been done on the inverse problem, i.e. the drawdown of floating solids
in agitated vessels. Thring and Edwards [41) investigated the effect of baming and clearance of
impeller from the bottom of the vessel, but gave no correlations of their results.

1. Denver flotation cell

2. Pipe impeller
3. Flattened cylindrical
impeller
4. Modified pipe impeller
5. Turbo aerator
~~--~'----~2----~3----4~--~5~--~6----~'~
Power consumption per unit liquid
volume Po/V [kW/m 3 ]

Figure 10. Flowrates of induced-gas for various gas-inducing agitators [42].

3.3. GAS-LIQUID DISPERSIONS

Air is dispersed in flotation cells either by induction or by using an external blower. The former
is usual in mechanical flotation cells, while the latter is common practice in flotation columns and
 221

stirred chemical reactors. Gas induction is achieved by a specific agitator design and a minimum
agitator speed. Joshi et al. [42] have reviewed the pertinent literature; Figure 10 illustrates tlx:
flowrate of gas induced by several impellers: it is most interesting to note the very low induced-gas
flowrates in Denver cells, as compared to those in stirred chemical vessels.
The mechanism of gas dispersion has been studied extensively. Van't Riet [43] demonstrated
that cavities of air are formed behind the agitator blades; he further distinguished vortex, clinging
and large cavities, depending upon the gas flowrate and agitator speed (Figure 11) : at constant
stirrer speed, these forms develop consecutively as the gas flow increases. As the gas leaves the
agitator blades, it breaks down into small bubbles and this is further enhanced by the presence of a
stator around the agitator.

dllC
dll"CCUoft
01_ .....•.
~~.~ ~
~
~ ~-:!:..
~
~
..
: -..........
::.

Vort.ca cavlLy CII"9i"9 ..."M.y

Q

~
~
~.
..
'"
~A'~
~
.'.
' ..

Figure 11. Cavity formation behind agitator blades [44].

There are two essential terms when discussing gas-liquid dispersion: flooding and loading.
Flooding occurs when gas is not dispersed by the agitator, but shrouds the shaft (Figure 12); this
usually happens at low rates of rotation. Loading on the other hand is achieved when gas is well
dispersed, at least above the agitator. The transition between the two regimes occurs at a certain
agitator speed, Np, which depends upon several operational as well as geometrical parameters, e.g.
vessel and agitator diameter, overall height of fluid in vessel etc. [45]. The effect of agitator
clearance from the bottom on the gas-liquid dispersion, in terms ofNp, was examined by Hudcova
et al. [46]; the higher the agitator, the better the air is dispersed and the flooding/loading transition
occurs at higher gas flowrates.

3.4. GAS-LIQUID-SOLID DISPERSIONS

So far, only two-phase systems have been considered. The introduction of solids in gas-liquid
dispersions, as is the case in flotation systems, causes several changes.
It is well known that solid particles will sediment, when gas is introduced in a solid-liquid
dispersion. Warmoeskerken et al. [47] investigated this in a stirred vessel and found that, although
the presence of solids (low concentrations, up to 5 wt%) did not affect the power consumption
222

...
° ~::-o:( ~<itI'o °
/'J ~o

I
It\=
o!o_

•
_J
-\io
(a) (b) (c) (d)

Figure 12. Effect of agitator speed on gas-liquid dispersion efficiency and flooding/loading transition
(a-c); effect of agitator size on agitator loading (c-d).

curves. the effect of gas rate was noticeable: when gas was introduced in a solid-liquid suspension
at the "just suspended" condition (agitator speed at NJs). some particles settled and this was
amplified by increasing the gas flowrate. At a given critical gas rate. QJs. massive sedimentation
occurred. They also showed that a correlation is possible between NJS and QJs. depending upon
operating conditions and cell configuration.
Two questions have to be answered when dealing with three-phases systems:
(a) what is the minimum power/agitation required to achieve an adequate phase mixing. so as to
obtain a satisfactory performance; and
(b) what are the scaling-up rules. so that knowledge obtained from small-scale experiments may
be used to design industrial-scale cells?
It is obvious that. often. machines are first installed and operated with little - if any - a priori
knowledge about their design; research into optimization usually comes later. when other problems
become important. This has also been the case with notation cells: they have been operating happily
for several decades and now that industry is facing problems (diminishing ores and grades. energy
crisis etc.). people start looking into the fundamentals of the process. so that an optimised and
more cost-effective process may be hopefully achieved. In the case of flotation. this is done by
investigating the role of turbulence in it.

3.5. TURBULENCE IN FLOTATION CELLS

Recently. attention has started being paid to the "microphenomena" occurring in the flotation cell.
as opposed to the "macroscopic" analysis. given so far.
The principal feature of turbulent flows is the presence of turbulent eddies. which are
responsible for the dispersion of the particles and bubbles and the efficient mixing of the phases.
Energy is transferred from the larger ones down to the smallest ones. where it is eventually
dissipated by the action of viscosity.
The eddies vary in size. having a maximum scale. A. which is of the order of the scale of the
turbulence-generating equipment. e.g. the agitator. down to a minimum value. A.. which was
defined by Kolmogoroff [48) as :

(6)

where v is the kinematic viscosity of the liquid and £ the local energy dissipation rate. The latter is.
according to Taylor. related to the turbulent velocity by the following equation:
 223

(7)

The maximum-size eddy scale, A, is related, according to Brodkey [49), to A. by:

(8)

The fluid velocity in a turbulent field may be written as a sum of two terms: a mean or time-
averaged velocity, U mean , and afluctuating velocity, u'(t) :

u(t) =U mean + u'(t) (9)

whith both U mean and u'(t) varying in the radial, axial and tangential direction in a cell. On the other
hand, for isotropic turbulence, the mean fluctuating velocity is related to the shear stress:
,2
t = - QU mean (10)

Measurements of turbulence, i.e. of mean and fluctuating velocities, are performed with hot-wire
or hot-film anemometry and lately with more modem techniques (Laser-Doppler Anemometry)
[50). Using these measurements, and calculating the lowest wave number ko, which is related to
the agitator diameter and the maximum eddy size:

(11)

and comibining these equations yields a semi-empirical rclation between the specific power input,
the agitator diameter and the fluctuating velocity:

(12)

where C is a constant. Laufhiitte and Mersmann [51) showed that a value of C = 6.25 correlates
reasonably well with experimental result~ and may be used for design purposes.
The specific power input may be correlated with operational variables, for a given cell design. lt
should be noted however that when dealing with three-phase systems, the power required to
counteract the sinking of the particles should also be taken into account [52):

Emean = Emean,flow + Emean,sink (13)

In small vessels, the power dissipated in the flow, Emean.flow, is predominant, but as the vessel
gets bigger, the second term, Emean.sink becomes the dominating one. Mersmann and Laufhiitte
[52] also found that in the scaling-up rules the ratio of panicle to vessel diametcr (dp/D) should also
be taken into account; the fact that all previous work on this subject has disregarded this ratio may
be responsible for the lack of agreement among the various correlations, often differing by a
factor of 70.
lt seems that little work has been done concerning turbulence in flotation cells. Fallenius [36)
224

considered the cell rotor as a centrifugal pump; thus the flow coming out of it may be seen as a
toroidal jet. He presented a theoretical analysis of the tutbulent fluctuation of velocity in the cell and
used operational data from Outokumpu and other manufacturers, mainly for sulphides flotation, to
show that turbulent velocity remains practically the same in each cell design, independent of cell
size.
Two- or three-phase turbulence measurements are not yet widely available; theoretical
considerations, however, indicate that solids in suspension reduce the turbulent fluctuations of
velocity [53], while dispersed air increase it [36]. Geisler and Mersmann [54] found that the local
specific energy dissipation increases slightly below the agitator, when solids (5 wt%; dp,mean =
715 !IDl) are added to the fluid; above the agitator, no increase was noticed.
Finally, several models have been developed to simulate turbulent flow in an agitated vessel.
Most are two dimensional, but lately three-dimensional models have been developed by Middleton
et al. [55] (considering the turbulence to be isotropic) and Iu et al. [56] (considering a non-
isotropic turbulence field).

4. Flotation Cell Hydrodynamics

The design and efficient operation of flotation cells requires knowledge of what is happening inside
the cell: how much time is allowed for the particles to mix with the air bubbles? Are there any
dead spaces, where matter gets trapped, or any bypassing, with either air or pulp leaving the cell
too soon? All these are cases of malfunctioning - termed non-ideal flow - and they lead to poor
performance. One question to add to the above ones is how thcse things change with cell scale: if
a design eliminates them in a small-scale, laboratory cell, are they still eliminated in the larger
ones? One tries to answer all these questions by performing measurements of Residence Time
Distributions (RID) in the cells.
Particles enter the cell, stay in it flowing around for some time, and leave - this applies equally to
all three phases. Since flotation is considered by most to be a first -order process, if one knows the
time spent by each particle in the vessel, or rather its distribution, then by combing these an
a priori calculation of the cell performance may be achieved. The development of RTD theory is
due to Levenspiel [57]; since then, new books have also becomc available, e.g. by Nauman and
Buffuam [58].
If one had the possibility of measuring the time spent in the cell by each individual particle
leaving it and classify the results in time intervals, a distribution of residence times would be
obtained. Following the Levenspiel notation, this distribution is called exit age or residence time
distribution, E(t); then E(t)dt is the probability that the particle residence time lies in the interval
(t, t+dt). Like all probabilities,

jE(t)dt = I (14)
o
Further, the probability that a particle has a residence time larger or smaller than a particular value,
there are two other functions: the probability F(t) that a particle has a residence time less than t :
 225

F(t) = jE(t')dt' (15)

o
or the probability Wet) that it has a residence time larger than t (often called the washout function) :

Wet) = jE(t')dt' (16)

It is obvious that

F(t) + Wet) =1 (17)

and that

dF - dW
E(t) = dt =---at (18)

E
of exit stream
older than 11

Figure 13. Residence time distribution E(t) for a fluid flowing through a vessel.

Once E(t) is known for a particular system, it may be combined with kinetic data to predict the exit
composition, Co :

Co,mean = JCelemenIE(t)dt (19)

1=0

The RTD may be experimentally determined with the stimulus-response technique. The stimulus
is an adequate amount of some tracer, which is introduced in the incoming stream, while the
response of the system is monitored at the exit, in terms of the tracer concentration. There are
various types of measurements : pH, conductivity, concentration, radioactivity etc. If the stimulus
is a instataneous pulse, then E(t) is obtained, whereas if it is a step change, then F(t) is obtained;
the one may be converted to the other one through eqn. 18.
From the experimentally-obtained RTD the mean, L, and variance, 0 2, are often calculated. These
may then be used for modelling purposes.
226

4.1. NON-IDEAL FLOW MODELLING

Often, the RTD of a given vessel is used to derive useful infonnation, so that a model describing its
behaviour may be set up. This model may then be used for predictive purposes. There are two
main lines of thought on this subject.

4.1.1. Tanks-in-Series Model. The RTD for a single well mixed vessel is very simple:

1
E(t) =- e·t/t (20)
t

where t is the mean residence time in the vessel (=V/Q, where V is the vessel volume and Q the
volumetric flow rate through it). More complex systems are modelled as a series of such vessels
and the RTD for N vessels is easily obtained:

(21)

where tj is the mean residence time in one of the vessels (which are here considered as having all
of them the same volume). Although initially eqn. 21 was developed for integer values of N, this
was later extended to real numbers [59] :

(22)

where f(N) is the gamma function:

(23)

Although real values of N lack any resemblance to a physical situation, this extension is useful for
characterising small departures from ideal behaviour.
A useful property of eqn. 22 is that its mean and variance are related to the value of N:

(24)

therefore by detennining t and 0 2 from the experimental curve, the value of N which closely fits it
is easily calculated.
Often a dimensionless representation of E(t) is preferred, since it pennits comparison of curves
obtained in different vessel sizes or operational conditions: if e =tit, then:

N(Ne)N-1
E(e) = (N-l)! c·Ne (25)

4.1.2. Axial Dispersion Model. This model assumes that mainly plug flow occurs in the vessel
under consideration, with some degree of mixing along it; this is also considered to be independent
 227

of position in the vessel. In this case, the model is described by a partial differential equation:

(26)

where u is the flow velocity, z the axial dimension and Da the axial dispersion coefficient, characte-
ristic ofthe degree of mixing. In dimensionless form, eqn. (26) is written as :

(27)

where e = tft = t uIL, and x = zIL. Pe is the Peclet dimensionless number (= uL/DJ and is the only
parameter of the model. Large values of Pe mean little dispersion, therefore flow resembles plug
flow; small values ofPe, on the other hand, mean large dispersion, Le. mixed flow.
Fitting the axial dispersion model to experimental curves is not so straightforward as the tanks-
in-series model, however. The solution of eqn. 27 depends upon the boundary conditions. Each
of the vessel ends is considered either open, i.e. the particles may flow back, or closed, i.e. the
particles flow only once through that end (either in or out). An analytical solution is obtainable only
for the open-open vessel (i.e. both ends are of the open type), but means and variances have been
calculated for all cases.

TABLE 2. Means and variances for various boundary conditions of the axial
dispersion model.
Boundary conditions cJ21t2

Open-Open 2 8
Pe + Pe 2

Open-Open, two measurements 2

f'o,(d21t 2) = -Pe

Open-Closed or Gosed-Open 2 3
Pe + PeZ

Closed-Gosed 2 2 [l-exp(-Pe)]
---z
Pe Pe

Figure 13 presents the theoretical E(t) curves for all three cases of boundary conditions, together
with the tanks-in-series model. It is obvious that there are not many differences between the four
curves and, what is more, the experimental uncertainties are often larger than the theoretical ones.

4.1.3. Multi-Parameter Models. The models presented so far incorporate a single parameter,
which makes them attractive to the user. However, they are often inadequate to represent the real
228

situation, especially in cases when regions of different types of flow co-exist in a vessel. For
example, in a mechanical flotation cell, the region around the agitator is obviously a well mixed
zone; this is not the case, though, for the upper more quiescent part of the vessel.
Thus, one has to resort to models having more than one parameter. Most of these consist of
several well mixed zones, arranged in series and/or in parallel. The number of such zones, their
volume, the amount of flow directed into and out of them etc. are examples of the parameters
involved. It should be noted that, although a complex model may be more realistic, the calculation
of parameter values by fitting the model to experimental results becomes lengthy and sometimes
difficult.

2 E
®

3 E

2 tiT 3

Figure 13. (A) Theoretical E(t) curves for a dispersive flow and comparison to the tanks-in-series model,
obtained with a pulse input of a tracer, for Pe 8 [601. Curves: 1, Open-Open; 2, Open-Closed; 3 : Closed-
Closed; 4 : Tanks-in-Series model (N=5). (B) Pe = 50, N = 98.

4.2. RTD MEASUREMENTS IN FLOTATION CELLS

4.2.1. Mechanical cells. Since the first flotation cells were mostly mechanically-agitated, they were
usually considered as well mixed; early experiments by Jowett [61], Woodburn and Loveday [62]
and Niemi [63,64] confirmed this. Lately, Koch [65] correlated the mean residence time and the
number of mixed zones-in-series with operating variables. Mehrotra and Saxena [66] demonstrated
 229

that part of the cell volume is in fact "dead" and correlated their results as well with operating
variables. Moys [67] used a three-parameter model to simulate the froth behaviour in a cell.
Finally, Frew [68] investigated the behaviour of a bank of cells and used a zones-in-series model to
describe the system, the single parameter being not the number of zones (which was fixed in this
case) but the amount of backmixing flow, Le. the amount of liquid flowing backwards between
adjacent cells, which is responsible for the mixing of their contents (Figure 14). Later, Wilson and
Frew [69] correlated this parameter to cell volume, aeration, cell agitation etc. A similar approach
has been followed by Ando et al. [70] for modelling the flow in a horizontal mixing vessel with
multiple impellers.

FT
I
~
C
T,
C
T.2
c..
T
G-
T5
-G Te

B, B3 B5

ce

Figure 14. Bank-of-flotation-cells model with backmixing flow [68].

4.2.1. Flotation columns. These are based essentially on bubble columns, for which there is a
great amount of work done. Dobby and Finch were among the first to do experimental work on
RTD measurements in flotation columns [71,72]; they used the axial dispersion model and
correlated the axial dispersion coefficient, Da , with the column diameter de and the superficial gas
velocity UG in the column:

(28)

Additional work on RTDs in columns was done by Mankosa et al. [73], Laplante et al. [74],
O'Connor and co-workers [75,76].
An important modification to eqn. 28 was suggested by Tinge and Drikenburg [77] : the
vertical misalignment of columns has an important effect on the value of the axial dispersion
coefficient and a correction should be added :

(29)

where Da is the truly-vertical axial dispersion coefficient and D a.8 the coefficient when the column
e
is misaligned by degrees. Finch and Dobby [5] calculated that a lcm/lOm misalignment raises
the axial dispersion coefficient by a value of over 4. Rice and Littlefield [78] suggested that a value
of 106 might be more appropriate in eqn. 29, but lack of further experimental verification makes
this proposition tentative.
Work by Finch and co-workers has shown that the axial dispersion coefficient is essentially the
same for both liquid and solid particles [5]. However, Mills and O'Connor [76] suggested that this
may be true only for laboratory-scale columns and that the correlation available underestimates the
230

solids mixing in larger-diameter columns; this has to be clarified by further experimentation.

The modelling of flotation columns has been based essentially on axial dispersion. A different
model, based on the tanks-in-series with backmixing flow was proposed [79] based on similar
models applied earlier in both mechanical flotation cells [68] and chemical reactors [37]; it is also
physically similar to the models proposed by Zehner [80] and Joshi and Sharma [81] for bubble
columns. Experimental data from RTD mea~urements of the liquid phase in a flotation column were
used to calculate the backmixing flow parameter and the optimum number of zones-in-series and to
relate them to the liquid and gas flowrates.

llliill.
I
fFmll' d.

(a) (b) (c)

Figure 15. Models of Zehner [78] (a,b) and Joshi and Sharma (c) [79].

5. References

1. Dolezil M., and Reznicek, J. (1990) Mineral processing 1989 - trends and developments,
International Mining 1(8), 37-52.
2. (a) Herodotus, Historiae, Book 4, 195. (b) Persian ms. 1877 (Elliott 176), dated 877/1472,
in the Bodleian Library, Oxford. See also: Ritter, H., Ruska, J. and Winderlih R (1935)
Eine persische Beschreibung der Fayence technik von Kaschan aus dem Jahre 700h11301d,
Istanbuler Mieneilungen .2-
3. Yoon RH., Adel G.T., Luttrell G.H., Mankosa M.J., and Weber A.T. (1988), Microbubble
flotation offine particles, in Y.A. Attia, B.M. Moudgil and S. Chander (eds.) 1nterfacial
Phenomena in Biotechnology and Materials Processing, Elsevier, Amsterdam, pp. 363-374.
4. Trahar, W.J. (1981) A rational interpretation of the role of particle size in flotation , Int. J.
Miner. Process . .8., 289-327.
5. Finch J.A., and Dobby G.S. (1990) Column Flotation, Pergamon Press, Oxford.
6. Kaya, M., Laplante, A., Blades A., Tan, c., Huls, B., Lacombe, P. Kemp, R, and Ayotte,
N. (1990) Industrial applications offroth washing in flotation , in G. Onal (ed.) Proceedings
III Int. Miner. Process. Symp. (Istanbul, 11-13 Sept. 1990), pp. 557-567.
 231

7. Orlov. S.A .• Ermakov. V.I.. and Shrarnrn. E.O. (1986) Industrial testing of a centrifugal
pneumatic flotation machine, Tsvetnye Metally 27, 92-93.
8. Orlov. S.A. (1986) Intensification of the flotation process in a centrifugal field. Tsvetnye
Metally27.107-109.
9. Matis, K.A .• and Mavros P. (1991) Recovery of metals by ion flotation from dilute aqueous
solutions, Separ. & Purif. Methods (in press).
10. Harris, e.e. (1976) Flotation machines. in M.e. Fuerstenau (cd.) Flotation, AIME, New
York. vol. 2. pp. 753-815.
11. Young, P. (1982) Flotation machines, Mining Mag. 146.35.
12. Wills. B.A. (1985) Mineral Processing Technology. Pergamon Press. Oxford. p. 426.
13. Van Deventer, J.S.J., Burger. A.J .• and Cloete F.L.D. (1988) Intensification offlotation
with an air-sparged hydrocyclone. J. South Afr. Inst. Min. Metall. 88. 325-332.
14. Wheeler, D.A. (1985) Column flotation - The original column. in S.H. Castro Flores and
J.A. Moisan (eds.) Froth Flotation, Elsevier. Amsterdam, pp. 17-40.
15. Hu. W., and Liu, G. (1988) Design and operating experiences with flotation columns in
China, in K.V.S. Sastry (ed.) Column Flotation '88, SME/AIME, Littleton, pp. 35-42.
16. Dolezil, M .• and Reznicek, J. (1986) Mineral processing 1985 - Trends and developments,
International Mining J(8), 49-59.
17. Rice. R.G., and Howell, S.W. (1986) Elastic and mechanics for membrane sparger, Amer.
Inst. Chern. Eng. J . .32... 1377-xx.
18. Xu, M. and Finch, J.A. (1989) Effect of sparger surface area on bubble diameter in flotation
columns, Can. Met. Quart. 2£, 1-6.
19. Hyde, G.R., and Stojsic, A. (1987) Electrochemical and column flotation technologies can
mean higher recovery at lower operating costs, Mining Engineering 39(7). 481-484.
20. Zabel, T. (1982) Flotation in water treatment, in K.J. Ives (ed.) The Scientific Basis of
Flotation, Proceedings NATO ASI (Cambridge, UK, 5-16 July 1982),29 pp.
21. Matis, K.A., Mavros, P.,and Kydros, C.A. (1991) A dissolved-air flotation micro-cell for
floatability tests with particulate systems, Sep. Techno\. (in press).
22. Redfearn, M. (1989) An overview of column flotation in Canada, in Proceedings Column
Flotation Workshop (University of Newcastle, NSW, Australia, 27-28 July 1989), TUNRA,
Adelaide, 17 pp.
23. Parekh, B.K., and Jordan, C.E. (1990) Flotation, Mining Engineering 42(5), 485-486.
24. Killrnayer, R.P., and Hucko, R.E. (1989) Interlaboratory comparison of advanced froth
flotation processes, SME Annual Meeting (Las Vegas 27/2-2/3), preprint no. 89-137.
25. Idlas, S.A .• FitzPatrick, J.A., and Slattery, J.e. (1990) Conceptual design of packed
flotation columns, Ind. Eng. Chern. Res. 29, 943-949.
26. Dell, e.e. (1978) Column flotation of coal- The way to easier filtration, Mine & Quarry 1,
36-40.
27. Dell. e.e. (1985) The Leeds flotation column. How it works and what it does, Leeds Univ.
Mining Assoc. Mag., 23-29.
28. Yang. D.e. (1988) A new packed column flotation system, in K.V.S. Sastry (ed.) Column
Flotation '88, SME/AIME, Littleton, pp. 257-265.
29. Karolat, W.H., and Hill, G.A. (1988) An agitated bubble column for use in potash flotation ,
Can. J. Chern. Eng. QQ, 1017-1020.
30. Idlas, S.A.. FitzPatrick, J.A., and Slattery, J.e. (1990) Conceptual design of stagedflotation
columns, Ind. Eng. Chern. Res. 29, 949-955.
31. Jameson, G.J. (1989 The Jameson column - A high intensity flotation celi, in Proceedings
232

Column Flotation Workshop (University of Newcastle, NSW, Australia, 27-28 July 1989),
TUNRA, Adelaide, 14 pp.
32. Bowen, RL. Jr (1986) Unravelling the mysteries of shear-sensitive mixing systems, Chern.
Eng., June 9, 55-63.
33. Barbery, G.(1982) Engineering aspects of flotation in the minerals industry: Flotation
machines, circuits and their simulation, in K.J. Ives (ed.), The Scientific Basis of Flotation,
Proceedings NATO ASI (Cambridge, UK, 5-16 July 1982),66 pp
34. Degner, V.R, and Treweek, H.B. (1979) Largeflotation cell design and development, in P.
Sornasundaran and N. Arbiter (eds.) Beneficiation of Mineral Fines, AIME, New York,
pp.816-837.
35. Schubert, H. (1985) On some aspects of the hydrodynamics offlotation processes, in K.S.E.
Forssberg (ed.) Flotation of Sulphide Minerals, Elsevier, Amsterdam, pp. 337-355.
36. n,
Fallenius, K. (1987) Turbulence in flotation cells, Int. 1. Miner. Process. 1-23.
37. Mann, R, Mavros, P., and Middleton, J.C. (1981) A structured stochastic flow modelfor
interpreting flow follower data from a stirred vessel, Trans. 1. Chern. E . .12, 271-278.
38. Mann, R., and Mavros, P. (1982) Analysis of unsteady tracer dispersion and mixing in a
stirred vessel using interconnected networks of ideal flow cells, Proceedings 4th Europ.
Conf. Mixing (BHRA, Noordwijkerhout, 27-29 April 1982), pp. 35-47.
39. Zwietering, Th.N. (1958) Suspending of solid particles in liquid by agitators, Chern. Eng.
Science.a, 244-253.
40. Nienow, A.W. (1985) The suspension of solid particles ,in N. Hamby, M.F. Edwards and
A. W. Nienow (eds.) Mixing in the Process Industries, Butterworths, London, pp.297-321.
41. Thring, RW., and Edwards, M.F. (1990) An experimental investigation into the complete
suspension offloating solids in an agitated tank, Ind. Eng. Chern. Res. 29, 676-682.
42. Joshi, J.B., Pandit, A.B., and Sharma, M.M. (1982) Mechanically agitated gas-liquid
reactors, Chern. Eng. Sci. 31,813-844.
43. Van't Riet, K. (1975) PhD Thesis, University of Delft.
44. Warmoeskerken, M.M.C.G., and Smith, J.M. (1982) Description of the power curves of
turbine stirred gas-liquid dispersions, in Proceedings 4th Eur. Conf. on Mixing (BHRA,
Noordwijkerhout, 27-29 April 1982), pp. 237-246.
45. Warmoeskerken, M.M.C.G., and Smith, I.M. (1985) Flooding of disc turbines in gas-liquid
dispersions: A new description of the phenomenon, Chern. Eng. Sci. 40, 2063-2071.
46. Hudcova, V., Nienow, A.W., Haozhung W., and Huoxing, L. (1987) On the effect of liquid
height on the flooding/loading transition, Chern. Eng. Sci. 42, 375-377.
47. Warmoeskerken, M.M.e.G., van Houwelingen, M.C., Frijlink, J.l., and Smith, I.M.
(1984) Role of cavity formation in stirred gas-liquid-solid reactors, Chern. Eng. Res. Des.
62, 197-200.
48. Kolrnogoroff, A.N. (1941) The local structure of turbulence in incompressible viscous fluid
for very large Reynolds numbers, Dok!. Akad. Nauk. SSSR 31.
49 Brodkey, RS. (1975) Turbulence in Mixing Operations, Academic Press, New York.
50. Calabrese, R. V., and Stoots, C.M. (1989) Flow in the impeller region of a stirred tank,
Chern. Eng. Progr. May, 43-50.
51. Laufhiitte, H.D., and Mersrnann, A. (1987) Local energy dissipation in agitated turbulent
fluids and its significance for the design of stirring equipment, Chern. Eng. Techno!. 10,
56-63.
52. Mersrnann, A., and Laufhiitte, H.D. (1985) Scale-up of agitated vessels for different mixing
processes, in Proceedings 5th Europ. Conf. Mixing (Wurzburg, 10-12 June 1985), pp. 273-
 233

284.
53. Neesse, Th., Schubert, H., Liepe, F., and Mockel, H.-O. (1977) Turbulenzmessungen in der
Mehrphasenstromungfest-fliissig eines Riihr-apparates, Chern. Tech. 2.2, 544-548.
54. Geisler, R.K., and Mersrnann, A.B. (1988) Local velocity distribution and power dissipation
rate of suspensions in stirred vessels, Proceedings 6th Europ. Conf. Mixing. (BHRA, Pavia,
24-26 May 1988), pp. 267-272.
55. Middleton, J.C., Pierce, F., and Lynch, P.M. (1986) Computations of flow fields and
complex reaction yield in turbulent stirred reactors and comparison with experimental data,
Chern. Eng. Res. Des. M, 18-22.
56. Ju, S.Y., Mulvahill, T.M., and Pike, R.W. (1990) Three-dimensional turbulent flow in
agitated vessels with a nonisotropic viscosity turbulence model, Can. 1. Chern. Eng. QJi, 3-
16.
57. Levenspiel, O. (1972) Chemical Reaction Engineering, Wiley, New York, 2nd Edition,
578 pp.
58. Nauman, E.B., and Buflham, B.A. (1983) Mixing in Continuous Flow Systems, Wiley -
Interscience, New York, 271 pp.
59. Buffllam, B.A., and Gibilaro, L.G. (1968) A generalization of the tanks-in-series mixing
model, A.I.Ch.E.J . .l.1.. 805-806.
60. Schweich, D. (1988) Some chemical engineering problems related to mixing and disorder, in
E. Guyon, J.-P. Nadal and Y. Pomeau (eds.) Disorder and Mixing, Kluwer Academic
Publishers, Dordrecht, pp. 57-84.
61. Jowett, A. (1961) Investigation of residence time offluid in froth flotation cells, Brit. Chern.
Eng. ll, 254-258.
62. Woodburn, E.T, and Loveday, B.K. (1965) The effect of variable residence time on the
performance of a flotation system, 1. S. Afr. Inst. Min. Metall. 65, 612-628.
63. Niemi, A. (1966) A study of dynamic and control properties of industrial flotation processes,
Acta Polytech. Scand. 48, 111 pp.
64. Niemi, A. (1966) On the dynamics of a pneumatic flotation cell, Acta Polytech. Scand. 49,
pp.5-39.
65. Koch, P. (1984) Charaketristika der Verweilzeitverteilungsfunktion im Trogf/otationsapparat,
Freib. Forsch. A719, 85-106.
66. Mehrotra, S.P., and Saxena, A.K. (1983) Effects ofprocess variables on the residence time
distribution of a solid in a continuously operated flotation cell, Int. J. Miner. Process. lQ,
255-277.
67. Moys, M.H. (1984) Residence time distributions and mass tansport in the froth phase of the
flotation process, ibid. 11, 117-142.
68. Frew, J.A. (1984) Backmixing in industrial flotation banks. I. Effect upon the detrmination
offlotation rate coefficients, ibid. U, 239-250.
69. Wilson, S.G., and Frew, J.A. (1986) Effect of operating variables upon liqUid backmixing in
a laboratory flotation bank, ibid.lQ, 281-298.
70. Ando, K., Obata, E., Ikeda, K., and Fukuda, T. (1990) Mixing time of liquid in horizontal
stirred vessels with multiple impellers, Can. J. Chern. Eng. 68,278-283.
71. Dobby, G.S., and Finch, J.A. (1985) Mixing characteristics of industrial flotation columns,
Chern. Eng. Sci.:ill, 1061-1068.
72. Dobby, G.S., and Finch, J.A. (1986) Flotation column scale-up and modelling, CIM Bulletin
l'i., 89-96.
73. Mankosa, M.J., Adel, G.T., Luttrell, G.H., and Yoon, R.H. (1988) Scale-up and design
234

aspects of column flotation. Proceedings Int. Syrnp. "Production and Processing of Fine
Particles" (Montreal. 28-31 August). 10 pp.
74. Laplante. A.R.. Yianatos. I.B.. and Finch. I.A. (1988) On the mixing characteristics of the
collection zone in flotation columns. in K.V.S. Sastry (ed.). Column Flotation '88.
SME/AIME. Littleton. pp. 69-80.
75. Goodall. C.M.• and O·Connor. C.T. (1989) Residence time distribution studies of the solid
and liquid phases in a laboratory column flotation cell. in Proceedings Int. CoIl. "Develop-
ments in Froth Rotation" (Gordon's Bay. Cape Town). vol. 2. 15 pp.
76. Mills. P.I.T .• and O'Connor. C.T. (1990) The modelling of liquid and solids mixing in a
flotation column. Min. Engng. 3. 567-576.
77. Tinge. I.T .• and Drikenburg. A.A.H. (1986) The influence of slight departures from vertical
alignment on liquid dispersion and gas holdup in a bubble column. Chern. Eng. Sci. 41.
165-169.
78. Rice. R.G .• and Littlefield. M.A. (1987) Dispersion coefficients for ideal bubbly flow in truly
vertical bubble columns, Chern. Eng. Sci. 42, 2045-2053.
79. Mavros. P.• Lazaridis. N.K .• and Matis. K.A. (1989) A study and modelling of liquid-phase
mixing in aflotation column. Int. J. Miner. Process. 22. 1-16.
80. Zehner. P. (1986) Momentum, mass and heat transfer in bubble columns. Part 1. Flow model
of the bubble column and liquid velocities. Int. Chern. Eng. 26. 22-28.
81. Joshi. J.B .. and Sharma, M.M. (1979) A circulation cell model for bubble columns, Trans.
I. Chern. E. Yl. 244-251.
2.3. CASE STUDIES OF FLOTATION CONTROL

D.J.McKEE
Julius Kruttschnitt Mineral Research Centre,
Isles Road
Indooroopilly, Qld 4068
Australia

ABSTRACT. Significant progress has been made in the development of automatic control systems
of flotation circuits since the first availability of on stream analysis data approximately 20 years
ago. A wide range of approaches has been tested, from simple feedback control loops to advanced
model based control strategies. Flotation control systems can be considered as either stabilising or
optimising in nature.
Flotation control has proved to be a difficult task, principally because the process is subjected to
disturbances which are not measurable in many cases, and because of inconsistent process
response to disturbances. There has also been considerable confusion over the formation of
realistic objectives for control systems.
An approach has emerged from the experience gained from many studies. It is crucially
important to specify realistic control objectives at the outset. In almost every instance, control
should commence with stabilising loops. Only after stable circuit operation has been achieved
should the next goal of process optimisation be attempted. The task of achieving effective
stabilising control can be quite challenging, and a range of techniques has been developed. The
most recent of these recognizes the role and skill of the flotation operator via an expert systems
approach.
The Julius Kruttschnitt Mineral Research Centre has been involved in the development of
flotation control systems for 15 years. The paper will present accumulated experience of the Centre
by outlining a selection of case studies which demonstrate different approaches to flotation control.

1. Introduction

In any assessment of recent developments in flotation technology, the topic of flotation control is
worthy of consideration. This is because the process is frequently subjected to a variety of
disturbances, which left unchecked, result in poor metallurgical performance of the circuit.
Traditional manual control methods, based on infrequent assay information and the generally
subjective observations of operators, were all that were available until the advent of a range of
instrumentation. particularly on stream analysis (OSA). approximately 20 years ago.
The operational and economic incentives offered by automatic control have resulted in numerous
flotation control studies over the last 20 years. The term studies. rather than completed control

235
P. Mavros and K. A. Matis (eds.},Innovations in Flota/ion Technology, 235-262.
© 1992 Kluwer Academic Publishers.
236

systems, has been used because the success rate, in terms of proven, lasting systems has not been
high. It has slowly been realised that flotation control is a difficult problem. What has emerged is
an approach to the development of flotation control systems. In this paper, aspects of that approach
are presented. A series of case studies of systems developed over many years by researchers at the
JKMRC follow. The studies, which apply to circuits containing conventional cells, have been
chosen to illustrate a range of approaches which have had varying degrees of success.

SCAVENGER

OSA

CLEANER

RECLEANFR OSA

Figure 1. Mount Isa Concentrator flotation circuit in 1976 showing OSA points.
 237

2. Early studies

The early flotation control trials carried out by the JKMRC have been fully described [1]. The trials
were successively carried out in the Copper Concentrator of Mount Isa Mines Limited, the lead
circuit of North Broken Hill Limited and the zinc circuit of Electrolytic Zinc Co., Rosebery. While
it is not the intention to describe the systems here, a number of important observations can be made
from these trials.
The Mount Isa Copper Concentrator was extensively equipped with OSA capability as indicated
in Figure 1. All reagent additions could be controlled from a computer. In addition, computer
control of the air additions was available to the initial sections of the rougher banks, the recleaner
bank, and parts of the cleaner and retreatrnent banks. Thus, there was a wide range of control
possibilities, with some degree of control feasible in all sections of the circuit.
In contrast, the instrumentation available at both North Broken Hill and Rosebery was limited
and from the outset it was recognised that control objectives in both these plants would have to be
more modest than at Mount Isa. In retrospect these limitations were an advantage, as attention was
focussed on defining very precisely control objectives which were within the scope of the
hardware. Because the hardware was available at Mount Isa "to control the whole circuit", the
tendency was to be less rigorous in specifying control objectives.
The issue of control system objectives has emerged as a critical one in flotation control work.
General objectives such as "stabilize the circuit" in fact mean very little. It is necessary to be quite
explicit in how this is to be done, e.g., control a tailing assay to less than a maximum value.
Secondly, objectives must be achievable, particularly in the metallurgical sense. This criteria means
that those specifying control loops must understand the metallurgical performance and
characteristics of the plant to a very precise degree.
Referring back to the Mount Isa system, the initial work was directed at controlling the rougher
xanthate addition. The reasons for this were reported by McKee et al. [2] as being:
- Indications from testwork that xanthate additions were frequently excessive.
- Adverse effect of excess xanthate on the cleaner circuit.
Frequently observed unstable grade and recovery performance of the roughers.
The OSA assays then available allowed calculation of the copper recovery within the first four
rougher cells and plant testwork indicated that the four cell recovery was very sensitive to xanthate
addition. A strategy evolved which provided for collector control to maintain a four cell recovery
set point selected by metallurgical personnel. When the system was under development and close
attention by technical personnel, this control strategy was effective and did result in reduced
xanthate consumption.
A problem with the initial system was the difficulty in selecting a suitable four cell recovery set
point, which in practice was found to move over the range 45-75% recovery of copper. Acceptable
total rougher recovery was of course far more important than a reduction in collector consumption.
The first control objective of attaining a set point for cell recovery was too imprecise.
This deficiency was quickly recognised and the strategy was modified so that the control
objective became the attainment of a specified rougher tail assay, as described by Lynch et al. [1].
While this modification removed the requirement to specify a four cell recovery set point, it only
transferred the problem to specification of a rougher tail set point. While such a task appears simple
initially, it is not necessarily the case. The attainable rougher tail is a value which is dependent on a
range of variables including head grade, grind, throughput and inherent metallurgical response.
There is no one fixed value for the tail set point. The difficulty in selecting a suitable tail set point
became a significant problem with the strategy.
238

Thus, for control purposes, the objective of a rougher tail set point is still not sufficiently
precise, and further, it does not recognize the metallurgical requirements of the total circuit. The
overall metallurgical objective of the Copper Concentrator was the attainment of maximum recovery
at a specified concentrate grade. A recovery-only rougher control objective is likely to be
incompatible with the total circuit recovery, as the rougher strategy may result in too Iowa rougher
concentrate grade to produce the required final grade target. Thus, the rougher recovery strategy
had two shortcomings:
- It was not compatible with the circuit metallurgical objective.
- It was difficult to specify an achievable rougher tail target regardless of grade constraints.
At North Broken Hill, the only means of control available was via computer manipulated
concentrate diverters installed on the single cleaner bank. At the time of the work, OSA information
was limited to the feed, concentrate and tailing from the circuit, and thus the control strategy
possibilities were restricted accordingly. However, it was possible to specify one control objective
which was fundamental to the metallurgical operation of the circuit, and which was achievable. The
operating requirement was to produce a lead concentrate containing a maximum of 3% Zn, as
measured by OSA. A simple single loop strategy was devised to satisfy this objective. While the
work did suggest refinements and additions, the basic strategy was capable of performing very
satisfactorily.
A similar situation existed in the zinc circuit of the Rosebery Cu-Pb-Zn concentrator, which had
diverters, installed on the final zinc cleaner bank. An OSA unit was installed on the zinc final
concentrate. The control objective was again explicit and achievable - attain a desired zinc content in
the concentrate. This objective was simply obtained by controlling the number of cells producing
final concentrate. The system had limitations, and it became apparent that attaining a set final
concentrate target was not possible if the grade of the rougher concentrate decreased excessively.
At the time of the work, there was no method available to control rougher grade, so a simple on-
line system was devised to adjust the final grade set point if the rougher grade altered substantially.
The North Broken Hill and Rosebery studies were important because they demonstrated that
useful and simple control systems could be developed, which were based on a minimum of
instrumentation and control hardware. The simplicity of the strategies was one of their major
strengths. The work also established the importance of strategies which were precise and set points
which could be precisely defined and were realistic in a metallurgical sense.
One consistent feature of JKMRC flotation control studies has been the emphasis on defining the
response of each flotation circuit to systematic changes in operating variables. In the case of Mount
Isa, this involved examining the effect of all possible control variables. For both North Broken Hill
and Rosebery, such work was more limited because of the practical realities of variables available
for control.
Based on the results of these early flotation trials, a general approach to the task of t1otation
control was formulated, and this approach is discussed in the next section.

3. An Approach to Developing Flotation Control Systems

A structured approach has been proposed for the development of automatic control systems for
t1otation circuits [3]. In order, the factors to be considered are listed below:
- Define the precise metallurgical objectives of the circuit.
- Isolate and understand the constraints present in the circuit.
Develop at least an initial understanding of the types of disturbances to be encountered.
 239

- Only after a thorough examination of the above factors. formulate an initial set of objectives for
the desired control system.
- Decide on the general control approach - stabilising. set point control or optimising.
- Finally. select a control approach and commence the development of strategies.
Some of the above points require amplification.

3. 1. METALLURGICAL OBJECTIVES OF THE CIRCUIT

A control system will not be successful if it does not suit the metallurgical objectives of the
particular circuit. For example. a strategy which seeks recovery maximisation as its objective will
not succeed if the real metallurgical requirement is to meet grade targets. It can be surprisingly
difficult to define in precise terms the real metallurgical objectives of a flotation operation. but
understanding these objectives is an essential prerequisite in formulating a suitable control system.

3.2. CIRCUIT CONSTRAINTS

Constraints. in the form of mechanical problems in a circuit. lack of instruments or imposed

operating requirements. can all modify what is practically possible with a control system. In many
cases. the constraints may be so severe that attempting to develop a control system is unrealistic
before the deficiencies are rectified.

3.3. TYPES OF DISTURBANCES

The principal reason for a control system is to combat the effects of process disturbances. It seems
self evident to state that at least the major disturbances should be identified before designing a
control system. but experience indicates that this is frequently not the case. A thorough understan-
ding of the types and effects of process disturbances may be extremely difficult to obtain and in a
practical sense. cannot be achicved until a reasonable amount of actual operating experience has
been obtained. This is one of the principal reasons why it is difficult. if not impossible. to design
flotation control strategies at the flowsheet stage.

3.4. INITIAL CONTROL OBJECTIVES

Experience has taught that the best approach in flotation control is to begin with simple loops and
gradually expand to more comprehensive (and complicated) strategies. It should also be realised
that despite careful consideration of the earlier mentioned factors. it is still possible to select
inappropriate control strategies. Early modification of strategies is often required.

3.5. GENERAL CONTROL APPROACH

Three control approaches were nominated: stabilising. set point and optimising. The purpose of
stabilising control is exactly what the meaning implies - the provision of loops which will maintain
selected process variables. usually within limits known to be realistic for most operating
conditions. Stabilising control aims to eliminate the large fluctuations in process behaviour which
often result from process disturbances. Set point control is merely a refinement of stabilising
control. where process variable set points are substituted for limits. The objective is essentially
stabilisation. but with the added refinement and complexity of seeking to maintain a process
240

FLOTATION SCAVENGER FINAL ~

FEED TAIL

REGRIND

CLEANER

FINAL TAIL
RECLEANER

FINAL
Figure 2. BougainviUe Copper Flotation Circuit in 1984. CONCENIRA1E
 241

variable at a desired set point. The challenge now is to be able to nominate realistic set points and,
most importantly, to recognize when set points must be altered and by how much. Many control
systems have foundered on this difficulty. Finally, optimising control aims to run a circuit so as to
achieve optimum performance, usually in terms of some economic criteria. In this case, the task of
the optimising system is to determine set points for the control loops.

3.6. SELECT A CONTROL APPROACH

There now appears to be reasonable agreement that the first control approach should be aimed at
stabilising a flotation circuit. This is certainly the experience of the JKMRC and at the Pyhasalmi
operation [4] which has the longest record of flotation control system development. The focus of
research in recent years has been on exploring ways of achieving stable operation in the light of
three major difficulties: recognising when changes in ore type (the most difficult of process
disturbances) occur; developing robust control loops which can accommodate long process delays
and variable process responses, and finally developing methods of selecting appropriate process set
points and limits.

3.7. SUMMARY

The concepts outlined in this section will be illustrated by describing a number of case studies of
flotation control system development which have been undertaken by the Centre over the last
decade.

4. Stabilising Control of a Rougher-Scavenger Circuit

4.1. THE FLOTATION CIRCUIT

During 1983/84 extensive flotation control studies were conducted at the now dormant
Bougainville Copper Ltd Concentrator. The Bougainville operation at the time of the work was a
large tonnage (approximately 130000 tpd), low grade (0.45% Cu) porphyry copper operation. The
operating strategy was one of maintaining a fixed maximum tonnage through single stage ball mill-
cyclone circuits which supplied the flotation circuit shown in Figure 2. Final concentrate grades
were maintained in the region of 28-30% Cu.

4.2. PRELIMINARY PLANNING FOR A CONTROL SYSTEM

The flotation circuit in 1982 was equipped with a Courier 300 OSA system, but with little other
instrumentation. However, as a result of a capacity expansion three new rougher-scavenger banks
were being installed and the decision was taken to instrument these banks comprehensively. The
operation was anxious to undertake some flotation control work and for practical reasons, the study
had to be directed at the new rougher-scavenger banks.
From a control objective consideration, there were a number of important factors:
- The constant tonnage philosophy resulted in a coarse and varying grind which directly affected
recovery, but the grinding philosophy meant that nothing could be done about this source of
recovery variation.
242

8 32
~

~
31 • ••
g 30
29
• ••••
•
• •
•• •• •
••
cJ
Z • • • • • •••
••••• •
0
U
28 •• • • •
• •••••••••• •

~
27
•
26
25 •
• •
13141516171819202122232425
ROUGHER CONe. GRADE %Cu
Figure 3. Relation between rougher and final concentrate grades at Bougainville Copper.

- It was established that rougher-scavenger recovery was principally dependent on grind size and
was not particularly sensitive to collector addition.
- The major flotation objective of obtaining a required concentrate grade was also not a direct
target for flotation control because there were no variables for manipulation in the cleaners.
Thus, most ofthc obvious control objectives were unavailable. Only two possibilities remained in
the rougher-scavenger area namely:
- Increase rougher-scavenger grade.
- Stabilising rougher-scavenger concentrate flowrates.
Froth depths in the scavenger cells were always low (2 cm typically) and frequently resulted in
large volumes of pulp reporting to the concentrate, suggesting volume flow control may be
beneficial. Secondly, historical OSA data indicated a general relationship between rougher and final
concentrate grade, as indicated in Figure 3. These data provided some incentive to investigate the
possibility of a control strategy which increased rougher concentrate grade [5].

4.3. DEVELOPMENT OF A CONTROL SYSTEM

Three parallel OK38 banks were instrumented as indicated in Figure 4. OSA measurements for the
feed, two concentrates and scavenger tail were for the combined three banks, as was the scavenger
volume flow. Level and aeration rate were individually measured and controlled to each group of
three cells in each bank (9 points in all for both variables).
A series of response tests was conducted to determine the effect of varying aeration rate and pulp
level to the rougher and scavenger banks. Figures 5 and 6 both indicate that while rougher grade
could be manipulated by changing the pulp level and froth depth, there was little effect on the
scavenger tail, or bank recovery. Similar response tests in the scavengers, Figures 7 and 8,
 243

WT - WEIGHT TRANSMITIER
XC - COLLECI'OR CON1ROL
FC - FROTHER CONTROL
OSA - COPPER ASSSAY MEASURMENT
LC - PULP LEVEL CONTROL
AC - AIR CONTROL
pH - pH MEASUREMENT
FT - FLOWRA TE TRANSMITIER

AC AC

ROUGHER SCAVENGER SCAVENGER

Figure 4. Instrumentation on Bougainville Copper rougher-scavenger banks.

showed a sensitivity of the scavenger concentrate assay and the associated volume flow, with little
effect on the scavenger tail. A particularly important observation during normal manual operation
of the scavengers was a clear link between variations in scavenger concentrate volume flowrate and
downstream upsets, as recorded by OSA, in the cleaner section.
Based on detailed observations from the testwork, the following control strategy was specified:
- Control the pulp level in the rougher cells to obtain a set point rougher concentrate grade. Pulp
level was selected as the control variable, rather than air, because in practice it was found
desirable to tune the aeration rate manually to each individual cell mechanism.
- Maintain the scavenger concentrate volume flow between limits by controlling the aeration rate.
Aeration rate was used because the narrow range of pulp level available in the scavengers made
adjustment of level difficult in practice.
244

0'5~
0.4~---",
AirFlowRate "-- .. • • • ....
0.3.L-------~::::=~~-

21
Concentrate Assay
20
8
~ 19

18~~-~~~--+---------------------

0.5~

I ~.
• •

0.4~-------------------------------
3
Scavenger Concentrate Assay

~O.08~
0.07~--------------------------------
TIME
(5 minute intervals)

Figure 5. Effect of varying air flow to the roughers at Bougainville Copper.

The final rougher strategy is shown in block diagram fonn in Figure 9. The strategy evolved
over some months, and the important addition in that time was an on line calculator for modifying
the rougher grade set point. It was known that the attainable rougher grade was a function of the
head grade, and a simple calculation based on (historical) relationships was used to alter the set
point in response to major head grade movements. The operator was also able to adjust the set
point at any time.
 245

21

19

17 Concentrate
Assay
15

13L-----------------------

0.4~_
~-..·F-e·:-d...
0.3~
;;::y
_________________________________

3
Scavenger Concentrate
Assay

0.1 Scavenger
Tail Assay

0.09 "-----...::::=---------
TIME
(5 minute intervals)

Figure 6. Effect of varying rougher pulp level at Bougainville Copper.

The control strategy in the scavenger banks was simply to maintain the scavenger concentrate
volume flowrate between nominated limits. The flow limits were selected on the basis of operating
experience. The scavenger aeration rate was selected as the control variable, with the scavenger
 .,
246

0'1
0

....
,•
II>
Air Flow Rate
rI:l
• •
'"II> 0.7
~
!:: ..... • .....
0.6
•
5 Scavenger Concentrate
Assay
4

3
~ 2

70
Scavenger Concentrate
60 Volume Flow

50

= 40
....~
'"II> 30
!::
~ 20

10

0.5L Feed Assay

~ O.4~ ......... ~r
0.7 +-a.......""'"

0.6 ....- - - - - - - - - - - - -
TIME
(5 minute intervals)

Figure 7. Effect of varying air flow to the scavengers at Bougainville Copper.

 247

0..
;J

::l . . ..
~
0 Pulp Level
a3
~
• • ~ • •
S • • /
u
20

2 Scavenger
8 Concentrate Assay
~
1.

70
Scavenger Concentrate
~
60

--t:'"
~ Volume Flow
11) 50
:.:3 40

Feed Assay

°0·.5 1t-____
4
=~=:::I::::::_::!t!::.wL'vA
J__.::
1

0.07 Scavenger Tail Assay

8
~
O.06L---------------------------

TIME
(5 minute intervals)

Figure 8. Effect of varying scavenger pulp level at Bougainville Copper.

248

I OOMPUlllR

I
I
I
I
L

Figure 9. The BougainviUe rougher control system.

level maintained at a fixed value. Because of the very low scavenger froth depths, it was far more
practical to bring the concentrate flow back within limits by small adjustments to the aeration rate.

4.4 SYSTEM PERFORMANCE

An evaluation of the system (for many months) indicated that the rougher banks under automatic
control did operate in a more stable marmer, with improvement in concentrate grade without loss of
recovery. A comparison of the controlled banks with others under manual operation is shown in
Table 1. It should be noted that the uncontrolled banks were of a different type.

TABLE 1 Comparatlve manual and automatic contro resu ts.

Manual Automatic
Feed Grade (% Cu) 0.395 0.395
Rougher concentrate (% Cu) 14.87 17.83
Rougher-scavenger recovery 80.8% 80.7%

Extended evaluation tended to support the above results, although it was noted that after the initial
success of the automatic systems, operators tended to mimic the controlled banks, with resultant
improvements to all operation. The best indicator of the success of the control was that the system
was extended throughout the entire rougher-scavenger circuit.
 249

NEW FEED

REGRIND

e
Legend

Fe, Sn assays

e Mass Flow

FINAL SULPHIDE CONCENTRATE

Figure 10. Initial sulphide flotation circuit at Renison Ltd.

Performance of the scavenger volume flow control was less successful because of problems in
measurement of the concentrate flow rate and interaction between the pulp level and aeration set
points. Selection of appropriate flow limits could also cause problems.
The Bougainville Copper Ltd experience indicated that a simple stabilising control system could
be developed after careful consideration of the metallurgical objectives of the circuit and the quite
severe constraints imposed by operating requirements and control hardware availability. The
problems to emerge from the study were associated with the following:
Specification of set points under some conditions.
- Problems in accurate volume flow measurement of concentrates.
"!l
1iQ.
~
=
.. Rougher/Scavenger o
.... Scavenger Air Flow (%) Concentrate Flow (tph) Scavenger Tail % Sulphur (OSA)
i"'
tr1 25 30 35 40 45 50 80 85 90 95 100 6.0 6.5 7.0
::;ll , , , , I , , ,
7.5

0
a...,
~s·
Qq
N
I»
=;. I~
~
~
0
50
<>
~
w
~ 18
Qq
=
~ >-l
!!! s·
n
f.
~
? w
I:;;

w
I~
 2S1

- Interaction between control variables (level and aeration in this case).

5. Stabilising Control Using Circulating Loads

5.1. THE FLOTATION CIRCUIT

The lKMRC developed a control system at the Renison Ltd sulphide flotation circuit during
1983/84 [6]. Renison treats a sulphide ore (pyrrhotite and pyrite) which contains 1.5% tin present
as cassiterite. After dense medium separation and grinding, the sulphides are floated to a waste
concentrate prior to the recovery of cassiterite by gravity concentration and flotation. The focus of
process control attention was the sulphide flotation circuit, shown in its 1983 form in Figure 10.
The objective of the circuit is to maximize sulphide recovery to the concentrate while maintaining
the tin content of the concentrate below upper limits. Any tin reporting to the concentrate is lost.
Conversely, high sulphide contents in the rougher tail adversely affect the subsequent gravity
circuit.
In practice, operators aimed to maintain the sulphide content of the rougher tail (inferred from an
iron OSA value) below stated limits, while not permitting tin in the final concentrate to exceed an
upper limit. Under manual control, operators directed most of their attention to the roughers,
making use of aeration rate adjustments. The Renison circuit is unusual in that over 60% of the
mass feed to the roughers is recovered as rougher concentrate. This explains why aeration rate is
an effective control variable, as it directly affects the mass recovery of concentrate. In contrast,
collector addition was less effective for control purposes.
The large mass recovery in the roughers resulted in significant amounts of fine liberated tin and
caSSiterite-sulphide composites reporting to the cleaners. The regrind was positioned to liberate
cassiterite from the composites and the cleaners rejected cassiterite to the tails, for entry back to the
rougher feed via a short residence time thickener.

5.2. CONTROL OBJECfIVES AND INITIAL STUDIES

Initial work sought to control the sulphide level in the rougher tail to a set point value by
manipulating the rougher aeration rate. A simple feedback loop was installed following a series of
response tests which showed the tail assay to be sensitive to the aeration rate, as indicated in
Figure II. The strategy appeared to work effectively. However, it operated for a very short time
because the circuit was to undergo expansion (increased throughput) and a new process control
computer system was to be installed.

5.3. THE REVISED CIRCUIT AND CONTROL STRATEGY DEVELOPMENT

The revised Renison circuit is shown in Figure 12. Changes included the addition of a scavenger
bank, a third cleaning stage and relocation of the regrind. Despite these changes, the operating
objectives of the circuit were unaltered, and it was felt that a similar control approach would be
effective.
A control loop, implemented on the aeration rate to the scavenger bank after testing indicated that
the scavenger tail sulphide assay was more sensitive to scavenger than rougher aeration rate.
Performance was not encouraging, as indicated in Figure 13. It became clear that low tail values
could only be maintained with increasingly high air rates. Investigation indicated that increases in
252

NEW FIlED

TO GRAVITY CIRCUIT

Legend

@ Fe, Sn assays

e Mass How

ANAL SULPHIDE CONCENTRATE

Figure 12. Revised sulphide flotation circuit at Renison Ltd.

the scavenger air increased the volume flow to the cleaners, and with no adjustment to cleaner
aeration rates, the load recycling from the cleaner tails back to the thickener and then the roughers
increased rapidly. A compounding effect occurred, with a need for still greater scavenger air.
Clearly, the magnitude of the circulating load was causing problems and attention was diverted to
the cleaners.
Response testing using aeration rate in the first cleaner, Figure 14, indicated the circulating load
effect. A decrease in cleaner aeration rate rapidly resulted in an increased circulating load and a
steadily increasing rougher tail assay. Return of the cleaner air to its previous position did not
return the circuit to the starting condition, but rather the circulating load persisted.
A revised control strategy was thus developed to control the circulating load, as shown in
 253

0
'"

~
c
.g
:a
-
'"<: ..,
0

-"
:;;:
OJ)
c
""
>
"
til

0
N

~ Original Selpoinl

SClpoinl Change t Setpoint Change

, '00
1 IQQ 12-
Time

Figure 13. Performance of scavenger air loops showing need for continuing increases in air addition to
maintain low tail assays.

Figure 15. Operators specificd two set points, the thickener underflow mass flow and density
values, which were understood to be important operating parameters. The first cleaner air was
adjusted to maintain the mass flow value.
The strategy was basically successful, but it underwent some modification. It was found
necessary to adjust the aeration rates to all three cleaner banks in unison. Also, it was found that
the density of the cleaner tail was a useful early indicator of an increasing circulating load. Finally,
254

<- .c

- ....
CIl.--..
8~
.0; '-'

f-< "
.... -§.
<)-
ih"
"CIl
0
~ "

.--.. ;;
..c::
§ ,...

---
~

~
0
G: ;::;
~ :::
0
u
:> :::
~
u
f!? :;;
0
~

~
0
12:2
.,a. -

---
"g'-'
::J ~ :;;
.... 0
.,t;::
;;;

--
C '"
<)'"
~ ~
.~ ::E
~
.....
F: ~

....<)
\ I
~~
~~
O~
'el< .,..
ti: '"

I i i i
00
12- 1 2~ 13Q!! 13~ 14QQ 14~
Time

Figure 14. Effect of varying cleaner air additions on circulating load and subsequent rougher and
scavenger performance.
 255

Thickener Underflow Thickener Underflow

Massflow (tph) Setpoint Density Setpoint

Calculated Thickener
Underflow Volume Flow Setpoint

Mass Controller: Volume Flow Controller:

Compares setpoint with actual Controls volume flow by
massflow and cascades a setpoint manipulation of thickener
to the First Cleaner Air Controller. underflow pump speed.

Figure 15. Circulating Load Control Strategy at Renison Ltd.

it was necessary to provide a form of dynamic compensation to the controllers to compensate for
the time delays present in the system. The final form of the mass flow controller is shown in
Figure 16.

5.4. SYSTEM PERFORMANCE

The circulating load control strategy proved to be successful at stabilising the sulphide circuit.
Earlier plans to reintroduce the scavenger aeration control were not implemented because it was
found that the control was not needed. A crucial factor in the success of the system was the
deliberate involvement of operators to interact with the system. In those days before expert
systems, the control system printed a series of messages advising operators when limits had been
reached. Final selection and modification of set points was in the hands of operators. The system
did not solve the problem of altering set points, but rather it directly sought to involve operators in
that role. The Renison system, which is operating effectively seven years after installation,
demonstrated conclusively the importance of circulating load control in stabilising circuits, and the
need for operator involvement.
 256

Thickener underflow
massflow setpoint

Massflow controller
with dynamic compensation
" Thicke ner underflow
for time between first
cleaner tail density
CONfROTI~ ... massflo w measurement

change and thickener

underflow massflow
response

First cleaner tail d ensity setpoint

Density controller with "

dynamic compensation
for time between first CONfROTr= ....... First cleaner tail dens
measurement
cleaner air change and
first cleaner tail
density response

First cleaner air setpoint

"
First cleaner CONfROIL......, ....... First cleaner air
measurement
ai r controller

, Signal to air contr01 valve

Figure 16. Block diagram of control loops at Renison Ltd.

The Renison control system also demonstrated that a successful control strategy has to be
tailored to the circuit in question. Control in the cleaners had proved effective in the Rosebery and
North Broken Hill systems mentioned earlier, but the similarities to the Renison strategy ended at
that point. It is also apparent that many of the general concepts outlined in Section 3 were found to
be significant in the Renison study.
 257

6. Improved Stabilising Control

6.1. INTRODUCTION

The preceding discussion on the Renison control system made reference to the need to provide
some form of dynamic compensation to standard proportional-integral (PI) control algorithms to
account for long time delays between changing a variable and the subsequent process response to
the change. This is only one of the problems involved in obtaining good control response in a
flotation system. A more serious problem is caused by a variability in the magnitude of process
response to a change in a control variable. For example, different ore types may require quite
different collector additions, and a loop which provides adequate response for one ore type may be
quite inappropriately tuned for another ore type. In this case, some form of self tuning controller is
required. Such controllers must be based on process models.

o Lead final concentrate grade (%Pb)

Predicted concentrate grade (%Pb)
55
"" 3rd cleaner air (m**3/sec) *100

50

45

40

35

30 L -__ ~ ____ ~ __ ~ ____ ~ ________ ~ ____ ~ __ ~~ __ ~ ____ ~ __ ~

3 5 7 9 II 13 15 17 19 2I 24

Time (7 minute intervals)

Figure 17. Actual and predicted lead concentrate grades in the Mount lsa lead circuit.

There is an increasing usage of model-based control approaches being applied in flotation

systems [7,8]. An example of the use of adaptive control of a cleaner circuit is summarised below.
258

6.2. AN ADAPfIVE CONTROL SYSTEM

Improved control response requires more accurate prediction of the behaviour of process outputs
(the variables to be controlled). Work associated with the lead cleaner section of the Lead-Zinc
Concentrator at Mount Isa Mines has indicated that a simple model of the following form was able
to predict lead grade one time step ahead [9].

y(t+I) = aoy(t) + aly(t-I) + bou(t) + blu(t-I) (1)

where
- yet) = Pb grade at time (t) - Pb grade at time (t-I)
- u(t) = Deaner air at time (t) - cleaner air at time (t-I)
- ai, bi =model parameters
- time interval = OSA interval.
Least squares estimation is used to calculate values of the model parameters at each time interval.
Agreement between the actual and predicted grades is shown in Figure 17. Other terms (process
inputs) may be added to eqn. I, and for the purposes of controlling the final concentrate grade
using the cleaner aeration rate, the lead cleaner feed grade was added as follows:

yet) = ao y(t-I) + al y(t-2) + bo u(t-2) + b l u(t-3) + Co v(t-3) (2)

where
yet) = set point - Pb grade at time t
- u(t) = air reference - Deaner air at time t
- vet) =feed reference - Deaner feed Pb at time t
The inclusion of the cleaner feed grade term adds a degree of feed forward to the prediction.
Given the ability to predict the concentrate grade, it follows that an actual value of the cleaner air
u(t), may be calculated which should minimize the error between the grade sct point and the actual
grade. In grade terms, the following calculations are performed.
- predict y(t+k+ I), i.e. the grade at t+k+ 1;
- establish an objective function J where

J =[ i
(y(t+k+l) - sp (t+k+I»)2 + L{ U~2 ) ] (3)

where: YsP is the grade set point, and u(t) the required air addition rate;
minimize J by differentiating with respect to u(t);
- set to zero and solve for u(t).
Model parameters are recalculated at each sampling interval and an updated value of the required
air addition is also provided. This air value then becomes the set point for a PI controller
manipulating the air addition.
The adaptive controller scheme for the cleaner banks is shown in Figure 18 and the response to a
quite severe process upset (a partial circuit feed shut down) is shown in Figure 19. Such
performance would not be possible without a model based control system. It has been reported
[10], that adaptive control techniques have been widely applied in the Mount Isa Lead-Zinc
concentrator with considerable success and good operating time. The important point to remember
here is that the objective is still stabilising control but with a much more robust control.
 259

I'I---~----~---mw----:=~~
G- AVERAGE

--~~----~~~--~----~~ 8 Pb - LEADASSAY
MEASUREMENT

o - AIR MEASUREMENT
AND CONTROL

SET POINT

SELF TUNING
I----<~ <DNm.OUllR 1----...
r-L----------I AIR SETPOINT
(BANKS I, 2 and 3)

Figure 18_ Self-tuning control system in lead circuit at Mt. lsa.

7. Expert Systems in Flotation Control

7.1. A STUDY

Reports on the application of expert systems for flotation control are now appearing in the literature
[11,12]. The Centre has been involved in only one expert system application to date, but the
experience gained does permit some useful observations about the potential of the technique.
A prototype expert system application for the flotation circuit of Mount Isa Mines Copper
Concentrator [13] was developed during 1988/89. The circuit is still similar to that shown in
Figure I, the only difference being the addition of columns to upgrade the retreatment concentrate.
The circuit has produced excellent metallurgical results (95%+ recovery at concentrate grades of
26% Co from a 3.5% Co head). Operating characteristics are generally well known. The circuit has
also been the scene of a number of flotation control system developments over 15 years.
The objective of the project was to develop a prototype expert system which mimicked the best
operators, and to gain experience with expert systems technology. The intention of the initial work
was not to develop a total control system.
Considerable effort was devoted to the task of developing an understanding of the way in which
operators "control" the circuit. In practice, operators develop a set of target grades for critical
streams in the circuit (from OSA values) and make adjustments to try to maintain these target
grades. The adjustments made are dependent on numerous factors and most importantly, are often
dependent on the extent to which past history of the plant (covering time periods of one to two
hours) influences operator decisions.
260

o Lead final concentrate grade (%Pb)

A Lead cleaner new feed (%PB)
55
o *100

45 HALF GRINDING SHUTDOWN

40

35

30

25

20

15
3 5 7 9 111315 171921232527293133353740
Time (7 minute interval)

Figure 19. Performance of the control system during a half grinding shut down.

The expen system received on line, real time data from the circuit. Its output was advisory only,
with messages to operators suggesting courses of action in the event of unsatisfactory overall
performance, or when intermediate target grades were not being obtained. One of the principal
difficulties confronting any flotation control system, or advisory expen system, is the appropriate
selection of intermediate target values. For example, this particular system recognised that the
rougher tail target grade was dependent on feed tonnage, but it did not take into account
dependencies on feed assay, grind size and the nebulous one of ore type. Any control or advisory
system must have the facility to vary limits or targets. In the Mount Isa case, variations were
implemented by logic within the system or dircctly by thc operator.
The original expen system is currently undergoing its second major modification. The first
changes were implemented to improve many of the diagnostic rules. Currently, effort is being
directed to provide a quantitative degree to advisory messages. For example, the aim is to expand a
message such as "increase the aeration rate to the roughers", to "increase the aeration rate to the
roughers by n %". Such enhancements have been seen as necessary to improve the level of advice
to operators. It remains too early to report on the overall success of the expert system application.

7 .2. SOME GENERAL COMMENTS ON EXPERT SYSTEMS APPLIED TO FLOTATION

The following comments are impressions rather than rigorous observations.

- An objective of producing a mimic of a good operator may not be the correct one. The system is
most unlikely to be able to handle all the upset conditions which occur, and unfortunately, failure
 261

to do so tends to degrade the effectiveness of the system in the eyes of operators.

- In some respects the advantage inherent in the approach that the system communicates in
language understandable to operators appears to make the same operator highly intolerant of
incorrect or inadequate advice from the system.
- The mimic approach depends enormously on knowing the corrcct value of intermediate circuit
limits, and effective specification of the limits remains a problem.
- The largest problem in flotation control is the correct selection of set points and limits.
Techniques such as adaptive control are quite capable of maintaining flotation circuit variables at
realistic set points.
- Perhaps the crucial role for expert systems in flotation is to provide a greatly improved means of
selecting set points. The knowledge and experience of the operator is still the only effective way
of recognising the occurrence of some disturbances, and in particular of ore type changes. A
focus on developing the expert system to accept operator experience for changing set points
appears to be a most attractive possibility.
- The ability of an expert system to present information in a manner which is intelligible to
operators, surpasses that of any other approach. In the Australian industrial environment, where
the trend is increaSingly towards rotation of operators throughout a plant, with the inevitable
reduction in extremely specialist flotation operators, this capability of the expert system will be
extremely useful.

8. Conclusions

All of the work described has been concerned with achieving control by manipulating physical
variables or reagents on the basis of assays values and steam flowrates. No attempt has been made
by the IKMRC to either measure or manipulate the chemical parameters which of course are
fundamental to the performance of a flotation circuit. Such work is beginning, based on chcmical
measurements using selective electrodes [14] and this approach undoubtedly will be crucial in
future developments.
The methods outlined in this paper form the basis for a systematic approach to flotation control.
Future developments are likely to be concerned with a number of parallel activities including the
following: improved stabilising control using more robust adaptive control techniques; the greater
involvement of operators to interact with the control system using expert systems type technology;
the development of chemically based control referred to above; and the re-emergence of optimising
control studies when effective stabilising control has been obtained.

9. Acknowledgements

The studies discussed in this thesis were largely the work of a succession of postgraduate students
at the IKMRC over many years. Their contribution in developing control systems designed for the
industrial environment has been very significant.

10. References

l. Lynch, A.I., Johnson, N.W., Manlapig, E.V., and Thome, e.G. (1981) MineralandCoal
262

Flotation Circuits - Their Simulation and Control, Elsevier, Amsterdam.

2. McKee, D.J., Fewings, J.H., Manlapig, E.V., and Lynch, A.J. (1976) Computer control of
chalcopyritejlotation at Mount IsaMines Ltd., in M.C. Fuerstenau (ed.) Flotation, AIME,
New York, vol. 2, p. 994.
3. McKee, D.J. (1991) Automativeflotation control- a review of20 years of effort, Presented to
Minerals Engineering '91, Singapore.
4. Miettunen, J. (1983) The Pyhasalmi Concentrate - 13 years of computer control, Proceedings
4th IFAC Symposium on Automation in Mining, Mineral and Metal Processing, Helsinki.
5. PI avina, P. (1982) Automatic control of rougher-scavenger flotation at Bougainville Copper
Ltd., JKMRC Report No MS-12-1982 (unpublished).
6. Casey, M.G. (1984) Automatic control of circulating load in aflotation circuit, M.Eng.Sc.
Thesis, University of Queensland (unpublished).
7. Hammoude, A., and Smith, H.W. (1980) Experiments in self-tuning controlofflotation,
Proceedings 3rd IFAC Symposium in Mining, Mineral and Metal Processing, Montreal.
8. Franklin, M.W., and Hendriks, D.W. (1985) Minimum variance control offlotation circuits
at Brunswick Mining, CIM 24th Annual Conference of Metallurgists, Vancouver.
9. Thomton, A.I. (1986) The application andfuture development of adaptive control to the froth
jlotation process, PhD Thesis, University of Queensland (unpublished).
10. Thomton, A.J. (1991) Cautious adaptive control of an industrial flotation circuit, Minerals
Engineering, (in press).
11. Kallioinen, J., Kesti E., Miettunen, J., and Mullen, I. (1987) A sophisticated information
system for process control, in Copper 87, CIM-Chilean Institute of Mining Engineers, Vina
derMar.
12. Harris, C.A., and Kosick, G.A. (1988) Expert system technology at the Polaris Mine,
Canadian Mineral Processors Annual Operators' Conference, Ottawa.
13. Russell, C. (1990) Development of a real time advisory process control expert systemfor the
flotation process, M.Eng.Sc. Thesis, University of Queensland (unpublished).
14. Heimala, S., Jounela, S., and Saari M. (1988) Flotation control with mineral electrodes, in
K.S.E. Forssberg (ed.) Proceedings XVlth Int. Min. Process. Congress (Stockholm).
2.4. MINERAL SURFACE COMPLEXATION IN FLOTATION

Z.-X. SUN and W. FORSLING

Division of inorganic Chemistry
Lulel1 University of Technology
S-951 87 Luled, Sweden

ABSTRACT. The equilibrium constants of surface chemical reactions are defined to describe
mineral surface complexation. An experimental technique is presented involving potentiometric
titrations of mineral suspensions supplemented by adsorption studies, solubility and zeta potential
measurements. Experimental data are evaluated using advanced computer programs giving
equilibrium constants of relevant surface reactions. The fundamentals of mineral surface
complexation and their meaning in flotation are presented and discussed and used to interpret
surface charge, as well as surface reactions with collectors, activators and depressants. Surface
speCiation and flotation prinCiples are correlated for a sulphide (e.g. galena) and oxide mineral (e.g.
quartz) system.

1. Introduction

The study of complexation occurring at the mineral/water interface offers the challenge of
understanding how interaction between the reagents and the surface affect their behaviour and
reactivity.
Mineral flotation is a process of surface modification; floatability is governed by the
hydrophobicity of surface complexes at the mineral surface. The pH of the mineral suspension is a
key variable in flotation, which indicate the importance of the proton transfer property. Thus,
Bronsted acid-base properties of surface and solution species are important to interpret mineral
surface reaction in flotation. The surface can also act as a Lewis acid or base by reacting with
ligand as anion collectors or activators, like metal ions, which are added to facilitate mineral
floatability .
All these reactions are well known; however, up till now, published equilibrium constants
including mineral surface speciation, relevant to flotation are scarce. Such eqUilibrium constants
can exclusively offer the possibility of a quantitative modeling of the mineral surface related
reactions occurring in flotation and to interpret their mechanisms. In order to obtain equilibrium
constants of surface complexation relevant to flotation, studies of sulphide and oxide mincrals have
been our research topic in past years.

263
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 263-281.
© 1992 Kluwer Academic Publishers.
264

Bearing in mind that in a flotation pulp all reactions are not at an equilibrium state the surface
equilibrium constants anyhow can provide new infonnation to predict the flotation behaviour of
different minerals.

2. Fundamentals

2.1. HYDRATION OF THE SURFACE

When mineral particles are immersed in aqueous solution, the surface will be hydrated, i.e. water
molecules will interact with the surface. Hydration of the surface binding site is considered to be
the first surface reaction in aqueous solution. The acidity of the hydrated surface functional groups
varies with their chemical properties.
The hydration processes of quartz, apatite and sulphide minerals may be presented as follows:

>Si+ +H 0 >SiOH Hydration

>0- z = >OH

>SiOH +H+ _ >SiOH z+

- >OH K l' Protonation
>OH

>SiOH >SiO- +H+ Kz' Dcprotonation

>OH = >OH

The protonation and deprotonation of surface hydrated group will induce the surface acidity and
alkalinity. The amphoteric properties of the surface originate from these reactions. At the hydrated
quartz surface, H+ is finnly bound to the strong base-surface 0- site (c.f. Oz- + H+ = OH-, log K
~ 36 [ID, and thus the protonation and deprotonation mainly take place at the surface silicon site.
At the surface of apatite the hydration process may occur like:

>Ca+ +H 0 _ >CaOH2+ Hydration

>PO- 2 - >PO-

Due to the weaker basicity, in the hydration of >PO-, (c.f. P043- + H+ = HPOi-, log K ~ 12) the
water molecules will mainly interact with the surface calcium sites, producing a zwitterionic
surface. The protonation and deprotonation reactions are

>CaOHz+ + H+ _ >CaOH z+ Kl' Protonation

>PO- - >POH

>CaOHz+ = >CaOH+ + H+ Kz' Deprotonation

>PO- >PO-

At sulphide mineral surfaces, for instance, galena, the hydration and proton transfer reactions
will perfonn as:
 265

>PbOH Hydration
= >SH

>PbOH +H+ K)' Protonation

>SH

>PbOH = >PbOH +H+ K2 ' Deprotonation

>SH >S·

The deprotonation is supposed to take place from the >SH sites due to their higher acidity (c.f.
pKHS· < PKoH·)·
All surface processes including the acid-base reactions, adsorption, dissolution, oxidation etc.
originate from the hydrated surfaces. Furthermore these reactions will induce changes in surfacc
charge as a result of complexation of positive and negative species.
The study of surface complexation, pioneered by Schindler [2,3] and Stumm [4], was laid on
very sound thermodynamic foundations. Equilibrium constants for surface complexation involving
the surface potential, 'Ij!, are given by

zF'Ij!
log K' =log K'(int) - RTlnlO (I)

where K'(int) corresponds to a pure surface chemical reaction with no surface charge (z,F, and R
have their usual meanings).
Evaluation of equilibrium constants of surface reactions means to define the reactions involving
the surface site as a component and introduce mass balance equations and surface charge
alterations.

2.2. SURFACE BINDING SITES

As the surface binding site constitutes one component in the expression of surface complexation,
the concentration of such sites is a key parameter in the evaluation of the equilibrium constants. In
the calculations, the concentration of surface units is used, i.e.

[>PbOH 1
>SH

in the case of fluorapatite and galena respectively.

2.3. MASS BALANCES AND SURFACE CHARGE

To describe the chemical reactions occurring in solutions, we assume the presence of three-
component equilibria of the general type

(2)
266

Applying the laws of mass balance to such equilibria. the conditions for the total concentration then
give:

(3)

(4)

(5)

Where b = [Mez+l• h = [H+] and c = [Ln-]. The last term in each of the mass balance expressions
above also includes possible two-component equilibria where either p. q or r equals to O.
To describe surface reactions. a surface component. >X • has to be added. The active surface
makes it convenient to assume the presence of a four-component equilibrium of the general form

(6)

Where X denotes the hydrated surface unit ~~H (M. A representing the surface metal ion and
anions respectively).
Applying the law of mass balance to these eqUilibria. the conditions for the total concentration
then give:

(7)

(8)

(9)

where b = [Mez+]. c = [Ln-], h = [H+] and s = [> X]. However. due to the very low concentrations
of dissolved surface anions or metal ions. practical equilibria may be restricted to a three-
component system

~Spqr (10)
or
pH+ + s>X + rLn· = Hp>XsL/nr·pl - Wprs (11)

or a two-component system

~sps (12)

The total concentration of surface binding sites is

[>X] = s + L~SpqrshPb'lcrss (13)

 267

where s denotes the free concentration.

Furthennore, surface complexation involves changes in the surface potential, which influences
the potential energy, ze'lj!, of ions in an electric field. The probability of finding an ion at a
particular point is proportional to the Boltzmann factor eze'ljl/kT, which has to be accounted for in the
equilibria above.

~Spqrs = Wpqrs(inl) e-(P+Qz-nr)1jJF/RT (14)

The surface charge density (mol dm-3) can be calculated using the expression:

(15)

If the specific surface area, s (m2 go!), and the solids concentration, a (g dm- 3), are known, and
by defining the charge gradient using a constant capacitance model, we obtain:

x.'Ij!sa
To =~ (16)

where x. is the specific capacitance (C V-I m -2) of the electric double layer at the mineral/water
interface; 'Ij! (V) and F (C moP) have their usual meanings.
Evaluation of experimental data means to find the sets of (p, q, r)-values with matching
equilibrium constants giving the lowest error square sums L(Hexp-Hcalc)2. Furthennore the surface
charge calculated using these parameters must be equal to the electrostatically calculated charge, To.

2.4. THE IMPACT OF SURFACE COMPLEXATION ON MINERAL FLOTATION

The surface complexation may offer the explanations of mineral floatability in several respects.

2.4.1. Surface reactions offlotation reagents. The study of surface complexation may provide a
better understanding of collector adsorption. The species which are responsible for mineral
flotation are usually binary surface-collector or ternary surface-metal ion-collector and surface-
ligand-collector complexes. A positively charged surface complex prefers a negatively-charged
collector and vice versa. In this case, the anion collector, acting as a ligand or a cation collector,
may serve as a Lewis acid to react with a positively or negatively charged surface species and fonn
hydrophobic surface ternary complexes.
Sometimes it is necessary to introduce activators to facilitate mineral flotation; surface
complexation plays a crucial role in the activation process. Activation may occur as an ion exchange
or adsorption process. The activators are reacting with the hydrated surface and fonn new surface
complexes, which provide the necessary conditions for collectors to interact with the surface.
Depressants are introduced to enhance mineral separation. The depressants react with surface
sites to fonn hydrophilic surface species and thus prevent collectors from reacting, by introducing
an increased surface charge. Depressants may also react with solution species and fonn solution
268

complexes and destroy the surface hydrophobicity, by dissolving the surface species.
Metal ion complexation in solution with organic depressants may offer some insights into
possible surface speciation in the corresponding mineral system. For example, in the Ca2+ -Alizarin
Red S system, both solution and surface speciation comprise the complexes:

OR
\
0- 1 Ca
Ca/
'0= o

2.4.2. Controlling surface properties. The surface properties are determined by surface
complexation and the hydrophobicity of the dominating surface species controls mineral
floatability. Both hydrophobic and hydrophilic surface species may form at the mineral/water
interface, thus the balance between surface hydrophobic and hydrophilic complexes governs the
mineral floatability. Mineral flotation is, at least initially, a process of surface complexation and so
the principles of chemical equilibrium are fundamentally suitable for describing the system. But
mineral flotation is also a kinetic process and, as such, equilibrium principles are often used to
describe all the phenomena observed in flotation, but the fundamental properties are thermo-
dynamically controlled.

3. Experimental Techniques

The studies of surface complexation in flotation systems involve : determination of the

concentration of surface binding sites by potentiomctric titrations, supplemented by adsorption,
solubility measurements, zeta potential measurements, and flotation experiments [5]. Experimental
data are evaluated by computer assisted calculations.

3.1. DE1ERMINATION OF SURFACE BINDING SI1ES

The concentration of surface binding sites is a key variable for identifying the surface complexation
models. This concentration may be determined either by potentiometric titrations [6] or by calcula-
tion using the specific surface area and crystal data [7,8].
Potentiometric titrations, a classical technique in solution chemistry, can be similarly applied in
 269

the evaluation of surface complexation.

Potentiometric titrations in conjunction with adsorption and solubility measurements may
provide information concerning surface proton reactions, metal ion adsorption and ligand transfer
properties. Experimental data can be treated by computer using suitable programs. The total
concentrations of all components are known and the free concentrations are determined by ion
selective electrodes or spectrometry, which makes it possible to identify the reaction models.

3.2. ADSORPTION

Adsorption studies are aimed at confirming the reactions of solution species with surface sites.
Residual concentrations of metal or ligand ions are measured to provide data for computer
treatment.

3.3. SOLUBILITY STUDIES

Solubility studies may be used to examine the extent of surface species dissolution and to obtain a
quantitative evaluation of transport phenomena. Such investigations may give critical experimental
data for establishing the composition of important solution species.

3.4. ZETA POTENTIAL MEASUREMENTS

Zeta potential measurement can be used to evaluate surface charge. When positively charged
surface complexes dominate, the surface will be positive. However, the ~ potential is the potential
at the shear plane, and it is different from the surface potential 'Ij!. Nevertheless, without specific
adsorption, the 'Ij! and ~ will have the same sign of charge.

3.5. FLOTATION EXPERIMENTS

F1otation experiments give quantitative data for mineral floatability under specified conditions. Its
thermodynamic mechanism is interpreted on the basis of surface complexation principles.

3.6. COMPUTER ASSISTED CALCULATION

Evaluation of data according to the constant capacitance, diffuse or Stem model was performed
using the computer program FITEQL by Westall [9J.
For the determination of equilibrium constants, this procedure comprises three main steps, i.e.
data input, computation, and testing for convergence. Data input involves specification of the
stoichiometric matrix of all equilibria involved in the system, including estimations of unknown K
values, and the component concentrations (total and free). Computation includes calculation of the
eqUilibrium concentrations from the model and the residuals YR = TOTR(calc) - TOTR(Expt) for all
speCified data. Testing for convergence comprises evaluating the goodness of fit for every
estimated unknown K in an iterative process until YR is minimised. After convergence is achieved
and the objective function is minimised, the program produces a report consisting of the input data
270

(for verification). residuals. derivatives. sum of squares of residuals. optimum parameter values.
and the standard deviations associated with the optimum values.
Using the determined equilibrium constants. the species distribution can be visualised after
calculation using computer program SOLGASWATER [10]. In its composition matrix. solutes are
divided into components and complexes. The components are the basic constitutional elements of
the complexes. Every species in the system is defined before the calculations commence. A
formation constant. Kf • is input for each species together with its composition based on the
components. Moreover. the quantity of each component included in the system is stated. The
quantity can be expressed as the total phases. although sometimes it may be advantageous to state
the activity of a free component. instead of the total concentration. e.g. pH. When the eqUilibrium
composition of the system has been calculated at the specified points. the results can be extracted in
the form oftables and graphs.

3.7. PREDOMINANCE AREA DIAGRAMS

The major species in certain concentration ranges can be illustrated using predominance area
diagrams. which are calculated by means of the computer program SOLGASW ATER. Relation-
ships between the concentrations of species as a function of independent variables can be visualised
from such diagrams.

3.8. DISTRIBUTION DIAGRAMS

The distribution diagram represents the fraction of different surface species plotted against 10gC or
-log[H+] of the suspension. Information about the change of species concentration with changing
solution 10gC or -log[H+] is clearly exhibited. Surface characteristics such as acid-base and ion
exchange properties are also reflected in the distribution diagram. By accounting for all tre
reactions taking place in the system. including those occurring both at the surface and in the bulk
solution. the distribution diagram can be specifically used in applied fields. like flotation. to
interpret the contributing phenomena. From the distribution of surface species. the surface charge
can also be estimated. as this originates from surface reactions.

4. Principles of Mineral Flotation Based on Surface Complexation

The principles of mineral flotation are largely understood. due to work by Targart et al. [11].
Wark [12]. Gaudin [13]. Fuerstenau [14] and De Bruyn [15]. However the chemical theory
extensively used in flotation as proposed by Taggart et al. [11] is based on solution reactions.
while surface chemical reactions take place at the monolayer of the mineral surface.
Ion exchange theory. as developed by Wark [12] and Gaudin [13]. is based on the monolayer
adsorption of xanthate at a sulphide mineral surface. If the concentration of surface binding sites
could be determined. a surface reaction constant would be obtained.
The application of E.D.L. (electric double layer) theory in mineral flotation [14] has been
successfully used to interpret some flotation phenomena. which implies the importance of surface
 271

charge in mineral flotation.

The equilibrium constants of surface reactions are too important to be neglected, since the
surface and solution reactions take place simultaneously. The study of surface properties must also
involve solution speciation, in order to clarify the distinction between their contributions from the
two phases considered. Bearing in mind the surface charge effect, surface chemical equilibria may
be defined.
The study of surface complexation has a profound impact on the understanding of mineral
flotation. The application in flotation of sulphide minerals and the most common gangue mineral
quartz will be discussed. Furthermore, the effects of surface complexation at the apatite/water
interface will be briefly commented.

4.1. SULFIDE MINERALS

Surface complexation at the sulphide mineral/water interface was studied with galena in an inert gas
atmosphere to avoid surface oxidation. An ion exchange process between metal ions at the surface
and protons in the bulk solution was discovered. The amphoteric and complex properties of the
sulphide mineral surface were also revealed. Surface chemical reactions are considered to be
important in sulphide mineral flotation system, and the relevant equilibrium constants are listed in
Table 1; similar reactions can also be found in sphalerite/water suspensions [7).

I num constants or reactions at the PbS -H 20 mte

TABLE 1. E!qUITb· . IT;ace [5)
No. Reactions 10gK error interval (30)
1 >SPb + 2 H+ = >SH 2 + Pb2+ 10.21 0.02
2 >SPb+ H+ = >PbSH+ 7.15 0.05
3 >SPb+H 2O = >SPbOH·+ W -10.2 0.09
4 >SPb + HS· = >SPbSH- 16.03 0.02
5 >SPb + HS· = >SPbS 2· + H+ 5.54 0.18
6 >SPb+ X· = >SPbX· 4.6 0.2
Here >SPb or >PbS represent the hydrated surface ~~H

It was found that under acidic and neutral pH conditions, the ion exchange eqUilibrium is the
dominating reaction, which controls the surface properties. Therefore the surface charge is
balanced by this reaction. At alkaline pH, however, the surface becomes more negative due to the
hydrolytic reactions of surface lead ions. Using all available eqUilibrium constants and including
both surface and solution reactions, a distribution diagram can be constructed according to the
results of the computer simulation. The mineral-galena, collector-xanthate, depressant-hydrogen
sulphide, proton and lead ions are considered to constitute the composition components of the
suspension (see Table 2) and the computer program SOLGASW ATER is used for the calculations.
The overall equilibria of the system determine the stability of surface complexes. Hydrophobic
complex enhance mineral flotation, otherwise depression occurs.
272

T ABLE 2. SspecIes consIdered III the Pb2+, H+, >SPb, HS- and X- system.
No. Species LogKf Stoichiometry matrix Ref.
1 PlJ2+ 0.0 1.0 0.0 0.0 0.0 0.0
2 H+ 0.0 0.0 1.0 0.0 0.0 0.0
3 >SPb 0.0 0.0 0.0 1.0 0.0 0.0
4 X- 0.0 0.0 0.0 0.0 1.0 0.0
5 HS- 0.0 0.0 0.0 0.0 0.0 1.0
6 >SPbOH- -10.21 0.0 -1.0 1.0 0.0 0.0 5
7 >SPbH+ 7.15 0.0 1.0 1.0 0.0 0.0 5
8 >SH2 10.21 -1.0 2.0 1.0 0.0 0.0 5
9 >SPbX- 4.6 0.0 0.0 1.0 1.0 0.0 5
10 >SPbSH- 16.03 0.0 0.0 1.0 1.0 0.0 5
11 >SPbS 2- 5.54 0.0 -1.0 1.0 1.0 0.0 5
12 HX 1.54 0.0 1.0 0.0 1.0 0.0 16
13 PbOH+ -7.7 1.0 -1.0 0.0 0.0 0.0 17
14 Pb(OH)z(~ -17.1 1.0 -2.0 0.0 0.0 0.0 17
15 Pb(OHh- -28.3 1.0 -3.0 0.0 0.0 0.0 17
16 Pbz(OH)3+ -6.4 2.0 -1.0 0.0 0.0 0.0 17
17 Pb:3(OH)42+ -23.9 3.0 -4.0 0.0 0.0 0.0 17
18 Pb4(OH)44+ -20.9 4.0 -4.0 0.0 0.0 0.0 17
19 Pl¥,(OH)g4+ -43.6 6.0 -8.0 0.0 0.0 0.0 17
20 S2- -13.9 0.0 -1.0 0.0 1.0 0.0 17
21 H2S(aq) 7.02 0.0 1.0 0.0 1.0 0.0 17
22 PbS(S) 13.6 1.0 -1.0 0.0 1.0 0.0 17
23 PbX 2(S) 16.7 1.0 -1.0 0.0 1.0 0.0 18
Here > represents a surface spccles.

The surface species distribution as a function of xanthate concentration in the absence of HS - is

visualised in Figure I, for pH 10 (a common value in flotation). As the xanthate concentration
increases, >SPbX- starts to dominate, the key reaction which leads to mineral flotation may be :

>SPb + X- = >SPbX- 4.6±O.2 (30) (17)

Thus the >SPbX- is considered to be the surface species responsible for flotation.
In the presence of HS-, however, the surface species >SPbX- decreases with increasing HS-
concentration (Figure 2), which may explain the depressive role of HS- in sulphide mineral
flotation. The reaction concerned may be

SPb + HS- = >SPbSH- 16.03±O.02 (30) (18)

 273

= .tOOE-OS =SPbX-
-/0.0
:::: .200
.IOOE-05

Figure 1. Distribution of surface species as a function of xanthate concentration. [>SPbh = 0.2 mM;
pH = 10; [Pb2+h ~ 0; [S2-h = o.

and
SPb + HS- = >SPbS 2 - + W 5.54±0.18 (3a) (19)

Since >SPbX- was predominant before HS- interaction, the reactant surface species start with
>SPbX- and the ensuing stepwise reactions are

>SPb+ HS- = >SPbSH- logK = 16.03 (18)

+ >SPbX- = >SPb+X- IO!2K = -4.6 (17)

>SPbX-+HS- = >SPbSH- + X- logK = 11.48 (20)

or
>SPb + HS- = >SPbS 2 - logK = 5.54 (19)

+ >SPbX- = >SPb+X- logK = -4.6 (17)

>SPbX-+HS- = >SPbS 2- + X- + H+ logK = 0.94 (21)

274

1.0
Fj =SPbX-
0.9
O.B =SPbSH-
0.7,
O.B
0.5 Tc/mM for Pb2+ = .IOOE-05
log a for H+ = -10.0
0.4 Tc/mM for =SPb = .200
Tc/mM for x- = 1.000
0.3
=SPbS2-
0.2
0.1 =SPbOH-

-I) -5 -4 -3
log[S2- 1T

Figure 2. Distribution of surface species as a function of [S2-h concentration. [>SPbh = 0.2 mM;
pH = 10; [Pb2+h ~ 0; [X-h = 1 mM.

The >SPbS 2 - species is unlikely to exist at nonnal pH conditions, and only dominates at very
alkaline pH over 10.5 (5). Thus, the surface species >SPbSH -is considered to be responsible for
the depression in the flotation of galena.
Since the flotation process is governed by a surface complexation, the concentration of reactants
is an important factor. The influence of the surface site concentration, [>SPb), which is
proportional to the particle size, has also been investigated. Figure 3 exhibits the surface species
predominance area diagram as a function of surface binding site concentration. Obviously, the
dominant surface species varies when the concentration ratio changes. At constant concentration of
HS- (0.1 mM) and X- (1 mM), the dominant surface species may be >SPbSH-, >SPbX- and >SPb
with increasing concentrations of surface binding sites. It is clear that the quantitative evaluation of
surface complexation in mineral flotation demands precise detennination of the surface binding
sites. In order to quantify the collector reaction

>SPb + X-= >SPbX-

[>SPb) and [X-) must be known. In flotation, grinding and particle liberation can affect the
concentration of effective surface functional groups and result in some uncertainty in [>SPb).
 275

=SPb

-3

=SPbX

cf
(f)
-4
II

=SPbSH

-5

-6 -5 -4 -3 -2
log S T

Figure 3. Predominance area diagram of surface spccies at varing [>SPbh and log[S2-] concentrations.
pH = 10; [Pb2 +h ~ 0; [X'h = 1 mM.

Besides those discussed above, some other surface reactions can also take place. For example,
the competition between OR and X- for the surface sites can be described as

>SPbX- >SPb+ X- logK = -4.6 (17)

>SPbOH-+H+ logK = -10.2 (22)

+ H++OH- logK = 14.0

>SPbX-+ OH- >SPbOH-+ X- logK = -0.8 (23)

When the pH increases, >SPbOH- will prevail and the flotation recovery of galena declines, which
is well known from the Barsky relations (19).
276

>SPbOH-+H+ >SPb+H 2O 10gK = 10.2 (22)

>SPb+ HS- >SPbSH- 10gK = 16.03 (18)

+ H2O H++OH- 10~K = 14.0 (24)

>SPbOH- + HS- >SPbSH- + OH- 10gK = 12.23 (23)

Obviously, from the equilibrium constants we may predict that HS - has a stronger affinity to the
galena surface than OH-, hence it is a more effective depressant to galena.

4.2. OXIDE MINERALS (E.G. QUARlZ)

4.2.1. Activation [20]. Quartz is a typical gangue mineral in many flotation systems. Although it
possesses hydrophilic surface properties in nature, the hydrophobicity can be induced either by
adding multivalent metal ions and anion collectors or a cation collector to the system [21].
In aqueous solution, there is a disturbed layer on the quartz surface [22], which resembles the
properties of amorphous silica. The complexation of amorphous silica and silica gel in aqueous
solutions have been widely studied, and the acid-base properties of amorphous silica with related
equilibrium constants were determined. Thc reactive part of aqueous quartz mineral is the hydrated
surfacc functional group. Thus, the reaction mcchanism of quartz resembles that of amorphous
silica, as the functional groups of both materials are the same.
The surface reactions in quartz flotation system by fatty acids can be considered as follows in the
absence of metal ions:

>SiOH = >SiO· + H+ 10gK= -6.8 (ref. [23]) (26)

and
RCOOH = RCOO- + H+ 10gK= -4.5 - 5.24 (27)

Direct adsorption of RCOO- at the surface is very unlikcly to take place. However, in the presence
of metal ions, surface reactions likc

>SiOH + Me 2+ = >SiOMe+ + H+ (28)

>SiOH + Me2+ + RCOO- + H20 = SiOMcOHRCOO- + 2 H+ (29)

or
~§ig~ + Me 3+ + RCOO- + H20 = ~~ignMeOHRCOO- + 3 H+ (30)

could take place, inducing surface hydrophobicity.

It is well known that calcium or magnesium ions can activate the quartz surface and reduce the
surface negative charge [20]. It was also noticed that, at optimum flotation conditions, the quartz
surface activated by calcium or magnesium ions was nearly neutrally charged. Rotation is a
process, in which the combination of quartz surface activator and collector contribute to the
 277

floatability. Both activators (like metal ions) and collectors (like oleate) have to affect the quartz
surface and their mutual contact is a prerequisite for quartz activation. Thus, it is reasonable to
assume that activation is a stepwise process, where the first step is neutralisation of the surface
charge according to the reaction:

>SiO- + Me2+ + H20 = >SiOMeOH + H+ (31)

where upon the surface >SiOMeOH complex reacts with oleate and forms a surface-metal hydrous-
oleate complex, which is responsible for quartz flotation, i.e.

>SiOMeOH + RCOO- = SiOMeOHRCOO- (32)

The validity of the reaction (32) for calcium and magnesium ions was tested by computer assisted
calculation, where an estimated equilibrium constant for reaction (32) was entered into the
composition matrix. The surface reaction constants were collected from those of amorphous silica
[23], and those of solution reactions were taken from ref. [17].

F;
1.0

0.9
•
• •
•
...
~
.. 100

90
0.8 80
0.7
• 70
•
~

~
0.6 =Si02 MgOHRCOO· 60
0.5 50
.,>~
0
u
0.4 40 ~
c
0.3 30 .c::>
"0
0.2 20 "0
LL
0.1 10
0.0 0
3 14
pH

Figure 4. Correspondance of surface species percentage to flotation recovery of quartz as a function of pH

in the presence of 2xlO- 4 M oleate and 1 mM [C032-).
(e) [Fe2 +)r- 5xlO· 5 M; pH=6; [>SiOHh = l.1xlO-4 M; [Ol·h = 2xlO-4 M;
(A) [Ca2+)r- lxlO- 3 M; pH=12.5; [>SiOHh = l.1xlO-4 M; [Ol-JT = 2xlO-4 M;
~) [Mg2+h= lxlO- 3 M; pH=!!; [>SiOHh = l.1xlO-4 M; [Ol·h = 2xlO- 4 M.

From Figure 4, it is obvious that the peak of flotation recovery corresponds to the maximum
fraction of the surface complex. Good agreement was obtained between the flotation recovery and
278

surface species >SiOCaOHOl- and >SiOMgOHOl- as well as (>SiO)zFeOHOl-, i.e.

>SiOH + RCOO- + Ca2+ + H20 = >SiOCaOHRCOO- + 2W log K= -9.5

>SiOH + RCOO- + Mg2+ + H20 = >SiOMgOHRCOO- + 2H+ log K= -8.25

~~igU + Fe3+ + RCOO- + H20 = ~i8MeOHRCOO- + 3 H+ log K= 8.6

It may be concluded that the estimated reactions (32) and (30) very likely exist and the surface
ternary species are possibly responsible for the activated quartz flotation.
In the presence of femc ions, quartz is activated at a slightly acidic pH. After trial and error tests
using computer simulation, the best estimated model was found to be the surface reaction (30). In
this case, two surface sites are needed for the surface complexation leading to quartz flotation.

4.2.2. Deactivation. In the presence of phosphates, reactions like

(33)

will occur in solution, changing the surface reaction equilibrium. The surface reaction

>SiOMeOHRCOO- = >SiO- + MeOH(n-l}+ + RCOO- (34)

or
>SiOMeOHRCOO- = >SiO-
>s.o >s.o-
+ MeOH(n-l)+ + RCOO- (35)

will be shifted to the right during the processing of the reaction (33) and consequently the quartz
losses its floatability. A distribution diagram of the surface species in Figure 5 shows that, as the
polyphosphate concentration increases, the surface ternary complex >SiOMeOHRCOO- or
~i8MeOHRCOO- decreases, some polyphosphate complex or precipatate forms in solution.

4.3. APATITE

Chemical equilibrium constants for the fluroapatite/water interface have becn established and the
amphoteric properties of hydrous fluorapatite surfaces revealed. The experimental data could be
explained by a model that assumes the surface equilibria

>P-O- + H+ = >P-OH 10gj3'lOl(inL) = 6.6±O.1 (30) (36)

>Ca-OH2+ - H+ = >Ca-OH 10gj3'llO(inL) = -9.7±0.1 (30) (37)

The acidity constants result in a value oflog[H+l pzc = -8.15, which shifts to 10g[H+l pzc = -7.13
upon exposition to atmospheric carbon dioxide.
The pHpzc indicated by eqn. 36 applies for suspensions that are protected from atmospheric
 279

1.0
0.9
Fj
0.8
0.7
0.6 (1) (=SiOhFeOHRCOO- (2) =SiOMgOHRCOO-

0.5
0.4
0.3
0.2
0.1
0.0
-6 -3 -2

Figure 5. Distribution of surface hydrophobic species percentage as a function of the concentration of

polyphosphate.
(I) [Fe2+1r- 5x10-5 M; pH=6; [>SiOHlT = l.lxlO-4 M; [OI-lT = 2xlO-4 M;
(2) [Mg2+1r- lxlO- 3 M; pH=ll; [>SiOHlT = l.lxlO-4 M; fonT = 2xlO-4M.
(3) [Ca2+1r- IxlO-3 M; pH=12.5; [>SiOHlT = l.lxlO-4 M; fOnT = 2xlO- 4 M;

Upon exposure of the systems to atmospheric CO 2 the pH was observed to be shifted down to a
value of 7.13 for a solids concentration of 1 gIl. This shift is a consequence of the following
reactions.

CO2(g) + H 20 = HC03- + H+ log K = -7.53 (38)

HC03- = C032- + H+ log K = -10.3 (39)

Ca2+ + HC03- = CaHC03- log K = 1.26 (40)

Ca2+ + C032- =CaC03 log K = 3.2 (41)

2 Ca2+ + HP0 42. + HC0 3- = Ca2HP04C03 + H+ log K = -1.33 (42)

Ca2HP04C03 = Ca2P04C03- + H+ log K = -8.3 (43)

280

A value of pH =7.12 was obtained by a computer simulation using the reactions above. which is
in excellent agreement with the direct observation.
The above information concerning the surface complexation at fluroapatite/water interface can
hopefully shed some light on fundamental studies of apatite flotation.

5. Summary and Conclusions

1. Potentiometric titration in conjunction with adsorption. solubility and zeta potential

measurements. by means of computer simulation. provides a new technique to determine the
surface chemical equilibrium constants.
2. Surface chemical equilibrium constants are the most imponant parameter in defining surface
reactions.
3. The mechanism of mineral flotation (e.g. galena. quartz) can be interpreted on the basis of
surface complexation. which explains the most important aspects of surface charging.
collection. activation and depression.
4. The equilibrium constants for surface complexation can be used for modeling surface-related
reaction process.

6. Acknowledgements

We wish to thank Professor Staffan Sjoberg and Lars Ronngren. Department of Inorganic
Chemistry. University of Umelt and Liu-Ming Wu. Division of Inorganic Chemistry. Lule!'!
University. for many helpful discussions and providing data used in this work. The work
described here forms part of a program financially supported by the Swedish Board for Technical
Development.

7. References

I. Lartimer.W.M. (1952) The Oxidation States of the Elements and their Potentials in Aqueous
Solution. 2nd ed .. Prentice-Hall. Englewood Cliffs. NJ. p. 34.
2. Schindler. P.W .. and Kamber. H.R. (1968) Die aciditiit von silanolgruppen. Helv. Chern.
Acta ~. 1781-1786.
3. Schindler. P.W .• and Gamsjager. H. (1972) Acid-base reactions of the Ti02 (anatase)-water
interface and point-of-zero-charge ofTi02 suspensions. Kolloid Z. u. Z. Polymere 250.
759-763.
4. Stumm. W .. Huang. C.P .• and Jinkens. S.R. (1970) Specific chemical interaction affecting
the stability of dispersed systems. Croat. Chern. Acta 42.223-244.
5. Sun. Z.-X. (1991) Ph.D. Thesis. Lulea University. Sweden.
6. Sun. Z.-X .. Forsling. W .. Ronngren. L.. and Sjoberg. S. (1991) Surface reactions in
 281

aqueous metal sulphide systems. 1. Fundamental surface properties of hydrous PbS and znS,
Int. J. Miner. Process (in press).
7. Ronngren, L., Sjoberg, S., Sun, Z.-X., Forsling, W., and Schindler, P.W. (1991) Surface
reactions in aqueous metal sulphide systems. 2. Ion exchange and acid/base properties at the
ZnS-H20 interface, J. Colloid and Interf. Science (in press).
8. Wu, L.M., Forsling, W., and Schindler, P.W. (1991) Surface Complexation of Calcium
Minerals in Aqueous Solution. 1. Surface protonation atfluorapatite-water interface, accepted
for publication in J. Colloid Interf. Science.
Wu, L.M., and Forsling, W. (1989) Surface complexation on fluorapatite, Proceedings 2nd
INMC (Beijing, China).
9. Westall, J. (1982) Reports 82-01 and 82-02, Chemistry Department, Oregon State
University, Corvallis.
10. Eriksson, G.A (1979) An algorithm for the computation of aqueous multicomponent,
multiphase equilibria, Anal. Chim. Acta ill, 375-383
11. Taggart, A.F., Taylor, T.e., and Knoll, AF. (1930) Chemical reactions in flotation, Amer.
Inst. Min. Metall. Engrs. Tech. Pub!. No. 312, pp. 3-33.
12. Wark, I.W. (1933) The chemical basis offlotation, Proceedings Australian Inst. Min. Metall.
90, 83-123.
13. Gaudin, AM. (1957) Flotation, McGraw-Hill, New York, 573 pp.
14. Fuerstenau D.W. (1984) Chemistry of flotation, in M.H.Jones and J.T.Woodcock (cds.)
Principles of Mineral Flotation - The Wark Symposium, The Aust. I.M.M., Parkvill, Vic.,
Australia, p. 7.
15. De Bruyn P.L. (1984) Some aspects of the physical chemistry of solid/water interface, in
K.J. Ives (ed.) The Scientific Basis of Flotation, Martinus Nijhoff, The Hague, p. 111.
16. Iwasaki, I., and Cook, S.R.B. (1958) The decomposition of xanthate in acid solution, J. Am.
Chern. Soc. 80, 825.
17. R.M. Smith, and Martell, AE. (1976) Critical Stability Constant, vol. 4, Plenum Press, New
York.
18. DuRietz, e. (1957) Xanthate analysis by means of potentiometric titration, Svensk Kern.
Tidskr. 69, 310-327.
19. G.Barsky (1934) Discussion, Trans. AIMME ill, 236-237.
20. Sun, Z.-X. (1981) M.S. Thesis, Central-South University of Technology, Changsha, China.
21. Fuerstenau, M.e., and Palmer B.R. (1976) Anionicflotation of oxides and silicates, in M.C.
Fuerstenau (ed.) Flotation, A.M. Gaudin Memerorial Volume, vol. I, pp. 148-196.
22. Ber, R.K. (1979) The Chemistry of Silica, John Wiley and Sons, New York.
23. Schindler, P.W. (1981) Surface complexes at oxide-water interface, in M.A. Aderson and
A.J. Rubin (eds.) Adsorption of In organics at the Solid/Liquid Interface, Ann Arbor Science
Publication, Ann Arbor, Mich., p. 1.
2.5. AMIDE-BASED TECHNOLOGIES IN FLOTATION AND
HYDROMETALLURGY OF PRECIOUS METALS

ST. GAIDARJIEV and S. SPASSOVA

Mineral Processing Department
University of Mining and Geology
Sofia 1156
Bulgaria

ABSTRACf. A process for recovering gold and silver and making quartz suitable for the glass
industry, comprising flotation, magnetic separation, leaching and sorption, is described. The
sorbent is a native ionite - low calorific brown coal which needs no regeneration. A two-stage
sorption is considered necessary for good gold recovery; the loaded sorbent is then subject to
burning for final precious metals recovery.

1. Introduction

Today the general awareness of the dramatical need for nontoxic environment cannot be overlooked
anymore. The careful analysis of Bulgarian precious metals ore basis suggests that most of the ores
are located in environmentally sensitive regions, i.e. they are very close to water supply sources
and to urban areas. Therefore it is obvious that special care has to be taken to prevent any eventual
pollution. This concern has stimulated interest in research of viable alternatives to the harmful
application of cyanide salts in flotation and hydrometallurgy. Besides this, some of the ores have
been found to be refractory to cyanidation.
One of the most studied alternative process is the thiourea leaching [1,2]. There are several cases
where it is known that thiourea (TU) is commercialy applied, however there are still many loose
ends which need to be tied down to achieve large-scale commercialisation. For instance, the
mechanism of gold complexation and oxidation is still debated, and the same is observed when
speaking about the sorption recovery as part of an integrated flowsheet development [3].
The problem of treating Bulgarian gold ores has never been in such a sharp focus as it is now,
because of the increased need for stabilising the national currency. On the other hand, there is a
recognised need for treating the huge masses of wastes and residues produced during the operating
period of the mills as a potential part of the global gold resources. This is because of the new
economical opportunity for processing these resources and of the more stringent environmental
regulations. References suggest also that today the main trend is to develop combined technologies
that include physical separation processes, flotation and hydrometallurgy for achieving maximum
efficiency in recovering valuable material from the concentrates.
So, our research has been focused at the aiming for scaling up of the amide-based combined
technologies, i.e. urea used as a collector mixture with xanthate, in order to achieve better recovery

283
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 283-289.
© 1992 Kluwer Academic Publishers.
284

of the precious metals and use of the thiourea leaching process as a strategy for ecological treatment
of gold ores, wastes and concentrates.
In the following, some of the results obtained from a research project for the complex treatment
of precious metals are presented briefly, containing flotation tailings and low grade gold ore
treatment by means of an amide-based combined technology.

2. Study of the Ore

The composition of the sample studied is given in Table I. The mineralogical study of samples
used for the research suggested that the degree of oxidising is quite uneven and that the main ore
minerals are present as lead minerals:
- galena 41 %, plumbojarosite 40 %, crocoite and pyromorfite 20 % ;
copper minerals: copper carbonate 25%, chalcopyrite 50% ;
secondary sulphides (covelitte, bornite) 25% ;
- zinc minerals: zincblende 69%, zinc silicate 31 %.
There are more than 40 secondary minerals in the ore while the most widely spread are the iron
hydroxides (goethite and hydrogoethite) and manganese hydroxides (pyrolusite, varnadite).
Precious metals are present as miargirite, native gold and silver, electrum, and custelite.

Fe
0.22 2.8

The phase composition of gold shows the following distribution: free gold 31 %, gold in units
50%, gold covered with iron hydroxides 12%, gold finely dispersed in quartz 5%, gold in
sulphides 2%.
From investigations performed in previous studies, it was found that the material treated for
precious metals recovery possesses unusual complexity and therefore a sophisticated flowsheet
seeking recovery of almost all valuable components - lead, gold, silver and quartz - must be used.
Taking into account the mineralogical characteristics and the forms in which gold and silver occur,
it was decided that flotation and magnetic separation could serve gladly for the purposes of
ensuring suitable technological conditions for the following leaching and for preventing a high
reagent consumption. The precious metals recovery in the flotation concentrate is between 45-50%.
A mixture of urea and xanthates is employed as flotation collector.
The flotation tailings are subject to magnetic separation on a roller induction separator with salt-
acid leaching of the non-magnetic fraction following. The purpose of the salt-acid leaching is to
unwrap the surface of the precious metals particles by dissolving the oxide ferrous coatings and to
produce in situ Fe03 - an oxidisin~ agent for the following TV leaching.
For the purposes of better understanding of the hydrometallurgical system, a thermodynamic
investigation on gold, silver and their alloy elect rum was carried out. The reactions with the
formamidine disulphide have al/:iO been studied. The most probable reactions in the system are as
follows:

I. gold and silver leaching:

 285

Au + FeCl3 + HCl + C2S2(NH2h(NHh = C2S2(NH2NHAu)(NHh +

+ FeCl2 + 2 HCl (react. 1)

Au + FeCl3 + HCl + CS(NHV2 = AuClCS(NHv2 +

+ FeCl2 + HCl (react. 2)

Ag + FeCl3 + HCl + C2S2(NH2h(NHh = C2S2(NH2NHAg)(NHh +

+ FeCl2 + 2HCl (react. 3)

Ag + FeCl3 + HCl + 3CS(NH2h = AgCl(CS(NH2hh +

+ FeCl2 + HCl (react. 4)

II. electrum leaching:

AuAg + FeCl3 + HCl + CS(NH2h = AuCICS(NH2h +

AgCl(CS(NH2hh + FeCl2 + 1/2 H2 (react. 5)

AuAg + FeCl3 + HCl + 2 C2S2(NH2h(NHh =FeC12 +

C2S2(NH2NHAu)(NHh + C2S2(NH2NHAg)(NH2) (react. 6)

According to the principles of theJTIlOdynamics [4-6], a thennodynamic state of the hydrome-

tallurgical system described by the equations shown above and characterised by means of the free
Gibbs energy is presented. The data are presented in Table 2.

. ald ata ~or d'ffi

TABLE2Thenno dlynamlc 1 erent reactIons 0 fTVI each'mg.
Reaction no. tJ.G IgK tJ.(jJ
Jlmol mV
1 -90.04 66.01 3.14
2 -75.84 55.50 4.46
3 -106.20 77.86 10.62
4 -48.24 35.64 6.08
5 -9.43 6.91 0.94
6 -59.54 43.45 5.95

where tJ.G is the free Gibbs energy, K is the reaction constant, and tJ.cp the redox potential.
Data presented in Table 2 as well as the reactions 1-6 show that the complex compounds with
fonnamidine disulphide C2S2(NH2h(NHh present in the fonns of C2S2(NH2NHAu)(NHh ;
C2S2(NH2NHAg)(NHh as well as the compounds obtained by means of joining reactions of gold
and silver salts with thiourea: AgCICS(NH2h ; AgCl(CS(NH2hh have a key function in the
leaching of gold, silver and their alloys.

3. Pretreatment Operations

In order to make the conditions suitable for the TV leaching, a pretreatment operation is required.
The salt-acid leaching has been conducted according to a factorial design and three main factors
286

have been examined: the degree of salt acid dilution, temperature and leaching time for agitated-cell
tests.
The parameters subjected to optimization were: minimum iron and maximum gold and silver
content in the liquid. Optimal values after resolving the equation system were found respectively:
degree of salt acid dilution 1:20, leaching time for agitated-cell tests 60 minutes, and tempera-
ture 60°C.
The cake obtained after salt acid leaching was treated by means of thiourea leaChing at an optimal
pH of 3.5, an optimal temperature and optimal TV concentration. Leaching takes place in agitating
vessels with a constant stirring machine. Temperature is maintained by the means of a thermostat.
Gold and silver in the material were determined by AAS. After two hours of leaChing, no gold was
detected in the filter cake. The cake material may be used either in the glass industry or as a racing
sand due to the low iron content -less than 0.2% and high Si02 content - 98.5 %.
The pregnant solutions afterTU leaching contained: Cu 24.19 ppm, Mn 270 ppm, Fe 2450
ppm, Zn 1480 ppm, Pb 540 ppm, Au 0.36 ppm, and Ag 4.97 ppm. The remaining elements in the
solution were in negligible quantities and were not of interest.

4. Sorption Experiments

Due to the lower precious metals content of the pregnant solutions, the highly expensive
activated coals were found to be unacceptable. Therefore it was decided to test the sorption
capabilities of different natural brown coals, which are less expensive and need no regeneration.

Sorption recovery [%]

100~------~~----~==~==~===r---!1

80

60 Pb
-+- Ag

--
-T- Cu
40
--Q-
Fe

Zn
20
~ Mn
--.A- Au

2 3 4
Time [hr]

Figure 1. Sorption recovery (stage I).

 287

Four main factors were examined to show the influence in the sorption process: coal type, coal
size (fraction), sorbent consumption (amount of coal in the solution) and sorption time. The
parameters subjected to optimization were minimum gold and silver content in the filtrate after
sorption. For this 20 ml aliquots were analysed by means of ICPAES. After optimization of the
equations describing the gold and silver sorption, the following normal scale parameters were
found:

Ag Au
A - coal size [mm] -0.5 -0.5
B - consumption [gil] 17 35.5
C - coal type Chukurovo origin
D - sorption time [min] 44 48

The technological criteria subject for optimization - silver and gold content in the solution - were:
Ag 0,078 ppm, and Au 0.27 ppm.
From these results (Figure 1) it is obvious that the silver sorption recovery according to the
model reaches 90 %; however, gold sorption recovery is quite unsatisfactory - in the range of 25
%. The sorption recovery of the remaining elements reaches the values of: Cu 89%, Fe 50%, Mn
47%, Pb 95%, and Zn 42%. This suggested that the lower gold sorption is probably due to the low
sorbent capacity of the native coals from Chukurovo. So in this context it was assumed that a two-
stage sorption could ensure satisfactory recovery values both for silver and for gold.

Gold concentration [ppm]

0.4r-------------~~~----------_.

0.3 1- ... :"' ............................................................ .

0.2

0.1

o~----~------~------~------~
o 2 :5 4
Time [hr]

Figure 2. Gold soprtion (stage II).

 288

-OOBmm
I Urea 10 gft
I K.X. 10gft
milling 6S%-O,07~mm
pine oil 10gft
~
Ro her

Cleaner Smin Ma netic se aration

. 1 40V ,6A,O.4T

Flot. product Nonmagn~. Magnetic fra~ion

HCl leaching 45min, 60 C HCl regeneration
~ TU3kgft I t
TU leaching 102ni1n
~
Filtration

Brown coal ~artz sand

"-t.
Two stage sorption
TU regeneration J
I. Two stage filtration
I t l
Ag- coal cone. Au-coal cone.
from the I stage from the 11 stage

Figure 3. F10wshcct for treating low grade ore containing precious metals.

The necessary time for the first sorption stage is estimated from kinetics experiments and it was
found sufficient for recovering silver, lead, copper, zinc and iron. After a predetermined time, the
loaded sorbent is removed and an amount of fresh one is added in the solution to recover the
remaining gold. After three hours of sorption, no gold was detected in the liquid phase (Figure 2).
The experiments from the factorial design were considered to show the influence of all
independent variables on gold and silver sorption. Based on the experiments, a technology with
 289

optimal conditions was designed. The flowsheet thus developed is shown in Figure 3. Table 3
illustrates the final technological results.

5. Conclusions

To recover gold and silver and to make the quartz product after leaching suitable for the glass
industry or for a racing sand, a complex technology seeking the recovery of almost all valuable
components and comprising flotation, magnetic separation, leaching and sorption is developed.

TABLE 3 T echno1oglc
. al results f rom 1ocked tests.
Product Assay [ppm] Leaching recovery Sorption recovery
[%] [%]
Au Ag Au Ag Au Ag
Filtrate * 2.6 92 86.6 6.8 14.5
Quartz sand * 1.33 8 13.4 8 13.4
Coalconc. 112 1156 - - 85.2 72.1
Feed 0.46 5.69 100 100 100 100
* not detected « 0.05 ppm)

The sorbent is a native ionite - low calorific brown coal which needs no regeneration. In order to
extract gold from the pregnant solutions to the utmost extent and due to the low sorption capacity of
the coal, a two-stage sorption was considered necessary for good recovery. The loaded sorbent is
subject to burning for final precious metals recovery.
Besides the low investment costs of applying such kind of sorbents, the environmental aspects
of the technology are evident, because these sorbents secure acceptable harmless characteristics of
the solutions, in the sence of their lake disposal after that.

6. References

1. Wall, N.C., Hornby, J.C., and Sethi, J.K. (1986) Gold beneficiation, Mining Magazine,
May, 393-399.
2. Shulze, RG. (1986) Thiourea leaching of precious metals, Journal of Metals 2, 112-115.
3. Gundiler, I.H., and Huyhua, H.I. (1986) Thiourea leaching of gold and silver, in
Proceedings of the Int. Symp. of Hydrometallurgical Reactor Design and Kinetics, 1MS:
AIME.
4. Heywood, R (1983) Thermodynamics of Equilibria, Mir, Moscow, pp. 220-280.
5. Westrell, C. (1980) Thermodynamics of Organic Compounds, Himia, Moscow, pp. 160-200.
6. Garrels, RM. and Christ, C.L. (1968) Solutions, Minerals, Equilibria, Mir, Moscow,
pp.88-123.
2.6. DETERMINATION OF FLOATABILITY DATA USING THE EMDEETM
MICROFLOT METHOD

M.W. CHUDACEK
M.D. Research Company Pty Ltd.
S Khartoum Road
North Ryde, NSW 2113
Australia

ABSTRACf. The discovery of the particle - bubble attachment/detachment equilibrium in a

vigorously agitated closed mineral/reagent system offers new research opportunities for the
flotation scientist. This mechanical/ thermodynamic equilibrium offers a potential link between
thermodynamic principles and the performance of continuous flotation systems. The set of
subsequent equilibria determined on the head sample by this method constitutes a mineral floatabi-
lity spectrum for a given ore/reagent system. The results relate to flotation rate constants in a batch
cell, so these data may be used for prediction of batch cell kinetics.

1. Introduction

Knowledge of the flotation process would improve substantially, if the link between fundamental
thermodynamic properties of the mineral orel reagent system and continuous flotation processes
could be determined or at least illustrated. The discovery of the existence of the particle - bubble
attachment! detachment eqUilibrium [1] and the description of the relationship between these
equilibria and the batch cell flotation rates [2] may constitute an opening for determination of the
above link.
It was also shown [3,4] that the equilibria do relate to the batch cell kinetics by their successful
prediction of batch flotation rate constants from test-tube equilibria data, and that they represent a
driving force controlling flotation also in open system, such as mechanical flotation cells. It is
conceivable that these equilibria may limit the maximum yield of cocurrent flotation systems such
as Bahr's or Jameson's cells. Once the pulp I air mixture in these cells reach an equilibrium mass
transfer to the bubble surface ceases and the gas phase must be separated and replaced by fresh air
for the process to continue. If the equilibrium values are low, more stages may be required for a
high recovery in the cocurrent processes. The objective of process optimisation then becomes to
maximise the equilibrium value.
As a reasonable prediction of continuous flotation systems from batch cell data has been possible
for some time, it may be possible to predict continuous system performance from test-tube
equilibria data. Therefore, once we manage to predict mechanical I thermodynamic equilibrium in a
rapidly agitated closed flotation system from thermodynamic and hydrodynamic data, the circle
may close and a limited prediction of continuous systems from these data may become possible.

291
P. Mavros and K. A. Matis (eds.),lnnovations in Flotation Technology, 291-294.
© 1992 Kluwer Academic Publishers.
292

2. Floatability Correlations

floatability data are obtained using the Microflot method. The pulp sample fills a test tube and,
after conditioning with the appropriate chemicals, the tube is vigorously agitated under preset
conditions. During agitation, a mineralised froth is formed, in which an eqUilibrium of particle-
bubble attachment/detachment is quickly established.
The most and the least floatable particles, as well as particles with intermediate floatabilities may
be separated by repeated equilibrations of previously separated froth [2]. This procedure should
yield a set of panicle samples with different floatabilities. If the head mineral sample was a very
narrow size fraction, the effect of the particle size is virtually eliminated.
An appropriate physical or chemical examination of the set of samples should identify the factors
determining the differences in floatabilities in the set of samples. Examination of another mineral
size fraction may yield a similar correlation, if the nature of the particle is similar to the one in the
previous size fraction.
Profound changes may also be found in the nature of the floatability spectra of a monosized
sample as the mean particle size decreases. This may point the different factors being at play for
very fine particles than for medium or coarse ones.

3. Thermodynamic Properties of the Equilibrium

Fichera [3] and Chudacek [1] have shown that equilibria can be manipulated by various
thermodynamic factors such as pH, concentration or mechanically by varing intensity of the
agitation. Once the effects of basic thermodynamic criteria on the equilibrium are known, there may
be a possibility of their prediction from first principles. Pure, simple and stable minerals in simple
and stable reagent systems would be a prime candidate for such an examination.

4. Prediction of Batch Cell Kinetics

It has been reported [5] that the first order kinetics single or double distributed models are adequate
for the batch cell prediction, if the heterogeneous nature of the feed is accounted for. The floatabi-
1ity spectra offer a very powerful descriptor of behaviour of flotation feed, determined on a very
small sample mass. Therefore, flotation behaviour even of the very rarest minerals in the batch cell
may be predicted from its spectra. In the case of pure minerals, only 0.1 g sample mass is required
for the test. The magnitude of the spectrum, as has been already stated, is a function of the
chemical environment. Therefore, flotation optimisation in terms of spectra means maximising the
flotation spectrum of valuable minerals while minimising the spectra of gangue minerals.

5. Practical Significance of Floatability Spectra

The magnitude and character of floatability spectrum points directly to the appropriate flotation
network as illustrated in Figure I, which shows the spectra for five mineral systems and the
matching batch cell responses. The two highest spectra (Cu ore and sphalerite) represent a fast
floating system with relatively small differences between the floatability of various components. A
simple flotation system would yield adequate separation and good recovery. The intermediate
 293

spectrum of quartz with the long tail would require a more complicated circuit if a high recovery is
the objective. Low spectra for Cu slag rougher and cleaner suggest that these mineral systems are
extremely slow floating and attainment of high concentrate grade is unlikely due to long flotation
time required. These spectra point to an inadequate chemistry or to a profound liberation problem.

MFI Spectra Blitch Cell Kinetics

l00~M~A~ __________________- - ,

eo eo

60

40
~ ZnCono. ~ ZnConc.

--- Quartz --- ClJI!fiz

20
- - OJ 8Iag (0) - - OJ 8Iag (0)
- - OJ 8Iag (R] - - OJ 8Iag (R)

o 02 [).4 0.6 O.B o 120 240 360 600

Mass fraction Time [secl

Figure 1. Floatability spectra (left) of some mineral/ore systems and (right) matching predicted batch
cell responses for eu ore, Zn concentrate, quartz in calcite ore, eu slag (cleaner) and eu slag (rougher).

6. Conclusions

The recently discovered equilibria of particle-bubble attachment! detachment open a new window of
perception for flotation reseach. The Microflot floatability spectra represent a powerfull descriptor
of flotation behaviour even for complex mineral systems. These spectra permit a prediction of the
batch cell kinetics from truly test-tube scale data.

7. References

I. Chudacek, M.W. (1990) A new quantitative test-tube floatability test, Min. Engng.l, 461-
472.
2. Chudacek, M.W., and Fichera, M.A. (1991) The relationship between the test-tube
floatability test aruJ batch cell flotation, Min. Engng.1., 25-35.
294

3. Fichera, M.A. (1990) The relationship between the batch cell kinetics and the test-tube
flotation test, MSc Thesis, University of New South Wales, Australia.
4. Chudacek, M.W. and Fichera, M.A. (1990) Prediction of batch cell kinetic from test-tube
scale data, Proceedings of the 2nd Workshop on Rotation of Sulphide Minerals (Lu1ea,
Sweden, June 18-21).
5. Fichera, M.A. and Chudacek, M.W. (1991) Batch cell flotation models - A review, Min.
Engng. (in press).
2.7. FROTH FLOTATION KINETICS - THEORY AND MODEL EXPERIMENTS

B. RADOEV and L. ALEXANDROVA

Department of Physical Chemistry
University of Sofia
1126 Sofia
Bulgaria

ABSTRACT. The problem of froth kinetics is treated as a transport phenomenon. Two coupled
equations of balance are fonnulated concerning the particles attached on the bubbles and the free
particles in the pulp. Special attention is paid to the mode of interaction between particles and gas
bubbles under dynamic conditions in a flotation cell. In contrast to the well known empirical first-
order chemical reaction rate model, the solution in the present paper predicts similar but not exactly
the same behaviour of the recovery rate. The adequacy of the theoretical assumptions is checked by
means of laboratory flotation experiments with glass ballotini. The floatability of the particles is
compared with their three-phase contact angles.

1. Introduction

It is well known that the first -order chemical reaction model seems to be one of the best flotation
process approximations. Many authors have pointed out the fact that flotation is affected by
collisions between mineral particles and gas bubbles. Thus, by analogy with the real gas kinetics
processes, it follows directly that a proportionality between mineral particles concentration c(t) and
their floating rate exists (see for example Finch and Dobby [1], also Ek in this volume) :

dc/dt= -k.c (1)

Here, the rate constant k (= J . E) depends on the flux J of particles colliding with bubbles
multiplied by an attachment efficiency coefficient E (0 ~ E ~ 1).
Strictly speaking, relation (1) describes the process at a given point r in the cell and for this
reason its integral:

c(r,t)
c(r,O) = 1 - exp( -kt) (2)

cannot be applied without additional analysis to the recovery kinetics

R(t) = 1 - exp( -kt) (3)

Here it should be reminded that R(t) is defined as the total amount of particles floated for time t
from the entire cell and its relation with the local concentration c(r,t) is not simply proportional but

295
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 295-300.
© 1992 Kluwer Academic Publishers.
296

a more complex one (see eq. 5 below).

Eqn. I does not take into account also the possibility of detachment of particles from bubbles
during their rising towards the froth cell surface (compare with eqns. 4 and 7).
The present communication aims to derive more rigorously the time dependence R(t) as well as
to compare some theoretical predictions with experimental data.

2. Statement of the Problem and Basic Equations

The transport of mineral particles through a pulp of bubbles obeys the following balances [2,3] :

(4)
de
(it = - C q,

where Y is the rising velocity of bubbles in the flotation cell (a constant), n(z,t) is the number of
attached particles per bubble, C is the bubble concentration in the pulp (a constant), q (= q+ - q_) is
the resultant particles-bubble interaction rate, where q+ is the attachment rate and q_ the detachment
rate, and c(t) is the concentration of suspended (unattached) particles.
Only the material transport which is vertical to the cell (z coordinate) is of interest for flotation
kinetics, and for this reason after averaging across the cell all variables depends only on z and t.
Another approximation in eqn. 4 is the supposed uniform distributions of both the suspended
particles c(t) and the bubbles (C = constant). This assumption is equivalent to a perfect pulp
mixing.
The attachment rate q+ coincides with the right-hand side of eqn. 1 and accounts for two
consecutive processes formulated by Scheludko et al. [4] as Before Contact State (BCS) and After
Contact State (ACS). The following processes dominate BCS (for more details see Schulze [5] and
Israelishvili [6]) :
- particles-bubbles hydrodynamics and van der Waals attractions;
- hydrophobic forces.
Capillary [7] and external dynamic forces such as gravity, turbulent acceleration [8] etc. act during
ACS.
The detachment rate q_ has been considerably less studied, and for this reason here we shall limit
ourselves only to some quite general considerations. Regardless of the various possible effects, q_
should be proportional to the amount of attached particles: q_ = n.L In the case of negligible
lateral (particle-particle) interactions the rate constant k_ should not depend on n but only on the
ratio of the acting forces (capillary forces I detachment forces).
There is a direct relation between the number of attached particles n(h,t) reaching the cell froth
surface (z=h) and the recovery rate (dRldt) :

: - = yeS n~~t) (5)

where S is the area of cell froth surface, No is the initial number of mineral particles

No = c(O) S h

and h is the height of the collection wne of the flotation cell.

 297

3. Solution of the Problem

A useful way to solve the system of eqns. 4 is the Laplace transfonnation method with respect to
time:

nez,s) = f n(z,t).exp( -st).dt (6)

o
The following boundary and initial conditions could be used for the solution of eqs. 4 [3]:

nez, t=O) = 0; c(z, t=O) = co; n (z=O, t) = 0 (7)

Note that the level z = 0 does not coincide with the bottom of the flotation cell. Close to the rotor
the flow is strongly turbulent, with high accelerations, so that the bubbles cannot rctain the
particles.
The Laplace image of R(t) follows from eqs. 4 and 8

V C k+ s + Kh (8)
s.R(s) = hs (s + K) [ 1 - exp (-s s + CVk)]

where

Despite the complicated fonn of eq. 8, several important conclusions could be drawn directly
from the Laplace image R(s):
Recovery R(t) is constrained in the limits [0,1] for the entire time interval : 0 = R(t=O) :0; R(t) :0;
R(t-> 00) = 1;
Although R(t) should represent a saturation curve, the Laplace-image from eqn. 8 does not
correspond exactly to an exponential time dependence R(t) as in eqn. 2. For the sake of
comparison here, the Laplace image of R according to eqn. 2 is given:

1
s R(z) = - k+S

For long time intervals (t> 1'), R(t) asymptotically tends to 1 in the following way:

1
R(t>1') = 1 - (I - -).exp(-Kt) (9)
K1'

where

As it should be expected, the time constant K (Le. the effective rate constant of the process)
depends both on the attachment and detachment rate characteristics.
298

4. Experimental

4.1. FLOTATION CELL

The experiments were performed with a mechanical laboratory flotation machine equipped with a
plexiglass cell of 0.5 I capacity and a rotor - stator system operated at 2500 rpm; the air flowrate
was 55 l/h and the mean bubble radius 500 1IDl.

4.2. MATERIALS

The material floated consisted of glass "ballotini" spheres of two different sizes: 100-160 IIDl and
160-200 ~m. The density of the suspension was Q = 40 gil or 3-4x106 particles per liter. The
collector was dodecylamine hydrochloride with a concentration of 10-5 M. The pH value was set
with a universal buffer (0.2M NaOH, 0.12M CH3COOH, 0.12M H3P04, 0.12M H3B04) and
controlled by a glass electrode. The ballotinis were cleaned before each experiment with hot
bichromate mixture and afterwards repeatedly washed with bidistilled water.

R( t)
1,O~~---------'--------------------------'

first flotation second flotation third flotation

0,8
100-160 fJm
0,6

0.4

160-200 ~m
0,2
II III
o,o~----------~------------~----------~

Time

Figure 1. Flotation kinetics curves R(t) for 3 successive procedures; duration of each procedure 5 min.

Flotation was preceded by agitation (about 10 min), after which air was fed into the machine and
the experiment started. The experiment lasted until the recovery rate (dR/dt) reached zero. The froth
product was removed by hand 15-20 times per minute, with due care not to scoop any liquid from
the pulp.
The effect of the hydrophobicity of the particles on their floatability was studied by a successive
flotation of tailings (Table 1). For this purpose, the particles remaining in the cell at the end of the
first (primary) flotation were agitated with a fresh solution at the same pH and collector
 299

concentration. The same procedure was repeated a third time after the second (secondary) flotation
as well.

TABLE 1. Results of flotation expenments.

Particle diameter Flotation Time const. t Ultimate recovery
[!IDl] [min] Roo
primary 1.0 0.48
100-160 secondary 1.1 0.46
tertiary 1.3 0.42
primary 1.9 0.19
160-200 secondary 2.0 0.16
tertiary 2.4 0.10

bNlN [-] llNlN [-]

80r-~~------------------, 80.-~~-------------------.

o
o ~
60 60

40 40

20 20

o~--~--~--~----~--~--~ o~--~--~--~--~--~--~
5 10 15 20 25 30 35 5 10 15 20 25 30 35
Contact argle [oJ Contact argle [oJ

Hydrophobicity of tailings Hydrophobicity of tailings

before primary flotation. before secondary flotation.

Figure 2. Particle distribution with respect to contact angle.

S. Results and Discussion

The data in Table 1 and Figure 1 show that at pH 6 and collector concentration 10-5 M the ultimate
recovery Roo was less than 1. This result could not be explained by the theoretical model proposed
here, according to which R~ = 1. Note that the finite detachment probability (k- =0) does not affect
Roo but only lengthens the process via the time constant (see eqn. 9).
The results from the successive flotations are somewhat unexpected. Non-floated particles from
300

tailings after agitation with fresh solution regain considerably their floatability. This peculiarity is
again observed after the secondary and the tertiary procedure.
The hydrophobicity of the glass surface was characterised by means of the three-phase contact
angle, 8. This showed a similar behaviour for the three successive tests. The mean value for the
pulp (averaged over more than 60 particles) after the first agitation seems to be 8 "" 22°, with
corresponding values for the concentrate and tailings 23° and 15.5°, respectively (Figure 2). After
the secondary treatment of the tailings, the contact angle increased from 15.5° up to 20, with 8 ""
21 ° in secondary concentrate and 18.5° in the secondary tailings.
The correlation between the contact angle and floatability observed here should not be considered
as an unambiguous one. This is seen from the contact angle distribution curves in Figure 2. The
ultimate recovery R~ (0.47 for the secondary flotation, from Table 1) means roughly that all
particles with 8 "" 20° should float. On the contrary, the overlapping with the two other
distributions - in the concentrate and the tailings curves - shows that particles in a quite broad
hydrophobicity range (15°:'; EJ:,; 25°) could be found both in the product and in the tailings.

6. References

I. Finch, I.A., and Dobby, G.S. (1990) Column Flotation, Pergamon Press, Oxford, chapt. 3.
2. Sastry, K. V.S., and Fuerstenau, D. W. (1970) Theoretical analysis of countercurrent flotation
column, Trans. AIME 247(1), 46-52.
3. Radoev, B., Alexandrova, L., and Tchaljovska, Sl. (1990) On the kinetics offrothflotaton,
lnt. I. Miner. Process. 2B., 127-138.
4. Scheludko, A., Tchaljovska, Sl., Fabrikant, A., and Radoev, B. (1971) Untersuchung zum
elementarakt der flotation, Freiberger Forschungshefte A484, 85-97.
5. Schulze, H.I. (1984) Physico-chemical Elementary Processes in Flotation, Elsevier,
Amsterdam, chapt. 4.
6. lsraelachvili, J.N. (1985) Intermolecular and Surface Forces, Academic Press, New York,
chapts. 7 & 8.
7. Nutt, C.W. (1960) The adhesion of solid particles to flat interfaces and bubbles, Chern. Eng.
Sci. 12, 133-141.
8. Schubert, H. (1974) Zur prozessbestimmenden rolle der turbulenz bei aujbereitungs-
prozessen, Aufbereitungs-Technik 2,501-520.
2.8. ELECTROLYTIC FLOTATION

K.A.MATIS
Laboratory of General & Inorganic Chemical Technology
Department of Chemistry
Aristotle University
GR-540 06 Thessaloniki
Greece

EXTENDED SUMMARY. This is a rather unconventional separation process, owing its name to
the bubbles generation method it uses, Le. the electrolysis of an aqueous solution.
Electrolytic flotation, (or electroflotation) has been extensively reviewed during a previous
ASI [1]; its development was mainly conducted in the USSR. Applications range from mineral
processing to effluent treatment [2,3] - see also Zouboulis et ai., in this volume.
There are several flotation differences between the fields of classical mineral processing on the
one hand, and water and waste water engineering on the other. These differences concern the
concentration (or pulp density), particle size, mixing characteristics, separation and economical
aspects.
The following applications have been investigated :
- suspended matter from industrial liquid wastes,
- emulsified oil from wastes (including power station oils),
- organic entrainment in copper solvent extraction,
- heavy metals (e.g. chromium from dilute aqueous solutions) and
beneficiation of salt type minerals and pyrites.
One of the problems, that has to be solved for an effective process design, is the electrolytic cell
(some electrode types are presented in Photographs 1-3). In fact, there are several alternatives, and
perhaps lead dioxide covered titanium electrodes constitute a reasonable solution with adequate
properties. This was found to be operating after two years of continuous operation under severe
conditions (in 0.1 % sodium chloride solution).
Membrane-separated electrodes should also be mentioned; these may separate hydrogen from
oxygen, since usually only one of the two gases is useful. The size of the bubbles produced
(having a mean diameter of 50 !-lm) has been measured (through photographs) by a specific camera
arrangement (photograph 4).
The process is generally operated with a current density of approximately 100 A m-2 at a low
voltage and a power consumption of 0.6 MJ m-3 (depending on the pulp conductivity).
It has been suggested that comparing electrolytic flotation to dissolved-air flotation - which is the
main technique applied in effluent treatment - the former was better for units having an area of 5
m 2 , or for flows up to about 14 m 3 S-I. This, in todays world of economic constraint, was
considered as a disadvantage which has precluded (at least up to now) the process from being
widely applied. Electrolytic flotation could be used in cases where air is difficult to dissolve; no

301
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 301-304.
© 1992 Kluwer Academic Publishers.
302

Photograph 1. Examples of electrolytic flotation cells constructed and tested. (Left) Horizontal,
stainless-steel perforated sheet; anode corrosion is apparent, after approximately 48 hrs of operation under
severe conditions. (Right) Bipolar cell with vertical plates.

Photograph 2. Examples of electrolytic flotation cells constructed and tested: stainless-steel grid on a
(horizontal) graphite anode; a tortuous path was also arranged for the effluent.
 303

Photograph 3. Examples of electrolytic flotation cells constructed and tested: stainless-steel expanded
mesh (pilot-plant unit).

high pressure pumps, pressurised vessels etc. are needed. In mineral processing, it has been
proposed to combine froth flotation and elecrolytic flotation.
Further, the energy costs of electrolytic flotation can be compensated by other means, such as
the lowering or complete elimination of collector addition. For instance, experiments were
conducted in our Laboratory, comparing electrolytic with dispersed-air flotation in a Hallimond
tube with pure magnesite fines, using a fatty acid as collector. The results obtained were
spectacular: in the pH range of 6-12, electrolytic flotation achieved recoveries over 90%, while
dispersed-air flotation ones were steadily less than 50%.
In conclusion, electrolytic flotation offers many advandages, in comparison with the other
flotation techniques, notably good coverage of the tank surface area, minimum turbulence,
operating conditions easily controlled, reliability and safety of the equipment and an electric field
gradient that aids coagulation, the so-called electrochemical effect. Finally, a large amount of very
small gas bubbles are produced and this was the reason why it has been suggested for fme minerals
processing [1].
304

Photograph 4. Gas bubbles electrolytically produced in a flotation cell. Distance of the two lines :
approximately 1 mm.

REFERENCES

1. Matis. K.A .• and Gallios. G.P. (1986) Dissolved-air and electrolytic flotation. in B.A. Wills
and R.W. Barley (eds.) Mineral Processing at a Crossroads. Martinus Nijhoff. Dordrecht.
pp. 37-70.
2. Zouboulis. A.I .• and Matis. K.A. (1989) Electrolytic flotation of chromium from dilute
solutions. Envir. Techno!. Letters lQ. 601-612.
3. Matis. K.A .. Gallios. G.P .• and Stalidis. G.A. (1988) Applicability of electrolytic flotation
for the recovery of carbonate fines. Chim. Chron. (New Ser.) 17. 29-5l.
3. Mineral Processing
3.1. FEATURES AND FLOTATION OF COMPLEX Cu-Pb-Zn SULPHIDES

LBAYRAKTAR
Demir Export A.S.
Izmir Cad. 2517
Kizilay - Ankara
Turkey

U.A. IPEKOGLU
Dokuz Ey/al University
Dept. oj Mining Engineering
Bornova - Izmir
Turkey

R. TOLUN
I.T.U.
Dept. oj Chemical and Metallurgical Engineering
Istanbul
Turkey

ABSTRACT. Problems arising during flotation of complex Cu-Pb-Zn sulphide ores are in general
of geological origin. The features of these ores are reviewed, stressing that mineralogical studies
extending to process mineralogy are prerequisite for a consistent flotation process. Some examples
affecting flotation are given. Flotation of such orcs often requires intensive testing to develop an
economically viable circuit configuration. Particular attention is paid to the use of depressants,
namely sulphur dioxide and sulphides. Some of the recent research results and industrial
experience on the production of selective copper and zinc concentrates are briefly discussed. A case
study of flotation of a complex Cu-Zn sulphide ore is outlined.

1. Introduction

Copper-lead-zinc production from primary sources has steadily increased, in line with the
increasing demand for these metals.
Complex copper-lead-zinc sulphide deposits, both developed and undeveloped, are of conside-
rable importance in terms of long-term supply of the metals. The reserve is large enough so that
complex sulphide deposits by themselves can theoretically sustain the world's requirements for
primary lead and zinc for several decades. Despite this abundance, exploration activities have been
intensively going on to discover new deposits because they have generally higher intrinsic value
than that of other sulphide deposits such as copper porphyries. It is obvious that if pyrite, being a
major constituent of these deposits, can be marketed then the whole deposit, composed of all
commercial minerals, will be utilised leaving no gangue minerals behind.

307
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 307-330.
© 1992 Kluwer Academic Publishers.
308

Although many excellent reviews of the complex sulphide deposits have been published so far,
most of them treat the subject broadly and deal with the general characteristics of these deposits,
ending with process options covering hydro- and pyrometallurgical routes [1-6]. Yet, the only unit
operation that enables mineral processors to produce selective concentrates of the metals from
complex sulphide deposits today is froth flotation, which has steadily been developed since it was
first applied to the complex sulphide deposits in the late 30's [7]. Several important articles have
appeared in the literature, specifically dealing with flotation [8-15]. It is fair to say that the progress
in the development of froth flotation employed in the treatment of complex sulphides has been quite
satisfactory. However, some obstacles generating from chemical and physical heterogeneties have
to be taken into account, when dealing with these ores. The orcs considered in this study have been
often described as: massive pyrite deposits related to volcanism; stratabound massive pyritic
sulphide deposits; stratiform sulphides of marine and marine volcanic association. Broadly, in the
eyes of a mineral processor they are massive, finely disseminated, refractory polymettalic ores.
This is why they are justly described as complex sulphides.
The objectives of this study are to review:
- the features of complex sulphide deposits;
- the problems encountered during the beneficiation stage; and
- recent developments in their flotation, with particular attention being paid to the depression of the
minerals using sulphur dioxide or sulphites for selective flotation.

I<'igure 1. A typical texture of a Siliceous Ore in quartz matrix(q), concentric pyrite (py) and chalco-
pyrite (cp).

2. Geology

Since many flotation problems in general are of geological origin, their genesis and types are
briefly described. The origin of the ores appears to be related to submarine exhalative or hydro-
 309

thennal activity associated with volcanism or seafloor fracture zones. In unmetamorphosed bodies
the sulphide minerals are often quite fine (minus 0.1 mm), but in metamorphosed ores the minerals
commonly recrystallise to fonn euhedral pyrite grains and bigger crystals of other sulphides which
are several milimeters across.
The unmetamorphosed volcanogenic ores are classified into different types [16-17] as described
below:
a) Siliceous Ore (Keiko type). This is the silicified part of the footwall and in many cases the
remnant structures of the acidic pyroclastic rocks and rhyolite can be observed, while in others
the ores consist completely of cryptocrystalline quartz and no traces of the original rocks can
be found. In the fonner type, pyrite is often large and euhedral while in the latter the mineral is
usually finely crystallised or rarely collofonn in texture. These ores apparently represent
feeder veins and stockworks and consist dominantly of pyrite, chalcopyrite and quartz with
only minor amounts of sphalerite, galena and tetrahedrite. The pyrite occurs as euhedral
grains, subhedral granular stringers and collofonn masses. A microphoto of a typical
"siliceous" ore is shown in Figure I.

Figure 2. Powdery Yellow Ure composed of pyrite (py) and chalcopyrite (cp).

b) Yellow type (Oko Type). This ore type consists mainly of pyrite with varying amounts of
chalcopyrite. Its colour is yellow, due to the dominant pyrite and chalcopyrite. Minor amounts
of sphalerite, galena, tetrahedrite, lead sulphosalts and trace amounts of electrum are dispersed
among the major sulphides. These ores are massive in many cases, but powdery orcs arc not
uncommon where the chalcopyrite content is relatively small. In unmetamorphosed bodies,
the copious pyrite is often quite fine « 100 ~), but in metamorphosed ores pyrite commonly
recrystallises to fonn euhedral grains which are several milimeters across. A microphoto of a
typical "yellow" ore is shown in Figure 2.
c) Black Ore (Kuroko Type). This is a polymetallic ore type which consists mainly of sphalerite,
galena, chalcopyrite, pyrite, fahlerz and baryte. A large variety of accessory constituents occur
310

in this ore and the mineralogy is rather complex. Bomite, electrum and silver minerals are the
common minor minerals. Most of the minerals are very fine, the larger grains of the major
minerals being of the order of 200-300 IJ.m in diameter, but most of the minor minerals are
much smaller, e.g. electrum grains are mostly 3-15 IJ.m. The ores are usually compact and
massive, but bedded, breciated and colloform structures are not uncommon. In ores
unmodified by metamorphism, pyrite occurs as from boids, rossetes, colloform bands, and
dispersed euhedral to subhedral grains. Chalcopyrite disease, i.e. the presence of chalcopyrite
as micron-sise blebs in grains of sphalerite, is not uncommon in black ores, indicating that
liberation is unlikely. In these cases, sphalerite will probably behave like chalcopyrite in
flotation, because of a considerable surface coverage. A typical texture of a "black" ore and
chalcopyrite disease are shown in Figures 3a and 3b, respectively.

Figure 3. (a) Fine-grained minerals of a Black Ore: chalcopyrite (cp), sphalerite (sp), galena (g), fahlerz
(I) and pyrite (py). (b) Chalcopyrite disease in sphalerite in a Black Ore.

d) Pyritic Ore (Cyprus Type). These ores are composed of massive banded to fragmental pyrite
with small amounts of interstitial chalcopyrite and other base metal sulphides. The pyrite is
present as friable masses of subhedral to euhedral, commonly zoned grains, as colloform
banded masses, and as framboids. Marcasite is admixed with the pyrite and often appears to
have replaced the pyrite. Chalcopyrite occurs as anhedral interstitial grains and as inclusions in
the pyrite; sphalerite occurs similarly but is less abundant. Secondary covellite, digenite,
chalcocite, and bornite occur as rims on and along fractures in pyrite and chalcopyrite.

3. Features of the Complex Sulphide Deposits

The complex sulphide deposits occur widely in all parts of the world. Although each ore body
posseses its own character, presenting different problems in mining and processing, nevertheless
 311

there are some similarities among the volcano-sedimentary complex sulphides. Their features
affecting their assessment are briefly reviewed.

3.1. MINERALOGY

Broadly, the mineralogy of complex sulphide ores is described as an association of chalcopyrite,

sphalerite and galena, generally finely disseminated in the dominant pyrite (80-90%). The
constituents of the gangue, if any, are baryte, quartz, carbonates and phylites [1,18-20]. The asso-
ciation of the main minerals as seen in Figure 3 is extremely intricate, the size of the valuable
minerals presenting a wide size range from extremely fine (a few /IDl) to fine (a few tens of /IDl).
Metallurgically significant amounts of impurities are often present in the sulphide matrix, of which
arsenic, antimony, bismuth and mercury are the most common and troublesome. Besides, in most
of the known deposits, the base metals occur with economic amounts of precious metals: gold and
silver.

3.2. INTRINSIC VALUE

The mean metal contents of the complex sulphide deposits are as follows [1,6] : Cu, 0.3-3% ; Zn,
0.2-10% ; Pb, 0.3-3% ; Au, 0-1 g/t; Ag, 30-100 glt ; S, 20-50%. The developed complex
sulphides have a high intrinsic value per tonne of are compared with almost all other metallic are
deposits. The combined values may be as high as $ 300 per tonne. No deposit of other type of
sulphides is known having a value of more than $ 100 per tonne. The high value of these ores
makes the complex sulphides attractive to exploit because of the potential revenue per tonne of are
mined and per tonne of concentrate transported.

~Massive are
~Stockwork are
E!1 Oocitic Tuff
[!] Tu ff o 100m

~ Basalt
~ Gypsum

Figure 4. Geological cross-section of the Cayeli ore deposit, massive complex Cu-Zn sulphide [19]
312

3.3. RESERVES

The reserves of copper-lead-zinc ores have been estimated to be more than 3000 million tonnes [3].
The metal contained in these reserves is over 38 million tonnes of copper, over 74 million tonnes of
lead and over 180 million tonnes of zinc. Although base metals have been very much in the shadow
of gold exploration, a number of significant discoveries has been made during the past decade. A
striking example is the deposits found in Tasmania, at Mount Read volcanics. The deposits range
in size from less than 100,000 tonnes to 110 million tonnes. Polymetallic, Pb-Zn-Cu-Au-Ag
deposits have a mean of 10 million tonnes averaging 21 % Pb+Zn, 160 gJt Ag and 31 glt Au [21].
It is believed that the addition of new reserves is almost equal to consumption. Hence, with the
mine production rate of 1989, the estimated reserves of complex sulphides will suffice for 5 years
for copper, 30 years for lead and 35 years for zinc. Furthermore, a substantial amount of the metals
is produced from complex orcs every year, e.g. the 1980 figures for free-world total metal
production show that 3.6% of the total copper, 21 % of total lead and 32% of total zinc are extracted
from copper-lead-zinc complex ore reserves [3J.

3.4. SIZE AND SHAPE

The complex are deposiL~ are relatively small. There are only a few deposits containing more than
one million tonnes of copper, e.g. Cayelli in Turkey; Vald'or in Canada etc. Most have between
50,000 to 500,000 tonnes, e.g. Scandinavian complex orcs; Lahanos in Turkey etc. A geological
cross-section of a typical massive complex Cu-Zn sulphide orc in volcano-sedimentary sequences
is seen in Figure 4. Although, in some cases, stockwork occurrences are seen beneath the massive
body, they are stratabound in origin with sharp diving lines between mineralised and barren zones
(as seen in Figure 4). Irregular distribution of lenses with a remarkably high metal content within
the body may frequently be seen.

4. Mineralogy and Texture of the Complex Sulphides and their Implications

From the foregoing information about the general characteristics of complex ores, it appears that a
very detailed ore microscopy must be applied to these ores to predict liberation size and potential
beneficiation problems, since studies in earlier stages related to ore genesis or paragenetic sequence
of mineral deposition etc. rarely extend to the prediction of beneficiation problems. The advantages
of conducting a detailed are microscopy during the development of an are body are various:
a) quantitative information regarding libcration size, which is one of the crucial criteria in
complex ores processing, is ohtained. The question of how fine should the are be ground may
thus be answered; at least the approximate fineness of the grind is determined.
b) the quality of concentrates to be produced may be predicted. For instance, copper concentrate
produced from complex ores containing substantial amounts oftetrahedrite and tennantite are
always contaminated by antimony and arsenic. This means that if their contaminant content is
higher than a certain limit imposed by the smelter, they may be subjected to penalties. From
the plant economics point of view, this is most important, since this penalty may be as high as
$50 to $75 per tonne of so called "dirty concentrates", containing substantial amounts of
arsenic, antimony, bismuth or mercury. Furthermore, these minerals make copper-lead
separation difficult, if cyanide is used as a depressant for copper, because they can hardly be
depressed by cyanide [12]. So, they may report to the wrong concentrate and this leads to
 313

another source of loss other than direct loss.

c) Finally, precious information regarding the floatability of minerals may be gathered by ore
microscopy, e.g. high iron and manganese contents of sphalerite impede its floatability and
this leads to activation and eventually recovery problems [22,23]. Another well known
problem, which is often ecountered in even slightly oxidised complex ores, is chalcocite rims
on pyrite, which increase iron content or reduce the grade in the copper concentrates. The
presence of secondary copper minerals (chalcocite and covellite) in the ore is of paramount
importance in process design, because these minerals are dissolved by sulphurous acid or
cyanide solutions. The presence of small quantities of covellite (0.1%) will impede zinc
depression by sulphur dioxide [11]. Examples are numerous since the convincing role of
process mineralogy has been well understood in process design and in plant operations of the
complex sulphides [22,24,25].

4.1. CLAY PROBLEM

It is not uncommon that the unmetamorphosed complex sulphides bear clay zones or bands within
the body. It is a well known fact that clay minerals spoil flotation for many reasons [59]. It is
worth mentioning briefly, however, that one of the most important points for a mineral engineer
designing complcx sulphides processes is that core samples, chipped pieces of the ore and the like
do not normally reflect clay problems. For instance, during drilling, a solution (in most cases
water) is circulated to cool down the bit and this washes off clay. So, the free samples of clay may
mislead the mineral engineer during flowsheet design. The clay problem may be overcome by
adding a washing stage before grinding. In fact, the washing operation will serve another purpose,
which is the elimination of water-soluble oxidation products (salts), occurring within the ore body
either before or after mining. The oxidation problem will be dealt with in the following section.

4.2. ULTRAFINES PROBLEM

Fine and intricate mineralogical texture of the complex sulphides presented in Figures 1-3 naturally
require fine grinding below 50 Ilm. This liberates the discrete particles but causes a substantial
reduction in selectivity and flotation rate; these are the sporadic problems encountered when trying
to achieve optimal economic balance between fineness of the grind and subsequent flotation
operation.
Fine grinding means high energy consumption. For instance, a complex Cu-Zn ore, given in the
case study below, consumes 21-23 kWh/t of energy at a fineness of 80% -45 Ilm. Furthermore, in
a new concentrator under construction, designed for treating a complex Cu-Zn sulphide ore, an
energy consumption of 26-28 kWh/t is expected at 90% -38 !lID from laboratory and pilot plant
studies.
Fine grinding inevitably produces ultrafines - which include all particles finer than 5-10 !lID
[26,27] - as defined in term of the size of particles, causing problems in flotation by virtue of their
fineness. Flotation rate of fines is low relative to that of intermediate size (30-100 Ilm). The
relationship between particle size and flotation rate constants of the minerals is shown in Figure 5.
In several plant operations, the relationship between particle size and flotation efficiency has
been investigated [29]. The pattern of notation behaviour of sulphide minerals was found to be
consistent. As seen in Figure 6, there is a plateau covering a range of sizes, over which the
efficiency is maximum whereas, at sizes larger or smaller than this, recovery decreases.
In general, it can be said that 10 Ilm is somewhat critical or very near to the critical size, below
 0 Zn 5
pH8
o Pb 5
pHS
t:. C u 5
pHS
"l "l Fe 52 pHS
• Cu Fe 52 pHI
100 mg KEX

5 10 20 50 100
AVERAGE PARTICLE SIZE(microns)

Figure 5. The relationship between particle size and flotation rate constants of several sulphide minerals
with excess collector [28].

100

90

>- 80
a:
lLJ
70
>
0
u 6
lLJ
a:
50 ........ CuFe52
'/- rr--o Zn 5
40 O--OPb 5

2 5 10 20 50 100 200 500

Particle Size, micron

Figure 6. Dependence of industrial flotation efficiency on particle size for several minerals (adapted from
ref. [29]).

which flotation effiCiency decreases. Owing to the fact that ultrafines show lower flotation kinetics
together with other unfavourable reasons [27J. significant losses of valuable minerals in the slime
fraction are commonly experienced. Table 1 shows metal distributions in various size fractions of
cleaner and final tailings of a complex copper-zinc sulphide ore. It is clearly seen that the -10
micron fraction is responsible for almost half of the copper losses in the final tail. A compounding
negative effect of ultrafines. apart from their slower kinetics. is that a substantial increase in
 315

entrainment of gangue has been observed. Figure 7 illustrates the relationship between particle size
and the recovery of gangue in the NBHC lead flotation circuit [29].

TABLE 1. Copper content in size fractions of flotation products of a massive complex

1 h'de ore
SUIPI!
Products Particle
of Copper Size Wt% Cu% Dist. %
Circuits (~)
+63 3.5 6.57 2.8
+38 18.0 5.98 12.9
2nd Oeaning +20 26.8 5.79 18.6
Tail +10 12.6 10.81 16.3
-10 39.1 10.56 49.4
100.0 8.35 100.0
+63 12.0 4.21 14.8
+38 20.5 3.96 23.7
1st Oeaning +20 34.7 2.74 27.8
Tail +10 13.2 3.16 12.2
-10 19.6 3.75 21.5
100.0 3.42 100.0
+63 4.8 1.96 5.2
+38 16.7 1.41 13.1
Final +20 30.0 1.27 21.2
Tail +10 17.5 1.39 13.5
-10 31.0 2.73 47.0
100.0 1.80 100.0
..
CondlllOns: 90 % - 38 ~m. gnndmg pH 10.5-11.0 WIth hme, flotatIOn pH 6.5-6.8 wIth
S02 (g), DTP+KAX 120 glt, MIRC 40 gIl, flot. time 13 min.

>-
:; 3
'>

..
o
U

tr 2

'"
:J
01
c:
o 1
C>

O~~-+-+ __ +-~ __ __
~ ~

2 3 5 10 20 50 100 200
Particle Size, micron

Figure 7. Relationship between particle size and recovery of gangue [29].

316

The magnitude of the effect obviously depends upon the volume of the ultrafines, i.e. on the
fineness of grinding required. The more finely disseminated and complex the sulphide ore the more
serious the effect.

4.3. OXIDATION PROBLEM

The most important characteristic of sulphide minerals is their instability conditions. They
transform slowly into stable oxidised minerals, e.g. PbS -> PbS04, Zns -> ZnS04 etc. Con-
siderable changes on the surface of sulphide minerals occur during their oxidation. The oxidation
of a sulphide mineral at low temperature is a process in which the metal atoms move into the
surrounding solution to become aqueous cations, accompanied by a stepwise decrease in the metal
to sulphur ratio of the remaining solid phase; when sulphur is finally left over, it undergoes a series
of reactions to be oxidised, finally giving sulphate ion. This expressions may be presented for a
simple sulphide, MeS, by the following reactions [30] :

MeS -> Me2+ + S + 2e-

4S + 60B" -> S2032- + 2 S- + 3 H2O
S2- -> S +2e·
2 S2032- -> S4062- + 2e·
S40 62- -> sol-
Furthermore, metal cations may precipitate on the surface of the mineral as MeSZ03, MeS04,
Me(OH)z, or Mee03 etc., if the appropritate conditions prevail. In the following series, some of
the sulphides are listed in a diminishing tendency toward oxidation:

FeAsS > FeSz > CuFeSz > ZnS > PbS> CuzS

Two adjoining solid phases are much more readily oxidised than a single substance. Thus, galena,
sphalerite and covellite oxidise from eight to twenty times faster in the presence of pyrite, because
of their conversion to oxidised compounds through electrochemical contact with the pyrite-oxygen
cathode. The oxidation species may increase with an increase in pH, with the concentration of
dissolved oxygen and with the time of minerals in the medium.
Sulphides of old geological formations are less readily oxidised than those from younger
formations. Since volcanogenic complex sulphide deposits under consideration are relatively
younger than the other simple sulphide deposits, e.g. the age is Miocen in Japan, Upper
Cretaceous in Turkey, etc. So, the unmetamorphosed complex sulphide deposits are known to be
very susceptible to oxidation. Washing tests with distilled water on Kuroko samples were carried
out so as to find out the intensity of oxidation. These tests indicated that heavy metal ions dissolved
to a large extent as seen in Table 2.
Although oxidation has a definite role in the surface phenomena and affects floatability when it
proceeds beyond a few tenths of surface layers, the resulting surface may not respond to sulphide
mineral flotation reagents. Furthermore, the oxidation products such as Cu 2+ and Pb 2+ have the
most deleterious effect on the selectivity between copper sulphides and sphalerite. Recently, a
series of solubility and flotation tests were carried out on a large number of samples of complex
massive Cu-Zn sulphides of volcanogenic origin [32]. Fresh samples were taken from the stops
just after blasting and the aged samples were exposed to the ambient conditions as received at least
a month or so later. Table 3 shows the results of these tests. It is interesting to note that tre
 317

solutions of neither the aged nor the fresh samples of the black orcs contain appreciable amounts of
copper. Although it was not proven experimentally, a possible explanation for the absence of
copper ions in the pulp of black ores may be that abundant sphalerite (approx. 30% of the ores)
absorbs almost all copper ions from the solution. Hence, the selectivity between copper sulphides
and sphalerite decreases notwithstanding all attempts to enhance the selectivity by using S02.

TABLE2 W as h·mg test resu ts 0 fK urok 0 samples [31]

Pulp Temp. Time Initial Final Concentration of dissolved species
density (0C) (hr) pH pH
(solids %) Cu Zn Fe TFe S04
20 20 0.5 5.6 4.5 46.3 197.4 41.9 51.25 1889.9
20 20 3.0 5.6 4.5 31.7 168.7 38.2 61.25 1894.6
40 40 0.5 5.6 4.5 13.3 216.9 140.1 150.5 2147.6

The amount of copper in the solution generated by the oxidation of yellow ores is much higher
than the amount of copper sulphate used in the industry for the activation of sphalerite, i.e.,
177 ppm Cu 2+ is roughly equivalent to 1400 g/t CUS04 5H20.
As seen in Table 3, sphalerite is oxidised faster than chalcopyrite, hence, the zinc content of the
solution is always higher than its copper content, regardless of the Cu/Zn ratio in the feed. The
grade and recovery of the copper concentrate deteriorate by ageing more than that of the zinc
concentrates. The effect of ageing and oxidation on flotation recoveries of copper, zinc and lead
concentrates produced from complex massive polymetallic ores were also elaborated in a recent
paper [15]. Figure 8 clearly shows that the most rapid deterioration in metal recovery occurred for
lead and that zinc was the least affected.

a.
:; 1600 100 VI
a. ~
o >~
o
eo cr..u
c

E
0-
a.
~
c
50 ~
e
-..
.~
u
~

c 40 ~
u
c 5 10 15 20 2S 30
o
U AGEING TIME IN DAYS

Figure 8. Effect of ageing of massive sulphide are on solubility of heavy metals and on flotation
recovery of indi vidual minerals [151.

From the mine and plant management point of view, these findings necessitate an action as rapid
as possible in the operation, in order to be able to produce marketable selective concentrates. It is
obvious that such ores should not be stockpiled more than a few days, otherwise metallurgical
 w
00

-~---- -- - - - - - - -- ~o--- ----~ - ---.-.---- - - ---------- -- -- --r--- -------. - -- - - - - --- ------

Black ore I Black ore II Yellow ore! Yellow ore II Yellow ore III

Cu=5.24%, Zn=19.10% Cu=3.75% Zn=21.60% Cu=6.05% Zn=2.20% Cu=5.42% Zn=1.68% Cu=6.74%, Zn=3.85%

AGED FRESH AGED FRESH AGED FRESH AGED FRESH AGED FRESH

Cu Zn Cu Zn Cu Zn Cu Zn Cu Zn Cu Zn Cu Zn Cu Zn Cu Zn Cu Zn
Solubility
tests! vvm 0.1 674 0.1 18 0.2 584 0.1 27 72 887 18 154 177 793 22 33 0.2 3 0.2 2

Cu-conc. 22.1 13.9 24.6 4.7 23.7 6.2 25.1 5.9 20.9 3.4 21.8 2.0 23.1 2.7 24.7 1.8 19.1 3.2 20.6 2.4

Zn-conc. 3.0 40.5 2.4 42.1 2.0 52.7 2.4 50.9 - - - - - - - - 1.6 47.7 1.9 48.3
Cu-Rec.2 (%)
in Cu-conc. 54.4 63.9 58.5 71.8 79.2 78.9 80.1 83.5 84.4 86.0
Zn-Rec. 2 (%)
in Zn-conc. 71.7 70.0 78.4 79.9 - - - - 39.2 41.3

1. A 100 g dry ground sample at the fineness of minus 38 microns was stirred in a beaker containing 200 ml deionized water for an hour at
25 oC. All equipment were glass or teflon.
2. Final concentrate recoveries of batch tests which were carried out at identical conditions for a given ore.
 319

results will be exacerbated. In the case of faster oxidation and/or when a stockpile has to be made
for a period of time, then a washing operation in the crushing section of the plant may be a remedy
to remove water soluble salts and primary slimes adversely affecting flotation, as it has been
employed in Demir Export's Harsit flotation plant shown in Figure 12.

S. Flotation

Flotation is still the most widely employed method for primary recovery of all base metal
sulphides. It is more than a century since flotation started being practiced in the mining industry.
So far major improvements have been witnessed in flotation reagents, flotation machines and in
plant control [35,36]; there are however some sufferings still to be alleviated, especially in the case
of complex sulphide flotation. Flotation itself is a complex process and unfortunately its complexity
is even more aggravated when it is applied to the complex ores.

TABLE 4. Grades and recoveries of the concentration produced from complex sulphides in
several concentrators in the world
Copper concentrate Zinc concentrate Lead concentrate
Concentrators Grade % R% Grade % R% Grade% R% Ref.
Cu :at Pb Cu :at Cu Pb :at Pb :at Cu Pb
Matsumine 21.0 5.0 4.6 75.4 55.0 1.3 2.5 91.0 52.0 6.3 5.2 66.9 31
(Japan)
Uchinotai 26.0 3.8 3.1 81.8 55.0 1.1 2.5 92.5 3.5 3.8 58.1 70.0 37
(Japan)
Woodlawn 19.3 7.7 10.1 44.2 50.6 0.6 3.5 67.2 30.1 12.2 3.7 56.2 38
(Australia)
Brunswick l 21.0 3.5 6.0 53.7 52.0 0.3 2.0 74.2 34.0 5.0 0.4 55.0 39
(Canada)
Garpenberg 19.8 10.7 11.8 78.7 53.7 0.3 2.8 79.4 58.5 9.2 0.7 67.3 40
(Sweden)
Meggen 2 - - - - 53.5 - 1.2 90.5 45.0 5.2 - 37.4 41
(Germany)
Magmont 30.2 0.4 5.6 76.4 59.6 0.3 1.6 75.3 76.5 1.5 0.4 96.1 42
Mill (USA)
Pyhasalmj3 25.0 4.4 - 91.2 50.4 0.3 - 88.6 - - - - 43
(Finland)
Harsit 3 19.6 2.5 1.9 84.0 47.6 3.6 - 56.0 - - - - 44
(furlcey)
I Bulk concentrates are not mcluded
2 Copper content too low, not recovered
3 Lead content too low, not recovered

In addition to the problems elaborated in the preceding section, one of the main problems
specific to complex sulphide ores arises from the similarity in flotation behaviour of copper
sulphides, galena, sphalerite and sulpho-salts. However, all these problems are serious enough to
hinder production of base metal concentrates only in a few cases today, because an adequate
320

repertory of specific depressants has been developed, including cyanide, chromate, sulphites and
some organic mixtures [14,15]. If adequate liberation is obtained without producing too much
ultrafines, selective concentrates suitable to current smelting and/or hydrometallurgical extraction
techniques can be profitably produced. However, grades and recoveries of the concentrates are
relatively poor in some cases, as seen in Table 4. Both grades and recoveries of the concentrates
show a wide range, owing to the fact that the intensity of deleterious effects elaborated in the
preceding sections change from one deposit to another. On the other hand, precious metals contents
of the copper and the lead concentrates are also variable. Nevertheless, they make a significant
contribution to the plant economics. For example, in the Uchinotai concentrator, the lead
concentrate contains 15.5 glt gold and 1500 glt silver [37] and, in Brunswick, the copper
concentrate contains 3000 g!t silver [39].
Common practice in flotation of the complex Cu-Pb-Zn ores is to selectively depress sphalerite
and iron sulphides, using S02(g) or sulphites (Na2S203, NaHS03 or Na2S20S), and float a
copper bulk concentrate. Application of sequential flotation of copper, lead and zinc is also possible
in some ores, in which galena is the slow-noating mineral [46]. After the bulk concentrate is
floated, the sphalerite contained in the tailings of the copper-lead notation is activated by copper
sulphate and floated in the usual manner. After zinc flotation, iron sulphides may be floated in
slightly acidic pulp, if there is a market for pyrite concentrate. Separation of copper from lead in a
cleaned bulk concentrate is accomplished by one of three processes generally referred to as the
cyanide, the chromate and the sulphur dioxide methods. The choice of method depends upon the
relative abundance of the copper and lead minerals. In cases where the amount of lead in the bulk
concentrate is more than twice the amount of copper, depression of the lead is usually preferred.
For the depression of the galena the use of sodium dichromate or S02 is most common. If the
amount of copper is more than the amount of lead, depression of the copper is carried out by the
use of sodium cyanide or the cyanide-zinc complex [8,10,45].
Flotation of complex copper and zinc ores is basically similar to that of Cu-Pb-Zn ores; the sole
difference is that there will be no copper-lead separation. All these techniques have been superbly
elaborated in the literature [8,10-12,14,45]. It is noteworthy to mention that due to the growing
concern about the environment, the use of cyanide for the depression of zinc and iron sulphides,
and the use of chromate for the depression of lead are becoming obsolete. For these purposes
sulphur dioxide or various sulphites (Na2S203, Na2S20S or NaHS03) are now widely used
throughout the world. However, the depression mechanisms of these chemicals have not yet been
fully understood. Brief information about the depression mechanism and the use of sulphur dioxide
or the sulphites are given bellow.

5.1. THE USE OF THE SULPHlTES - THEORY AND PRACTICE

The use of sulphoxy compounds for sphalerite and pyrite depression was practised as early as
1919 in a lead-zinc concentrator in Utah [47]. Since then, alkaline sulphite and other related sul-
phoxy compounds, bisulphite, metabisulphite and gaseous sulphur dioxide have been widely used
[8]. Sulphite, whether added as sodium sulphite (Na2S03), sodium hydrogen sulphite (NaHS03),
sodium bisulphite (Na2S203), sodium metabisulphite (Na2S20S), or S02.gas or solution, their
species in flotation pulps are thought to be dissolved sulphur dioxide, S02 (aq) (not sulphurous
acid), hydrogen sulphite (HS03-), sulphite (S03 2-), disulphite (S20S2-), thiosulphate (S2032-) and
sulphate (S042-)with peroxomonosulphate (SOS2-) being of possible interest [48]. As shown in
Figure 9, the pH of the solution determines the equilibrium distribution of the species between
S02, HS03- and S03 2- and any of these can be the predominant species in a flotation pulp.
 321

50 2 -
75 3

c:
0

~ SO
~
IL

~ 25
0
~

4 6 8 10 12 14
pH

Figure 9. Proportion of species present as a function of pH [48].

The mechanism of depression by sulphoxy compounds has been studied and several possible
mechanisms have been suggested. Gaudin said: "One possible function of sulphur dioxide in
depressing activated and collected sphalerite is to reduce the cupric-ion concentration. The entire
problem of the mechanism of depression by sulphoxy compounds needs study and it promises to
be difficult" [49]. Summarising pyrite depression, M.e. Fuerstenau stated that sulphites are more
reducing than the dixanthogen-xanthate couple; thus, formation of dixanthogen, which is
responsible for pyrite flotation, is inhibited [50]. Galena depression by sulphitcs and chromate ion
was investigated in detail [51]. The reactions of sulphite ion with xanthates and the depressing
effect of S02 or sulphite ion on sphalerite and pyrite were also studied [52,53]; the findings
relevant to the complex Cu-Pb-Zn sulphide flotation are summarised below:
a) Oxidation of galena is required to enchance the depressing effect of sulphite ion. The
depressing action by sulphite ion arises primarily from the deposition of lead sulphite films,
which render the galena surface hydrophilic.
b) In the presence of sulphite ions, decomposition reactions of xanthates take place, depending
upon the pH and the dissolved oxygen concentration of the pulp. The rate of decomposition is
very rapid at neutral pH (pH 6-6.5). Although the exact reaction involving sulphur and
xanthate species are complex, however, the following generalised reaction was proposed [52]:

After xanthate had been decomposed, sulphite ions were oxidised to sulphate ions.
c) The selective depression of sphalerite by SOz is related to the relative rates of xanthate
absorption in comparison to the rate of xanthate decomposition.
d) The depressing action of sulphite on sphalerite activated by copper sulphate is due to the
formation of hydrophilic zinc or calcium sulphite films on the mineral particles in the presence
of zinc sulphate or lime.
e) Formation of insoluble heavy metal sulphite salts and subsequent reduction in the heavy metal
322

ion concentration in the pulp, thus preventing unintentional activation.

1) Limited collector absorption occurs by control of the redox potential. According to a bench
scale study [54], depression action of sulphur dioxide on the sulphide minerals decreasses in
the following order: Pyrite, Sphalerite, Galena, Chalcopyrite. The depressing effect of
sulphur dioxide and sodium sulphite on the sulphide minerals was found to be the same [54].
In practice, sulphur dioxide or the sulphites are widely used in the following stages of
flotation of Cu-Pb-Zn or Cu-Zn sulphide ores.

5.1.1. Primary Copper-Lead or Copper Flotation. The purpose of using sulphur dioxide or
sulphite in this stage is to depress sphalerite and pyrite. Table 5 shows some test results concerning
the effect of S02 gas on the Cu-Pb bulk flotation of Kuroko [31]. In the test, the extent of S02
injection was measured by the pH variation of the pulp. After the S02 injection, the pH was
regulated with lime addition.
Table 5 shows that acidifying the pulp by introducing S02 gas was sufficient only for sphalerite
depression. For instance, at pH 4.0, 82% of the sphalerite and 49.8% of the pyrite were
depressed. To depress the pyrite efficiently, neutralisation of the pulp with lime was required. The
best flotation results were obtained by acidifying the pulp up to pH 4.0 by introducing S02 gas
first and then by neutralising the pulp with lime.

TABLE 5. Roatability of copper minerals, galena, sphalerite and pyrite in the Cu-Pb
bulk flotation using S02 injection [31].
pH pH Distribution of the minerals in the Cu-Pb bulk conc.
S02 gas Ca(OHh Cu minerals Galena Sphalerite Pyrite
3.0 5.0 86.0 71.6 18.3 25.2
3.0 6.0 89.6 81.8 25.9 18.9
3.0 7.0 89.4 89.5 34.1 21.3
4.0 5.0 87.4 71.2 18.4 27.8
4.0 6.0 90.2 85.2 29.2 22.2
4.0 7.0 89.8 86.0 31.6 19.5
5.0 5.0 87.9 70.9 17.1 40.9
5.0 6.0 86.5 73.4 18.5 35.4
5.0 7.0 85.8 73.0 19.9 29.3
4.0 - 88.0 65.2 18.1 50.2
3.0 - 86.9 75.2 17.9 54.6
Pulp density: 34% solids, 24 °C: , AF 208 (100 gil) and DF 250

5.1.2. Copper-Lead Separation. The copper-lead bulk concentrates generally contain considerable
quantities of zinc and iron sulphides. It was found out that increasing the pulp temperature was an
effective means of depressing galena and floating chalcopyrite [55]. The separation procedure
commonly employed is the starch, dextrin or guar-S02 method in a heated pulp up to 70 'C by
steam injection. The ability of this technique to depress successfully galena and sphalerite from
copper concentrates in the industry has been well proven [55-57].
It is believed that desorption of collector from the galena surface and preferential oxidation of the
surface are caused by heat, whereas chalcopyrite is not affected.
 323

5.1.3. Reverse Flotation in Zinc Circuits. Zinc concentrates contaminated by readily floatable iron
sulphides (pyrite and pyrrhotite) may be upgraded by reverse flotation, which is achieved by
conditioning the pulp with S02 in the pH range of 4.5-5.0 at an elevated temperature (70-80"C) to
depress sphalerite completely. After conditioning, iron sulphides are then floated with xanthates
and the non-float is the upgraded zinc concentrate. Industrial application of the reverse flotation is
quite common. For example, it has been successfully employed in Meggen, Germany and in
Folldal, Norway [58].

6. A Case Study of Flotation of a Complex Cu·Zn Ore

A typical volcanogenic sedimentary, finely disseminated and massive complex copper and zinc
sulphide ore was discovered in the Eastern Black Sea region of Turkey in 1988. The drill core
samples were delivered to the company's mineral processing laboratory. A composite sample
containing 6.74% Cu, 3.85% Zn, 0.95% Pb and 92 g/t Ag was prepared for flotation studies.
Before the flotation tests, an intensive mineralogical investigation was carried out so as to predict
potential processing problems. The intricate association of the ore minerals are clearly seen in
Figure 10.
Chalcopyrite, the major economic copper mineral, is almost completely associated with and
included in pyrite and/or sphalerite, which shows the typical texture of complex copper-zinc
sulphide ores. From the point of view of sufficient liberation, the study of ore mineralogy indicated
that grinding finer than 45 11m was required.
Solubility tests were also carried out before flotation experiments, in order to find out the degree
of oxidation, which determines reagent scheme and facilitates the interpretation of flotation test
results. It was fortunate that a negligible oxidation was determined as seen in the preceding section
(Table 3, Yellow Ore III).
Batch flotation tests were carried out in a Denver D-12 laboratory flotation machine. S02 was
injected into the flotation machine through the air feeder. A standard flowsheet was applied for
flotation tests, Le. sphalerite and pyrite were depressed before copper flotation, then sphalerite was
floated from the tailings of copper flotation.

6.1. THE FINENESS OF GRIND

To assess the fineness of grind, a range of possible grinds between 92% and 62% -45 Ilffi were
tested under identical flotation conditions. Results are presented in Figure II.
As seen in Figure II, the highest recovery being 82% was obtained with a copper grade of
19.2% Cu at a fineness of 80% -45 11m. As a finer grinding was rendering a higher copper grade,
the copper recovery decreased considerably. The reason for obtaining the highest recovery at
comparatively medium grind was due to the fact that interlocked particles, exposing enough chalco-
pyrite surface for flotation, were floated and also copper losses increased with finer grinding, as
described in the preceding section.

6.2. AMOUNT AND TYPE OF DEPRESSANTS

A special attention was paid to depression of the sphalerite and the pyrite, since success of the
process depended upon efficient depression of these minerals. The depressants which were tested
324

Figure 10. Finely disseminated massive complex copper and zinc sulphide ore. Chalcopyrite (cp), sphale-
rite (sp), pyrite (py).

22 100 6
u u
c: c:
0
0 21 u 5
'"
01
/o-.~
. 0 u
c: ,/
...c: c:

.
'c0 20 90 4 0

..
./

-;:;
p v

;; ~
---
-....
19 3 0.
.<:; c: 0.
.A-----t'- 0

..
u

.
....
.<:;
18
..- " "
.-,," ----A-_ 80 .
>-
~

>
2
...
.<:;

..
0
~
,,- ,
.S
c: 17 0--0 Cu '/, u
;;!
b---8 Cu
~

recovery
'/- 16 cr--o Zn '/,
:>
ON
C

:> 0 u
u
92 IlO
F1neness of Flot. Feed, Percent minus 4S microns

Figure 11. Effect of the fineness of grind on copper grade and recovery.
 325

Cu Cone. Zn Conc.

Figure 12. Complex Cu-Zn sulphide flotation plant flow sheet (Demir Export, Turkey). I-Ore bin, 2-
Jaw crusher, 3-Double-deck screen, 4-Cone crusher, 5-Spiral class, 6- Settling tank, 7-Fine ore bin, 8-Rod
mill, 9- Spiral classifier, lO-Ball mill, 11- Hydrocyclone, 12, 16, 17-Conditioners, 13-Cu-Rougher cells,
14-Cu-scv. cells, IS-Cu cleaning cells, 18-Zn-Rougher cells, 19-Zn scv. cells, 20-Zn cleaning cells, 21-
Cu- conc. thickener, 22,24-Drum filter, 23-Zn-conc. thickener, 2S-Tailing pond.
326

for the sphalerite and the pyrite were lime, S02, Na2S205, NaCN, NaCN + ZnS04 .
Results obtained for comparing the effectiveness of the depressants are pressented in Table 6.
Better results were obtained with lime-S02, lime-Na2S20S, lime-S02, ZnS04 and lime-Na2S205-
ZnS04 combinations as far as grade and recovery of the copper concentrate are concerned. It is
interesting to note that the depressing effect of S02(g) and of Na2S205 at neutral pH (6.5-6.7)
appeared to be the same and the addi tion of cyanide did not seem to decrease the zinc content of the
copper concentrate.

TABLE 6. Effect of depressants on grade and recovery of the copper concentrate.

Depressant Dosage Addition Rougher Copper Rougher Concentrate
(g/t) points flotation Gi Zn Pb Ag R
pH % % % ppm %
Lime 3450 Mill 11.9 15.7 6.2 3.0 - 77.9
SOz - Conditioner 6.5 18.5 3.6 2.1 174 84.3
NaCN + ZnS04 150+300 Mill 7.2 16.4 6.8 2.6 163 89.6
Lime 5300 Mill 6.7 17.5 3.2 2.5 167 90.2
SOz - Conditioner
Lime 3200 Mill 6.5 16.4 3.7 2.8 168 90.6
Na2S20S 1500 Conditioner
Lime 4000 Mill
NaCN 250 Mill 8.0 17.4 5.7 2.2 - 76.3
SOz - Conditioner
Lime 3300 Mill
NaCN +ZnS04 150+300 Mill 6.5 16.2 3.2 2.1 184 89.1
SOz - Conditioner
Lime 3300 Mill
Na2S20S 1600 Conditioner 6.7 16.9 3.3 2.3 181 89.8
ZnS04 300 Conditioner

Therefore the use of cyanide was ommitted in further studies. The results show that no
depressant com bination allows the zi nc content of the copper concentrate to be reduced lower than
3% Zn. This is due to the fine intergrowns between chalcopyrite and sphalerite as microscopic
studies indicated. The lead content of the copper concentrates was consistently lower than 3% Pb
and the high Cu/pb ratio, together with a very fine dissemination of galena, did not allow the
production of a lead concentrate economically.
Having studied other parameters, such as the amount and type of collector, flotation pH and
kinetics, etc., conditions to be experimented in the plant were determined. Metallurgical results
obtained under optimum conditions are presented in Table 7.
Relying on the promising results of the bench scale studies, as shown in Table 7, an existing
concentrator having a nominal capacity of 400 tpd was altered to process the ore. The actual
flowsheet of the plant is given in Figure 12. A period of two weeks was sufficient to achieve the
objective, thanks to a great deal of data accumulated during intensive bench scale studies, which
consisted of over 100 flotation tests. The results obtained in the plant trial are given in Table 8.
As seen in Table 8, the efficiency was vastly improved when zinc sulphate was used in
conjunction with sodium metabisulphite. Since then, the sphalerite and the pyrite have been
succesfully depressed with a combination of lime-Na2S205-ZnS04. For instance, over 6000
 327

tonnes of copper concentrate containing 19.6% Cu, 2.5% Zn, 1.9% Pb with a recovery of over
80% was produced under the aforementioned conditions. The use of Na2S20S instead of S02 (g)
appeared to be advantageous, since a special investment is not required and it is easy to handle.
Unfortunately the intricate ore texture is responsible for relatively low recovery of the zinc
concentrate.

UfgJC results 0 b
TABLE 7 B ench sc alemetall'al ' d und er optImum con IUons.
tame
Products \\t Assay Distribution
% OI Zn Fb Ag OI Zn Fb Ag
% % % (J!)t.) % % % %
Cu-Conc. 28.5 19.1 3.2 2.2 168 84.4 25.6 54.4 47.4
Cu Scv. Cone. 3.1 5.1 6.0 0.7 167 2.5 5.3 2.0 5.1
Cu Middling I 2.3 3.1 4.1 5.7 141 1.1 2.7 11.7 3.2
Cu Middling II 0.9 4.6 6.1 1.5 179 0.6 1.6 1.2 1.6
Cu Middling III 1.6 16.6 1.5 4.1 189 4.1 0.7 5.9 3.0
Zn-Conc. 2.4 1.6 57.7 0.3 89 0.6 39.2 0.6 2.1
Zn Scv. Conc. 1.0 3.3 18.9 2.1 144 0.5 5.3 1.9 1.4
ZnMiddling 1.5 3.4 24.8 0.7 163 0.8 to.5 0.9 2.4
Tailing 58.7 0.6 0.6 0.4 58 5.4 9.1 21.4 33.8
Feed 100.0 6.5 3.2 1.1 101 100.0 100.0 100.0 100.0
..
Condlttons : Gnndmg pH 12.0 (hme). fineness of the gnnd : 92 % mmus 45 !-Lm. 5 mm
conditioning at pH 6.5-6.7 by S02 injection. Phosocreso] E 110 g/t and MIBC 20 glt stage
addition. Cu-rougher f1ot. 13 min. Cu-scv. flot 3 min. Cu-cleaning pH are 5.S. 5.0 and 4.0
respectively. Zinc f1ot. conditioning at pH 12.2 for 8 min. activation with 500 glt CuS04 5H20.
rougher flotation with 35 glt Phosocresol E and 5 glt MIBC for 5 min. cleaning f1ot. 2 min.

TABLE 8. Effect of different depressants on grades and recoveries of the copper and
. concentrates- A. plant
the zmc 1 tn'al
Depressants Lime Lime +Na2S205 Lime+Na2S20S+ZnS04
0800gft) (1800 gJl) (400 gJl)
pH 11.5 - 11.8 6.7 - 6.8 6.7 - 6.8
Assay % Recovery Assay Recovery Assay Recovery
Products % % % % %
OI Zn OI Zn OI Zn OI Zn OI Zn OI Zn
Cu-Conc. 20.9 5.6 82.6 57.2 21.8 3.9 82.3 30.0 21.2 2.4 84.0 22.9
Zn-Conc. 3.2 43.3 0.7 24.2 3.9 48.1 1.7 42.9 3.6 47.6 1.8 56.0
Tailing 1.4 0.6 16.7 18.6 1.2 1.0 16.0 27.1 1.3 0.8 14.2 21.1
Feed 6.2 2.4 100 100 5.7 2.8 100 100 6.5 2.7 100 100
..
CondItions: Aot. feed 80-8S% -45 !-Lm. 30-32% sohds. Phosocresol-E 100-120 g/t, KAX 40-S0 gil. 3
MIBC + 1 DF250 3S-S0 glt.

7. Conclusions

Despite the fact that complex Cu-Pb-Zn sulphides show serious metallurgical problems. due to
their mineralogy and textures, their high intrinsic value make them attractive for their exploitation.
328

Intensive mineralogical studies and flotation test~ must be carried out so as to find out the most
feasible and consistent plant flow sheet, since there are quite a few ways to produce various
concentrates, such as all selective, selective + bulk and all bulk flotation.
Although the depressing effects of the sulphites and S02 (g) have not been fully understood yet,
they are the most widely used reagents in flotation of the complex Cu- Pb-Zn sulphides. Very
important findings have been published so far, which are extremely helpful for process designing
and plant operations.
Parallel to the pace of improvement in flotation technology as a whole, metallurgical problems of
the complex sulphide will eventually become less challenging.

8. Acknowledgements

The authors would like to express their gratitude to Mr. A.S. Kanpak, General Manager of Demir
Export A.S., Ankara, Turkey, for the permission to publish this work.

9. References

I. Cases, J.M. (1980) Finely disseminated complex sulphide ores, in M.J. Jones (ed.) Complex
Sulphide Ores, IMM, London, pp. 234-247.
2. Barbery, G., Fletcher A.W., and Sirois, L.L. (1980) Exploitation of complex sulphide
deposits: a review of processing options from ore to metals, ibid. pp. 135-150.
3. Imre, U., and Castle, J.F. (1984) Exploitation strategies for complex Cu-Pb-Zn ore bodies , in
M.J. Jones (e.d.) Mineral Processing and Extractive Metallurgy Symposium. Kunming
(China), IMM, London, pp. 473-489.
4. Barbery, G. (1985) Recent progress in the flotation and hydrometallurgy of sulphide ores, in
K.S.E. Forrsberg (ed.) Flotation of Sulphide Minerals, Elsevier, Amsterdam, pp. 1-38.
5. Barbery, G. (1986) Complex sulphide ores : processing options, in B.A. Wills and RW.
Barley (eds.) Mineral Processing at a Crossroads, Martinus Nijhoff, Dordrecht, pp. 157-194.
6. Gray, P.M.J. (1984) Metallurgy of the complex sulphide ores, Mining Magazine ill, pp.
31-321.
7. Powell, C.R (1978) Asarco Incorporated, Buchans Uni, in D.E. Pickett (ed.) Milling
Practice in Canada, CIM Special Vol. 16, Montreal, pp. 157-158.
8. McQuiston, F.W. (1958) Pb-Zn-Cu : How flotation's most diffiCUlt separation is being done,
World Mining 2,41-44.
9. Dckok, S.K. (1972) Differential flotation of copper-zinc at Prieska Copper Mines (pty)
Limited: a preliminary report, J. South Afr. I.M.M. I, 305- 321.
10. Wills, B.A. (1984) The separation by flotation of copper-lead-zinc sulphides. Mining
Magazine 1. 36-41.
11. Konigsmann. K. V. (1985) Flotation techniques for complex ores. in A.D. Zunkel, RS.
Boorman. A.E. Morris and R.J. Wesely (eds.) Complex Sulphides, The Met. Society of
AlME, Warrendale, pp. 5-19.
12. Norrgran, D.A., and Armstrong, R.O. (1985) Developing a selective flotation concentration
technique for polymetallic ores, ibid .. pp. 37-54.
13. Ek, C.S. (1985) Selective flotation of different complex pyritic ores, ibid., pp. 83-99.
 329

14. Bulotovic, S.M., and Wyslouzil, D.M. (1985) Selection of reagent scheme to treat massive
sulphide ores, ibid., pp. 101-137.
15. Bulotovic, S.M., and Salter, RS. (1990) Some aspects of recent improvements in treatment
and separation of refractory polymetallic ores, SME Annual Meeting, preprint no. 90-25,10
pp.
16. Sato, T. (1974) Distribution and geological setting of the Kuroko deposits, in S. Ishihara, K.
Kanehira, A. Sasaki, T. Sato and Y. Shimazaki (eds.) Geology of Kuroko Deposits, The
Soc. of Min Geologists of Japan, Tokyo, pp.I-9.
17. Craig, J.R., and Vaughan, D.J. (1981) Ore Microscopy and Ore Petrography, John Wiley
and Sons, New York.
18. Shimazaki, Y. (1974) Ore minerals of the Kuroko type deposits, in S. Ishihara, K. Kanehira,
A. Sasaki, T. Sato and Y. Sf azaki (eds.) Geology of Kuroko Deposits, The Soc. of Min
Geologists of Japan, Toky( 111-322.
19. Altun, Y. (1978) Geology <-J .? Cayeli-Madenkoy copper-zinc deposit and the problems
related to mineralisation, Bull. of thc Min. Research and Expnl. Inst. of Turkey .82, 10-23.
20. Altun, Y. (1990) Ore minerals and their genesis of polymetallic deposits in Giresun-Gorele
and Tirebolu (the Eastern Black Sea Region), Bizim Buro Basimeviu, Ankara (in Turkish
with English abstract).
21. Moon, C.J., and Khan, M.A. (1988) Mineral exploration, Mining Annual Review, p. 133.
22. Liebenberg, W.R (1970) Mineralogy and the meetallurgist, Minerals Sci. and Engng. 2. (1).
23. Harris, P.J., and Richter, (1985) The influence of surface defect properties on the activation
and natural floatability of sphalerite, in K.S.E. Forrsberg (ed.) Flotation of Sulphide
Minerals, Elsevier, Amsterdam, pp. 159-174.
24. Petruk, W. (1987) Applied mineralogy in ore dressing, in B. Yarar and Z.M. Dogan (eds.)
Mineral Processing Design, Maninus Nijhoff Publisher, Dordrccht, pp. 2-36.
25. Petruk, W. (1991) Applied mineralogy and materials balancing procedure: evaluation of
flotation concentrators, in this volume.
26. Trahar, W. J., and Warren, L.J. (1976) Thefloatability of very fine particles - a review, Int.
J. Miner. Process. 3,,103-131.
27. Warren, LJ. (1984) Ultrafine particles in flotation, in M.H. Jones and J.T. Woodcock (eds.)
Principles of Mineral Flotation, The Australasian (MM, Parkville, pp. 185-213.
28. Trahar, W.J. (1981) A rational interpretation of the role of particle size in flotation, Int. J.
Miner. Process . .8., 289-327.
29. Lynch, A.J., Johnson N.W., Manlapig E.V., and Thome, C.G. (1981) Minaral and Coal
Flotation Circuits, Elsevier, Amsterdam.
30. Tolun, R (1987) Chemistry of sulphide mineral flotation, in B. Yarar and Z.M. Dogan (eds.)
Mineral Processing Design, M. Nijhoff, Dordrccht, pp. 37-75.
31. Wakamatsu, T. and Shibata, J. (1989) Progress and outlook of complex sulphide ore
(Kuroko) processing in Japan, in K.V.S. Sastry, and M.e. Fuerstenau (eds.) Challenges in
Mineral Processing, SME, Littleton, pp. 130-141.
32. Bayraktar, I. (1985-1990) Reports on several complex Cu-Pb-Zn and Cu-Zn sulphide ores,
not in public domain.
35. McKee, DJ. (1991) Case studies offlotation control, in this volume.
36. Rao, S.R, and Forssberg, K.S.E. (1991) Electrochemistry in the plant, in this volume.
37. Maeshiro, I. (1980) Recovery of valuable metals from "black ore", in M.J. Jones (ed.)
Complex Sulphide Ores, IMM, London, pp. 315-324.
38. Roberts, A.N., Bums C.J., and Cameron A.W. (1980) Metallurgical development at
330

WoodlawnMines,Australia, in M.J. Jones.(ed.) Complex Sulphide Ores, IMM, London,

pp.128-134.
39. Hendricks, D.W., and Ounpuu, M. (1985) Modifications to the zinc flotation circuit of
Brunswick Mining and Smelting Corp Ltd. , in A.D. Zunkel, RS. Boorman, A.E. Morris
and RJ. Wesely (eds.) Complex Sulphides, The Met. Society of AIME, Warrendale,
pp.271-29.
40. Broman, P.G., Hultqvist, J., and Marklund, U. (1985) Experiencefrom the use ofS02 to
increase the selectivity in complex sulphide flotation , in K.S.E. Forrsberg (ed.) Flotation of
Sulphide Minerals, Elsevier, Amsterdam, pp.277-291.
41. Heide, B. (1990) Meggen concentrators, Germany, private communication.
42. Sharp, F.H. (1976) Lead-Zinc-Copper separation and current practice at Magmont Mill, in
M.e. Fuerstenau (ed.) Flotation, SME, New York, vol. 2, 1215-1231.
43. Outokumpu Technical Export Division, Concentrators of Outokumpu Oy Finland.
44. Bayraktar, I. (1989) Bench and Plant Scale Studies on a complex Cu-Zn ore, not in public
domain.
45. Wills, B.A. (1989) Mineral Processing Technology, Pergamon Press, Oxford, 5 tho Edition.
46. Williams, S.R., and Phelan, J.M. (1985) Process development at Woodlawn Mines, in
ref. [10), pp. 293-304
47. Pallanch, R.A. (1928) The role of sulphites in the differential flotation plants of the U.S.
Smelting, Refining and Mining Co., in Flotation Practice, AIME, New York, pp. 78-81.
48. Jones, M.H., and Woodcock, J.T. (1984) Applications of pulp chemistry to regulation of
chemical environment in sulphide mineral flotation, in M.H. Jones and J.T. Woodcock (eds.)
Principles of Mineral Flotation, The Australasian IMM, Parkville, pp.147-183.
49. Gaudin, A.M. (1957) Flotation, McGraw Hill ,New York, 2nd. edn.
50. Fuerstenau, M.C. (1982) Sulphide mineral flotation, in RP. King (ed.) Principles of
Flotation, S. Afr. Inst. of Min. and Met. Monograph Series No.3, pp. 159-182.
51. Shimoizaka, J., Usui S., Matsuoka I., and Sasaki H. (1976) Depression of galena flotation
by sulphite or chromate ion, in M.e. Fuerstenau (ed.) Flotation, SME, New York, 1, pp.
393-413.
52. Yamamoto, T. (1980) Mechanism of pyrite depression by sulphite in the presence of
sphalerite, in M.J. Jones.(ed.), Complex Sulphide Ores, IMM, London, pp. 71-78.
53. Misra, M., Miller, J.D., and Song, W.Y. (1985) The effect of S02 in the Flotation of
Sphalerite and Chalcopyrite, in K.S.E. Forrsberg (ed.) Flotation of Sulphide Minerals,
Elsevier, Amsterdam, pp.175-196.
54. Mukai, S., and Matsubara, H. (1968) Effect of sulphurous acid on the floatability of sulphide
minerals, Fusen 34, pp.25-28 (in Japanese, English abstract only).
55. Kubota, T. (1970) The effect of pulp heating on floatability of galena and a practical
application to Cu-Pb separation, J. Min. & Met. Inst. of Japan .8Q, 31-34 (in Japanese,
English abstract only).
56. Schnarr, J.R. (1978) Brunswick Mining and Smelting Corp., in D.E.Pickett (ed.) Milling
Practice in Canada, CIM Special Vol. 16, Montreal, pp. 158-161.
57. Twidle, T.R and Engelbrecht, P.e. (1984) Developments in the flotation of copper at Black
Mountain, J. S. Afr. I.M.M., June, 164-178.
58. Bayraktar, I. (1988 and 1990) Visit reports of Folldall, Norway and Meggen, Germany
Concentrators, not in public domain.
59. Parsonage, P.G. (1985) Effects of slime and colloidal particles on the flotation of galena, in
K.S.E. Forrsberg (ed.) Flotation of Sulphide Minerals, Elsevier, Amsterdam, pp. 111-139.
3.2. INTERACTIONS OF ANIONIC COLLECTORS IN FLOTATION OF
SEMI-SOLUBLE SALT MINERALS

K. HANUMANTHA RAO and K.S.E. FORSSBERG

Division 0/ Mineral Processing
Lulea University o/Technology
S-95J87 Lulea
Sweden

ABSTRACf. Our recent studies on the interaction of oleate with salt-type mineral surfaces have
been summarised and discussed in the light of the published literature. The mechanism of oleate
adsorption on these minerals was evaluated based on adsorption, zeta potential and infrared studies
as well as theoretical thermodynamic calculations. In contrast to earlier knowledge, a quantitative
evaluation of oleate adsorption in systems forming a monolayer coverage as in the case of calcite
and a bilayer formation in the cases of fluorite, apatite, scheelite and baryte, preceding Ire
precipitation of metal soap is presented. The monolayer coverage is shown to correspond to a
condensed state of oleate with a molecular coverage area of 33 A2 (liquid crystal state). Depending
on the surface potential and its magnitude at basic pH values, oleate is either chemisorbed on the
surface calcium or monocoordinated through counter sodium and calcium ions for the monolayer
filling. The results have been discussed in terms of the existing adsorption models.

1. Introduction

The sparingly soluble minerals such as calcite, fluorite, apatite, scheelite and baryte are largely
concentrated by flotation processes. Generally, fatty acids are used as collectors in these systems.
For effective separations, selective adsorption of the collectors at the mineral/water interface is
essential. The mechanism of the adsorption of fatty acid based collectors on these minerals has
been widely studied and a comprehensive review of the subject was published by Hanna and
Somasundaran in 1976 [1]. This has been updated by Finkelstein very recently [2]. Even though,
chemisorption, surface precipitation and bulk precipitation are thought to be the important
reactions, a clear understanding of the chemistry of fatty acid interaction does not exist. The
adsorbed species extend to several layers due to surface and bulk precipitated phases in these
systems. thus making the interpretation of the shape of traditional adsorption isotherms difficult.
However, recent studies indicate that it is still possible to differentiate between adsorption and
precipitation reactions and the shape of the isotherms can be explained in terms of the existing
monolayer models [3-10].
The authors intend to summarize the findings of their recent investigations in the light of the
results of previous studies.

331
P. Mavros and K. A. Matis (eds.), innovations in Flotation Technology, 331-356.
© 1992 Kluw(!r Academic Publishers.
332

2. Theory

Several theories have been proposed to explain the nature of adsorption of the long-chain
surfactants from aqueous solutions. The earliest analysis is due to Gaudin and Fuerstenau [11,12].
These authors claim that the surfactants are adsorbed at low concentrations as individual counter
ions and the two dimensional lateral aggregation (hemi-micelles) between the adsorbed long chain
molecules in the Stem layer takes place above a given critical equilibrium concentration. The
increase in surface coverage is described by the formation of patches.
An entirely different approach was considered by Cases and coworkers [13-15]. They have also
assumed the association of adsorbed ions and suggested a two dimensional condensation on the
surface due to the importance of lateral bonds. Moreover, they have considered the potential
energies and entropy of the adsorbed layer, as well as the heterogeneity of the adsorbant surfaces.
Because of surface heterogeneties, the two dimensional condensation initially takes place at the
homogeneous domains that are formed with the most energetic sites, at the smaller values of
solution concentrations followed by successive two dimensional condensation on less energetic
homogeneous domains. Hence, in the monolayer range, the isotherm exhibits either a series of
small vertical steps or a smooth curve depending on the distribution of surface energies. Should the
formation of a bilayer be feasible, it is always characterised by a vertical step on the isotherm, since
the second layer occurs on a surface which is rendered homogeneously energetic due to the
formation of a first layer.
An admicelle hypothesis was proposed [16), which states that surfactant aggregation producing
bilayered structures (admicelles) forms locally on the surface due to heterogeneity in the same
manner as micelle formation. These admicelles occur on a given patch of a heterogeneous surface
at a critical solution concentration and for concentrations less than the critical value, aggregates do
not exist on that patch of the surface.
The formation of admicelles beyond the critical solution concentration should mean that the
normal-normal bonding of the hydrocarbon chains takes place. If the polar head has a high affinity
for a particular solid, the surface can be completely covered before the critical solution
concentration is reached and, the plateau for a monolayer coverage may appear. On the other hand,
if the surface is not completely covered, the voids on the surface may be covered with bilayered
structures along with the formation of a second layer on the previously covered surface. Hence, the
admicelle hypothesis is a natural extension of the two-dimensional condensation model [17]. In
any case, if the bilayered structures are only dependent on the normal-normal bonds of the
surfactant, the critical solution concentration of the particular surfactant should be the same
irrespective of the solid substrates used. This has not been tested so far experimentally, to the best
knowledge of the authors.
Recently, the calculations of Ananthapadmanabhan and Somasundaran [18] for a number of
mineral-reagent systems showed that the conditions for the precipitation of metal-surfactant
complexes in the interfacial region are to be satisfied before the same occurs in the bulk solution.
On the basis of a chemical equilibrium that prevails for the bulk interactions along with the
activities of the interfacial region, they found that the marked increase in the flotation response and
the adsorption in a number of mineral-surfactant systems correlate with the conditions for the
precipitation in the interfacial region, leading them to suggest that the so-called chemisorption is in
fact a process of surface precipitation. Should this be the case, the earlier models are not likely to
be explained or accounted for by the surface precipitation process.
 333

3. Chemistry of Oleate in Relation to Adsorption

Fatty acids are hydrolysable surfactants and the equilibrium constants for the dissociation of oleic
acid in aqueous solutions have been reported by several authors [19,20], which are in fairly good
agreement. Thus, the chemistry of soluble ionic oleate species in dilute aqueous solutions seems to
be well understood.
For mechanistic considerations, concentration-pH domain are chosen, where conditions of
solution devoid ofliquid oleic acid and NaH(OLh prevail and calcium oleate is formed [21]. Oleate
ions being the predominant species at basic pH values, only studies that have been carried out in
this pH region have been considered. However, the concentration range of oleate studied in most
of the cases exceeds the solubility limit of calcium oleate, while considering the dissolved calcium
ions from the minerals. The reported solubility product of calcium oleate fall in the pK range of
12.4-15.74 [22-28] and because of the ambiguity of this value, it is difficult to predict the stage at
which the precipitation occurs. Moreover, due to the complexity of the multi phase system, it is not
certain that the precipitation takes place immediately after the conditions of solubility are satisfied.
The need for supersaturation for the nucleation of a new solid phase from the solution species was
pointed out [29]. Cases [15] emphasised that there will be no confusion between adsorption and
precipitation since the isotherms indicate an infinite vertical step following precipitation under
conditions of excess ions in the solution.
The cross-sectional area of oleate in a vertical orientation is reported to be 20.5 A2 [30]. The
alkyl chain in the condensed state (liquid crystal) corresponding to the lamellar phase is reported to
be 33 A2 [31]. Thus, if the chain is incorporated in a hydrated crystal state or precipitated phase,
the monolayer coverage can be expected at 8 ~ol m- 2 adsorption density. On the other hand, if
the oleate is assumed to be in a liquid crystal state in the adsorbed region because of the two
dimensional condensation on the surface, the monolayer coverage corresponds to only 5 ~ol m-2 .
In essence, the adsorption density results should be assessed for both the cases of oleate being in a
hydrated crystal state and a liquid crystal state in the adsorbed region.
The applicability of the infrared spectroscopy for an adsorbed species at a mineraVwater interface
involves the removal of the aqueous phase to reduce the solvent opacity before examination. Many
authors have expressed doubts as it is not an in situ technique and the sample preparation methods
may themselves lead to surface alterations [1,2]. At basic pH values, most of these minerals show
negative zeta potential and hence the double layer is composed of counter calcium ions. In this
situation, if the oleate is monocoordinated through counter calcium ions rather than on the surface
of calcium, the oleate and calcium ions are expected to condense on the surface as calcium oleate
due to drying. On the other hand, if the adsorption is only on the surface of calcium, no changes in
the adsorbed oleate structure is to be expected. Therefore, infrared correlations should be made
with caution.
The methods of oleate analysis such as radio tracer technique, UV spectrophotometry, infrared
spectrophotometry and differential thermal analysis have been summarised by Yap et al. [32]. The
disadvantages of using a radio tracer technique have been outlined. The main disadvantage appears
to be that the solution of oleic acid in the presence of 14C decomposes rapidly. If the tagged oleic
acid also decomposes, the adsorption results will be drastically affected as the specific activity of
the tracer would still remain and is counted as the activity of the oleic acid. In contrast to all other
methods, the UV spectrophotometric method of Gregory [33] is thought to give a higher sensitivity
and better accuracy, though the amount that disappears is only measured.
334

4. Theoretical Equilibrium Calculations

Marinakis and Shergold [27] calculated the equilibrium conditions in the pulp liquid saturated with
a mineral to which sodium oleate is added. These calculations arc based on the mass balance of the
solubility products of the mineral, oleic acid and metal-soap. Similar calculations involving
scheelite and myristate were also reported earlier [34]. Recently, the SOLGASWATER software
was used for the theoretical calculations of the thermodynamic chemical equilibria in the pulp liquid
[6-9]. This program offers a method for calculating the equilibrium compositions in a multi phase
system. The general principles applied in this program for mineral pulps have been documented
[35].

5. Mineral-Reagent Interaction

5 .1. CALCITE-OLEATE SYSTEM

5.1.1. Adsorption studies. Most of the adsorption studies on calcite were carried out in the
alkaline pH region. Iskra et at. [36] observed that the adsorption of oleate remains constant in the
pH range of 8.6-12.0. It was also found that the adsorption is independent of pH between the
values of 9 and 12 [27]. However, Gutierrez [37] observed a pH dependent adsorption with a
sharp maximum between pH 10.2 and 11.5.

w,-----------------~------,-_.
(A) pH Reference
09.3 [41
:It 10.0
A 11.0
D 10.0 [40]

10'0-::---------------------------,
(B)

pH Reference
- 10.0 [41]
o lO.O [27]
"" 9.0 [39]
• 10.0 [37]

~~~,-~--~~~--~~~~1044--~~ 10 .'::10'"~~-'-'"::IOT'~~'"""::IO"'-~~~~IOS'~~~I0--2

EQUILlliRIUM OLEATE CONe. (moles II) SODIUM OLEATE EQUILIBRIUM CONe. (mol/!)

Figure 1. (A) and (8): Adsorption isotherms of oleate on calcite.

Somasundaran [38] observed a sharp increase in the slope of the adsorption isotherm above
3xlO-5 M equilibrium oleate and this increase was attributed to the precipitation of calcium oleate.
A similar increase at 4x 10-5 M equilibrium oleate was reported by Gutierrez [37], suggesting the
precipitation of calcium oleate. The isotherms as reported elsewhere [4,39] also show an increase
at 2x10- 5 M equilibrium oleate (Figure 1). The isotherms at different pH values. as shown in
 335

Figure lA, indicate a break at a 5 !Jl1101 m-2 adsorption density, prior to the increase in the slope of
the isotherms. A plateau at 5 !Jl1101 m-2 was observed by Sadowski [40], preceding the infinite
increase in the slope of the isotherm (Figure I A). The increase in the slope of the isotherms after
the plateau is ascribed to calcium oleate precipitation.
Zimmels et al. [41] proposed a stepwise adsorption model, consisting of multilayers in stages
including a direct monomer, a submicelle and a micelle adsorption due to their varying affinities for
adsorption at different oleate concentrations. A chemical interaction for the adsorbent-adsorbate
bond was considered with a kind of chemisorbed precipitate on the surface. A similar chemical
interaction at the surface of the calcite was suggested by Szezypa et al. [42] and Fuerstenau et al.
[43] for sodium laurate.
Marinakis and Shergold [27], from considerations of thermodynamic equilibria, correlated the
abstraction of oleate by calcite with the preCipitation of calcium oleate for the range of oleate
concentrations studied. The experimental and theoretical oleate concentrations for the preCipitation
of calcium oleate agree only beyond an oleate adsorption density of 5 !Jl11ol m- 2 [4]. However, a
pK value of 15.40 for calcium oleate is considered for theoretical computations in the first case,
whereas a pK of 14.04 is used in the later, which is estimated from the solution conditions.

5.1.2. Infrared studies. Peck [44] reported that oleate chemisorbs on calcite by the carbonyl
absorption band which has shifted to 1587 cm -1. However, the exact position of the band was
stated to be uncertain owing to the highly interfering carbonate absorption. Recently, Antti and
Forssberg [45] also found it difficult to obtain any information about the bonding of the surface
oleate species using differential Ff-IR and after subtracting the calcite spectrum in the spectrum of
adsorbed oleate species.

5.1.3. Zeta-potential studies. Electrostatic adsorption of oleate below the i.e.p (pH 8.2) of calcite
and chemisorption above that have been suggested since the zeta potentials are observed to change
continuously with the change in the oleate concentration studied [38]. A decrease in the zeta
potential of calcite in the presence of oleate above the i.e.p. was reported [4,46], suggesting a
specific adsorption of oleate on calcite.
The influence of oleate on the calcite zeta potential at pH 10 as obtained by several authors is
shown in Figure 2A. Initially, all of the investigations show a decrease in the zeta potential and
between lxlO-5 and Ix10 4 M oleate concentration, the potential remains constant, while above
IxlD4 M oleate, the zeta potential again decreases and reaches a minimum potential of -50 mY.
The electrophoretic mobility of calcium oleate with respect to pH [47], and for the system
containing 5x10 4 M potassium oleate with varying concentrations ofCa(N0)h [48] are shown in
Figure 2B. The electrophoretic mobility of calcium oleate is constant from pH 8 to 11, and has a
value of -3.9 cm 2 V -1 s-2. As the concentration of calcium changes, the mobility changes,
indicating that both Ca 2 + and oleate are potential determining. However, at equimolar
concentrations, as shown by the broken line in the figure, the mobility has the same value. This
value corresponds to a zeta potential of -50 mY. Hence, the minimum zeta potential of calcite in the
presence of oleate corresponds to that of calcium oleate.
If the abstraction of oleate is in the form of precipitated calcium oleate and multilayers are formed
on the surface as suggested by several authors, the zeta potential should decrease with the increase
in oleate concentration, at least up to -50 mY. Since this is not the case, adsorption of oleate on the
surface has been suggested for the initial decrease in zeta potential and precipitation of calcium
oleate in the bulk solution between 1x I 0-5 and I x 10-4 M oleate concentration. The three dimensio-
336

nal growth of calcium oleate on surface or adsorption of ealcium oleate precipitate on top of the
hydrophobic layer was thought to be the reason for a sudden decrease in the zeta potential
equivalent to that of calcium oleate [4 J.

(B)
o Mobility as a function of pH [4 7J

~ ~~~~~~O;:~::i~:~;aO~:~:~mg
25,------------------, > -I concentrations of Ca(NO]) 2 [48J
(A)
x Rao e[ al [41
[] Marinakis et al [271 E.
6
o
Zimmels et al [41]
Mishra [46] § -2
iii
~
u
~ -3
'" 0
9
~ 6 6
o 6

~ 4 -------------- 6 6

-5 \--3--0--+---;---0\--7-+--;;---~IO,,--!II
pH

-4 -3.5 -3.0
LOG. ~10LAR COI\'c. OFC<lG",03) 2
INITIAL OLEATE CONe. (moles /1)

Figure 2. (A). Zeta-potential of calcite as a function of initial oleate concentration at pH 10.

(8) Electrophoretic mobility of calcium oleate.

5.1.4. Discussion. The adsorption isotherms either showed a break or a elear plateau at
5 ~mol m- 2, before the slope increased. The adsorption density of 5 ~mol m- 2 corresponds to a
monolayer capacity for a two dimensional condensed state of oleate in the adsorbed region (liquid
crystal state). The initial decrease in zeta potential can thus be attributed to a monolayer surface
coverage. In fact, under specified conditions of zeta potential measurements, the initial decrease in
zeta potential is shown to correspond approximately to an oleate adsorption density of 5 ~mol m- 2
[4]. The infinite increase in the slope of the isotherms alter the plateau level can be ascribed to a
precipitation step of calcium oleate. Hence, the formation of a monolayer preceding to the
precipitation of calcium oleate can be the mechanism for adsorption of oleate on calcite.

5.2. FLUORITE-OLEATE SYSTEM

5.2.1. Adsorption studies. The adsorption of oleate from 5x1O- 5 M solution was reported to be
maximum at pH 9, while at higher values it decreases rapidly [36]. For the same concentration of
oleate, a maximum adsorption was found at pH 8 [37]; it was also found that the maximum uptake
of oleate occurs around mid pH region [491. Thus, the oleate adsorption on fluorite seems to be pH
dependent.
Earlier results on the oleate adsorption on fluorite [50] are shown in Figure 3A. The adsorption
density levels off at 10 ~mol m- 2, which is considered to be just above the monolayer coverage,
when dry CaF2 is added into oleate solution. On the other hand, when olcate solution is added into
a suspension of CaF2 equilibrated with water, the measured adsorption density exceeds three
monolayers. However, it is interesting to note that below 10 ~mol m- 2 adsorption density,
 337

irrespective of whether the dry CaF 2 is added to the oleate solution or oleate is added to a
suspension of CaF2, the adsorption density with respect to equilibrium oleate is the same.

40,---------------, 12r(::":"B:-)- - - - - - - - - - - - - - - - ,
pH = 10 • With US Treatment
(A) 0 ~~~~~ea~~~~~ion
T= zooe a Wilhout US Treatment
.6 ~~~~I~~~~~ to 0-----0 <"~l
pH= 10 / E

/
I >- 8
/
f-
/ Vi
z
20
, /
/0
3 6

~)':(
Z
o
t;: 4

10 " .0. I t;
o c,++ , '\~r Monob",
C
<:

o 4 6 8 JO 0.5 1.0 50 100

OLEATE CONCENTRATION 104 M EQUILIBRIUM CONCENTRATION OF OLEATE ( mg/l )

I<'igure 3. Adsorption isothenns of oleate on fluorite at pH 10: (A) Bahr et al. LSD], (B) Hu et al. [28J.

Gutierrez [51] observed a maximum in the adsorption isotherm at pH 9 and this was thought to
be due to the physical adsorption of precipitated calcium oleate on the fluorite surface or on top of
the hydrophobic layer of chemisorbed oleate. No adsorption maximum was observed by Texeira
et at. [52] at the same pH value, instead, the isotherm was found to level off at 30 !lIDol m-2 with
equilibrium oleate.
Giesekke and Harris [53J reported the distribution of oleate in contact with fluorite. As the initial
oleate concentration increases, both the amount of oleate adsorbed on the fluorite and the bulk
precipitation increases simultaneously. The results are explained in terms of the multilayers of
adsOlbed oleate in the form of calcium oleate.
The adsorption isotherm presented by Marinakis and Shergold [27] begins at an adsorption
density higher than required for a monolayer coverage and their results have been correlated to the
abstraction of calcium oleate precipitate. Hu et at. [28] have distinguished three regions of oleate
adsorption, a chemisorbed plateau at low oleate concentrations (oleate in horizontal position, 92.5
A2), a surface precipitation of calcium oleate where the adsorption density attained a maximum,
and a third region where the adsorption density decreases at high equilibrium concentrations from
the results as shown in Figure 3B. The chemisorbed horizontal monolayer is supported by the non-
removal of the adsorbed oleate by the ultrasonic treatment. However, in a later study, they
observed a relationship between the measured total heat and the surface coverage from the
microcalorimetric and the FTIR adsorption density measurements [54]. Here, the results suggest
the transition of the chemisorbed monolayer to surface precipitation, for an oleate molecular
coverage area of 34.2 A2 (liquid crystal state). Hence, the chemisorbed plateau of monolayer in a
horizontal position of oleate as stated earlier is uncertain.
Adsorption isotherms of oleate on natural and synthetic fluorites at pH 10 are shown in Figure
4A. The natural fluorite isotherm shows a plateau level at 10 ~mol m· 2 before a steep increase in
the adsorption density.
The composition of the pulp liquid after adsorption is shown in Figure SA. This figure shows
that the negative logarithm of the ionic product of calcium and the square of the equilibrium oleate
338

30
(A)

E
"0
[
20
..
Cl

"'"'"
0
""
Cl
<:
'"~
OJ pH Reference
..J
0 Naturalfiuorile Synthcticfluoritc + 10.0 [55J
c 1.0 % Solids 0 O.OS % Solids
II; to.o [551
010.0 {27J
x 0.10% Solids
.t. 0.20% Solids

10" I
10- 10-0 10- to 10-3
EQUILIBRIUM OLEATE CONe. (moles II) EQUILIBRIUM OLEATE CONCENTRATION ( molesll )

Figure 4. Adsorption isotherms of oleate on fluorite at pH 10: (A) Rao el al. [7,8]; (B) Dobias el al.
[55]; Marinakis and Shergold [27].

are constant at 14.39 for initial oleate concentration between 3xlO-5 and 6xlO- 5 M. This range of
constant ionic product corresponds to the vertical step immediately after the plateau in the
adsorption isotherm. Hence, this is characterised as the precipitation region of calcium oleate with a
solubility product of 4.07xlO- 15 moP 1-3 (pK=14.39).

(B) THEORETICALCLTRVES,pK=14.40
Tc forCa. 2+", 1.2", 10 ~M
1O·"r--------------------,16 Te for F-,.1.4 Jt iO'~M
(A) TcforH~C01" 1.0", IO-SM
ThCOR:licd
OLEATE LEFT

x Fluoride
o Calcium

INITIAL OLEATE CONe. [moles II) INITIAL OLEATE CONe. (moles II)

Figure 5. (A) Analysis of liquid phase as a function of initial oleate concentration; (B) Comparison of
theoretical and experimental oleate concentrations for Ca-soap formation. Analysis and experimental results
are for the studies on natural fluorite as shown in Figure 4A.
 339

The theoretical and experimental oleate concentrations for the calcium oleate solid phase have
been compared and the values correspond to one another only after 3xlO-5 M initial oleate
concentration (Figure 5B)_ The important observation is that the correlation between the
experimental and theoretical oleate concentrations from 3x 10-5 to 6x 10-5 M of initial oleate, is the
same as that for the assumed precipitation step in the isotherm after the plateau level of 10
!lIl101 m -2.
In the case of the synthetic sample, a break at 10 !-lmol m- 2 is once again observed in the
adsorption isotherms at three different solidi liquid ratios as can be seen in Figure 4A. At low solid
concentrations (0.05 and 0.1 % solids), the increase in the adsorption density after the plateau level
is ascribed to the precipitation of calcium oleate. The solubility product of calcium oleate
corresponding to these precipitation regions is estimated to be around pK 14.35, from the analysis
of the pulp liquid after adsorption [8]. An interesting observation is that at high solid concentration
(0.2% solids), the isotherm is seen to level off at 10 !-lmol m- 2 similar to that of the isotherm
reported earlier [50] (Figure 3A). A similar isotherm was also recently reported by Dobias et al.
[55] in the presence of EDT A, as shown in Figure 4B.

5.2.2. Infrared studies. Peck and Wadsworth [56] reported a shift in the peak from 1562 cm -I to
1555 cm- I after washing and assigned the peak at 1555 cm -I to surface calcium oleate which is
thought to be obscured by the peak at 1562 cm- I for sodium oleate after adsorption. The bands at
1562 and 1555 cm- I were assigned to physically-adsorbed sodium oleate and surface calcium
oleate, respectively [57]. However, the same authors in a subsequent investigation considered the
peak at 1562 cm- I for the surface calcium oleate and the physically adsorbed sodium oleate and
suggested that it is not possible to distinguish between them [58]. Kellar et at. [59] characterised
the monocoordinated calcium oleate (chemisorbed) by an absorption band at 1549 cm -I by using
in situ internal reflection FT-IR spectroscopy. Double peaks corresponding to calcium oleate were
observed in other studies [60,61].
The diffuse reflectance IR spectra for the adsorption results (0.1 % solids, Figure 4A) are
presented in Figure 6A [8]. The absorbance band that occurs between 1557-54 cm -I, when the
adsorption density is below 10 !-lmol m- 2, is assigned to the surface calcium oleate (1: 1). Above
10 !-lmol m- 2 adsorption density, the spectra show absorbance bands at 1576 and 1540 cm- I
corresponding to calcium oleate. These spectra further substantiate that calcium oleate is precipita-
ted after the plateau in the isotherms. Similar IR spectra are reported for the adsorption results at
0.2% solids up to 10 !lIl101 m- 2 adsorption density.
Figure 6B shows the spectra where the adsorption studies have been made while changing the
order of oleate addition. At low oleate concentrations, the spectra are found to be the same,
irrespective of whether solids are added to an oleate solution or oleate is added to a fluorite
suspension. Moreover, the intensity of the surface calcium oleate band (1555 cm- I ) is unchanged.
The bands corresponding to precipitated calcium oleate species appear only after an initial oleate
concentration of 2.2x 10-4 M. This implies that oleate presumably adsorbs on the surface prior to
the nucleation of a new solid calcium oleate phase, since the solubility product is exceeded several
times before the actual formation of calcium oleate. However, the growth of adsorbed layer as
studied by diffuse reflectance IR spectra in other cases showed that the surface is filled with
sodium and calcium oleates from the beginning of the surface coverage [10,62]. Here,
monocoordination of oleate through counter sodium and calcium ions is suggested for the
monolayer filling rather than on surface calcium.

5.2.3. Zeta-potential studies. The zeta potential curves of fluorite in the presence of sodium oleate
 340

showed that the oleate has a pronounced influence on the surface potential. A build up of negative
charge between about 20-60 mV at pH above 5 was reported [63]. Similar zeta potential results are
shown in Figure 7 A, suggesting a high affinity of oleate for the mineral surface.

INITIAL OLEATE

(A)

i
t:.:

0.30)
1602)

;j

1900 laoo 1700 1000 1500 \400 1300 1200 1900 1800 1700 IWG I~OO '400 1300 ;:00
WAVE:-.IUMBERS (em I) WAVE:-':UMBERS (em,i)

Figure 6. Diffuse reflectance FT·IR spectra [8]: (A) At different adsorption densities (!-lmol m- 2) for the
isotherm at 0.1 % solids shown in Figure 5A; (B) At different initial oleate concentrations.

20,--------------------,
(B)

:;
-5 x Raoe[ a\ 17]
o R:lOe[ al 18J
...l -10
<: o MannakJs el a.l (271
1=
:;: Z
1= 10
"'
z ~
"'
f-
~ ·30
~"1
~
<: -2 N
t;; -40
N -3
-4
·50
-5
-60
7 JO II 12 10· WS 10' 10-3
pH INITIAL OLEATE CONe (moles II)

Figure 7. Effect of oleate on zeta potential of fluorite: (A) At different initial concentrations as a function
of pH [8]; (B) At pH 10 as a function of initial concentration.
 341

The results on the effect of sodium oleate on the zeta potential of fluorite at pH 10 are presented
in Figure 7B [7,8,27]. With increased initial oleate concentration, the decrease in the zeta potential
is obselVed to take place in two stages and the minimum potential reached in all the cases is found
to be the same. The first stage is correlated to a successive oleate adsorption up to a plateau level
(10 lIDlol m-2) of the isotherms as shown in Figure 4A [7,8]. The second decrease up to -50 m V is
thought to be the growth of a three dimensional condensation of calcium oleate on the surface. The
change in the zeta potential accounts for the increase in the surface charge density and is linked to
the change in the molecular area of the adsorbed molecule. This hypothesis is supported by the fact
that the constant zeta potential of -50 m V correspond to the zeta potential of calcium oleate, based
on the adsorption results and the IR spectra [8].

5.2.4. Discussion. Although the shape of the isotherms and the adsorption densities vary from
one worker to the other, a few similarities can be noted. The levelling off of the isotherms at
10 /-lmol m- 2 or a plateau at this stage, prior to a steep increase in the adsorption density, are
obselVed in certain cases.
This adsorption density was explained as a bilayer formation, using a molecular coverage area of
33 A2 for a liquid crystalline state of oleate in the adsorbed region [7,8]. The isotherm, showing a
steep increase in adsorption density after the formation of the bilayer, is attributed to the
preCipitation of calcium oleate or perhaps the growth of calcium oleate on the surface. The results
of Bahr et al. [50] can also be explained with a bilayer formation for the isotherm, which levels off
at 10 /-lmol m- 2. However, the increase in the adsorption density after the plateau level, when
oleate is added to fluorite suspension, can be considered to be the precipitation of calcium oleate
(Figure 3A). While using EDTA, the frec calcium ions might form a complex with EDT A and
hence are not available for calcium oleate precipitation. Hence, the isotherm levels off at the bilayer
formation of 10 lIDlol m-2 (Figure 4B).
Although chemisorption of oleate by infrared studies is suggested by several authors, a good
transition of singlet absorption of surface calcium oleate to a doublet absorption of calcium oleate
was obselVed only in one study that could be due to the positive zeta potential of the sample
employed at the adsorption pH of 10 (Figure 7A). In this case, no strong counter calcium at tre
interfacial region is expected and the oleate might have the possibility of making a chemical bond
with the surface calcium. In such cases, where the fluorite samples show a negative potential at the
adsorption pH, oleate can be monocoordinated through counter calcium ions. Therefore, the
adsorbed structure can be transformed into a surface compound formation during the sample
preparation (drying), although oleate is monocoordinated through counter calcium ions for the
monolayer filling during the adsorption sequence.
Figure 6A also shows the appearance of two shoulders at 1574 and 1540 cm- I , for 6.02 to
10.39 lIDlol m- 2 adsorption densities. These adsorption density levels are in the region of bilayer
and the two shoulders correspond to the absorption bands of calcium oleate. As such, these two
shoulders are assigned to calcium oleate involving the oleate in bilayer and the counter calcium ions
that saturate the head group of this layer. Furthermore, the spectra in the case of the isotherm at
0.2% solids (Figure 4A) showed the absorbance bands of surface calcium oleate (1555 cm- I ),
calcium oleate (1576 and 1540 cm- I ) and sodium oleate (1560 cm- I ) in the region of the isotherm
where it levelled off at 10 !-lmol m- 2 [8]. In this region, the intensities of these bands are not
influenced by concentrations of oleate. Below the 10 /-lmol m- 2 adsorption density, only the
surface calcium oleate band is obselVed. The conclusion is that these bands are due to the
formation of calcium oleate and sodium oleate involving the oleate ions in the second layer and
calcium and sodium ions saturated on the head group of this layer during the sample preparation.
342

Monocoordination of the oleate through counter sodium and calcium ions (when the sample
exhibits negative potential) or chemisorption of oleate on the surface calcium (when the sample
exhibits positive potential), followed by a bilayer formation through a tail-tail bonding, preceding
the growth of calcium oleate on the substrate can be concluded as the adsorption mechanism of
oleate on fluorite.

5.3. APATITE-OLEATE SYSTEM

5.3.1. Adsorption studies. Oleate adsorption on apatite from 1.87xlO-4 M solution as a function
of pH was reported by Moudgil et al. [64]. The maximum adsorption of the oleate occurs at a
pH 8.5. While studying the adsorption of oleate at alkaline pH values, Rao et al. [5] observed that
the maximum uptake of oleate is at a pH 8.0.
Moudgil et al. [65] reported isotherms at two different solid loadings and indicated three regions
for oleate adsorption involving molecular adsorption, surface and bulk precipitation of calcium
oleate. However, as the sample was very porous and the adsorption density reached from lxlO-7
to lxl0-3 M, residual oleate concentration was only 1.11 ~ol m- 2, the accessible surface area is
stated to be limited by the rate of diffusion of the reagent into fine pores [2]. Therefore, these
results give no information on the adsorption process.

25
(A)
pH
25
"x 8.0
9.0
(B)
0 10.0
20 Solids (0/0)
lJ. 11.0
0 0.1
20 0.2
NE
NS "
I>. 0.3
-0 c 0.4
•
E -0
15
.:'0 E 0.5
Cl ~ 15 <> 0.6
'"
""0'"
Ci 1 0.7
~
crl + 0.8
'"Cl
..:
~
0 t 0.9
10
~
..:
'"Ci
..: 10
+
'"
..J
0 ~
..:
~
...l
0

EQUILIBRIUM OLEATE CONe. (moles / I)

EQUILIBRIUM CONC. (moles /1)

Figure 8. Adsorption isotherms of oleate on apatite: (A) at different pH values; (8) at different solidi
liquid ratios at pH 10 [5).

The adsorption isotherms of oleate on fluorapatite at different pH values are shown in Figure 8A
[5]. At pH 8, an infinite vertical line occurs at an equilibrium oleate concentration of 9xlO- 6 M,
 343

characterising precipitation of calcium oleate at an adsorption density slightly higher than

5 Ilmol m-2 . The solubility product of this three dimensional phase is estimated to be pK 14.5.
The isotherms at pH 9, 10 and 11 showed two plateau levels, one at 5 Ilmol m-2 and the other at
10 Ilmol m- 2 . A similar shape of the isotherm is observed, irrespective of whether the solids are
added to oleate solution or oleate is added to an apatite suspension. The shape of these isotherms
have been attributed to a bilayer formation, with a liquid crystal state of oleate in the adsorbed
region.
Evidence of oleate adsorption up to 10 Ilmol m- 2 was also obtained while studying the
adsorption at different solid/liquid ratios. Figure SB shows the results where the adsorption density
is observed to be independent of the solid/liquid ratio up to the formation of the bilayer, after which
it is found to be dependent on the solid content. The dependence of the solid content on oleate
abstraction after the bilayer formation is considered to be due to calcium oleate precipitation.

5.3.2. Infrared studies. Fuerstenau [66] reported the formation of surface calcium oleate on the
apatite with infrared spectroscopy (the results are not published). The IR spectrum of apatite treated
with 7xlO-4 M oleate at pH 9.5 shows the absorbance bands of sodium oleate (155S cm- I ) and
calcium oleate (1535 and 1570 cm- I ), indicating both the species on the surface [67].
The FT-IR spectra of the apatite floated at pH S and 10, while using an oleate concentration
equal to a monolayer capacity (5 Ilmol m- 2), show a characteristic peak at 1550 cm- I regardless of
pH [3,45]. The double peak, characteristic of calcium oleate, at 1571 and 1540 cm- I in the
spectrum of the unfloated material (oleate concentration somewhat higher than that required for a
complete double layer is used in the flotation) support the precipitation in the bulk phase. Double
layer formation is also confirmed from flotation recovery, as the maximum recovery was obtained
at a monolayer coverage, with a sharp drop in the recovery with an increase in the oleate
concentration. The occurrence of only one peak shows that it is a 1: 1 calcium oleate complex that
builds up the monolayer.

5.3.3. Zeta-potential studies. Most of the zeta potential studies of apatite in the presence of oleate
showed an increase in the negative charge with increasing oleate concentrations at pH values above
5 [46,47,63]. Because of the increase in the negative charge of apatite by many units above the
i.e.p., chemical interaction between the carboxylate ion of oleate and surface calcium is suggested.
The results obtained by Mishra et al. [47] are shown in Figure 9A. They concluded that the
decrease in zeta potential above pH 7 is due to the adsorption of oleate ions and not due to the
precipitation. This conclusion is based on the calculation of the pH region of precipitation, taking
into account the dissolved calcium. Similar zeta potential curves for fluorapatite in the presence of
oleate as shown in Figure 9A were also reported [5].
Figure 9B shows the zeta potential of apatite as a function of initial oleate concentration at basic
pH values. From these results, Mishra [46] suggested the specific adsorption of oleate due to the
similarity of surface charge, whereas the increase in negative zeta potential at pH 10 was correlated
[5] to a successive oleate adsorption and bilayered structure as per the isotherm at pH 10 (Figure
SA). The constant zeta potential beyond 2x 10 4 M oleate is thought to be due to the preCipitation of
calcium oleate.

5.3.4. Discussion. The isotherms shown in Figure SA indicate a bilayer formation for a two
dimensional condensation of oleate on a heterogeneous surface. Also, it can be seen that the
isotherms for bilayer formation make a vertical jump because of the condensation on a
344

homogeneous surface rendered by the first layer. However, Finkelstein criticised the significance
of these results [2] by stating that the solubility of calcium oleate is exceeded at all points on the
adsorption curves and the experimental method did not distinguish adsorption from precipitation of
calcium oleate in the solution phase. It is difficult to accept this judgement, since it is based on a
calcium oleate pK value of 15.6. It is not likely that a change in the solid/liquid ratio will yield the
same adsorption density at a particular equilibrium concentration, if precipitation is indeed taking
place below 10 !IDlol m-2 (bilayer). The FT-IR spectra and flotation studies support the bilayered
structure of oleate on apatite [3,5,45J.

Or--------------~
(B)

-10
;; pH 10.0
E
::J -20
::;
f-
Z -30

~
c..
-< -40
o
~ oRao et al [5]
c.o. Mishra [46]
-50

-5 3\------>4-->---,6!-.......7----!;---;9\--~\O~---,Jll -60
10- 10-6 10-' 10 4 10-3
pH SODIUM OLEATE ADDITION (moles / I)

Figure 9. (A) Electrophoretic mobility-pH curves of apatite at different concentrations of oleate [47);
(8) Zeta potential of apatite versus initial oleate concentration at basic pH values.

The adsorption behaviour at pH 8 forming only a monolayer coverage prior to the precipitation
is thought to be due to a higher level of calcium (3.5x 10-5 M) in the pulp. If we accept a pK value
of 14.5 for calcium oleate as per the pulp conditions at pH 8, then an oleate concentration greater
than 1.77xlO-5 M is needed for precipitation at pH 10, considering the dissolved calcium ions of
1. Ox 10- 5 M at this pH [5J. At 1. 77x 10-5 M oleate, the isotherm at pH 10 indicates a monolayer
coverage and once the surface is covered, the calcium ions in the pulp may possibly be less due to
the inhibition of dissolution. Hence, bilayer formation is feasible prior to precipitation.
The sample, which shows a surface calcium oleate (I: I) from the infrared spectrum, is slightly
negative (-5 mY) in the entire alkaline pH region [3]. Hence, oleate is adsorbed on the surface
calcium rather than via counter calcium ions. However, in the case of the spectrum [67] showing
sodium and calcium oleates after conditioning in a 7x 10-4 M oleate solution, it could be that
precipitation takes place in this concentration range or that the samples possibly have a high
negative potential at pH 9.5 thereby allowing the adsorption of oleate through counter sodium and
calcium ions.
 345

In conclusion, the results indicate that a two dimensional condensation of oleate on a

heterogeneous surface gives rise to a bilayered structure of oleate on apatite, prior to the
precipitation of calcium oleate. The adsorption density for a monolayer coverage corresponds to a
liquid crystal state of oleate with a molecular coverage area of 33 A2.

5.4. SCHEELITE-OLEATE SYSTEM

5.4.1. Adsorption studies. The adsorption of sodium myristate on scheelite for two different
solid/liquid ratios at pH 10 was reported [34]. These results suggested that, when a small amount
of solid was added, the uptake was not limited to a monolayer, but corresponded to a
multimolecular layer on average. From the experimental and theoretical mass balances, they
correlated the abstraction of m yri state with the precipitation of calcium soap.
Rybinski and Schwuger (68) suggested a double layer formation of oleate on the scheelite from
the adsorption and calorimetric studies, while using a molecular coverage area of 20.5 A2. An
isotherm which levels off at an adsorption density corresponding to the bilayer has been reported
(Figure lOA). However, the exchange enthalpy seems to correlate up to 5 Ilmol m- 2 adsorption
density rather than to a monolayer coverage (8 1J.ID01 m-2).

35r-------------------~~------__,
(B) pH: 10
Spectroscopic method
A pH adjusted WIth NaOH
30 Radio tracer method
o pH adjusted with NaOH 0
)( pH adjusted wlth
carbonate buffer
25

(A)
800
50 ~ 20
"-
Monolayer Bilayer -600 _ o
Ul
2 ~
c I
I o
15
5 30 I
I -400 ~ '"o
I
""~
'- lOV-__ t~_- ___ t~~n~I~"' ____ -
>< 20 Bilayer
10
-200 <J"5
""
Ul

:5
o I I ' I , I

0.5 I 1.0 I 1.5 2.0

: ex 10 3 (~olll)
I I
, ['0 1'S! 20 25 ~O-7 10- 5 10 4 10"3

EQUILIBRIUM OLEATE CONe. (moles II)

Figure 10. (A) Adsorption isotherm and exchange enthalpy as a function of adsorption density of oleate
on scheelite at pH 10 [68); (8) Adsorption isotherms of oleate on scheelite at pH 10 [6).

Two methods of oleate analysis were used [6) to determine the adsorption isotherms as shown
in Figure lOB. The adsorption isotherm using the UV spectroscopic method shows a linear
increase in the amount adsorbed versus eqUilibrium oleate concentration up to 10 Ilmol m-2. After
346

this, a steep increase in the adsotption density can be seen. These adsotption results have been
compared with the theoretical values for the precipitation of calcium oleate as shown in Figure
IIA. The theoretical values suggest that the precipitation starts at IxlO-5 M initial oleate
concentration and is almost complete around IxlO-4 M oleate concentration, since the amount of
oleate left after this concentration increases proportionally with the increase in the initial oleate
concentration. On the other hand, the experimental results of oleate left in the solution show a
constant value between IxlO-4 and 2xlO-4 M of initial oleate, after which it increases with the
increase in the initial oleate concentration.

(A) 25
Thewe(ica.i Ellpe:nment.aJ
OLEATELEFI' o (B)
Radio tracer method
pH::::9 (adjusted with borate buffer)
20 o Solids allowed to settle
after equilibrium
Me Ie Suspension cennifused
after equilibrium
~
.:; 15
0
UJ

'"'"s;:
5 0 10

"'I-'"
5o
z
~
'"
UJ
to 6
~

lO":Wr--~~-'-':"lO,",O~---'-'-'-"lU:':.',...--'~'-W."'l":<r';;-'-'--'-'--""
, , ~IJ) ~\J7 1,," 10' liT' w3
INITIAL OLEATE CONe. (moles II) EQUILIBRIUM OLEATE CONe. (moles II)

Figure 11. (A) Comparison of theoretical and experimental oleate concentrations for Ca-soap formation
(exptl. results are from Figure lOB and of the isotherm while using spectroscopic method for oleate
analysis); (8) Adsorption isotherm of oleate on scheelite at pH 9 [6].

If the region at which the constant oleate left is due to precipitation, then the oleate consumed for
precipitation is around I x I 0-4 M. This concentration of oleate nearly corresponds to the amount of
oleate needed for the precipitation of dissolved calcium ions (5xlQ-5 M at a pH value of 10). These
results suggest that the onset of calcium oleate precipitation begins at an initial oleate concentration
of IxlO- 4 M, which corresponds to an equilibrium oleate of 1.56xlO-5 M. At this eqUilibrium
oleate concentration, the isotherm exhibits a steep increase in the adsotption density. Thus the steep
increase in the slope of the isotherm can be attributed to calcium oleate precipitation. A solubility
product of pK 14.44 is estimated while considering the analysis of pulp liquid after adsorption.
The shape of the isotherms are found to be the same when the radio tracer technique is used for
oleate analysis (Figure IDB). In the presence of carbonate, the surface calcium is expected to
complex with the carbonate species and for an exchange reaction with oleate ions, a high
concentration of oleate is probably required. This explains the shift in the adsotption isotherm to a
higher equilibrium oleate concentration in the presence of a carbonate buffer. The infinite increase
in the slope after 10 lIDlol m-2 characterises calcium oleate precipitation.
 347

Figure lIB shows the oleate adsorption at pH 9 using a borate buffer solution and when the
residual oleate concentration is measured after allowing the particles to settle and the liquid
centrifuged. It shows a plateau around 10 lIDlol m-2 before the precipitation step of calcium oleate.
However, when the pH is adjusted with NaOH, the shape of the isotherm is found to be similar as
shown in Figure lOB [6]. In the presence of borate, the free calcium ions are not available for
precipitation due to complexation with the borate species. This explains the plateau that precedes
the precipitation step of calcium oleate.

5.4.2. Infrared studies. The infrared spectrum of scheelite treated with sodium oleate
(5.86xlO- 3 M) at pH 9.5 was reported [69]. Although chemisorption of oleate is suggested, the
spectrum exhibits a calcium oleate absoroance bands.
The growth of the adsorbed layer as studied by the diffuse reflectance IR spectra showed the
bands corresponding to sodium oleate (1561 cm- I) and calcium oleate (\539 and 1577 cm- I) from
the beginning of the surface coverage [6]. From the adsorption studies, the precipitation of calcium
oleate is explained after 10 !-lmol m- 2, whereas the spectra showed that the surface is filled with
sodium and calcium oleates even in the monolayer range. This discrepancy is explained by the
negative zeta potential of the mineral surface and hence, monocoordination of oleate through
counter sodium and calcium ions that exists at the interfacial region is considered rather than the
growth of surface precipitate at low oleate concentrations and in the monolayer range.

5.4.3. Zeta-potential studies. As scheclite is negatively charged in the entire pH range and the zeta
potential decreases in the presence of oleate above the pH values of 6, oleate is possibly
specifically adsorbed on scheelite [6]. The gradual decrease in zeta potential with increasing oleate
concentrations at pH 10 is shown to correspond to a bilayer formation. However, at high oleate
concentrations, the solids are thought to be covered with a layer of calcium oleate as the constant
zeta potentials at these concentrations corresponds to that of calcium oleate.

5.4.4. Discussion. In general, three regions of oleate adsorption on scheelite have been described
from the shape of the isotherms as shown in Figure lOB [6]. The first region is that where there is
a linear increase in the amount adsorbed at low eqUilibrium oleate concentrations. In almost all
cases of the adsorption isotherms, this linear increase is shown only up to 10 !-lmol m- 2, and in
one case, a plateau was observed at this adsorption density (Figure 118). Hence, the first region of
adsorption is considered to be due to a bilayer formation, assuming the liquid crystal state of oleate
in the adsorbed region because of the two dimensional condensation.
The second region of oleate adsorption is characterised by an infinite increase in the slope of the
isotherms. This region corresponds to the three dimensional growth of the calcium oleate on the
substrate. The results indicate that there is no formation of calcium oleate bulk precipitate and even
if it has been formed, the solids have abstracted all of it.
The third region of adsorption is the zone where there is a decrease in the adsorption density at
high equilibrium oleate concentrations. Formation of bulk precipitate is shown apart from the
surface precipitate in this region. The decrease in the adsorption density at high oleate
concentrations is possibly due to the redissolution of the precipitate caused by the micellar state of
oleate at these concentrations.
In summary, monocoordination of oleate through counter sodium and calcium ions for the
monolayer filling followed by a bilayer formation through tail-tail bonding, prior to the growth of
calcium oleate precipitate on the substrate is proposed as the adsorption mechanism of oleate on
scheelite.
348

5.5. BARYTE-OLEATESYSTEM

Plitt and Kim [70] reported the quantitative infrared adsorption studies of oleate on baryte. The
adsorption of oleate from a 1xl 0·3 M solution as a function of pH showed a maximum at pH 3 and
decreased as the pH increased up to 9. Above pH 9, a constant adsorption of oleate is shown
which is mostly in ionic form.
The adsorption isotherm of oleate on baryte showed a plateau slightly below the amount required
for a vertically oriented monomolecular coverage (Figure l2A). The authors suggested a thick
multilayer adsorption model where the oleate species lie parallel to the solid surface. This model is
based on the fact that the isotherm in the adsorption range shows more than a monolayer coverage
(using 20.5 A2 molecular area for oleate), and the floatablity of baryte is reduced under conditions
of high collector concentrations. The reduced floatability was thought to be due to the detachment
of the mobile outermost collector species.
The shape of the isotherm as presented in Figure l2A, can be explained successfully by
invoking the idea of a two dimensional condensation of oleate on the surface. The surface area of
the mineral was reponed to be 2.47 m 2 g.l. Hence, a monolayer coverage is expected (in the
adsorbed region, oleate is in a liquid crystal state, i.e., a molecular area of 33 A2) at an oleate
adsorption density of 1.24xlO· 5 mole g'!. The first plateau corresponds to the same adsorption
density indicating a monolayer coverage. The second plateau around 2.7xIO· 5 mole g'!
corresponds to a bilayer formation and hence, a reduced floatability at high oleate concentrations is
observed.

6 (A)

L
MONOLAYER COVERAGE
USING 20.5 A2 PER (B)
OLEATE MOLECULE
DISPERSED
----- -----------------------

HEMATITE
NotNO J
Z~lO-JM

PERCENT SOUDS
o 0.05
pH95 pH JO
6. 0.10
o 0.040

la' la l
OLEATE CONCENTRATION (moles II) EQUILIBRIUM CON CENTRAnON (moles II)

Figure 12. (A) Adsorption isotherm of oleate on baryte at pH 10 [70]; (8) Adsorption isotherms of
oleate on hematite al various pH values [32]

The measured abstraction of oleate by baryte is compared with theoretical solubility values [27].
Despite the discrepancy between the results, the abstraction of oleate is stated to be due to the
precipitation of metal soap. This conclusion is based on similar studies with calcite and fluorite,
where a good agreement was obtained between the theoretical and experimental values.
 349

Peck [44] reported the chemisorption of oleate on baryte by infrared studies, where the anti-
symmetrical carbonyl absorption band appears at 1543 cm- 1 compared to 1515 cm- 1 for the
corresponding band for barium oleate. No evidence for bulk soap precipitation was observed at
pH 9.5, instead an equilibrium between chemisorbed oleate, physically adsorbed sodium oleate
and sodium oleate in the solution was shown.
As baryte shows negative zeta potential at basic pH values [27], monocoordination of oleate on
either surface barium or through counter sodium and barium ions can be suggested for the
monolayer filling. This is supported by the presence of both the absorbance bands of sodium oleate
and barium oleate in the infrared spectrum. Moreover, the adsorption of oleate from sodium
sulphate solution shows a decrease in the chemisorbed oleate species but an increase in the
physically adsorbed sodium oleate [56]. This implies that the mineral surface is saturated with
sodium ions (as counter ions in the double layer) in the presence of excess sodium ions.

6. General Discussion

Finkelstein [2] critically reviewed the recently published papers on the adsorption of oleate on
sparingly soluble calcium minerals. He observed that there is no consistency in the abstraction of
the oleate by any of the minerals. The results vary widely from one mineral to the other and
between two investigations on the same mineral. An attempt to correlate the published adsorption
results quantitatively has not been successful; a qualitative feature of the oleate adsorption is
discussed and several shortcomings for the surface precipitation mechanism are presented.
From the available evidence, it is believed that the first molecules of the oleate adsorb on the
surface of the calcium minerals and not by a precipitation of calcium oleate at the surface.
However, for a mechanism based on chemisorption, the available evidence is not sufficient.
Evidence for the chemisorption of the oleate has now been presented. In the synthetic fluorite-
oleate system, the growth of the adsorbed layer as studied by the infrared spectra showed that it is
a surface calcium oleate (l: I) that builds up the monolayer coverage, irrespective of whether the
solids are added to oleate solution or oleate solution is added to fluorite suspension (Figure 6). The
transition of singlet absorption of surface calcium oleate to doublet absorption of calcium oleate
was observed only around 10 !lmol m- 2. The solubility of calcium oleate is exceeded several times
in these studies, but, still the spectra indicate only the presence of surface calcium oleate species up
to 10 !lmol m-2 adsorption density. In some studies, the infrared spectra indicate the adsorption of
sodium and calcium oleate species from the beginning of surface coverage and no surface calcium
oleate band [7,62]. A similar spectra were obtained in the scheelite-oleate system [6]. The
difference between the studies showing only the surface calcium oleate band and, sodium and
calcium oleate bands is attributable to the nature of the sample employed. The sample had a positive
potential in the first case and a negative potential with the latter at the adsorption pH value of 10.
Thus, the adsorption of the oleate with the samples showing high negative potentials can be
described as a monocoordination through counter sodium and calcium ions that exists at the
interfacial region. In these cases, the infrared spectra can only show the crystalline calcium oleate
species as calcium and oleate ions could condense as calcium oleate during the sample preparation
(drying). The body of the literature showing the calcium oleate bands at low oleate concentrations
and in the monolayer range could be due to the compound formation (because of monocoordination
of oleate via counter calcium ions) during samples drying rather than the growth of the calcium
oleate precipitate on the surface. It is also worth while to point out that the distribution of the
adsorbed oleate species as a function of pH [44] shows the presence of sodium oleate species in
350

the basic pH region. This implies that oleate is monocoordinated through counter sodium ions as
opposed to physical adsorption of sodium oleate as described previously. since both sodium and
oleate ions are in ionic form. Similarly. oleate can be monocoordinated through counter calcium
ions and in this case. the IR spectra cannot represent the true adsorbed oleate structure for the
reasons described above. Hence. depending upon the surface potential and its magnitude. the
interaction of oleate to the mineral surfaces can be described as follows:

-CaOH + ·OOCR -Ca+ ·OOCR + OH·

-CaOH + Na+ ·OOCR + OH· -CaO Na OOCR· + H20
-CaOH + Ca2+·OOCR + OH· = -CaO Ca OOCR + H20
-CaOH2+ + ·OOCR -Ca+ ·OOCR + H20
-CaOO· Na+ +-()()CR -CaO Na OOCR. where 6 < I
-Cao'l· Ca2++ ·OOCR -CaO Ca OOCR, where [) :;; I

From the results presented and the discussions followed. it is apparent that precipitation is not
taking place in the monolayer range. Should this be the case. the nature of oleate adsorption cannot
be different from that of the other oxide minerals. where precipitation problems are not
encountered. Thus. the monolayer models developed for the insoluble oxide minerals would also
be applicable to the systems reviewed.
The shape of the isotherms show some similarities. despite the varying adsorption results; they
either show a plateau or level off at 10 I-lmol m· 2 (Figures 3A,4.8.llB and 12). Moreover. an
initial break at 5 I-lmol m- 2 adsorption density was also observed in some cases (Figures IA.8A
and l2A). These breaks correspond to a mono and bilayer formation for the two dimensional
condensation of oleate on the mineral surface (liquid crystal state. 33 A2). However. the isotherms
showing a steep increase in adsorption density after 51-lmol m- 2 (monolayer) in the case of calcite.
and after 10 I-lffiol m-2 (bilayer) in the cases of fluorite. apatite and scheelite can be described due to
calcium oleate precipitation. These precipitation regions have been characterised with a calcium
oleate solubility product of pK '" 14.4.
Hall and coworkers (71) from an analysis of the changes in the water-adsorption properties of
fluorite and calcite on which varying quantities of oleate were adsorbed. concluded that the
coverage of the surface by oleate is equivalent to 70% of a closely packed. vertically orientated
monolayer (molecular coverage area 20.5 A2) and therefore the molecules probably pack with a
density equivalent to that of a condensed liquid film. If one uses'" 33 A2 for the molecular
coverage area. the results can be explained with a monolayer capacity. Recent results [54) also
suggest an oleate packing area of 34.2 A2. Thus. the present shape of the isotherms discussed.
support the two dimensional condensation model and the assumption of oleate as being in a liquid
crystal state in the adsorbed region corresponding to a molecular coverage area of 33 A2.
The plateau at a bilayer formation could also be seen from the isotherms in the presence of borate
and EDTA (Figures lIB and 4B). Here. the isotherms show a clear break at 10 I-lmol m- 2
adsorption density. The dissolved calcium ions form a complex in the presence of borate and
EDT A. and hence. are not available for precipitation. In the absence of borate and EDTA. the
plateau is masked due to the precipitation of calcium oleate that follows immediately after bilayer
formation. This implies that if there is no precipitation reaction in the presence of oleate. the
isotherms should level off at 10 I-lmol m- 2. It is really the case with the isotherms of oleate on
hematite [32]. Figure 12B shows these isotherms between the range of pH values 7.7 to 10. where
they tend to a plateau at an oleate adsorption density of 10 I-lmol m- 2. Even though the authors
indicate that this adsorption density appears to be more than that required for a closed packed
 351

monolayer, it correlates with the bilayer formation if one assumes a two dimensional condensation
of oleate on the surface, i.e., a molecular coverage area of 33 A2. If we assume that this adsorption
density is a bilayer formation, then the results in Figure 12B are self explanatory, i.e., (a) at the
plateau level, the particles are well dispersed; (b) the particles start dispersing at and above 5 ~ol
m- 2 at pH values 9, 9.5 and 10, a clear indication of the building up of bilayer formation; (c) the
slope of the isotherms increases after 5 !lmol m- 2 at pH values 9, 9.5 and 10, which is the
condensation of oleate on a homogeneous surface as per the present discussion; (d) more
importantly, it has been mentioned that after treating the sample with cyclohexane, etc., the
particles at the plateau level tend to be hydrophilic, whereas in the flocculated region, they are
highly hydrophobic, which is a clear indication of a bilayer formation at the plateau value of
1O~0Im-2.
The isotherms describing abstraction-by-precipitation curves are based on the correspondence
between the experimental oleate concentrations to that of the theoretical values for the formation of
calcium oleate solid phase. The theoretical calculations are dependent on the chosen pK value of
calcium oleate, the value of which is ambiguous due to the complexity of the present multiphase
system. It should be noted that Marinakis and Shergold used a pK value of 15.41 [27], although
they obtained a value of 14.1 by nephelometric titrations. Similarly, a pK value of 15.4 for calcium
myristate was used against the literature value of 15.0 [34). Moreover, as the measured
abstractions are more than a nominal monolayer in these cases, it was argued [2) that the
correspondence between the experimental and theoretical values is irrelevant to the question of the
mechanism of formation of the first adsorbed layer. Using the SOLGASWATER software for the
theoretical computations, the theoretical and experimental values were shown to correspond to one
another only up to the precipitation regions of the isotherms [4,6-8). Here, a calcium oleate pK of
14.4 was used in the theoretical calculations, which is estimated from the analysis of pulp liquid.
The results also indicate that a change in solid/liquid ratio has no effect on the shape of the
isotherm up to a bilayer formation (Figure 8B). In another case, the shape of the isotherm was
found to be dependent on the solid content, even though, only the adsorption phenomenon is
indicated up to 10 !lmol m- 2 (Figures 4A and 6). The dependence on the solidi liquid ratio within
the adsorption region could possibly be explained by considering the adsorption models. The
models describe the growth of the adsorbed layer with the formation of patches initially at surface
sites which are most energetic, followed by less energetic sites. It is possible that a change in the
solid/liquid ratio could lead to a different distribution of surface energies. This can be realised due
to the same amount of calcium in the solution at eqUilibrium, irrespective of the solid content. This
means that the dissolution of calcium per unit area will increase with a decrease in solid content and
therefore a different distribution of surface energies can be expected.
The adsorption isotherms also show that with increasing solid contents, the precipitation step
after the bilayer formation takes place at a high equilibrium oleate and the extent of precipitation
decreases. At a high solids content, the isotherms level off at a bilayer formation (Figures 4A and
8B). This can be attributed to the adsorption of calcium ions on the hydrophilic layer (bilayer).
The adsorption of counter ions on the second layer has been considered by Harwell et al. [16) in
the admicelle hypothesis, and in their recent studies they showed a near 100% counter ion
adsorption [72). As the solid content increases in the pulp liquid, the availability of the hydrophilic
layer for calcium ions to adsorb on is greater and hence the reduction in the bulk precipitation.
When the surface area is sufficiently high to adsorb all the dissolved calcium ions, no precipitation
is expected, which is the cause of the isotherms levelling off at the bilayer formation. Hence, the
adsorption maximum due to preCipitation is dependent on the solid content, which possibly
explains the different abstraction levels reported in the literature. Based on the results, the
352

following adsorption models at low and high solid concentrations are proposed: at low solid
concentration:

Dilute adsortJcd -> Formation of patches in the liquid crystal

state of the layer state issue of the two dimensional condensation
on heterogeneous surface

-> Bilayer formation directly -> At very high concentrations. possibility of three
on the monolayer or through dimensional condensation on surface and
admicelles (the layer stays change over of the adsortJcd layer to the
in liquid crystal state although hydrated crystal state
the outer point of the layer is
mainly saturated with Ca2+)

At high solid concentration. the adsorbed layer is limited to the bilayer mainly saturated with Ca 2+
ions.
Although. a correlation of zeta potential curves of increasing initial oleate concentrations to that
of the isotherms with equilibrium oleate concentrations is rather questionable. the results do
indicate a close comparison for a monolayer coverage in the case of calcite [4] and a bilayer
formation in the case of fluorite. apatite and scheelite [5-8]. Such a correlation can be justified by
the very low solid content generally used in the zeta potential studies.
In conclusion, the authors suggest a monolayer coverage in the case of calcite and a bilayer
formation in the cases of fluorite, apatite, scheelite and baryte, prior to the precipitation of calcium
oleate, as the adsorption mechanism of oleate on these minerals. The adsorption density at
monolayer coverage corresponds to a condensed state of oleate with a molecular coverage area of
33A2.

7. Acknowledgement

The authors are grateful to the Swedish Mineral Processing Research Foundation (MinFo) for
financial support.

8. References

1. Hanna, H.S., and Somasundaran, P. (1976) Flotation of salt-type minerals, in M.C.

Fuerstenau (ed.) Flotation, A.M. Gaudin Memorial Volume, AIME, New York. vol. 1,
pp.197-272.
2. Finkelstein. N.P. (1989) Review of interactions injlotation of sparingly soluble calcium
minerals with anionic collectors, Trans. IMM.!:2l!. 157-177.
3. Hanumantha Rao, K., Antti, B-M., Cases, I.M., and Forssberg, K.S.E. (1988) Studies on
the adsorption of oleate from aqueous solution onto apatite. in E. Forssberg (ed.)
Developments in Mineral Processing, Proceedings XVIth Int. Miner. Process. Congress.
Elsevier. Amsterdam, vol. lOA, pp. 625-636.
4. Hanumantha Rao, K., Antti, B.-M .. and Forssberg, K.S.E. (1988/89) Mechanism of oleate
interaction on salt-type minerals: Part I. Adsorption and electrokinetic studies of calcite in the
 353

presence of sodium oleate and sodium metasilicate, Coil. & Surf. 34 227-239.
5. Hanumantha Rao, K., Antti, B.-M., and Forssberg, K.S.E. (1990) Mechanism of oleate
interaction on salt-type minerals: Part II. Adsorption and electrokinetic studies of apatite in the
presence of sodium oleate and sodium metasilicate, lnt. 1. Miner. Process. 2.8., 59-79.
6. Hanumantha Rao, K., and Forssberg, K.S.E. (1991) Mechanism of oleate interaction on salt-
type minerals: Part lIl. Adsorption, electrokinetic and diffuse reflectance FT-IR studies of
scheelite in the presence of sodium oleate, Coll. & Surf.~, 161-187.
7. Hanumantha Rao, K., Cases, J.M., de Donato, P., and Forssberg, K.S.E. (1991)
Mechanism of oleate interaction on salt-type minerals: Part lV. Adsorption, electrokinetic and
diffuse reflectance FT-IR studies of natural fluorite in the presence of sodium oleate, J.
Colloid lnterf. Sci. (in press).
8. Hanumantha Rao, K., Cases, 1.M. and Forssberg, K.S.E. (1991) Mechanism of oleate
interaction on salt-type minerals: Part V. Adsorption and precipitation reactions in relation to
the solidI liquid ratio in the synthetic fluorite-sodium oleate system, J. Colloid lnterf. Sci. (in
press).
9. Hanumantha Rao, K., and Forssberg, K.S.E. (1991) Adsorption of oleate species on
sparingly soluble calcium minerals from aqueous solutions: A comparision of theoretical and
experimental thermodynamic chemical equilibria, Proceedings 3rd lnt. Symp. Beneficiation
and Agglomeration (Bhubaneswar, India, January 16-18).
10. Hanumantha Rao, K., and Forssbcrg, K.S.E. (1991) Adsorption of oleate from aqueous
solution onto fluorite, Proceedings XVIIth Int. Miner. Process. Congress (Dresden).
11. Gaudin, A.M., and Fuerstenau, D. W. (1955) Quartz flotation with cationic collectors, Trans.
AIME 202, 441-501.
12. Somasundaran, P., Healy, T.W., and Fuerstenau, D.W. (1964) Surfactant adsorption at the
solid-liquid interface - Dependence of mechanism of chain length, J. Phys. Chern. 68,
3562-3566.
13. Predali, J.J., and Cases, J.M. (1973) Thermodynamics of the adwrption of collectors, in
M.J. Jones (ed.) Proceedings Xth Int. Miner. Process. Congress, IMM, London, pp. 473-
492.
14. Cases, J.M. (1979) Adsorption des tensio-actifs a [,interface solide-liquide: Thermodyna-
mique et influence de L'heterogeneite des adsorbants, Bull. Mineral. 102,684-707.
15. Cases, J.M., Levitz, P., Poirier, J.E., and Van Damme, H. (1986) Adsorption of ionic and
nonionic surfactants on mineral solids from aqueous solutions, in P. Somasundaran (ed.)
Advances in Mineral Processing, SME Publication, Littleton, Col., pp. 171-186.
16. Harwell, J.H., Schechter, R., and Wade, W.H. (1985) Recent developments in the theory of
surfactant adsorption on oxides, in J.M. Cases (ed.) Solid- Liquid Interactions in Porous
Media, Technip, Paris, pp. 371-409.
17. Cases, J.M. (1989) Thermodynamics of the adsorption of surfactants, in F.I.A. (ed.)
Kurzjassungen der Vortrage und Posterbiertrage- 2. Internationale Fachtagung Fortschritte in
Theorie und Praxis der Aujbereitungstechnik, Freiberg, pp. 341-356.
18. Ananthapadmanabhan, K.P., and Somasundaran, P. (1985) Surface precipitation of
inorganics and surfactants and its role in adsorption andflotation, Coil. & Surf. D, 151-167.
19. lung, R.F. (1976) Oleic acid adsorption at the goethite-water interface, M.Sc. Thesis, Univ.
of Melbourne.
20. Ananthapadmanabhan, K., Somasundaran, P., and Healy, T.W. (1980) Chemistry of oleate
and amine solutions in relation to flotation, Trans. AIME 2QQ, 2003-2008.
21. Jung, R.F., lames, R.O., and Healy, T.W. (1988) A multiple equilibria model of the
354

adsorption of oleate aqueous species at the goethite-water interface. J. Colloid Interf. Sci.
122,544-549.
22. Du Rietz, C. (1958) Fatty acids in flotation, in Proceedings Congress "Progress in Mineral
Dressing" (Stockholm, 1957), Almquist and Wiksell, Stockholm, pp. 417-439.
23. Yoke. J.T. (1958) The solubilities of calcium soaps. J. Phys. Chem.~. 753-755.
24. Irani, R.R. and Callis, C.F. (1960) Metal complexing by phosphorus compounds. I/.
Solubilities of calcium soaps of linear carboxylic acids, ibid. 64. 1741-1743.
25. Fuerstenau, M.C., and Elgillani, D.A. (1966) Calcium activation in sulphonate and oleate
flotation of quartz. Trans. AIME m. 405-413.
26. Du Rietz. C. (1975) Chemisorption of collectors inflotation. in Proceedings XIth Int. Miner.
Process. Congress (Cagliari), Instituto di Arte Mineraria e Preperazione dei Minerali, Rome,
pp. 395-403.
27. Marinakis. K.I .. and Shergold, H.L. (1985) The mechanism offatty acid adsorption in the
presence offluorite, calcite and baryte. Int. J. Miner. Process. H. 161-176.
28. Hu, J.S .• Misra, M., and Miller, J.D. (1986) Effect of temperature and oxygen on oleate
adsorption by fluorite. ibid. ll. 57 -72.
29. Kitchener, J.A. (1984) Thefrothflotation process: past, present andfuture - in brief. in K.J.
Ives (cd.) The Scientific Basis of Flotation. Martinus Nijhoff Publishers, The Hague,
pp. 3-51.
30. Parekh. B.K.. and Aplan, F.F. (1985) The critical surface tension of wetting of minerals
coated with collectors, in Proceedings XVth Int. Miner. Process. Congress (Cannes). Societe
de l'Industrie Minera\e, vol. 2, pp. 3-15.
31. Belle, J., and Bothorel. P. (1977) Thermodynamical study of the phospholipid chain
structure in mono and bilayers. Nouv. J. Chim.l(4). 265-267.
32. Yap, S.N .• Mishra, R.K., Raghavan, S., and Fuerstenau, D.W. (1981) The adsorption of
oleate from aqueous solution onto hematite, in RH. Tewari (ed.) Adsorption from Aqueous
Solution, Plenum Press, New York. pp. 119-142.
33. Gregory, G.R.E.C. (1966) The determination of residual anionic surface active reagents in
mineral flotation liquors. Analyst 21, 251-226.
34. Atademir, M.R, Kitchener, J.A., and Shergold, H.L. (1981) The surface chemistry and
flotation of scheelite. ll. Flotation "collectors", Int. J. Miner. Process. ~ 9-16.
35. Forssberg. K.S.E., Antti, B-M., and Palsson B.l. (1984) Computer assisted calculations of
thermodynamic equilibria in the chalcopyrite-ethyl xanthate system, in M.J. Jones and R.
Oblatt (eds.) Reagents in the Minerals Industry, IMM, London, pp. 251-264.
36. Iskra, J., Gutierrez, C., and Kitchener, J.A. (1973) Influence of quebracho on the flotation
offluorite, calcite, hematite and quartz with oleate as collector, Trans. IMM 82. C73-C78.
37. Gutierrez. C. (1979) bifluence of calcium ion and quebracho in the oleate flotation offluorite
and calcite. Trans. AIME 2QQ, 1918-1924.
38. Somasundaran, P. (1969) Adsorption of starch and oleate and interactions between them on
calcite in aqueous solutions, J. Colloid Interf. Sci. n(4). 557-565.
39. O'Connor, C.T., Dunne. RC .• and Martalas. A. (1987) The adsorption of oleate and the
guar-based gum, acrol LG-21, onto apatite and calcite. CoIl. & Surf. 27, 357-365.
40. Sadowski, Z. (1990) The effect of dispersant reagents on the sodium oleate adsorption at the
salt minerals-water interface, 8th Int. Symp. Surfactants in Solution, Gainesville, USA, June
10-15.
41. Zimmels, Y., Lin, I.J., and Friend J.P. (1975) The relation between stepwise bulk
association and interfacial phenomena for some aqueous surfactant solutions, ColI. Poly. Sci.
 355

253,404-421.
42. Szczypa, J., Chibowski, St., and Kuspit, K. (1979) Mechanism of adsorption of sodium
laurate by calcium carbonate, Trans. IMM, March, CII-C13.
43. Fuerstenau, M.C., and Miller, J.D. (1967) The role of the hydro carbon chain in anionic
flotation of calcite, Trans. AIME ill, 153-160.
44. Peck, A.S. (1963) Infrared studies of oleic acid and sodium oleate adsorption on fluorite,
baryte and calcite, Rep. Invest. U.S. Bur. Mines 6202, p. 16.
45. Antti B-M., and Forssbcrg, K.S.E. (1989) Pulp chemsitry in industrial mineral flotation.
Studies of sUrface complex on calcite and apatite sUrfaces using FTlR spectroscopy, Miner.
Engng. 2, 217-227.
46. Mishra, S.K. (1982) Electrokinetic properties andflotation behaviour of apatite and calcite in
the presence of sodium oleate and sodium metasilicate, Int. J. Miner. Process. 2, 59-73.
47. Mishra, R.K., Chander, S., and Fuerstenau, D. W. (1980) Effect of ionic surfactants on the
electrophoretic mobility of hydroxyapatite, Coli. & Surf. 1, 105-119.
48. Matijevic, E., Leja, J., and Nemeth, R. (1966) Precipitation phenomena of heavy metal soaps
in aqueous solutions, 1. Colloid Interf. Sci. 22,419-429.
49. Pugh, R.J. (1986) The role of the solution chemistry of dodecylamine and oleic acid
collectors in the flotation offluorite, Coil. & Surf.~, 19-41.
50. Bahr, A., Clement, M., and Luther, H. (1968) Effect of some electrolytes on the flotation of
fluorspar with sodium oleate, Z. Erzbergbau Metallhuttenw. 21 (I), 1-8.
51. Gutierrez, C. (1976) Presence of a maximum in the adsorption isotherm of oleate on fluorite.
Influence of calcium and quebracho, in Proceedings 7th Int. Congress Surface Active
Substances (Moscow), pp. 638-647.
52. Texeira, O.e.G., Oliveira, J.F.D., and Boas, R.C.V. (1988) Surface chemistry andflotation
of scheelite, fluorite and apatite with anionic collectors, in S.H. Castro Flores and 1.A.
Moisan (eds.) Froth Flotation, Elsevier, Amsterdam, pp. 175-184.
53. Giesekke, E.W., and Harris, P.J. (1984) A study of the selective flotation of fluorite from
calcite by the use of a single bubble stream microflotation cell, in L.F. Haughton (ed.) Mintek
50: Int. Conf. Miner. Sci. Tech., Council for Minerals Technology, Randburg, vol. I, pp.
269-277.
54. Miller, J.D., Hu 1.S., and Jin, R. (1989) Thermochemistry of oleate adsorption at the
fluorite/water interface, ColI. & Surf. 42,71-84.
55. Dobias, B., Orthgiess, E., and Schieder, D. (1990) The determination of anionic surfactants
in dilute aqueous solutions, in Proceedings 8th 1m. Symp. Surfactants in Solution,
Gainesville, USA, June 10-15.
56. Peck, A.S., and Wadsworth, M.E. (1964) Infrared study of the depression effect offluoride,
sulphate, and chloride on chemisorption of oleate on fluorite and baryte, in Proceedings 7th
Int. Miner. Process. Congress, Gordon and Breach, New York, vol. I, pp. 259-267.
57. Lovell, V.M., and Finkelstein, N.P. (1969) The influence of preadsorbed oleate ions on the
adsorption of water vapour by insoluble ionic crystals, Nat. Inst. Metall., Johannesb., Rep.
504.
58. Lovell, V.M., Goold, L.A., and Finkelstein, N.P. (1974) Infrared studies of the adsorption
of oleate species on calcium fluoride, Int. J. Miner. Process. 1, 183-192.
59. Kellar, 1.1., Cross, W.M., and Miller, 1.D. (1990) In-situ internal spectroscopy for the study
of surfactant adsorption reactions using reactive internal reflection elements, Sep. Sci. &
Techn. vol. 25.
60. Iskra, 1., and Kielkowska, M. (1980) Application of ATR technique to fluorite-oleate-
356

quebracho systems, Trans. IMM 89, C87-C90.

61. Hu, 1.S., Misra, M., and Miller, 1.0. (1986) Characterisation of adsorbed oleate species at
the fluorite sUrface by FTIR spectroscopy, Int. 1. Miner. Process. lli, 73-84.
62. Sivamohan, R, de Donato, P., and Cases, 1.M. (1990) Adsorption of oleate species at the
fluorite-aqueous solution interface, Langmuir 2,637-643.
63. Pugh, R, and Stenius, P. (1985) Solution chemistry studies and flotation behaviour of
apatite, calcite andfluorite minerals with sodium oleate collector, Int. 1. Miner. Process . .1l,
193-218.
64. MoudgiJ, B.M., and Chanchani, R (1985) Beneficiation of complex phosphate ores from
south Florida by two stage conditioning process, in Proceedings XVth Int. Miner. Process.
Congress (Cannes), Societe de l'Industrie Minerale, vol. 3, pp. 357-366.
65. Moudgil, B.M., Vasudevan, T.V., and Blaakmeer, 1. (1987) Adsorption of oleate on apatite,
Miner. Metall. Process., February, 50-54.
66. Fuerstenau, M.C. (1982) Semi-soluble salt flotation, in R.P. King (ed.) Principles of
Flotation, S. Afr. Inst. Min. Metall., Johannesburg, pp. 199-213.
67 Johnston, 0.1., and Leja, L. (1978) Flotation behaviour of calcium phosphates and
carbonates in orthophosphate solution, Trans. IMM December, C237-C242.
68. Rybinski, W., and Schwuger, M.J. (1985) Kalorimetrische Untersuchungen zur adsorption
von tensiden an mineralen, Henkel Referate 21, 94-99.
69. Changgen, L., and Yongxin, L. (1983) Selective flotation of scheelite from calcium minerals
with sodium oleate as collector and phosphates as modifiers. II. The mechanism of the
interaction between phosphate modifiers and minerals, Int. 1. Miner. Process.lQ, 219-235.
70. Plitt, L.R., and Kim, M.K. (1974) Adsorption mechanism offatty acid collectors on baryte,
Trans. AIME 256, 188-193.
71. Hall, P.G., Lovell, V.M. ,and Finkelstein, N.P. (1970) Adsorption of water vapour on ionic
solids containing preadsorbed sodium oleate, Part I. Calcium fluoride; Part II. Calcium
carbonate, Trans. Faraday Soc. Q2 (574), B 1520-BI529; ibid., B2629-B2635.
72. Bitting, D., and Harwell, 1.H. (1987) Effects of counterions on surfactant surface aggregates
at the alumina/aqueous solution interface, Langmuir.3., 500-511.
3.3. FLOT ATION OF SALT-TYPE MINERALS

G.P. GALLIOS and K.A. MATIS

Laboratory of General and Inorganic Chemical Technology,
Department of Chemistry, Aristotle University ofThessaloniki
GR-540 06 Thessaloniki
Greece

Abstract. Salt-type minerals flotation is discussed, based mainly on experiments on magnesite,

dolomite and calcite. Among the parameters affecting flotation, the following are examined: crystal
structure, solubility of minerals, surface charge, collectors, organic and inorganic modifiers,
adsorption, contact angle and floatability. The selective separation by flotation of salt-type minerals
from each other and from silicates is also investigated.

1. Introduction

Salt-type minerals are important raw materials for the metallurgical, chemical, construction and
agricultural industries. Many millions of tormes of these minerals are consumed each year.
Examples are phosphates or apatites Ca2(P04)6(OH,Cl,Fh, baryteBaS04, fluorite CaF2, gypsum
CaS04.2H20, magnesite MgC03, calcite CaC03, dolomite CaC03.MgC03, scheelite CaW04,
etc. A large proportion of these minerals is produced by flotation processes. A typical flowsheet of
(fatty acid) flotation plant is given in Figure 1 [1].
Magnesite is an important economic nonmetallic mineral, since it is the main source of
magnesium oxide, which is widely used as a refractory material. It is one of the important products
of Greek mineral wealth. Proven, indicated and inferred ore reserves in Greece amount to 50, 65.5
and 80 million tormes respectively [2].
Magnesite ore deposits usually contain a variety of gangue minerals, mostly other carbonates
(calcite, dolomite etc.), silicates and oxides. From the various beneficiation techniques, that have
been tested for the separation of magnesite from gangue minerals, flotation is the most important,
because it is applicable to most of the ores and can give better concentrates. However, the flotation
of magnesite has not been studied extensively.
Cationic flotation has been used to separate magnesite from silicate gangue [3-7] . The silicates
are floated away and the depressed fraction becomes the magnesite concentrate (reverse flotation).
Primary, secondary and tertiary amines, quaternary ammonium salts and admixtures of them are
used as collectors. This technique is currently used in a beneficiation plant in Mantoudi, Euboea,
Greece [6]. The silicates are removed by flotation at a neutral pH, with a mixture of primary and
quaternary amines, fuel oil and pine oil; the flotation product is actually the tails. The pulp is first
passed from a thickener, a disc filter and a rotary dryer. Afterwards it is calcined in a multiple

351
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 351-382.
© 1992 Kluwer Academic Publishers.
358

r RUN OF KINB ORB

I
I CRUSHING

r
GRINDING
I
r
CLASSIFICATION
I
I FINES
I
I SCRUBBING
I
CLASSIFICATION
I
I I
pH REGULATORS
KODIFIERS 11
CONDITIONING II COLLECTOR ADDITION

------------------------------- --------------------_._--------

r
-.
1
ROUGHER FLOTATION

I UNDERFLOW
I I FROTH PRODUCT I
r FROTH PRODUCT I II UNDERFLOW
I
~ SCAVANGER FLOTATION I ~ CLEANER FLOTATION

------------------ --------------------------- ---------------

r TAILS
I I CONCENTRATE
I
Figure 1. Flowsheet of a typical salt-type mineral processing plant.

hearth furnace, transported to a press and the briquettes are burned in a shaft kiln. Finally, dead-
burned magnesite is produced which is used in the refractory industry.
When dolomite and/or calcite is a major gangue component, anionic flotation by fatty acids as
collectors is used [8-11]. The application of modifying agents is necessary in order to achieve
selectivity. Usually dolomite is depressed and magnesite is recovered in the froth. When both
silicates and carbonates occur in the ore, a cationic flotation of the silicates by amines, followed by
an anionic flotation of the depressed fraction by fatty acids has been proposed [12,13]. The most
systematic work on the flotation of magnesite was carried out by Predali [14-18] on Czechoslova-
 359

kian and Yugoslavian samples and by Gallios [19] on Greek samples.

Calcite and dolomite often occur as gangue minerals in nonmetallic ores. They are usually
components of apatites and their flotation behaviour has also been studied quite well. In a recent
review, Finkelstein has critically analysed the literature on all aspects of flotation of apatite, calcite
and fluorite [20].

2. Experimental

In this work, experiments were conducted with magnesite and dolomite mainly from the Gerakini
(Chalkidiki) and Neraida (Kozani) areas, both in northern Greece. Their chemical analysis showed
a purity over 99% for both minerals. After crushing and grinding, the -180 +125 I.I.IIl fraction was
taken by wet-sieving.
The greek magnesite, as inspected under the microscope (x60), is mainly crypto-crystalline, with
a lot of cavities on the surface and a much higher specific surface area, compared to dolomite; the
latter was crystalline and free of cavities. Magnification revealed numerous ultrafines (with size less
than 10 1.I.IIl) on larger magnesite particles, which were easily removable from the surface.
Sodium oleate was used as collector. On certain cases, a commercial tall oil fatty acid was used
(Acintol FA-I, kindly supplied by Arizona Chemical Co.). According to the company, its chemical
analysis was 34% non-conjugated linoleic acid and 44% oleic acid.
Various modifiers were applied: sodium pyrophosphate (Nll4P207.lOH20), sodium hexameta-
phosphate (Na6P6018), sodium silicate (Na20.xSi02, with x=3-5), sodium fluorosilicate
(Na2SiF6), sodium salts of carboxymethyl cellulose and Calcon (C20H 11 (OHhN2S03Na, known
also as Eriochrome Blue Black R).

• Co
00
a

0= b = 6.42 A
C=17.06 .
Figure 2. Crystal structure of calcite.
360

The floatability experiments were carried out in a modified Hallimond tube; adsorption studies
followed a common practice [21]. while the S-potential was measured by the method of
electrophoretic mass transfer [22].

3. Results and Discussion

3.1. CRYSTALSTRUCfURE

Magnesite and dolomite belong to the calcite minerals group. They crystallise in the rhombohedral
system. The crystal structure of calcite (Figure 2) is considered to be derived from that of NaCl, in
which the sites of Na+ are taken by Ca2+ and these of the Cl- ions are taken by planar C032-
groups. The planes of all the C032- are perpendicular to the 3-axis system. Each cation is coordina-
ted in the structure to six oxygen atoms of six different C032- groups. The bonds between the
cations and the anion radical are ionic, while the bonds within each complex C032- are covalent.
In magnesite, all the Ca2+ cations are replaced by Mg2+ cations, while in dolomite only about
half of them are replaced. Because of the large difference in the ionic radius of Ca 2+ and Mg2+
(about 25 %), dolomite has an ordered structure and lower symmetry than magnesite. Substitutions
of Ca2+ and Mg2+ by other cations may also occur.
Moving further in the salt-type minerals family and mainly in the apatite series, a quite complex
nature of their chemical and structural constitution is observed. In Table 1 the names and types of
common apatites are shown.

TABLE I Chemical types of certain apatites 123].

Name of mineral Type
Hydroxyapatite Cas(P04)30H
Fluorapatite Ca5(P04)3F
Francolite CalO(C,P04)6F2
Dahlite CalO(C,P04)6(OH)z
Dehmite (Ca,Na,K)lO(P04)6(OH)z
Lcwistonite (Ca,K,Na)}o(P04)6(OH)z
Wilkeite Ca}o(S,Si,C,P04)6(OH)z
Ellestadite Ca}o(Si,S,P04MCI,F,OH)z

Since any fracture (due to grinding, for example) breaks bonds that have at least a partial ionic
character, the minerals have strongly hydrophillic surfaces and collectors must be used in order to
float them.

3.2. SOLUBILITY AND SURFACE CHARGE

Solubility properties of minerals are of importance in flotation, because of their role in determining
both the chemical composition of the aqueous phase and the charge characteristics of the interface.
Salt-type minerals are characterised by solubilities that are lower than those of simple salt minerals
(such as halite and sylvite), but higher than those of most oxides and silicate minerals. In general,
the solubility can be expressed as follows:
 361

(1)

(where M stands for certain cation, as Mg2+, Ca2+, Ba2+ and A for an anion, like F, C032-, S04 2-,
W042-, P043-)_
The distribution of various species as a function of pH for dolomite, in a system which is open
to atmosphere, have been calculated using the equilibrium constants as shown in Table 2 and
presented in Figure 3. Here, the mass and charge balances have been taken into account and the
stoichiometric release of calcium and magnesium ions during dissolution has been assumed. The
total solubility of dolomite is given by the sum of the concentrations of all the Mg and Ca
complexes in the liquid phase. However, only two Mg and two Ca species are critical to the
solubility: for pH values below 9.2, only Ca2+ and Mg2+ concentrations are significant; above this
pH, (MgC03hq and (CaC03)aq are the solubility-controlling species_

TABLE 2. Hydrolysis equations ofmagnesite* (the first 11) and dolomite* (aU).
MgC03 (s) <=Z MgC03 (aq) (1)
MgC03 (aq) <2 Mg2+ + C03 2- (2)
C03 2-+ H2O <2 HC03- + OH- (3)
HC03- + H2O <2 H2C03+ 0H - (4)
H2C03 <2 C02 (g) + H2O (5)
Mg2++ HCOf <2 MgHC03 (6)
MgHC03+ <2 H+ + MgC03 (aq) (7)
Mg2++OH- <2 MgOH+ (8)
MgOH++ OH- <2 Mg(OHh (s) (9)
MgOH++OH- <2 Mg(OHh (aq) (10)
Mg(OHh (aq) <2 Mg(OHh (s) (11)
CaC03.MgC03 (s) <2 Ca2+ + Mg2+ + 2C032- (12)
Ca2++ HC03- <2 CaHC03+ (13)
CaHC03+ <2 W + CaC03 (aq) (14)
Ca2+ + OH- <2 CaOH+ (15)
CaOH++OR <2 Ca(OHh (aq) (16)
Ca(OHh (aq) <2 Ca(OH)z (s) (l7)
* pKI = -4.51
The solubilities of dolomite and magnesite as calculated from thermodynamic data, and shown in
Figure 3, are high. However, the minerals do not dissolve very fast [14]. The effect of
conditioning time on the final pH of the solution (containing magnesite and dolomite) has been
investigated [25]. It was observed that the rate of change of the pH was greater for dolomite than
that for magnesite. The presence of sodium oleate changed the final pH to 9.5, from approximately
9.0, and this was attributed to the capture of Ca2+ in solution by oleate ions.
Because of the dissolution of the minerals, the fracture of the bonds and the adsorption of
different species from the solution, the surface of the minerals is charged. A study of surface
charge, usually expressed as ~-potential, can give useful information for the mechanism of collector
adsorption on minerals and possibly the way the different reagents, which are used as modifiers to
obtain selectivity, act in flotation_ However, care is needed in comparing values obtained from
362

literature: for instance. Brandao [26] reported that the zero point of charge of magnesite range from
2.0 up to even 11.5. depending on the method of measurement and origin of the samples.

Or-~~~~.-----,.--------~-------.

-2
-4
""""
;--..

6 -6~~====~~~~~~======~
u -8
c
0-10
U
'c];-12
o
--l
-14
Co(OHh..
-16~----~-----------------------------1

-18~~-U--~-L--~~~--~-L~~~~--~~
o 2 3 4 5 6 7 8 9 10 11 12 13 14
pH

Figure 3. Equilibrium concentrations of different species of dolomite in aqueous solution.

-160
r-.
>
E -140
'--'

"0 -120
:;:;
....,~ -100
0
n.. -80
....,0 -60
(I)
N
-40

-20

05 6 7
pH

Figure 4. Zeta potential measurement with (lines 1 & 2) and without (lines 3 & 4) carboxymethyl-
cellulose. Reprinted with permission from ref. [28]. Copyright 1988. Elsevier Science Publishers.
 363

The point of zero charge of magnesite and dolomite was set at a pH of 6.0 - 6.5 and both
minerals are negatively charged in the pH range investigated of 7.0 to 12.0 [27]. It is noted that 1;-
potential measurements at lower pH values were not possible, because of the minerals solubility
which caused a fast shift of pH.
The effect of modifying reagents in the flotation of magnesium carbonates, using a commercial
collector, was examined [28]; I;-potential measurements were also taken. Carboxymethyl cellulose
(and sodium hexametaphosphate) increased the negative charge of dolomite, but did not really
change that of magnesite (Figure 4). It is probable that the action of these modifiers is due to
adsorption on sites where calcium exists; as the surface of the minerals is negatively charged in the
pH range studied, the carboxymethyl cellulose may be adsorbed specifically on dolomite.

3.3. COLLECfORS AND ADSORPTION

Two types of collectors are generally used : (i) anionic, like fatty acids, alkylphosphates,
alkylsulphates and alkylsulphonates, leading to a classical flotation of salt-type minerals, and
(ii) cationic collectors, like the primary, secondary, or tertiary amine and quaternary ammonium
salts, leading to an inverse flotation.
Collectors hydrolyse in aqueous solutions. An example is given in Figure 5 for sodium oleate;
oleic acid is a weakly acidic insoluble compound, which forms very soluble salts with monovalent
alkali metal ions. The species that have been considered in the hydrolysis reactions are the oleate
ion RCOO-, oleic acid RCOOH, acid-soap (RCOOhH- and oleate-dimer (RCOOh2-.

-2
-3
RCOO-
-4
........
........ -5
~
(RCOO)~-
'-"
-6
()
c -7
0
() -8
'-"
0\
0 -9
...J
-10
-11
-122~~3---4~~5~~6---7~~8~~9---1~0--~1-1--1~2~
pH

Figure 5. Activities of various oleate species as a function of pH of solution, for a total oleate
concentration of 30 ppm.
364

The development of industrial flotation reagents over the past seventy or so years has been very
much an empirical process. Few flotation reagents, even today, can match the standards of purity
required in other sectors of chemical industry. It was ascertained [29] that impurities have
frequently proved beneficial and, when attempts have been made to prepare purer flotation chemi-
cals, it was found that such reagents did not perform as well as their counterparts.
A typical example of this are fatty acid collectors. It has been found that refined fatty acids alone
are inferior collectors to neutral tall oils and this is due to the important role apparently played by
the resin acids. Oleic acid, which is often applied due to its double bond reactivity, is known
however to oxidise. A possible chemical mechanism of oleic acid oxidation on mineral particle
surfaces was reported [30] (Figure 6). However, IR studies of the adsorption of oleic acid on
different salt-type minerals have not shown any oxidation products and it was proposed that the
double bond does not participate in the adsorption mechanism. If this is the case, it is very difficult
to explain the higher collector effectiveness of oleic acid in comparison to other unsaturated and
saturated fatty acids.

CH3(CH2)7-CH=CH-(CH2)7COOH
O2 1 Autooxidation

I ~~"
CH3(CH2)6CH2-CH-CH-(CH~)7COOH

+ H"
CH 3 (CH 2 ) 6CH2-CH-CH-(CH2)7COOH

j h-OH

- H"
CH 3 (CH 2 ) 6-CH=CH-CH-(CH2) 7COOH

1 h-OH

j
CH3(CH2)6-CH=~7H-(CH2)7COOH

0" +""

I'd
CH3(CH2)6-CH-CH-CH-(CH2)7COOH

+."

CH 3 (CH2)7CH-C\-iH-(CH2) 7 COOH
o

Figure 6. Mechanism of oleic acid oxidation.

The solution pH did not seem to have any substantial influence on the adsorption of collector on
magnesite, for various collector [21] (Figure 7); similar results were obtained with dolomite.
However, pH affected significantly the flotation of magnesite.
The adsorption results, taking into account specific surfaces for magneSite and dolomite (1.93
 365

and 0.0442 m 2g. 1 respectively). for varying amounts of pyrophosphate (added as modifier) are
shown in Figure 8. Collector adsorption on dolomite increased when the pyrophosphate
concentration decreased. while it did not affect adsorption on magnesite. This may be attributed to
the adsorption of pyrophosphate on dolomite specifically on calcium sites.

60

"-"'50 50
E e
a.
a.
-0 0 0 0 0 0 0 ....
""""'40 40
-
&I 1:1 iI B 1:1 e 0 &I
0 &I .....

----
830 30
• • • • * *
"~ 20 '" '" •
20
~
0
a a D 6 6
• a
0
III
10
«
" 10 X
. R X X
X

"
U A at

06
7 8 9 10 11 12 13
pH

Figure 7. Effect of different concentrations of adsorbed sodium oleate on magnesite at variable pH values.
Reprinted with permission from ref. [21J. Copyright 1989. Gordon and Breach Science Publishers Inc.

400 2800

"'"
"'"
N
N

.,...
1\1 2600 '~
'~ 300 0>
0> :l...
;J... "-'
"-'
2400
0
0200 U
(,)
2200 '0
v
'0
v .0
'-
.0 0
0
Vl
100 D
2000 '0
Vl

'0 c(
c(

00 1800
200 400 600 800 1000 1200
Initiol Tspp Cone. (ppm)

Figure 8. Adsorption experiments of 50 ppm sodium oleate on the carbonates in the presence of sodium
pyrophosphate at pH 9.5 .
366

The very large difference in specific surface area between the two minerals. despite nominally
similar size ranges. is attributed to the numerous ultrafine particles (with size less than 10 11m)
which exist on the larger magnesite particles (see photograph).

Photograph. Magnesite particles observed through a microscope.

These ultrafine particles detached during conditioning and affected flotation. Fine particles are
known to affect flotation in different ways; the processing of fine carbonate minerals by novel
flotation techniques has been the subject of recent work by the authors [28.31].
It is known [32] from IR studies that the adsorption of oleate on salt -type minerals is of chemical
nature (chemisorption) in the neutral and basic region of pH.
In a study of oleate on calcite. fluorite and barite [33]. it was found that the removal of oleate
from solution could be attributed to the precipitation of calcium or barium oleate on the surface.
Somasundaran et at. [34] proposed the surface precipitation of surfactants and inorganics on
mineral solids. too. Its role in adsorption and flotation was examined and its difference from the
surface reactions was discussed. The adsorption of fatty acids on magnesium carbonates has been
studied extensively by Predali [14] and Brandao [35] and was found to be of chemical nature in the
neutral and basic region of pH.

3.4. FLOATABILITY - CONTACf ANGLE

The effect of pH on the floatability of magnesium carbonates was investigated [27]. Acomparison
for two extreme collector concentrations is presented in Figure 9. Even the reagents used for pH
adjustment may influence the flotation of minerals. An example of this was the flotation of calcite.
 367

100 '-'-o.-+.~ lJ. ~'~lJ.,-O-D

~.-.A.'-n-'~ -4- ' -.
.J",. ./ -0' r M
/ O".~
80

........
~ M
60
~
Q)
>
0 40 D
U
Q)
0:::

20

04
5 6 7 8 9 10 11 12 13
pH

Figure 9. Effect of pH on flotation of minerals by 10 ppm sodium oleate (solid lines) and 40 ppm
(dashed lines).

fluorite and apatite by oleate where the reagent used for pH adjustment (sodium hydroxide or
carbonate) changed flotation of the minerals [36].
Service water contains significant quantities of calcium and magnesium ions, which are
detrimental for fatty acid flotation, because they produce insoluble salts that decrease the available
collector. This was studied by calcium sulphate addition and some of the results are shown in
Figure 10 [25] . The chemical reaction of oleate with calcium is quantitative and for 40 ppm of
Ca2+ approximately 600 ppm of sodium oleate are needed. However, the flotation rate of magnesite
was approximately 40 % with 250-500 ppm oleate. This shows that the precipitation of oleate by
calcium ions does not mean that it will be ineffective as collector, although it may well mean that its
efficiency will be lowered. The presence of colloidal calcium on fluorite surface was reponed by
Gaudin [37].
Contact angle measurements usually serve as a measure of the hydrophobicity of mineral
surfaces, although this - as far as flotation is concerned - has been disputed in the past [38]. Figure
11 illustrates magnesite contact angle measurements and recovery versus pH, using sodium oleate
(30 ppm) as collector; at higher concentrations (100 ppm), the obtained curve was entirely
different, reflecting probably the formation of a polymolecular adsorbate structure.
Forssberg et al. [39], among many others, examined the pulp chemistry in calcite anionic
flotation and modelled oleate adsorption using theoretical equilibrium calculations. The aim of that
study was the cases where calcite occurs as an impurity, such as apatite and scheelite flotation.
The use of the modified Hallimond tube for floatability tests was investigated several years ago,
doing experiments in two flotation cells; one was a Leeds Autoflotation cell, and the other one a
modified Hallimond tube; the anionic flotation of pure calcite was chosen for the tests and results
are illustrated in Figure 12 [40]. Distilled water was used in the Hallimond micro-cell. The
-180 +125 IJ.I1l particle size range was selected for the Hallimond tube, while -300 IJ.I1l particles
were used in the larger cell. Although different collector concentrations were needed in order to
368

100 MAGNESITE

80
,--...
~
'-"
60
>.
I...
Q)
>
0 40
u
Q)
a:::
20

100 200 300 400 500 600 700 800

Collector Concentration (ppm)

Figure 10. Use of hard water in magnesite flotation; effect of collector concentration at pH 10.0: 1 =
40 ppm Ca2+; 2 = 200 ppm Ca2 +. Reprinted with permission from ref. [25]. Copyright 1989, Elsevier
Science Publishers.

60 100

,....., 50
~

40 tR
~
0'1 60 i::'
C Q)
0« 30 >
..... 0
u 40 u
Q)
.....
0
20 0:::
c
0
U 20
10

05
6 7 8 9 10 11 12
a
pH
Figure 1I. Correlation of contact angle and flotation recovery of magnesite against pH at sodium oleate
concentration of 30 ppm. Reprinted with permission from ref. [27]. Copyright 1989, The Institution of
Mining and Metallurgy.

obtain the same recovery in the two cells, the results were similar, indicating that the Hallimond
tube can be effectively used to study the flotation behaviour of minerals and then scale up the
results.
 369

0 200 400 600 800 1000

100

80
,.-...
~
'-'
60
>.
....
(I)
>
0 40
U
(I)
0:: Cells
20
* Hallimond
... Leeds
00 100 200 300 400 500
Collector Concentration (g/ton)

Figure 12. Influence of commercial collector concentration on calcite flotation in two bench-scale cells
(top the laboratory, bottom the micro-cell).

100

80

-----
~
'-'
60
....(I)>. M
>
0 40
U
(I)
0::

20 c
D
C

0
4 5 6 7 8 9 10 11 12 13
pH

Figure 13. Effect of hexametaphosphate (concentration 20 ppm) on carbonates flotation by 20 ppm

sodium oleate.
370

3.5. MODIFIERS (INORGANIC AND ORGANIC)

Modifying agents are added in flotation circuits to achieve separation, possibly selective, of salt-
type minerals. Sodium hexametaphosphate belongs to the group of cyclic polyphosphates [41] and
has a very high negative charge (more than the pyrophosphate). It was found that this reagent
depressed dolomite anionic flotation in the alkaline region of pH (Figure 13); it increased also
slightly magnesite flotation in the pH range 8 to 10, where a minimum existed otherwise.
The upgrading of magnesite deposits is often achieved through cationic silica flotation; however,
when the ore contains substantial amounts of dolomite, the bulk of it remains with magnesite. It is
noted that the ratio CaO/Si02 affect significantly the quality of the product; the best refractory
properties are obtained with a ratio of 2:1.
Sodium hexametaphosphate increases the negative charge of dolomite but does not change the
zeta potential of magnesite [19]. This may be attributed to the adsorption of this modifier in calcium
sites on the surface of dolomite [23]. However, it is also possible that its action is related to the
solution rather than the solids, because it is known that phosphate and polyphosphate ions can
bond calcium from solutions, giving complexes or insoluble salts; they can even redissolve
insoluble salts of calcium. It was reported that sodium orthophosphate, pyrophosphate and hexa-
metaphosphate did not adsorb on fluorite, scheelitc and calcite, while the calcium concentration in
solution increased; hence, calcium was selectively dissolved from mineral particles surface [42].
Anyway, the calcium sites available for the collector decrea~e and dolomite is depressed.

o
AMORPHOUS SILICA
-1
-----
~-2
~-3~------------------~~--~------------~
Si(OH)4
g -4
o
U-5
~

8' -6
.-J
-7
-8
-95~--6~~7~~~8~~9---1~O---1~1---1~2~~1~3--~14

pH

Figure 14. Activities of different species of silicates as a function of the pH of the solution.

Sodium silicate is one of the most common modifiers, generally acting as a depression agent,
e.g. of silicates in the flotation of scheelite [43]. Its hydrolysis reactions may be represented as in
Figure 14 [44]; in large concentrations, sodium silicate is polymerised [33].
 371

Sodium silicate caused a slight activation of magnesite, as seen in the pH-recovery curve (Figure
15). In the case of dolomite, sodium silicate was a strong depressant, for pH values under 9.0
approximately.

100

80
~

~
'-"
60
>.
L
v
>
0 40
u
v
a:::
20

0
4 5 6 7 8 9 10 11 12 13
pH

Figure 15. Effect of sodium silicate (concentration 120 ppm) on carbonates flotation by 20 ppm sodium
oleate.

100 D

80
~

~
'-"
60
>.
L
v
>
0 40
u
v
a:::
20

0
4 5 6 7 8 9 10 11 12 13
pH

Figure 16. Effect of sodium fluorosilicate (concentration 100 ppm) on flotation of carbonates by 40 ppm
sodium oleate.
372

Sodium silicate adsorbs on calcite, dolomite and apatite, as found from l;,-potential measurements
[45]. It is, however, possible that in salt-type minerals, a salt like calcium silicate may be formed
either straight on the mineral surface by chemisorption or in solution and adsorbed on the
surface [46].
Sodium fluorosilicate has also been applied as modifier, acting as a depressant [25], although its
flotation behaviour is basically different from that of sodium silicate; the presence of fluorides
catalyses its polymerisation to colloidal silicate [44]. The l;,-potential of magnesite and dolomite
was shifted to more negative values in the presence of sodium fluorosilicate [47].
A further comparison of the two magnesium carbonates is shown in Figure 16. In the pH region
of about 9 to II, the recovery of magnesite using sodium fluorosilicate increased by 35%.
The above modifiers are more or less common. Another reagent, Calcon, which is used in
analytical chemistry as an indicator in the volumetric determination of calcium by EDT A, was tested
(Figure 17), due to its specific reaction with calcium [25]. It did not affect the flotation of
magnesite (no calcium in the structure) at pH 7.5 to 11.5; however, dolomite was strongly
depressed at pH 8.0 to 11.5, with a greater action at a pH of about 10.0, where recovery dropped
from 95% to 25% approximately.

100
D

80
M
,--...
~
'-"
60
>.
L-
v
>
0
u 40
v
a::
20

0
4 5 6 7 8 9 10 11 12 13
pH

Figure 17. Effect of Calcon (concentration 30 ppm) in the flotation of magnesium carbonates by 30 ppm
oleate.

3.6. SELECTIVE SEPARATION

The selective separation of salt-type minerals both from silicates and from other salt-type minerals
is of interest; the former is much easier.
A two-stage process was studied by Baranovskii [8] in the laboratory, but for magnesite ores
with low CaO content and later scaled up, with flotation of silicates using kerosene and oil in the
first stage and flotation of magnesite in the second one. This was applied also elsewhere [12], with
the addition of amines in the preliminary flotation of gangue (talc, etc.). Amines were also applied
 373

in separation problems with sepiolite. a hydrophobic mineral [4].

Figure 18 presents some laboratory results for a magnesite/quartz mixture (size range -250 +100
!J.Ill) in a 250 g Denver flotation cell [48]. High recoveries (over 90%) in acidic pH (4-6) were
noticed and also a quite good grade of the concentrate (less than 2% quartz content).

100 10.0
9.0
~
80 8.0
>.
I-
CI) 7,0 ~
>
0 .....
u 60 6,0 c
CI)
~
0:: 5,0 c
0
...,CI) 40 4,0 U
'iii
CI)
C
3,0 .....
N
I-
0'1 0
0 20 2.0 :::J
~ a
1.0
03 0,0
4 5 6 7 8 9 10 11 12 13
pH

Figure 18. Flotation by sodium oleate of a binary mixture magnesite/quartz.

100 1 -e-'-'-'l-._
ll'"-'",
\
80
\

!R
.........
"ii
60 i
:>.
I-
\
Q) 0\
> \
0
() 40 P
Q)
0::
2
20

04
5 6 7 8 9 10 11 12 13
pH

Figure 19. Use of sodium silicate in the flotation of dolomite by 20 ppm of collector: I=without
modifier; 2=120 ppm silicate. Reprinted with permission from ref. [25]. Copyright 1989. Elsevier Science
Publishers.
374

Figure 19 shows the application of sodium silicate in dolomite flotation [25); the modifier was
indeed a strong depressant (at pH less than 9.0). This could be attributed to silicic acid, which
dominates in solutions in the pH range 4 - 10. Results for binary mixtures of magnesite/dolomite

100 100

~
~ 90 80
....>.
v v
-0
2 >
....
0 0
<.:> 80 60 (,)
v
a:::
2
.iii 2
v 70 40
VI
C
0'1
v
C
0 0'1
::::!: 60 20 0
::::!:

5040 0
50 60 70 80 90 100
Initial Magnesite Content Sl5

Figure 20. Influence of initial content in magnesite of the binary mixture on the grade in magnesite (I)
and recovery of magnesite in the concentrate (2); sodium oleate 20 ppm, sodium silicate 240 ppm and pH
11.0. Reprinted with permission from ref. [25). Copyright 1989, Elsevier Science Publishers.

100 ~_a-1
~'-'~'-'-C-.'--o-.g
..... . ~.

;
.~ \
\
D\ ;
80 Dj
\ ;
,,--.. i i
I!R
'--" i \ i 2
>.
60 i \ i
i
I...
Q)
> i
\
0\ i.
0 .0.
\0 I
U
40
Q) \ A

.
n:::
0\
20 ,.~.
\.
-.,._.-0-'- -o_._·r 4

0
4 5 6 7 8 9 10 11 12 13
pH

Figure 21. Effect of carboxymethylcellulose on carbonates flotation by 20 ppm sodium oleate: (I & 2)
without modifier, (3 & 4) 100 ppm cellulose. Reprinted with permission from ref. [49). Copyright 1989,
Marcel Dekker Inc.
 375

are presented in Figure 20. An interaction of the minerals is observed, if the flotation results are
compared with values calculated theoretically from experiments with a single mineral. A similar
interaction was also reported for the calcite/apatite system [24] .
Further results from single-mineral flotation experiments with sodium carboxymethyl cellulose
are given in Figure 21 [49]. The modifier acted as a strong depressant for dolomite in the pH range
from 7.0 to 12.0. With the same conditions, only a slight activation action was observed for
magnesite.
The adsorption study showed a significant decrease of oleate adsorption on dolomite. This
organic modifier is well known [50] to be ionised completely and possesses a high negative
charge. From l;-potential measurements it was found that carboxymethyl cellulose increases the
negative charge of dolomite in the basic region of pH. So, it probably acts antagonistically to oleate
for calcium sites on dolomite surface.

100 -o-t:Jt:"~

80
. \. .\
.~

..-..
I!R 2
'-'
50
~
CI)
>
0
U
40 ...........
CI) 0....
0::: D""'o- 4

20

°4~--~5--~5--~7~~8----9~--1~O---1~1---1~2--~13~
pH

Figure 22. Effect of pyrophosphate on carbonates flotation by 30 ppm sodium oleate: (l & 2) without
modifier, (3 & 4) 900 ppm pyrophosphate. Reprinted with permission from ref. [49]. Copyright 1989,
Marcel Dekker Inc.

Pyrophosphate addition was also studicd (Figure 22). The depressive action was more
noticeable with dolomite than with magnesite at pH greater than 9.0 [49]. Pyrophosphate, in
contrast to hexametaphosphate, belongs to the group of polyphosphates having a linear chain [51].
It generally acts as s strong dispersing agent in pulps of fine mineral particles. It is also a
polyelectrolyte with a high negative charge and forms characteristically chelates with metal cations.
Figure 23 shows some further work on simple binary mixtures, without any modifier addition
other than sodium hydroxide (for pH adjustment) [27]. A quite promiSing separation of magnesite
from dolomite may be achieved by reverse flotation at pH 11.0, using only small quantities of
sodium oleate as a collector, with magnesite remaining in the non-floated fraction.
The responses of the two carbonate minerals studied at the various separation conditions were
dominated by a rapid decrease in the solubility of the calcium carbonate component of dolomite
376

TABLE 3 Selecuve separation of salt-type minerals

Mineral flotation details - admixtures Ref.
Baryte Reverse or direct flotation - 36 studies at CANMET. fluorite and Celestite 52
(SrS04).
Fatty acid adsorption, precipitation of metal soap. fluorite and calcite. 46
Sodium silicate (22 ppm) to depress calcite and fluorite. 53
Quebracho as a depressant for calcite and citric acid for fluorite (review) 54
Fatty acids, alkaline pH - Existence of a calcium carbonate layer on fluorite
surface (IR studies) 55
Depression of barite by chromate. fluorite and celestite. 56
Fatty acid flotation of barite from silicates ore - Role of sodium silicate. 57
Alkyl-sulphates or alkylsulphonates, pH 9.5-11. Quartz, fluorite and
carbonates. 23
fluorite Oleate flotation, adsorption - Calcite (review). 58
fluorite/oleate system, surface carbonation, IR (adsorption), temperature
effect, oxygen effect, activation, concentration effect, double-bond reactivity. 59
Influence of quebracho and calcium ion, oleate - Calcite. 60
N-alcylamino caIboxylic acids as collectors - Calcite. 61
Lattice ions in solution - Apatite, calcite 62
Solution chemistry, sodium oleate, flotation behaviour, microelectrophoresis
- Apatite, calcite. 63
Phosphates Fatty acid, two-stage conditioning process, FfIR studies, surface 64
precipitation - Dolomite.
Cationic collectors, non-polar hydrocarbons (Part 1), effect of particle size,
abrasion and pH (Part 2) - Dolomite. 65
Alizarin Red S as a modifier, oleate, acid wash stage, adsorption controlled
by OH· groups - Calcite. 66
Oleate, isoelectric point, species distribution, pH, alteration of surface
mineralogies - Calcite. 24
Dissolved mineral species, flotation of each mineral in various supernatants
prepared by mixing with water, addition of Ca(N03)z, KzC03 and K3P04 -
~~ ~
Sodium silicate, gum arabic and sodium tripolyphosphate, zeta potential, ions
released - Calcite, dolomite. 45
Depression of phosphate minerals with HzSiF6 (anionic flotation of
carbonates) and cationic flotation of silicates. Characterization of complex,
low-grade phosphate rock. 68
Cationic, surfactants, adsorption - Siliceous, calcareous and siliceous-
calcareous phosphates. 69
Sedimentary ores, electrochemical phenomena, properties of the calcareous
guangue, reverse flotation of phosphate oolites, collector type (4 parts) -
Calcite. 70
Depressants (dipotassium hydrogen-phosphate, etc.), stagewise flotation of
low-grade ore, liberation size, adsorption - Pyrite, calcite and other
carbonaceous impurities. 71
 377

TABLE 3 (cont)
Phosphates Conditions, flowsheets, thennal treatment, slimes, acid medium, magnetic
(cont.) separation, solubilization, flotoflocculation. Carbonates, quartz-glancolite (a
silicate), phosphatized sands. 72
Scheelite Slimes, around 50% recovery, fatty acids, sodium silicate, five types of ores,
Uludag plant - Pyrite, magnetite. 73
Oleic acid, sodium sulphosuccinate modifiers, electrophoretic mobility,
adsorption - Apatite, fluorite. 74
Water-glass, temperature (333 K), fatty acid (petrov's method) - Calcite. 75
Fatty acids, depression of silicates by sodium silicate, hot condiotioning, IR
study - Hornblende, quartz, feldspar, etc. 43
Phosphates as modifiers, selective flotation of sheelite by oleate from pure
minerals (part I), mechanism of interaction between modifiers and minerals
(part 2) - Auorite, calcite, garnet and quarts. 42

with rising pH, and a much stronger affinity between oleate ions and calcium ions than between
oleate ions and magnesium ions. This affinity is based on the relative solubility of the respective
hydroxides.

20 100

~
16 80 >.
'-
v
'"'
~
>
0
'-'12 60 u
v
~ a::
E
0 8 40 ~
"0 VI
v
0 c
(J"I
4 20 0
:::i:

00 0
5 10 15 20 25 30 35 40
Collector Concentration (ppm)

Figure 23. Influence of collector concentration on the grade of dolomite and recovery in the tails of
magnesite with binary mixtures of 20 wt% initial dolomite; pH 11.0.

Finally, a review of the various salt-type minerals studies, particularly from the point of view of
selective separation, are presented in Table 3.

4. Concluding Remarks

In this group of industrial ores, solubility properties are of particular importance. The separation of
378

minerals, such as apatite and scheelite, by flotation from other salt-type minerals (such as calcite) is
extremely complex, Owing to the similarity between their physicochemical properties. Often it is
not achieved, particularly if the ores are of microcrystalline type.
Flotation is a very flexible and effective beneficiation process. Even minerals such as magnesite
and dolomite with very similar mineralogical and physicochemical properties may be separated by
flotation. From our experiments is was found that :
- in single mineral tests, magnesite and dolomite showed some differences in floatability;
- the various modifiers applied (sodium hexametaphosphate, sodium pyrophosphate, sodium
silicate, sodium fluorosilicate, Calcon and carboxymethylcellulose) had different action on the
two minerals, which was related to the presence of calcium on dolomite surface;
- it is possible to obtain a good separation of the two minerals in artificial mixtures;
- there is an interaction of the two minerals in the mixed feed, which results in better selectivity
when small quantities of sodium oleate (collector) are used;
- when modifiers were applied, the interaction between the two minerals caused a poor grade of
the concentrate; and
- finally, magnesite and dolomite act antagonistically towards the collector.

S. References

1. Baarson, R.E., Ray, C.L., and Trcweek, H.B. (1962) Plantpracticeinnonmetailicmineral

flotation, in D. Fuerstenau (cd.) Froth Flotation - 50th Anniversary Volume, SME/AlME, N.
York, pp. 427-453.
2. Spiliadis, Th.S., and Fragiskos, A.Z. (1979) Magnesite or whitestone, Conf. Metal. Corp.
(Athens), pp. 318-324.
3. Clemmer, J.B., Doerner, H.A., and DeVaney, F.D. (1940) Experimental flotation of
Washington magnesite ores, AIMME, Technical Publication No. 1148.
4. Emrullahoglou, EO. (1984) Ph.D. Thesis, Anadolu Univ., Eskisehir (Turkey).
5. Das, T.K. (1977) Flotated magnesite - Production and use in refractory industry, Interceram
1,236-238.
6. Gambopoulos, Th., and Nestoridis, A. (1976) Separation of Magnesite from its Contami-
nants by Reverse Flotation, US Pat. 3,976,251.
7. Karantzavelos, G.E. (1984) Brief description and evaluation of the magnesite flotation
processes - Development of a new flotation process, Min. Metal. Proc., May, 69-70.
8. Baranovskii, N.I. (1967) Flotation beneficiation of Semibratsk magnesite, Ogneupory
(Refractories) 12, 17-19 (Trans!.).
9. Bron, V.A., Diesperova, M.I., Stepanova, I.A., Kukuruzov, A.P? Bugaev, N.F., Simonov,
K.V., Luzin, A.G., Baranovskii, N.I., and Yur'eva, N.A. (1970) Obtaining dense powder
from Satkinsk magnesite beneficiated by flotation, Ogneupory 1, 1-5 (Transl.)
to. Komlev, A.M., and Potapenko, V.E. (1972) Study of the role of reagents in magnesite
flotation, Ogneupory 2., 22-24 (Transl.).
11. Bugaev, N.F., Simonov, K.V., Mezentseva, A.N., Bocharov, L.D., Kolbosov, A.G.,
Bron, V.A., Alekseev, V.V., Baranovskii, N.I., and Potapenko, V.E. (1975) Powders and
refractory products from flotation-concentrated magnesite, Ogneupory 2., 6-15 (Trans!.).
12. Jepsen T.L.B. (1968) Controlled Reagent Introduction in a Magnesite Ore Concentration
Process, US Pat., 3,383,057.
13. Jepsen T.L.B. (1972) Flotation concentration of magnesite with emulsified collector reagents,
 379

US Pat., 3,383,057.
14. Predali, J.J. (1973) Adsorption des Sels d' Acides Gras a l'lnterJace Solide Uquide et
Flottation des Carbonates, Science de la Terre, Memoire 27, Universite de Nancy.
15. Predali, J.J. (1969) Flotation of carbonates with salt offatty acids: role ofpH and alkyl chain,
IMMTrans.11i, CI40-CI47.
16. Blazy, P., Cases, J.M., Houot, R., and PredaJi, J.J. (1969) Valorisation de magnesite
sedimentaire, Rev. Ind. Miner. a, 5,419-430.
17. Predali, J.J. (1970) Etude theorique et experimentale de laflottation d'un melange binaire
dolomite-magnesite de Kosice (Tchecoslovaquie), Proceedings 9th Inl. Miner. Process.
Congress (prague), pp. 241-249.
18 Cases, J.M., Pred ali , J.J., and Blazy, P. (1967) Contribution a ['etude du systeme
magnesite-oleate de sodium, e. R. Acad. Sci. Paris, 02717-02720.
19. GalJios, G.P. (1987) Ph.D. Thesis, University of Thessaloniki, Greece.
20. Filkelstein, N.P. (1989) Review of interactions in flotation of sparingly soluble calcium
minerals with anionic collectors, Trans. IMM, C2li, 157-CI77.
21. Matis, K.A., Gallios, G.P., and Zouboulis, A.I. (1989) Adsorption offatty acids onfine
mineral particles as a purification method of industrial wastewaters, Toxicol. & Environm.
Chern. 20-1,425-435.
22. Gallios, G.P., Matis, K.A., and Birda E.S. (1987) A study of the zeta potential of
magnesium-containing minerals for their flotation, Tech. Chron. r::J.., 3-4, 21-51 (in Greek
with English extended summary).
23. Hanna, H.S., and Somasundaran, P. (1976) Flotation of salt-type minerals, in M.e.
Fuerstenau (ed.) Flotation - A.M. Gaudin Memorial Volume, SME/AIME, N. York, pp.
197-271.
24. Somasundaran, P., Amankonah, J.O., and Ananthapadmanabhan, K.P. (1985) Calcite-
apatite interactions and their effects in selective flotation using oleate, in Proceedings XV Int.
Miner. Process. Congress (Cannes), pp. 244-254.
25. Matis, K.A., and Gallios, G.P. (1989) Anionic flotation of magnesium carbonates by
modifiers, Int. J. Miner. Process.~, 261-274.
26. Brandao, P.R.G., and Poling, G.W. (1982) Anionic flotation of magnesite, Can. Metal1. Q.
21,211-110.
27. Matis, K.A., Gallios, G.P., Stalidis, G.A., and Hollick, C.T. (1989) Flotation of magnesite
and dolomite by fatty acids, Trans. IMM ~, C99-C105.
28. Matis, K.A., Balabanidis, Th.N., and Gallios, G.P. (1988) Processing of magnesium
carbonate fines IJy dissolved air flotation, Coil. & Surf. 29,191-203.
29. Lovell, V.Metall. (1982) Industrial flotation reagents, in R.P. King (ed.) Principles of
Flotation, S. Afr. Inst. Min. Metal!., Johannesburg, pp. 73-89.
30. Matis, K.A., and Gallios, G.P. (1988) Flotation of salt-type minerals by fatty acids, Tech.
Chron. a, 4, 59-85 (Greek with English extended summary).
31. Matis, K.A., and GalJios, G.P. (1986) Dissolved-air and electrolytic flotation, in B.A. Wills
and R.W. Barley (eds.) Mineral Processing at a Crossroads, Martinus Nijhoff, Dordrecht,
pp. 37-70.
32. Peck, A.S., and Wadsworth, M.E. (1963) Infrared studies of oleic acid and sodium oleate
adsorption onfluorite, barite and calcite, US Bureau of Mines, RI 6202.
33. Marinakis, K.I. (1980) Ph.D. Thesis, Imperial College, London.
34. Ananthapadmanabhan, K.P., and Somasundaran, P. (1985) Surface precipitation of
surJactants and inorganics in minereral solids and its role in adsorption and flotation,
380

Proceedings XV Int. Miner. Process. Congress (Cannes), paper I1I.4.

35. Brandao, P.R.G. (1982) Ph.D Thesis, University of British Columbia, Canada.
36. Klassen V.I., and Mokrousov. V.A. (1963) An Introduction to the Theory of Flotation.
ButelWorths. London. p. 303.
37. Gaudin, A.M. (1957) Flotation, McGraw-Hill, N. York.
38. Wark. I. (1984) Contact angles and flotation research, in M.H. Jones and J.T. Woodcock
(eds.) Principles of Mineral Flotation - The Wark Symposium. Australasian IMM, Parkville.
pp. 1-6.
39. Antti, B.M .• and Forssberg. E. (1989) Pulp chemistry in calcite flotation. Modelling of oleate
adsorption using theoretical equilibrium calculations. Minerals Engg. 2. 93-109.
40. Matis. K.A .• Balabanidis. Th.N .• and Gallios, G.P. (1984) A study of the floatability of
calcite. in Proceedings 9th Nat. Chern. Conf.. Greek Chern. Assoc .• Athens, pp. 695-702.
41. Leja, J. (1982) Surface Chemistry of Froth Flotation, Plenum Press. N. York, p. 618.
42. Changgen. L.. and Yongxin, L. (1983) Selective flotation ofscheelitefrom calcium minerals
with sodium oleate as a collector and phosphates as modifiers. II. The mechanism of the
intaraction between phosphate modifiers and minerals. Int. J. Miner. Process. ill. 219-235.
43. Auge, P., Bahr, A., and Koser, H. (1975) Selective depression of silicates in scheelite
flotation with fatty acids, Proceedings XI Int. Miner. Process. Congress (Cagliari), paper 26.
44. Shergold. H.L. (1982) Flotation in mineral processing. in K.J. Ives (ed.) The Scientific Basis
of Flotation, Proceedings NATO ASI (Cambridge), preprint.
45. Parsonage, P .• Mclven, D .• Healey, A.F., and Watson. D. (1984) Depressantfunction in
flotation of calcite. apatite and dolomite, in Proceedings Conf. "Reagents in the Mineral Indus-
try" (IMM. Rome), pp 33-40.
46. Marinakis. K.I., and Shergold, H.L. (1985) Influence of sodium silicate addition on the
adsorption of oleic acid byfluorite, calcite and barite,lnt. J. Miner. Process. 14. 177-193.
47. Salatic. D .• and Milic M. (1978) Electrokinetisches potential und selective flotation von
magnesit-und dolomit mineral en • Freib. Forsch. 5.21, 205-213 (Ger.).
48. Balabanidis. Th.N. Personal Communication (unpublished data).
49. Gallios. G.P., and Matis. K.A. (1989) Floatability of magnesium carbonates by sodium
oleate in presence of modifiers. Sep. Sci. & Techno!. 24, 129-143.
50. Ref. [41], p. 326.
51. ibid., p. 758.
52. Andrews. P.R.A.. and Collins R.K. (1988) Beneficiation studies, production and current
interest of barite, celestite andfluorite ores in Canada, in E. Forssberg (cd.) Proceedings XVI
Int. Miner. Process. Congress. Elsevier. Amsterdam. pp. 1427-1438.
53. Eigeles, M.A. (1957) Selective flotation of non sulphide minerals. in Proceedings IV Int.
Miner. Process. Congress, Almquist and Wiksell, Stockholm, pp. 591-609.
54. Andrews. P.R.A.. and Collins. R.K. (1989) Beneficiation of barite. A review of processing
studies at CANMET. Mining Engg .• June. 457-462.
55. Bahr, A., Clement. M .• and Surmatz, H. (1968) On the effect of inorganic and organic
substances on the flotation of some non-sulphide minerals. in Proceedings VII Int. Miner.
Process. Congress (Leningrad), paper S-II.
56. Jowett, H. (1974) Depression of barites by chromate. Powder Technology lQ, 157-158.
57. Tsailas. D.P.. Catrakis, S.D., and Frangiskos, A.I. (1968) The role and the mechanism of
the action of sodium silicate in the flotation of barite, in Proceedings VIII Int. Miner.
Process. Congress (Leningrad), paper D-17.
58. Fuerstenau. M.C., Miller J.D .• and Kuhn M.e. (1985) Chemistry of Flotation. SME/AIME.
 381

N. York, p. 127.
59. Miller, J.D., Wadsworth, M.E., Mistra, M., and Hu J.S. (1984) Flotation chemistry of the
fluorite/oleate system, in M.H. Jones and J.T. Woodcock (eds.) Principles of Mineral Flota-
tion - The Wark Symposium, Australasian IMM, Parkville, pp. 31-42.
60. Gutierrez, C. (1979) Influence of calcium ion and quebracho in the oleate flotation offluorite
and calcite, Trans. AIME 226,1918-1924.
61. Baldauf, H., Schubert, H., and Kramer, W. (1985) A new reagent regime for the flotation
separation of fluorite and calcite, in Proceedings XV Int. Miner. Process. Congress
(Cannes), pp. 222-231.
62. Schulz, P., and Dobias, B. (1985) Effect of lattice ions in the selective flotation of salt-type
minerals, in XV Inti. Miner. Process. Congr., Cannes, pp. 16-27.
63. Pugh, R., and Stenius, P. (1985) Solution chemistry studies and flotation behaviour of
apatite, calcite andfluorite minerals with sodium oleate collector, Int. J. Miner. Process. 15,
193-218.
64. Moudgil, B.M., Vasudevan, T.V., Ince, D., and May, M. (1988) Adsorption of oleate on
dolomite and apatite, in Y.A. Attia, B.M. Moudgil and S. Chander (eds.) Interfacial Pheno-
mena in Biotechnology and Materials Processing, Elsevier, Amsterdam, pp. 285-303.
65. Soto, H., and Iwasaki, I. (1986) Selective flotation of phosphates from dolomite using
cationic collectors. I. Effect of collector and non polar hydrocarbons, Int. J. Miner. Process .
.1.2, 3-16.
66. Fu, E., and Somasundaran, P. (1986) Alizarin Red S as aflotation modifying agent in calcite-
apatite systems, Int. J. Miner. Process. U, 287-296.
67. Ananthapadmanabhan, K.P., and Somasundaran, P. (1984) Role of dissolved mineral
species in calcite-apatite flotation, Miner. & Mctal. Process., May, 36-42.
68. Rule, A.R., Kirby, D.E., and Dahlin D.C. (1977) Recent advances in beneficiation of
western phosphates, Proceedings SME Fall Meeting, St. Louis, Missouri, preprint.
69. Hanna, H.S. (1975) The role of cationic surfactants in the selective flotation ofphosphate ore
constituents, Powder Technology 12,57-64.
70. Smani, M.S., Blazy, P., and Cases, J.M. (1975) Beneficiation of sedimentary moroccan
phosphate ore, Trans. AlME 258,168-182.
71. Rao, D.V., Narayanan, M.K., Nayak, V.B., Ananthapadmanabhan K., and Somasundaran,
P. (1979) Flotation of low grade mussorie phosphate ore, in Proceedings Int. Symp.
"Resources Engineering and Technology", vol. 2, pp. 1-11.
72. Ratobylskaya, L.D., Klassen, V.I., Boiko, N.N., Baskakova, M.I., and Smimov, V.M.
(1975) Development and industrial introduction of new concentration process for
phosphorites of complex mineral composition, in Proceedings XI Int. Miner. Process.
Congess (Cagliari), pp. 1148-1172.
73. Atak, S., Gurkan, V., Guney, A., Cakir, M., Ozturk, N., and Akyol, F. (1988) Applicability
of slime flotation to Uludag tungsten plant, in Y. Aytekin (ed.) Proceedings II Int. Miner.
Process. Symp. (Izmir), pp. 187-196.
74. Texeira, O.C.G., Oliveira, J.F., and Boas, R.C.V. (1985) Surface chemistry and flotation of
scheelite,fluorite and apatite, with anionic collectors, in S.H. Castro and J. Alvarez (eds.)
Froth Flotation, Elsevier, Amsterdam, pp. 175-184.
75. Glembotskii, V.A., Klassen, V.I. and Plaksin, LN. (1972) Flotation, Primary Sources, N.
York, p. 202.
382

Visiting a flotation plant

3.4. VIBROACOUSTIC IMPROVEMENTS OF FROTH FLOTATION

ST. STOEV. L. KUZEV. M. METODIEV and SlIT. DJENDOV A

Higher Institute of Mining and Geology
1156 Sofia
Bulgaria

ABSTRACT. The vibroacoustic technique is used in various industrial processes in the chemical
industry. building. robotics. agriculture etc. Some important economic advantages of this method
were proved. e.g. the high relative productivity. the high quality of products. lower energy and
water consumption, convenience for automation, etc. In mineral processing, the vibroacoustic
technique is mainly used in discharging hoppers, feeding. transportation. sieving, grinding, ore-
washing and table concentration. Vibroacoustic crushers. flat surface concentrators, cementators,
extractors, dryers etc. have already been adopted and introduced in practice.
In separation processes. however. vibroacoustic techniques have been insufficiently applied. A
number of investigations have shown the positive effect of vibroacoustics on flotation. The aim of
this paper is to acquaint the mineral processing specialists with results (obtained by the authors) in
studying various vibroacoustic effects on the separate stages of the flotation process, i.e.
preparation of the pulp and reagents, formation and behaviour of the bubbles, secondary
concentration in the froth layer, desorption of the reagents, joint accomplishment of grinding and
flotation. A description is also given of two designs of vibroacoustic flotation machines. The
technological and economic results obtained justify the increasing interest on the part of researchers
and technologists in the vibroacoustic method as a means of improving the froth flotation
technology.

1. Introduction

Froth flotation is a process widely used in mineral separation today. It is of particular importance
for the minerals economics of many countries. Thanks to the efforts of a large number of
researchers, this process is being continuously improved and now it has become possible to use it
in the separation of practically all mineral mixtures.
In order to improve the technological and economic indices of flotation, various studies are still
being carried out. One way of achieving this purpose is the use of effects caused by vibroacoustic
excitations, which affect positively the accomplishment of the flotation mineral separation. The
term "vibroacoustic" is used because in practice the mechanical vibrations are inseparable from the
acoustic wave process excited by them.
Various authors have shown that the separate elements of the flotation process can be favourably
affected either by vibrations or by the acoustic wave process or by the joint manifestation of these
two physical phenomena. The conditions for a positive effect. both within the low and high

383
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 383-407.
© 1992 Kluwer Academic Publishers.
384

frequency ranges, have been determined. No matter how positive are the results of the vibro-
acoustic effect within the ultrasonic range, the requirements in terms of apparatus and high energy
consumption essentialy restrict the practical application of high frequency vibrations. Therefore, the
effects of the low frequency range were studied. The present paper does not claim to draw general
conclusions; these are given in ref. [1]. Certain results are presented, which orientate towards
perspective trends.
The lack of rotating parts in vibrating machines allows grinding and flotation to be performed in
a single mechanical unit. Such a device was designed and tested, using real baryte-iron ore,
yielding good results, i.e. it reduced flotation time, increased the yield of concentrate and enhanced
recovery.
Good results were also obtained in the laboratory flotation of baryte-iron ore, when the flotation
chamber vibrated together with the pulp in it (total vibration). Investigations under such conditions,
however, are more theoretical in nature because the total vibration (though possible) does not seem
very probable, as far as economic and sanitary considerations are concerned, i.e. higher energy
consumption and noise protection requirements.
In view of these considerations, pulp vibration will be more efficient if the vibrating devices are
placed in the pulp itself. This was proved by using two design versions of the vibrating flotation
cells - vibrovalve and auto pulsation. The vibration of the pulp together with air bubbles in the cells
occurs outwards, thus excluding any energy consumption for vibrating the cell itself and damping
the noise. Tests carried out with baryte-iron ore have yielded results which characterise these
flotation devices as promising.
Positive vibroacoustic effects may also be obtained within existing flotation machines by exciting
vibrations of the bubbles by means of insonification of the air supplied in the flotation cells. Tests
carried out with talc, baryte, coal and copper ore have shown a several per cent increase in recovery
with improved concentrate grades.
The secondary concentration of floatable component in the froth allows to obtain higher grade
middlings and end products. It was found that adequate vibration of the froth gives better results
than water spraying, with the additional advantage of obtaining more dehydrated froth. The
laboratory experiments were confirmed under production conditions by vibrating the froth in the
lead cycle of the Kardjali processing plant.
The investigation also proved the positive effect of the combination of regrinding of the bulk
copper-molybdenum concentrate with desorption of the reagents under conditions of vibration
impact. Results also improved by vibrosteaming of the concentrate.
Good results were finally obtained in the vibroemulsification of reagents for coal flotation, as
well as in the insonification of reagent solutions used in the flotation of baryte-iron, copper and
polymetallic ores. The results have been implemented at the Tvarditsa (coal), Kremikovtsi (baryte-
iron ore), Mirkovo (copper ore) and Kardjali (polymetallic ore) processing plants, in Bulgaria.

2 _ Technological Aspects

The effects of vibroacoustics are examined here either on flotation, in general, or on individual
components of the flotation system; some devices for generating the desired high frequency
vibrations are also discussed.
It was found that by means of vibrations it is possible:
- to facilitate the attachment of mineral particles to air bubbles;
- to increase the recovery of valuable components;
 385

- to improve selectivity and, correspondingly, achieve a rise in the quantity of valuable

components in the concentrate;
- to increase the flotation rate;
- to fonn air bubbles with appropriate dimensions;
- to achieve a desorption of reagents from the mineral surfaces;
- to emulsify flotation reagents;
- to destroy viscous and resilient froths, etc.

2.1. INTEGRATION OF VIBROGRINDING WITH FLOTATION

This integrated technology showed a number of advantages: power consumption is reduced,

slurrying and regrinding are avoided (thereby reducing losses of valuable component); the
unnecessary grinding oflarge amount of rock mass is also avoided.
Figure I shows a vibromill flotation device designed for laboratory tests. It consists of a lower
part, which vibrates and causes grinding and an upper part, which stabilises the pulp motion. The
two parts are elastically connected. The air needed for flotation is fed through a sparger (perforated
pipe) located at the bottom of the cell.

Figure 1. VibromiIJ - flotation setup. I, damping screen; 2, elastic band; 3, airpipe; 4, vibrator.

Vibration parameters for these tests were set as follows: frequency (f) 25 Hz and amplitude
(A) 3 mm. The grinding media consisted of 15 mm (40%) and 9 mm (60%) balls. Flotation
experiments were conducted with baryte from the Kremikovtsi ore deposit, having a baryte content
of 27.17%, at reagent conditions similar to those of industrial plants.
Some of the results concerning the effcct of processing time and pulp density are given in
Figure 2. Experiments were perfonned with dense pulps; the results obtained were less positive,
compared to results for pulp densities nonnally used in flotation. An improvement was observed,
however, in the presence of vibrations, achieving a 30-40 % shorter flotation time; if the grinding
time is included, then the total time for treatment is reduced by 3-4 times, compared to the separate
running of the processes. The experiments showed that with only one rougher flotation in the
integral treatment, in spite of the reduced flotation time, the waste content of the baryte was the
same as that obtained in conventional flotation with scavenger operations [1]. Such a method of
386

75 Baryte grade [%]

70

66···
2 min
15 min
60 ..... .
7 min

55~----~----~----~----~----~--~
25 30 35 40 45 50 55
pulp density [%]

Figure 2. Grade of baryte product as a function of pulp density and flotation time (f= 25 Hz. A= 25 mm).

treatment is advisable for ores with irregular dissemination of the valuable component. The air
passing through the grinding media is well dispersed.
Vibration of the flotation pulp is observed practically in all flotation machines because of
inbalance of the moving parts or because of turbulent motions. The vibrations are random and of
low intensity.

2.2. TOTAL VIBRATION (VIBRATION OF THE FLOTATION CELL WITH THE PULP)

Experiments with vertical linear vibrations were carried out. as shown in Figure 3. Baryte from the
Kremikovtsi ore deposit was ground down to 0.074 mm (70%) and was floated at 32% pulp
density. The process selectivity. estimated according to Beloglazov's index [2]. increassed because
of vibrations by 23-52%.
The results are similar for total vibration of the pulp in cleaner baryte flotation. The rise of
frequency from 80 to 120 Hz or of the amplitude from 0.2 to 0.6 mm helped in increasing the
baryte content from 57.0% (without vibrations) up to 63.2%. the recovery being reduced from
90.3% down to 87%. It is possible that during the pulp vibration at vibroaccelerations of 3 to 10 g.
along with the removal of silicate and ferrous minerals. less-strongly attached baryte-bearing
particles or middlings are also removed from the bubble surfaces. The increased selectivity is due
both to the self-cleaning of the bubble surfaces from gangue components and to the destruction of
the flocs in the pulp. thus enabling the particles to float individually. Results of baryte flotation
(concentrate. grade and recovery) as a function of the frequency and amplitude of vibrations are
given in Figure 4.
Analysing the reasons for baryte losses in the tailings. it was found that in the tests carried out
without vibrations the baryte is lost with the fines. whereas after intense vibrations baryte is lost
with the coarse particles. This phenomenon is accounted for by the easier overcoming of the energy
barrier on the bubble surfaces by small particles. in the presence of vibration effects and the easier
inertial removal of the coarser particles from the bubble surfaces. In studying the effect of collector
consumption. it was found that the positive effect of vibrations decreased with increasing collector
consumption. Thus. at a collector consumption of 0.5 kg/to the difference in contents is 5.5% while
at 1.25 kg/t the difference goes down to 2.0%.
 387

5
4
3
2

Figure 3. Laboratory flotation cell vibrating vertically. I, vibrator; 2, mounting platform; 3, rod;
4, rubber seal; 5, disk.

Similar tests with total pulp vibration were carried out for a Pb-Zn ore. At 100Hz frequency and
0.4 nun vibration amplitude, the Pb content improved by 3.0% and the Zn content by 2.4%, with a
decrease in the recovery for Pb by 0.8 % and for Zn by 0.2%. Selectivity improvement for an
amplitude of 0.4 mm was 48.4%, while for A =0.6 mm it went up to 60.5%.

2.3. FLOTATION WITH 'SOUND' TREATED BUBBLES

The physical characteristics of the bubble surfaces, during flotation, have an essential effect on the
adhesion of the minerals. Destruction and thinning of the hydration layers decrease the energy
barrier, as a result of which the act of adhesion is facilitated and accelerated. That is why the bubble
generation using "sound-processed" air is of particular interest. The aim is to set the bubble
surfaces in vibration, thus achieving an improvement of the particle adhesion and the cleaning of
the bubble surfaces from mechanically-attracted hydrophilic particles and weakly attached
aggregates.
When the bubbles vibrate intensively, it is possible to remove hydrophobic particles from their
surfaces as well. In conventional flotation machines, where the bubbles are formed by turbulent
vortices, finite vibrations are generated but these are of very short duration and do not affect
practically the final flotation results. For maintaining optimal pulsation of the bubbles in the pulp, a
continuously acting source of energy is needed, which causes sign-alternating pressure of the
interface between the liquid and gaseous phases from the pulp or the gas side. As a result of
modulating the pressure, the conditions of existence of the bubbles change - when increasing the
pressure from the pulp side, the bubbles shrink and vice versa, i.e. the surface vibrates. The use of
388

52 Baryte yield [%]

0,2

0,4
0,8

0,8
40L---~----~----~-----L-----L--~
60 80 100 120 140 160
frequenoy [Hz]

eo Baryte grade [%]

50~--~-----L----~----~--~~--~
60 80 100 120 140 160
frequenoy [Hz]

ge Baryte reoovery [%]

0,4

8e~--~----~----~-----L-----L--~
eo 80 100 120 140 160
frequenoy [Hz]

Figure 4. Influence of frequency and amplitude of vibrations on the yield, grade and recovery of baryte
concentrate [3].
 389

sound-treated air is particularly expedient when it is necessary to act only upon the bubbles and not
upon the whole flotation system.
The behaviour of non-treated air bubbles fonned without a frother and sound-modulated air (by
a plate generator), was studied by filming with a 7500 frames/second cine camera (Fastex WF4).
In the presence of "insonified" air, non-linear vibrations were found on the surface of the bubbles.
The various sides of 0.2-4.0 mm bubbles vibrate with different frequencies and amplitudes.
Surface vibrations were established, having frequencies from 80 up to 3240 Hz at amplitudes from
0.1 up to about 0.6 mm. In tracing the behaviour of the bubbles, it was found that while the large
bubbles started emerging immediately, the small ones moved very slowly. This delay of the small
bubbles is very useful because it brings about an increase in the level of aeration in the pulp,
thereby improving the bubble mineralisation.
Many bubbles have considerable irregularities and surface waves are observed. The particles
stuck to the irregular bubble surfaces slide because of the different inclination of the separate
surface zones, as a result of which the poorly attached particles are repelled. Besides, bubbles of
that type are dispersed resulting in further cleaning of unwanted particles. The surface waves
probably affect the hydrated layers at the liquid-gas interface. For bubbles with a diameter of over
3 mm, the velocity of the surface waves is about 150-600 mm/s at a frequency of 300-900 Hz, and
a length of the semi-wave of 0.2-0.7 mm. For smaller diameter bubbles, surface waves were not
observed.
When "sound"-modulated air was used, better dispersion was observed. The bubble dispersion
rate grew with the increase of the sound frequency and the air discharge rate. The effects of using
"sounded" air in flotation were tested by means of a laboratory airlift flotation chamber (Figure 5).
Plate and membrane whistles (Figures 6 and 7 respectively) were used as sound generators.
Bubble sizes obtained in the laboratory cell ranged from 0.2 to 10 mm, corresponding to a
frequency range of 328 to 16400 Hz. The frequency calculations were made according to the
fonnula:

f'" r = 328 (1)

where f is the frequency of vibration, and r the radius of the bubbles (in centimeters).
1

3
~----4

Figure 5. Pneumatic flotation laboratory cell: 1, air manifold; 2, pipe; 3, sound generator; 4, mesh.
390

TABLE 1. F"requency ch aractenstlcs 0 fth e Insom

. 'fiIed rur.
.
Generator design Intensil't in dB at freQuencies [Hz)
31.5 63 125 250 500 1000 2000 4000 8000 16000
Plate 34 36 33 34 45 49 58 64 51 47
Membrane 58 56 62 63 78 74 63 53 51 44

4
'3
A 5

6
2
1

sound trea.ted
a.ir
Figure 6. Plate sound generator: I, pipe; 2 and Figure 7. Membrane sound generator: 1,6, pipes;
4, parts forming air channel; 3, vibrating plate. 2, lower side of sound chamber; 3, screens; 4, upper
side of sound chamber; 5, rubber membrane.

The frequency characteristics of the insonified air in the laboratory flotation ceil, for the two
sound generators, are given in Table 1. These characteristics were measured with a sound analyser
at 1 m distance from the cell. When the sound waves pass through a finite medium, the amplitude
of vibrations is calculated by the formula

(2)

where P s is the sound pressure [bar], Q the air density [g/cm 3), v the velocity of sound in the air
[cm/s) and f the frequency of vibrations [Hz). For the large bubbles, the vibration amplitude
calculated is about 0.6 mm, whereas for the small ones it is about 0.01 mm.
Comparative tests were carried out with coal, baryte, talc and copper ore in order to investigate
the effect of sound-treated air in rougher flotation, without selecting optimal conditions. The data
from baryte flotation at a solid:liquid ratio of 1:3 and rougher flotation of 5 min are given in Table
2. Results for talc flotation at a solid:liquid ratio of 1:4 are given in Table 3; results for coal
flotation at a solid:liquid ratio of 1:5 are given in Table 4; results for copper ore flotation at a solid
to liquid ratio of 1: 3.8 are given in Table 5.
 391

TABLE 2 Results of baryte flotation

Yield BaS04 BaS04
content recovery
(%) (%) (%)
Hotation with insonified air 26.8 65.8 43.9
Conventional flotation 25.4 64.5 4l.5

TABLE 3 Results of talc flotation

Yield Talc Talc
content recovery
(%) (%) (%)
Hotation with insonified air 25.23 60.98 45.00
Conventional flotation 22.28 58.18 34.10

TABLE 4 Results of coal flotation (concentrates)

Yield Ash Recovery of
content combustibles
(%) (%) (%)
Hotation with insonified air 34.80 7.85 4l.90
Conventional flotation 25.50 8.05 30.60

TABLE 5. Results of copper ore flotation.

Yield Cucontent Cu recovery
(%) (%) (%)
Hotation with insonified air 6.28 19.28 69.45
Conventional flotation 5.82 18.20 66.20

2.4. SECONDARY CONCENTRATION UNDER VIBRATION

The secondary concentration of the valuable component in the concentrate is known and is
exploited by sprinkling with water. Difficulties arise from the exact dosage of water required and
the necessity to provide a means of drop dispersion. However, if the froth is intensively shaken,
the bursting and merging of the bubbles leads to a secondary concentration.
The effects of vibrations on the froth were studied with coking coal, baryte, pyrite and talc,
directly in the flotation machine and separately from it. A 15 W electromagnetic vibrator, operating
at a frequency of 100 Hz and 1 mm amplitude was used to act upon the froth in a 3 litre flotation
machine. An inertia vibrator with f = 45 Hz and A = 4 mm was used for impact on the froth
separated from the cell.
Figure 8 shows data from coal flotation at the Devina Mine with a coal ash content of 26.4% and
a particle size below 0.5 mm. The ash content decreased by 2% on average, irrespective of
collector consumption. For establishing the effect of size in the vibration cleaning of the froth, size
fractions +0.25 -0.5 mm with 22.6% ash, +0.1 -0.25 mm with 25.7% ash and -0.1 mm with
33.1 % ash were floated separately. Vibrations helped in decreasing the ash content in the fine and
coarse size fractions.
Vibration of the froth removed from the flotation cell showed that, at low frequencies and high
392

amplitudes, intensification of the secondary concentration of valuable component in the froth was
achieved.
When pyrite and baryte froth was subjected to vibroaction, the content of the valuable
component in the froth decreased, which implied that the vibration effects may influence essentially
the flotation results.

Ash content [%]

14~~~~~~---------------------------,

12

10

6~--~----~--~----~--~----~~
100 200 300 400 500 600 700
011 consumption [g/t]

Figure 8. Relationship between ash content in coal concentrate and reagent (pine oil) consumption :
1, without froth vibrations; 2, with froth vibrations. Solid to liquid ratio of 1:4, 1 min agitation time and
3 min flotation time.

2.5. VIBRODESORPTION

Efficient reagents desorption from mineral surfaces can be achieved under low frequency
vibroactions, too. The process of desorption takes place as a result of the attrition of the particles
(autogenous vibration friction), which destructs the reagent films. This process can be effiCiently
combined with vibration regrinding of the bulk concentrates.

TABLE 6 Results from vibrodestruction (Cu-Mo bulk concentrate).

Solid : liquid Ratio of changes in xanthate concentration
without desorbent water glass 1200 t/g
4 min 8 min 12min 4 min 8 min 12 min
1: 0.6 4.30 2.90 2.00 19.2 21.3 21.5
1: 0.75 3.87 1.80 1.50 17.6 19.2 19.9
1: 1 2.90 1.40 1.40 12.1 14.0 14.7

The results of the tests on vibration destruction without grinding media showed low efficiencies.
In fact, there is a desorption, but the reverse process of sorption of the individual reagents is also
activated; this is obvious from the results given in Table 6. When a desorbent is added, a
considerable increase in the quantity of xanthate released during attrition is determined in relation to
the quantity of xanthate released in the liquid phase after agitation. The water glass hydrophilised
the released mineral surfaces from xanthate and the latter remained in the liquid phase. In
 393

combining the vibration effects with several exchanges of the liquid phase, a satisfactory cleaning
of the surfaces is achieved.
The vibration of a CuoMo bulk concentrate, aiming at the desorption of reagents, may be
combined with regrinding which, from a technological point of view, is useful as the particles are
uncovered and the selection is more fully realised. Results from such vibration effects combined
with regrinding for desorption of the reagents by the surfaces are given in Table 7.

TABLE 7. Results on vibrodesorotion (attrition and gnnding).

Solid : liquid Ratio of change in xanthate concentration
without desorbent water glass 1200 'ilIt
21.4 Hz 16.5 Hz 21.4 Hz 16.5 Hz
1: 0.75 30.9 18.8 39.8 28.4
1:1 12.6 8.9 18.8 13.1
1:2 5.4 3.2 9.8 6.8

Experiments were performed at f = 21.4 Hz (A = 2.8 mm) and at f = 16.5 Hz (A = 3.1 mm).
The CuoMo bulk concentrate was reground to 90% -0.08 mm. The amount of desorbed xanthate
increased in comparison with the vibration attrition. The general analysis of the results showed that
there is an optimum time of vibroattrition in the presence of desorbents. Longer times increased
slightly the amount of reagent released. From the bulk CuoMo concentrate, with one basic rougher
operation, a froth was obtained containing 0.82% Mo at 78% recovery, i.e. vibroattrition competed
with steaming.
Investigations were carried out to determine the effect of vibrations on the thermochemical
destruction of the reagent coverage of the CuoMo bulk concentrate particles. The tests were
performed according to the schema in Figure 9. Steaming was performed in a cylindrical container
with a double shell, in which hot water circulates at 75-80 0c. The bulk concentrate contained
0.34% Mo and 8.8% Cu. Lime was added at a rate of 2.34 kg/t. The agitation was mechanical,
with a mixer mounted on the platform of a vibration stand, providing operation with and without
vibrations. The mechanical agitation caused suspension of the particles in the whole volume.
The vertical linear vibrations affected the diffusion layer of the particles, the physico-chemical
mass exchange being altered. The absence of active mechanical contacts among particles was
indirectly proved by the grain size analysis of the samples treated. The contents of the -0.074 mm
class treated by steaming and vibrosteaming differed only by 1-3%. Without vibrations, the highest
recovery of Mo - 42.4% with a 0.9% content - was obtained at 90 min steaming, whereas for 10
min steaming at 50 Hz frequency (A = 1 mm), the recovery was 50.7% with a 0.54% content of
Mo in the concentrate. Combining the thermochemical and the vibration impact (for 40 min), a
recovery of 73.5% was obtained with a 8.4% Mo content.
Decreasing the processing time and improving the indices are factors making vibrosteaming a
promising technique for selective flotation of Cu-Mo bulk concentrates.

3. Vibration Flotation Machines

Two types of vibration flotation machines, vibrovalve (4) and autopulsation (5), were designed at
the Higher Institute of Mining and Geology (in Sofia).
394

Figure 9. Laboratory installation for vibrosteaming tests : 1, frame; 2, mixer; 3, container; 4, vibro-
stand; 5, buffer container; 6, heater; 7, thermometer; 8, pump.

o
o
lEE 1

Figure 10. Vibrovalve flotation machine: 1, disperser; 2, pipe; 3, valve; 4, side openings; 5, ring-
shaped channel; 6, ring-shaped slot
 395

3.1. VIBROVALVE FLOTATION MACHINES

The vibrovalve flotation machine is pneumo-mechanical, with a disperser vibrating in a vertical

direction (Figure 10) at a frequency of 25-45 Hz and amplitude of 2.5-4.0 mm. Suction of the pulp
from the bottom part of the chamber and its circulation through the side openings of the disperser is
provided by a valve. The air, fed through the carrying tube, is distributed in a ring-shaped channel
in the lower part of the disperser and enters through a ring-shaped slot in the pulp where, under the
action of vibrations, is dispersed into bubbles of 1-2 mm diameter without a frother.

3.2. AUTOPULSATIONFLOTATIONMACHINE

The autopulsation flotation machine is pneumatic (Figure 11). As air is fed through the inlet
channel at 103-105 Pa, the pressure in the chamber rises and when it overcomes the pressure of the
water column on the plate and the spring, the valve opens and part of the air enters the chamber and
disperses in bubbles having a diameter of 2-5 mm. The pressure in the chamber falls, the plate
closes the opening and the cycle starts again. Thus, autovibrations of the valve at a frequency of
30-100 Hz and amplitude of 0.1-0.5 mm occur. The parameters of the plate vibrations depend on
the air pressure, the ratio of the chamber volume to the diameter of the inlet opening, the spring
characteristics, the plate size. A zone of weak turbulence is created in the lower part of the vibration
machine, caused by air leakage, which helps to prevent the sedimentation of the solid phase.

-.n=. __ ._._. =tL

o 0 0 0
o 0

Figure 11. Autopulsation flotation machine: I, inlet channel; 2, chamber; 3, plate; 4, spring.

3.3. OBSERVATIONS AND RESULTS

The vibrovalve flotation machine allows for fine dispersion of the air. Under the action of
vibrations, the gases dissolved in the water are released, the air introduced is dispersed and in the
presence of a frother, the diameter of the air bubbles decreases down to 0.2-\'0 mm.
Circulation of the pulp through the aerator during the valve vibration provides continuous motion
396

of the particles in the lower part of the chamber, thus preventing their settling down at the bottom.
In relation to the processes occuning when setting up the motion and the aeration of the pulp,
three zones can be distinguished in the vibrating flotation machines: bottom zone, intermediate zone
and froth layer zone. For the vibrovalve flotation machine, these zones are characterised by the
following features:
in the bottom zone, active circulation of the pulp occurs under the action of the vibrating
disperser. Air bubbles are created, the solid particles and the air bubbles move in opposite
directions. In the upper part of this zone, where the circulating flow comes out of the disperser,
the solid particles and the air bubbles meet at right angles.
- in the intermediate zone, upward flow at low turbulence levels acts from the bottom zone. The
mineralised and free bubbles emerge in a fairly quiescent medium, which improves additional
mineralisation.
- in the froth layer zone (in the pulp under the froth layer), a flow in the direction of the discharge
from the cell is created under the action of the vertical circulating flow.
In the whole volume of the vibrovalve flotation machine, the liquid phase, the solid particles, the
air bubbles and the froth layer are under the action of vibrations created by the vibrating disperser.
These three zones have different parameters :
- the bottom zone is characterised by a weak turbulence, which depends on the air pressure and
the elastic indices of the spring, as well as by the small height. The air bubbles in this zone meet
with the settling particles of the solid phase at an angle of 90°-130°.
- the intermediate zone is the longest one. Here, an upward flow is created under the action of the
emerging air bubbles and a downward flow by the sedimenting solid particles.
- the froth layer zone is subjected only to the impact of vibrations caused by the plate and
distributed through the whole volume of the flotation machine.
The solid particles found in the vibrating aqueous medium are drawn into motion, vibrating at a
frequency equal to that of the liquid. The amplitude of vibration, however, depends on the natural
frequency of vibration of the particles, on their diameter and density. as well as on the liquid
viscosity. When the natural frequency of vibration of the solid particles coincides with the
frequency of vibrations applied or their harmonics. their amplitude is biggest. the difference
between the phases of particle vibration and the medium is Cl'p - Cl'm -> 90 (where Cl'p is the
phase of particle vibration and Cl'm is the phase of medium vibration) and the particle sedimentation
rate is the lowest [6]. Because of the varying particle diameter and their different natural vibration
frequencies. their amplitude is different, ranging throughout the whole spectrum from 0.3 up to
1.0 Am (where Am is the amplitude of the liquid vibration).
The air bubbles dispersed in the pulp. under the action of the vibrations applied. vibrate as well
with their own frequency. Because of their lower density, the difference in the oscillation phases
between the air bubbles and the liquid is Cl'p - Cl'b -> 90° (where Cl'b is the phase of bubble
vibration). Under the action of vibrations. their rate of formation decreases. This and the amplitude
of vibrations of the air bubbles depend on their diameters and the frequency of vibrations applied.
In some cases the air vibration amplitude can be larger than the liquid vibration amplitude.
The decrease in the rate of formation of air bubbles allows for increasing the time of possible
contacts with the solid particles. By changing the frequency and amplitude of vibrations, it is
possible to regulate the formation time and respectively, to increase the degree of pulp aeration.
The oscillating motion of the liquid is accompanied by fluctuations of the pressure in the vibrated
volume. The volume of the air bubbles in the liquid periodically increases and decreases. at a
frequency equal to the frequency of vibrations applied. Therefore, the application of vibration.
apart from creating intensive circulation of the pulp in the flotation cell and the fine air dispersion.
 397

100

~3
Recovery [%]
80
60
40
(a) 20
100 0
Grade [%]
80
eo ~ 2
40
~___ 1
20 ----. II< 3
0
0 2 4 6 B 10
Time [min]

Recovery [%]
100
~ 80

~ 60

X
40
(b) 20
100 0
Grade [%]
80
eo
40 +--~ 2
20 3~ II<
0 '"
....~
0 2 4 6 8 10
Time [min]

Reoovery [%]
=It 100

~
80
60
40
(c) 20
100 0
Grade [%]
80
eo
40
20
~ ..........
0
0 2 4 6 B 10
Time [min]

Figure 12. Flotation results with vibrating and non·vibrating machines (1, vibrovalve flotation cell;
2, autopulsation flotation cell; 3, Mekhanobr flotation cell). (a) pulp density 40%, grain size 100%
-0.074 mm); (b) pulp density 60%, grain size 100% -0.074 mm); (c) pulp density 40%, grain size 100%
-0.030 mm).
398

also decreases the settling rate of the solid phase. the rate of fonnation of air bubbles and causes
volumetric oscillations of the air bubbles. Depending on the natural frequency of vibration of the
solid particles and the air bubbles. the ampli tudes of their vibration. the phases of vibration
between the particles and the liquid. between the air bubbles and the liquid. between the particles
and the solid phase which can reach up to 180°. are changed by absolute value.
This vibration effect changes the conditions for the adhesion of the particles to the air bubbles.
for additional mineralisation and for the release of accidentally attached particles. The air bubble
oscillation increases their effective diameter. de stabilises the structure of the hydrated layer. thus
increasing the possibility for encounter between the particle and the bubbles and decreases the
energy threshold. The difference in the phase of oscillation between the particles and the air
bubbles causes wide changes in their velocities. increasing the possibility of smaller particle
adhesions. The difference in the phases of oscillation allows also for the contact between a particle
and a bubble. even if their motions are parallel.
As the mineralised air bubble flows all the way up and into the froth layer. its oscillations and the
differences in the oscillation phases of the particles and the bubbles assist in releasing mechanically
attached gangue particles. thus raising the concentrate grade.
Comparative laboratory tests were carried out with the vibrating flotation machines described
above and a Mekhanobr mechanical flotation cell. Swedish lead ore and baryte ore from the
Kremikovtsi ore deposit were used. The results of these experiments are given in Figure 12. As
far as recovery of the valuable component is concerned. the vibrovalve and the mechanical flotation
machines show similar results. remaining the same at different liquid-solid ratios.

TABLE8M atnx 0 fth e planne d expenment.

Vibration frequency Vibration amplitude Air discarge
X]o s-1 x2. mm X3. 1/h
Basic level X o 27 2.6 150
Units of variation 3 0.2 30
Upper level x+ 30 2.8 180
Lower level x- 24 2.4 120
No. of experiment
I - - +
2 + - -
3 - + -
4 + + +

In relation to the grade of concentrate. the vibrovalve and the autopulsation flotation machines
show better results. by approximately 10-20%. as compared with the conventional mechanical
flotation machine. The vibrovalve flotation machine. at a high degree of recovery. increases the
content of valuable component in the concentrate by approximately 8%. even at a 60% pulp
density. In the test with the vibrovalve flotation machine, it was found that the vibration mechanism
desribed above is adequate.
On the basis of these results, the production of baryte concentrate was studied. Parallel
laboratory tests were perfonned in a Mekhanobr cell. Three factors, viz., frequency of vibrations,
amplitude of vibrations and air discharge were optimised according to the Box and Wilson method
[7]. A semi-replica of the complete factorial design was used as a matrix for the planned experiment
(see Table 8). The Gaudin selectivity index was chosen as a criterion for efficiency:
 399

S = ~(1 - 8) (3)
8(1 - ~)

where ~ is the content of useful mineral in concentrate, and 8 the content of useful mineral in the
tailings. Additional determinations of the valuable component recovery were done. To determine
the error, four tests were carried out under the same conditions The average results are given in
Table 9. In the additional experiments for gradient motion, the highest result (S = 24.72) was
obtained for a 49.3% baryte content and a 93.1 % recovery. Cleaning operations were performed
under this regime and their results are given in Table 10.

TABLE 9. Results of experiments carried out according to the factorial design

(y : yield, ~ : grade, R : recovery).
Expe- Concentrate (%) Tailings (%) Sele-
riment ctivity
no. index
v ~Ba ~Fe RBa v BBa BFe RFe S
1 53.46 47.59 3.10 92.79 46.54 4.24 15.96 81.80 20.51
2 56.22 48.00 2.88 94.03 43.78 3.94 16.40 81.60 22.63
3 52.84 48.80 2.77 90.07 47.16 6.04 15.73 83.52 14.83
4 55.29 45.26 3.32 93.62 44.71 3.82 15.84 79.42 20.82

TABLE 10 Results of baryte concentration

Flotation 1 Concentrate (%) I Tailings (%)
machine I y I BBa I BFe I RBa I y I BFe
Rou~er flotation
Vibrovalve 153.21 149.28 12.44 193.08 146.79 116.07
Mekhanobr 157.37 143.36 14.21 192.77 142.63 117.17
With two cleaning operations
Vibrovalve 138.45 156.60 10.22 I 161.55 116.18
Mekhanobr 138.12 155.36 10.76 I 161.88 115.51

The results from the laboratory experiments show the possibilities of the vibrovalve flotation
machine in the rougher and particularly in the cleaning operations, where for a given raw material a
standard concentrate is obtained with one cleaning operation less, as compared with the mechanical
flotation machine. The advantages of the vibrovalve flotation machine are as follows:
- high recovery as compared to the recovery in mechanical flotation machines;
- an ability to operate efficiently at pulp densities of up to 60 %;
- increased content of valuable component in the concentrate by 3-8 %;
- decreased energy consumption by approximately 25 %.
This improvement results from the influence of the following factors of vibration impact:
- active circulation of the pulp in the bottom zone of the flotation machine, which prevents creation
of stall zones and settling of solid particles;
- very good dispersion of the air and generation of equal-sized bubbles, combined with releasing
from water a certain amount of dissolved air;
- feeding of the dispersed air into the zone of active turbulence; and
400

- vibration impact on the three-phase system in the whole volume of the flotation cell.

4 . Treatment of Flotation Reagents

4.1. VIBROEMULSIACATlON OF REAGENTS

One of the methods to improve the efficiency of flotation is to prepare reagents properly before
feeding them into the pulp. The insoluble or not readily soluble flotation reagents present a serious
problem in this respect. The efficiency of using such reagents is considerably increased when they
are used as an emulsion. This can be achieved either by the use of emulsifying devices or by
adding emulsifying substances. By introduction of an original emulsifying device and emulsifica-
tion of kerosene and pine terpentine in coking coal flotation, the yield of concentrate increased by
8% and the ash content decreased by 0.5%.
Further improvement of the flotation process can be achieved if the emulsification of the reagents
(collector and frother) is accomplished simultaneously. Laboratory experiments were conducted
with reagents emulsified separately and as a mixture. The conditions for the experiments were
selected by statistical planning. The following regime was found to be optimal: frequency of
vibrations in the emulsifying device 35 Hz, amplitude 0.9 mm, water pressure 6 atm, terpentine
dosage 100 g/t, kerosene dosage 1200 gil, flotation time 2 min at 2% emulsion.
Two tests with simultaneous emulsification and two tests with separate emulsification of the
reagents were carried out. The simultaneously emulsified reagents give more stable emulsion
(lifetime longer than 24 h) with the size of the oil drops being under 5 Ilm. The results of the
experiments are shown in Table II. The experiment performed with separate emulsification of the
reagents differs from the one mentioned above only in the time of flotation being 6 min instead of 2
min and the vibration amplitude being 1.9 mm. The emulsification of the reagents in a mixture
allows a 7% increase in concentrate yield, the grade being also slightly improved.

T ABEL 11. Comparison of results of experiments with simultaneously and

separateIy emul SI'fiIed fl otation reagents.
Products Simultaneously emulsified Separately emulsified
Yield (%) Ash content, Ad Yield (%) Ash content, Ad
at dry base (%) at dry base (%)
Concentrate 48 11.94 41 12.24
Tailings 52 62.08 59 59.87
Feed 100.0 38.01 100.0 40.34

The emulsification of non-polar reagents is widely applied. It is known that the use of neutral
oils as collectors allows improved flotation of fines and aggregates, thus increasing the recovery of
valuable components in floating sulphide minerals. They are used in the flotation of copper
sulphide and oxide ores, copper-zinc, copper-lead, copper-molybdenum, etc.
The investigations also aimed at improving the selectivity of a zinc-pyrite concentrate flotation by
using emulsified neutral oils. For this purpose, laboratory tests were carried out with various non-
polar reagents under a flotation regime similar to the one used at the processing plant. Four types
of oil, differing in viscosity and chemical composition, were studied. The experiments were
conducted according to a bulk flotation flowsheet. Four experiments were carried out with each oil
 401

under the same regime. The results are summarised in Table 12. The best flotation results were
obtained with MMO-12 type oil. When this was used, the quality of the bulk concentrate was
similar to that obtained in the other experiments, but the lead and zinc recovery in the bulk
concentrate was 96.5% and 96.8% respectively, considerably higher than the recovery obtained
with the other oils.

TABLE 12. Results of bulk flotation of Pb-Zn ore under laboratory conditions using
various neutral oils (50 glt) 1.
Experi- Neutral oil Product Content (%) Recovery (%)
mentno. Lead Zinc Lead Zinc
1 MMO-12 Concentrate 5.89 5.50 96.54 96.83
2 M8D Concentrate 5.67 5.74 94.75 95.77
3 Kerosine Concentrate 5.76 5.43 94.40 95.39
4 Waste oil Concentrate 6.06 5.30 91.58 94.13
..
The expenments were earned out under the supervIsIOn of Mr. II. Zapartov, supenntendent
of the Kardjali processing plant.

The use of neutral oil for improving the selectivity of the zinc-pyrite flotation was necessary
because of the anal ysis, that showed the zinc losses in the pyrite concentrate in 1979 to be 1.2%
and in 1980 to be l.3%. For this purpose, a screen-and-chemical analysis was made of the bulk
zinc-pyrite concentrate and the pyrite concentrate obtained after stage flotation. The results are
given in Table 13. These results show that the zinc minerals in the pyrite concentrate are
concentrated in the slime fraction and in sizes of over 0.20 mm. It is natural to expect that the use
of neutral oil as an additional collector for zinc minerals will improve the recovery of fine classes.
The production experiments were carried out by feeding emulsion ofMMO-12 neutral oil in 50 gjt
quantity bulk concentrate, 20 gjt of xanthate also being fed during the emulsification of the oils. An
aqueous emulsion (0.1 %) was prepared; this was fed into a tank from where it was distributed into
a rougher zinc flotation.

TABLE 13. Screen analysis of zinc-pyrite and pyrite concentrate.

Size (mm) Zinc-pyrite concentrate Pyrite concentrate
Yield Zinc Recovery Yield Zinc Recovery
(%) (%) (%) (%) (%) (%)
1 +0.20 15.70 14.19 13.59 5.22 l.69 16.02
2 0.20 - 0.15 16.40 12.34 13.25 6.96 0.47 6.00
3 0.15 - 0.10 13.37 14.96 13.10 2.61 0.38 l.80
4 0.10 - 0.074 7.86 15.92 8.19 11.30 0.43 7.05
5 - 0.074 46.65 1.20 50.87 73.91 0.51 69.13
Total 100.00 15.28 100.00 100.00 0.55 100.00

The samples collected over a period of three months showed a decrease in the zinc content in the
pyrite concentrate by 0.4% and in the recovery by 0.9%. The zinc recovery in the zinc concentrate
increases by 0.6 up to 1.0%. It was also found that the losses of zinc in size +0.20 mm did not
decrease, i.e. the increased zinc recovery is due to the improved flotation of the slime sphalerite
particles when neutral oil emulsion is used.
402

4.2. BARITE ORE FLOTATION USING INSONIFIED SOLUTIONS OF FLOTATION

REAGENTS

When solutions of flotation reagents are insonified by ultrasound of a given intensity and for a
given time, a considerable improvement of the flotation is achieved [I]. The insonification of a
xanthate solution, for instance, led to the oxidation of part of it to dixanthogen, thus increasing its
flotation activity.
Because of the high cost of ultrasound, sound frequencies were used for treating solutions of
xanthate, potassium oleate, water glass and frother emulsions, for which the existing data
undoubtfully point to improving their flotation effects. For this purpose, a vibration mixer was
used (Figure 13) which operated at 50 Hz frequency and 2 mm amplitude of the vibrator. The
vibrating band was perforated with conical openings, thus helping to generate small streams
perpendicular to the band.
The first part of the tests was directed towards improving the baryte ore flotation by using an
insonified solution of potassium oleate and water glass. Tests were canied out under laboratory
conditions to determine the optimal insonification time of a 50% solution of potassium oleate and
water glass. The tests were assessed according to the efficiency criterion, E% [8]; Figure 14 and
Table 14 illustrate the results obtained.

-.l.
- ~t
1 1
10 0 01
1
10 0 0 1
1 0 0 0 1 2
10 0 1
1 °1
10 0 0
1 1
1° ° 0
1 1
LO_.!!.. _oJ

~ 500 ~I

I.. 700
~I
Figure 13. Vibroband mixer: I, container; 2, conical openings; 3, vibrator (dimensions shown in the
figure are related to the mixer operating at the KremikovL~i plant).

For both reagents, the optimal insonification time can be taken to be 2 min, for which the
insonification of water glass leads to an efficiency rise from 52.8% up to 62.2%, whereas in
potassium oleate insonification efficiency rises from 50% up to 61 %. The differences in the zero
test (zero-time - no insonification ) are due to a difference in the feed ore samples.
The simultaneous insonification for both reagents has been studied together with five more
factors by means of statistical planning of the experiment. Optimal results were obtained for a
3 min insonification time of potassium oleate solution, 1400 glt consumption, 20 °C pulp tempe-
 403

TABLE 14. Effect of insonification time of a 5 % solution of potassium oleate and

I on baryte ore fl otatlOl"'
water glass
Insonification time BaS04 BaS04 E(%) BaS04 BaS04 E(%)
(min) content recovery content recovery
(%) (%) (%) (%)
Water glass Potassium oleate
0 61.92 89.85 52.83 56.79 84.50 50.01
1 - - - 67.61 83.46 57.60
2 69.03 86.14 62.22 68.88 85.32 61.03
3 63.00 90.34 57.03 68.32 83.41 59.17
4 64.75 89.23 57.94 - - -
rature, 11 min flotation time, 500 glt solid content, 30 glt consumption of alkyl sulphate and 4 min
insonification time of water glass. Under this regime, the BaS04 content in the concentrate was
73.7% at a 87.1 % recovery and a 66.4% efficiency. It is important to note that efficient flotation
was achieved without heating the pulp, as it is now at the existing plant.
The positive results of the laboratory experiments served as a basis for applying the new
technique under industrial conditions. The results of the industrial test (three shifts) are given in
Table 15. The industrial tests confirmed the laboratory test~.

TABLE 15. Rotation results with the use ofinsonified potassium oleate in
baryte flotation!
Products Shift results according
to BaS04 content (%)
Shift I Shift II Shift III
with insonification of reagent
Feed 45.01 21.42 41.75
Concentrate, rougher flotation 61.35 67.65 73.29
Tailings, rougher flotation 23.31 7.91 19.13
Recovery (%) 77.75 71.42 73.32
Concentrate yield, rougher flotation 57.04 22.61 41.77
Tailings yield, rougher flotation 42.96 77.39 58.23
Efficiency (%) 46.58 64.76 59.27
without insonification
Feed 45.85 34.32 43.84
Concentrate, rougher flotation 62.94 73.01 71.50
Tailings, rougher flotation 30.24 21.12 19.13
Recovery (%) 64.97 54.12 76.95
Concentrate yield, rougher flotation 47.12 25.44 47.18
Tailings yield, rougher flotation 52.88 74.56 52.82
Efficiency (%) 44.14 48.53 58.52
The expenments were performed at the Kremlkovtsl Proeessmg Plant

The technological indices obtained during the three testing shifts are higher. For approximately
the same grade of the concentrate, the recovery increases considerably for the first two shifts,
404

65 Efficiency [%)

Water
....."........ I

45~------~------~--------~------~
o 2 3 4
Time [min)

Figure 14. Influence of insonification time of 50% solutions of potassium oleate and water glass on the
baryte flotation (I, potassium oleate; 2, water glass).

whereas for the third one the results differ slightly, which is probably due to the higher quality of
feed in the test carried out without activating the oleate.
The insonification device is installed near the flotation front and does not cause any
inconvenience to the operating personnel.

4.3. FLOTATION OF NON-FERROUS METAL ORES BY USING INSONIFICATED

AQUEOUS SOLUTIONS OF FLOTATION REAGENTS

Xanthate and frothers (pine oil, AEA - mixture of higher alcohols) were also insonified and were
used for the flotation of copper and lead-zinc orcs. The results, showing the effect of an insonified
solution of xanthate on cupro-pyrite ore flotation, are given in Table 16.

TABLE 16. Effect of insonification of a xanthate aqueous solution on the concentrate

quality in cupropyrite ore flotation
No. Insonification time Content (%) Recovery (%) Efficiency (%)
(min) ClI S O! S O! S
1 0 3.75 5.79 82.95 77.10 76.52 71.23
2 3 3.48 5.90 85.58 77.20 75.86 71.05
3 5 2.81 4.90 85.93 78.58 77.12 70.66
4 7 5.76 9.93 88.59 84.69 83.34 79.88
5 10 4.42 7.58 76.13 74.14 71.23 69.52
6 15 4.96 7.94 73.64 69.04 68.86 64.69

The table shows that the optimal insonification time for both elements is 7 min, the copper
content increasing from 3.7% up to 5.8%, and the recovery - from 82.9 up to 88.6 %. With pyrite,
the content of S increases from 5.6% up to 9.9% and the recovery from 77.1 up to 84.7%. In the
lead-zinc ore flotation, it was established that the optimal insonification time of the xanthate
solution is different for the two metals (Table 17).
 405

TABLE 17. Effect of insonification time on xanthate solution on lead-zinc

ore flotation recovery (%)
Products Retention time (min)
0 3 6 9 12 15
Lead concentrate 83.92 85.99 86.84 87.47 87.40 93.31
Zinc concentrate 94.14 95.99 95.91 94.88 93.36 97.04

For the lead minerals the optimal retention time proved to be 12 minutes, the lead recovery
increased by 3.6% and the grade by 2.2%. In zinc flotation the highest indices were obtained for a
15 min retention, the recovery increased by nearly 3% and the grade was improved by 0.1 %.
The reason for improved flotation of various minerals turned out to be the oxidation of part of
the xanthate to dixanthogen. This was determined by using different methods, e.g. ultraviolet and
infrared spectroscopy, electron-paramagnetic resonance studies, etc. The dixanthogen content of
the solution, for various retention times, is given in Table 18.
It was found that the higher flotation indices are due to the higher recovery of fine classes.
Besides,the flotation kinetics were also improved, the recovery after the first 2 min being 66.6%
(when xanthate was insonified), whereas with the untreated reagent the recovery was 46.5% [9].

TABLE 18. Dixanthogen content in xanthate solutions for various

retention times.
Retention time (min) Dixanthogen content (%) Extinction (241 nm)
0 1.04 0.141
5 2.57 0.348
7 2.94 0.398
10 3.51 0.475
15 3.29 0.445

The effect of sound treatment of frother solutions (pine oil and AEA) is also favourable. The
optimal insonification time in these cases ranges from 3 to 5 min. In the case of pine oil the 5 min
treatment leads to an increase of copper content from 3.5% up to 3.6% and the recovery from
82.7% up to 83.2% [10]. The effect on pyrite recovery is also positive, the increase of the sulphur
content and its recovery varying within the same range as with copper.
The insonification of the AEA frother affects considerably the grade of the concentrate and the
recovery of copper and sulphur for a 3 min treatment (Figure 15); the effect of a treated solution on
the recovery of molybdenum and silver is also favourable.

5. Conclusions

The investigations carried out with the originally designed vibration devices established their
abilities for positive vibroacoustic impacts on various elements of the flotation process. The results
obtained allow the drawing of the following basic conclusions:
- the joint running of the processes of grinding and flotation in a vibration machine leads to
considerable improvement of the technological and economic parameters. The good results come
as a consequence of the formation of less slime, improved air dispersion, regular distribution of
reagents and cleaning of the bubble surfaces from undesired impurities. These positive effects
406

Grade [%] Recovery [%]100

_ _-±:------4--:..:ReCOvery Ou 90

80
Recovery 8
10
Grade Cu
70
8
6
4
2
OL---~--~--~--L---~--~--~

o 2 3 4 t5 e 7
Insonlflcation time [min]

Figure 15. Influence of insonification time of emulsion with AEA-80B on copper-pyrite flotation.

are due not only to the collisions of the grinding bodies but also to the acoustic emission
gencrated in the colissions and propagating in the medium;
- there are vibration regimes which intensify the process of secondary concentration in the froth
layer and promote the efficient desorption of the reagents from the particle surfaces in the bulk
concentrate. These regimes usually require low energy consumption for vibration. The higher
energy consumption is harmful, since the very intensive vibrations destroy the froth or regrind
the bulk concentrate.
- the vibrovalve flotation cell designed allows fine, regular dispersion of the air in bubbles,
thereby making the efficient flotation of finely ground (even slime) material possible. The
acoustic emission inside the flotation cell facilitates the attachment of the slime particles to the
bubbles. The autopulsation flotation cell, because of the character of turbulent flow, is more
suitable for flotation of coarsely-ground material.
- with the conventional pneumo-mechanical and pneumatic machines, the results can be improved
provided the air, before dispersing in bubbles, is subjected to a "sound treatment". The vibrating
bubble surfaces obtained with such air are freed of the particles of the undesired component, in
this way obtaining higher grade concentrates with increased recovery;
- the implemented vibroacoustic emulsification and insonification of the reagent solutions have
shown that these processes can be readily adopted in industrial practice and can improve the
flotation of coal, industrial minerals and orcs of ferrous and non-ferrous metals. The elaborated
vibroband mixer-insonifier can replace successfully the existing agitators.
- the maChine-building industry has already adopted the manufacture of various reliable vibrators
and vibration machines. The vibroacoustic devices described in this paper arc simple in design
and may be recommended for implementation. Parallel to the practical application, more detailed
theoretical studies on processes causing vibroacoustic effects on flotation systems should be
conducted in future.
 407

6. Acknowledgements

The developments described here were carried out in the laboratory of "Vibroacoustical
Technologies" at the Higher Institute of Mining and Geology. The laboratory was founded with the
very helpful moral and financial support of Prof. N. Videnov. We express our great thanks for
that help.

7. References

1. Stoev, St. (1979) Vibroacoustic Technique for Mineral Processing, Technika Publ. House,
Sofia.
2. Glembotskii, V., and Klassen, V. (1973) Flotation, Nedra, Moscow.
3. Georgiev, R., and Stoev, St. (1973) Possibilities for improving the quality of flotation
concentrates with vibrations, NIITHERMET, 1, pp. 123-131.
4. Stoev, St. A vibrodisperser - A mixer for flotation machines, Bulg. Patent No. 23266.
5. Stoev, St. Pneumohydraulic vibrator, Bulg. Patent No. 20653.
6. Metodiev, M., and Stoev, St. (1978) Effect of liquid phase vibrations on sedimentation of
fine mineral grains, Annals Higher Institute of Mining & Geology (Sofia) 2.2(IV), pp. 77-84.
7. Shoupov, L. (1972) Applied Mathematical Methods in Mineral Processing, Nedra, Moscow.
8. Pavlov itch, V., Fomenko, T., and Porartcheva, L. (1966) Determination of Results from
Coal Preparation, Nedra, Moscow.
9. Djendova, Sht., and Mekbandjiski, V. (1985) Copper ore flotation by insonifjed xanthate
solution, Annals Higher Institute of Mining & Geology (Sofia) 1, pp. 229-237.
10. Djendova, Sht., and Mekbandjinski, V. (I 986) New concepts for vibroacoustic treatment of
xanthate solutions, Proceedings ISBA (Bhubaneswar, India).
3.5. THE INSTITUTE OF GEOLOGY AND MINERAL EXPLORATION (IGME)
OF GREECE AND ITS RESEARCH ACTIVITIES ON FLOTATION

D. MALLIARIS
Mineral Processing Department
Institute o/Geology arui Mineral Exploration
GR-1l5 27 Athens
Greece

ABSTRACT. The Institute ,of Geology and Mineral Exploration is a research institute in the
geological, hydrogeological, geotechnical and mining field aiming at higher rates of development of
Greece in these fields. The task of the Mineral Processing Department is the beneficiation study of
Greek minerals and ores. Its activity covers a wide range of ore beneficiation, either in laboratory
or in pilot plant scale equipment; results are recorded in mostly non-confidential publications.
Scientific collaboration with other countries is encouraged.

1. Institute of Geology and Mineral Exploration (IGME)

IGME was established 40 years ago. The main task of the Institute is the geological study of the
country and the research of the underground wealth (minerals and underground water). The target
of the Institute is the achievement of higher rates of development of Greece in the field of minerals
and orcs and, also, in the field of geosciences. The Ministry of Industry, Energy and Technology
finances its activities and controls its operation.
The main fields of research are: the knowledge of the geological structure of Greece, the
research and economic evaluation of minerals and ore deposits, the research and study of
underground water deposits, the research and study of energy raw materials, the geotechnical
research and studies, and the study of environmental geology problems.
Its headquarters is situated in Athens, with branches in six big cities of Greece. IGME is divided
into Departments, according to the object of the research: Geology, Mineral Processing,
Geophysics, Geochemistry, Ore Deposits, Mineralogy, Energy raw materials, Mining Research,
Feasibility studies, Chemistry, Hydrogeology, Engineering Geology, and a few others more.

2. The Mineral Processing Department

The task of the Mineral Processing Department is the beneficiation study of Greek minerals and
ores, applying conventional or advanced technological methods. These methods are: physical,
physico-chemical or chemical separation and extractive metallurgy. The research usually takes place
in laboratory-scale equipment but in several cases, tests have also been performed in pilot-plant

409
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 409-417.
© 1992 Kluwer Academic Publishers.
410

installations. By its research activity, the Department contributes to :

- the essential increase of the recoverable reserves of the deposits under research,
- the economic improvement in the exploitation of certain deposits,
- the establislunent of criteria for the possibility of potential economic exploitation of minerals and
ores, and
- the general orientation of the mining research.
The Department is divided into three sections: the metallic ore section, the mineral and industrial
ores section and the hydrometallurgy section, with an ore crushing subsection.

3. The Equipment of the Department

The Mineral Processing Department laboratory has at its disposition the following instruments and
mechanical equipment :
ore preparation laboratories: crushing and grinding machines, heavy liquids laboratory, size
analysis laboratory (cyclones, cyclosizers, screens, ultrasonic screens etc.), liberation measure-
ment and representative sampling equipment).
ore separation laboratories : gravimetric separation machines (jigs, shaking tables, duplex tables,
spirals, micropanners etc.), electromagnetic separation machines (Eriez, F.I.S. etc.), physico-
chemical separation machines (flotation, dispersion, flocculation etc.), hydrometallurgical
separation machines (leaching apparatus, ion exchange, solvent extraction etc.), desliming,
sedimentation, cycloning, dewatering apparatus.
analytical equipment : z-potential measuring instrument, chemical analysis instrument
(SYRANO), instrumentation measuring viscosity, whiteness, plasticity, porous index, wetness,
ion exchange behaviour, dilatometer behaviour.
specific laboratories for particular minerals: perlite laboratory (preheating, expansion, com-
pression strength), clay minerals laboratory, gold extraction and estimation laboratory.
pilot plants: stationary and mobile pilot plants for the processing of alluvial ores, stationary pilot
plant for hydrometallurgical application methods, flotation pilot plant, dense media separator,
cross belt high intensity magnetic separator.

4. Scientific Collaborations of the Department

Collaborations have been established between IGME and foreign institutes and universities, that
resulted in :
- funding: joint research projects financed by the E.E.C.,
- bilateral cooperation: these collaborations belong to the category of cooperation between the two
countries,
- institutional collaboration between IGME and foreign organisation (either universities or
institutes or mineral processing companies).
Examples of such collaborations are:
a. between Greece and Bulgaria, in the field of processing and estimation of : industrial ores,
e.g. perlite, kaolin, silica sands, clays etc. (in the period 1980-1988), mixed sulphide ores
(1985-87), stibnite (1985-87), bacterial leaching (1983-89).
b. between Greece and Czechoslovakia, in the fields of: industrial ores (1984-86), non metallic
raw materials (1989), hydro metallurgical processing of mixed sulphide ores and scheelite
 411

(1985-87).
c. between IGME and I.T.M. - C.N.R. (Italy), in 1980, 1984 and 1990.
d. between IGME and K.B.F.I. (Hungary) in 1989.
e. between IGME and 3 institutes of the USSR: the Institute of Mineralogy, Geochemistry and
Crystallochemistry of Rare Elements of Moscow, VITR-VPO of Leningrad, and TSNIGRI
(Central Scientific Research Geological Prospecting Institute) in Moscow (1987).
f. between Greece and France, in the field of uranium (in 1984), and with B.R.G.M., in
particular (1987, 1989, and 1990).
g. between Greece and Belgium, with the Free University of Brussels (in 1980).

5. Research at the Mineral Processing Department

5.1. LABORATORY

The research activity covers a wide range of minerals and industrial ores of Greece, such as :
i. metal ores:
gravimetric and hydrometallurgical processing of alluvial gold,
hydrometallurgiCal and bio-hydrometallurgical processing of primary gold,
flotation and hydrometallurgical processing of Pb-Zn mixed sulphide ores,
gravimetric, bio-hydrometallurgical and leaching processing of manganese ores,
gravimetric, electromagnetic and flotation processing of chromite ores,
flotation processing of stibnite,
gravimetric and magnetic processing of iron ores,
flotation processing of copper ores,
gravimetric processing of scheelite,
hydrometallurgical processing of uranium and rare earths.
ii. industrial ores:
gravimetric,magnetic and hydrometallurgical processing of feldspars and silica sand,
clay minerals for use in the ceramic industry,
gravimetric and magnetic processing of dounite,
perlite processing,
gravimetric and hydrometallurgical processing of phosphates,
magnetic processing of talc,
gravimetric and hydrometallurgical processing of kaolin,
processing of beach sands, graphite, gypsus, white calcite, diatomite, cyanite, bentonite,
hydrometallurgical processing of lignite.

5.2. PILOTPLANT

In the past, the Mineral Processing Department was essentially a laboratory department, but it has
been increasingly involved in the scaling-up of the results originally obtained. In the last decade,
six different ores were tested in pilot plant installations, mostly in outdoor activity, close to the
deposit under research.

5.2.1. Alluvial gold. The Department operated 3 hydrogravimetric pilot plants for the processing,
extraction, measurement and estimation of gold alluvial deposits, one stationary pilot plant based
412

on screening-classification-sluicing-shaking tabling research procedure, and one mobile pilot plant

based on scrubbing-trommelling-coning-jigging-shaking table procedure, for the assesment of
big diameter drill holes material.

5.2.2. Phosphate ores. The Department operated a stationary gravimetric pilot plant.

5.2.3. Uranium ores. The Department operated a stationary hydrometallurgical pilot plant, based
upon leaching, ion exchange and solvent extraction procedure.

5.2.4. Manganese ores. The Department operated a pilot plant for the gravimetric preconcentration
and concentration of the ore.

5.2.5. Pb-Zn sulphide are. The Department operated a pilot plant including crushing-sampling
section and flotation section,

5.2.6. Feldspars and silica sands. The Department operated a pilot plant including classification-
scrubbing-preconcentration and magnetic separation units.

6 . Flotation in the Mineral Processing Department

The following ores have been tested by flotation techniques in the laboratory of IGME.

6.1. Pb-Zn MIXED SULPHIDE ORES

6.1.1. Essimi. From an exploratory drift in Essimi (in the Evros river area, close to the Greek-
Turkish border), the feed ore had an assay of 0.4% Pb and 2% Zn. The differential flotation
procedure was the conventional one. The recovery of both valuables was high enough, at about
90%. In spite of the success, the project was terminated because the ore body proved to be too
small [1].

6.1.2. Thassos. Another test was done with an ore sample from the Thassos island. The ore feed
was partly oxidised with a head analysis 3.3% Pb, 5% Zn and 11 ppm Ag. Tests were performed
both by differential and bulk flotation techniques (Table 1). The feasibility study, however, gave
negative results.

TABLE 1 Flotation results of the Evros ore [2]

Recovery %
PbS Zns
Differential flotation 80 65
Bulk flotation 90 85

6.1.3. Polykastro. A third ore sample, from the Polykastro area exploratory drift, was also tested
(Table 2). Unfortunately, the feasibility study resulted again in non-economical numbers.

6.1.4. Molar. A fourth ore sample, from an exploratory drift in the Molar area, in Southern Greece
 413

TABLE 2 Flotation results of the Polykastro ore [3]

Grade Recovery
Pb% Zn% M,'i!/t Cd ppm Cu% %
Feed 2.0 4.4 9 276 0.15 -
PbS concentrate 55.0 65 78
Zns concentrate 54.0 0.4 77

was also tested (Table 3). This project was carried out by IGME (Greece) in collaboration with
ITM-CNR (Italy) and BRGM (France). Because of the fine dissemination of the ore, the relatively
high oxidation of lead and zinc minerals (0.7% Pb and 1.2% Zn) and the ore's low grade in
anything else except Zn, bulk flotation was considered as appropriate for its beneficiation.

TABLE 3 Flotation results of the Molal ore [4]

Pb% Zn% Agppm Cd ppm
Feed 1.1 6.8 28 400
Concentrate grade 4.0 48.0 170 3200
Concentrate recovery % 82.0 53.0 60 80

The laboratory results were duplicated in pilot plant scale. Once again, the ore body was
estimated too small for industrial exploitation.

6.2. BARYTE

6.2.1. Milos. A baryte ore sample from the island of Milos was tested. A successful degree of
liberation was obtained at 200 mesh. Anionic flotation was applied, in pH 10, using EDT A or
sodium silicate. The floatability of baryte was found to be adversely affected by the presence of
soluble salts (cations) in the pulp, because of:
- flocculation of the pulp particles
- undesirable activation of the gangue, and
- wasteful consumption of the collector.

TABLE 4 Flotation results of the Milos baryte ore [5]

BaS04 % Si02% specific gravity
Feed 82.0 14 4.0
Concentrate grade 95.5 - -
Concentrate recovery 96.0 - -

6.2.2, Mykonos. Another baryte ore sample from the island of Mykonos was tested. A successful
degree of liberation was obtained at dgo% -0.075 mm (Table 5).

TABLE 5. Flotation results of the Mykonos baryte ore [6].

BaS04 % Si02% specific gravity
Feed 60.0 33 3.1
Concentrate grade 93.4
Concnentrate recovery 92.3 4.2
414

The method applied was an anionic flotation in alkaline media after activation of baryte by
Pb(N03h-

6.3. STffiNI1E

The main metallic minerals, contained in the sample from the Lachana area, were stibnite, pyrite,
marcasite and arsenopyrite. The gangue mineral was silicate. The method applied was gravimetric
preconcentration and anionic flotation.

TABLE 6 Rotation results of the Lachana stibnite ore [6,7)

Sb% As%
Feed 0.5-2.7 0.01-0.1
Concentrate grade 50 0.80-1.3
Concentrate recovery 90 -

6.4. COPPER

A copper ore sample from the Chalkidiki area was tested. The method applied was anionic flotation
in alkaline environment (PH= 11).

TABLE 7 Rotation results of the Chalkidiki ore (8)

Cu% Aug/t Agglt
Feed 0.6 1.7 2.4
Concentrate grade 23.0 68.0 23.0
Concentrate recovery [%) 90.0 98.0 24.0

6.5. INDUSTRIAL MINERALS

6.5.1. Granite. A sample of granite, from the Sithonia area of Northern Greece, containing
feldspar and silica sand, was tested. The method applied was a combination of magnetic separation
and flotation.

TABLE 8 Rotation results of the Sithonia ore (9)

Fe203 Si02 K20+ Na20
Feed % 1.0 72.0 6.0
Concentrate grade % 0.11 80.0 8.0

6.5.2. Quartzitic sandstone. A sample from the Argos Oresticon area was tested. The method
applied was, once again, a combination of magnetic separation and flotation.

TABLE 9 R otatlon resu ts 0 fth e Argos orestlcon ore (10)

Fe203 % Si02%
Feed 1.0 85.0
Concentrate grade 0.05-0.2 > 95.0
 415

6.6. CHROMITEORES

Cationic flotation gave poor results. Anionic flotation was however successful, using fatty acids in
alkaline media without preliminary desliming and after dispersion-selective flocculation of the
gangue:
a. From the Kerasitsa exploratory drift. From the mineralogical point of view, the ore consisted
of 11-14% chromite, 60-70% serpentine and 10-20% olivine. The liberation degree was
found to be at d95% -0.125 mm 80% (Table 10).

TABLE 10 Flotation results of the Kerasitsa ore [11]

Cr20J % Fe20J % SiD2% Cr/Fe %
Feed % 10.0 7.0 30.0 -
2nd cleaner concentrate grade 35.0 80.0 - <3
RouEher tail grade 3.5 - - -

b. A preconcentrate from the southern part of the deposit, which is now under exploitation.
From the mineralogical point of view, the ore consisted of 40% chromite, 55% serpentine and
5% olivine. The same procedure as previously described was followed, because cationic
flotation proved again to be unsuccessful. No scavenging flotation stage was applied. So, the
final results are even better (Table 11).

TABLE 11. FlotatIOn results 0 f the preconcentrate ore [12).

Cr203 % Fe20 3 % Si02 %
Feed 32 11 15
Rou~er concentrate grade 55 - -
Rougher tail grade 6 - -
Rougher concentrate recovery 93 - -
Rougher tail recovery 7 - -

c. From the so called "south area" deposit. From the mineralogy point ofvicw thc ore contained
19-22% chromite, 65-72% serpentine and less than 5% olivine. A 77% liberation degree was
achieved at d95% -0.115 mm (Table 12).

TABLE 12 Flotation results of the "south area" ore [13]

Cr203 % Fe203 % Si02 %
Feed 17 8 25
Rougher concentrate grade 45 - 7
Rougher tail grade 2-5 - -
Rougher concentrate recovery 90 - -
1st cleaner concentrate grade 48-52 - 6

d. From the Xerolivado area deposit. The ore feed contained 25-30% chromite, 60-68%
serpentine and 5-10% olivine. A 78% liberation degree was achieved at d95% -0.17 mm
(Table 13).
416

TABLE 13 Flotation results of the Xerolivado area ore [14].

Cr20:3 % Fe203 % Si02%
Feed 22 9 22.0
Rougher concentrate grade 48-49 - 7.5
Rougher tail grade 3-S - -
Rougher concentrate recovery 91 - -
1st cleaner concentrate grade >SO - < S.O

The conclusions of the research worl< on chromite flotation are :

- cationic flotation is not successful for Greek ores;
- olivine floats together with chromite. So, anionic flotation gives excellent results for chromites
with serpentine gangue, but medium quality results in chromites contained more than IS%
olivine;
- there is a strong relation between the free cations of the pulp and the results of the anionic
flotation, due to the gangue activation. These free cations may arise from the mineralogy of
chromium spinnels or gangue and from the water used [IS].

7. Publications

All the research work of the scientists and engineers of the Mineral Processing Department is
recorded in two categories of publications whilst are published internally and are available for
purchase:
- an integrated beneficiation study of a specific ore from a metalliferous area (in Greek language
with English summary)
short technical notes (Greek only).
There have been more than SOO such publications during the 33 years since the Mineral
Processing Department was formed. Most of the publications are non-confidential.

8. Acknowledgements

The author would like to express his gratitude for the advice and support provided by the General
Director ofIGME, Dr. V. Andronopoulos and by the Head of the Mineral Processing Department,
Ms. M. Grossou. He would like also to thank Dr. M. Patronis for his cooperation.

9. References

1. Grossou, M., Bassios, D.G., and Dimou, E.K. (1981) Beneficiation study of Pb-Zn mixed
sulphide ore from Essimi area, N. Greece, IGME (Athens), Metallurgical Researches, No. 30
(in Greek language with English summary).
2. Adam, K.D., and Malliaris, D.G. (1986) Beneficiation study of Pb-Zn mixed sulphide ore
from Marlou area, Island of Thassos, N. Greece (IGME, Athens), Metallurgical Researches,
No. 39 (in Greek language with English summary).
3. Grossou, M.G., Charalabides, P.K., and Malliaris, D.G. (1984) Beneficiation study of Pb-
 417

Zn mixed sulphide ore from Polykastro area, N. Greece (IGME, Athens), Metallurgical
Researches, No. 35 (in Greek language with English summary).
4. Patronis, M.E., Grossou, M.G., Marabini, A.A., Alesse, V., Huang, P.e., Prevosteau, 1.-
M., El Kaliobi, F., and Gateau, e. (1991) A process design study for afinely disseminated
partially oxidized complex sulphide ore from the Molar area, S. Greece. Synthesis report for
E.e. contract No. MAIM-0051-C(AM) (to be published by E.E.C.).
5. Tsailas, D.P. (1996) Study on the mechanism of the action and of the contribution of sodium
silicate during the flotation of baryte, (IGME, Athens), Geological and Geophysical research,
vol. XI(1), (in Greek language with English summary).
6. Grossou, M.G .• and Charalabides. P.K. (1984) Beneficiation study of stibnite ore from the
Lachana area, N. Greece (IGME, Athens. in Greek language).
7. Malliaris. D.G .. and Monachos, 1.D. (1978) Beneficiation study of stibnite ore from the
Lachana area, N. Greece, by flotation. NTUA. Athens (in Greek language).
8. Grossou. M.G. (1976) The process design study for a low grade porphyry copper ore from
the Chalkidiki area, Imperial College of Science and Technology. London.
9. Chalkiopoulou. P.O .• and Kaklamanis, N.D. (1988) Beneficiation of feldspar from the
Sithoma area by magnetic separation andflotation (lGME. Athens. in Greek language).
10. Chalkiopoulou. P.O .• Kaklamanis. N.D .• and Boudolos. V.G. (1988) Beneficiation of
quartzitic sandstone from the Argos Orestiko area (IGME. Athens. in Greek language).
11. Malliaris. D.G. (1985) Contribution to the flotation processing of disseminated chromite ores,
Part 1 (NTUA-IGME. Athens, in Greek language).
12. Malliaris. D.G. (1986) Contribution to the flotation processing of disseminated chromite ores,
Part 2 (NTUA-IGME. Athens, in Greek language).
13. Malliaris, D.G. (1986) Contribution to the flotation processing of disseminated chromite ores.
Part 3 (NTUA-IGME, Athens, in Greek language).
14. Malliaris. D.G. (1987) Contribution to the flotation processing of disseminated chromite ores.
Part 4 (NTUA-IGME, Athens, in Greek language).
IS. Malliaris, D.G. (1991) Beneficiation study of disseminated chromite ores by flotation
(NTUA-IGME. in Greek language with English summary; to be published).
16. Malliaris. D.G. (I979) A process design study for the recovery of Barium Sulphate from a
baryte are, Island of Milos, Greece, Imperial College of Science and Technology, London.
3.6. THE INFLUENCE OF MULTIVALENT CATIONS ON THE
FLOATABILITY OF SCHEELITE, FLUORITE AND CALCITE

S.RAATZ
Research Institute of Mineral Processing
Freiberg
Germany

ABSTRACT. The physicochemical properties of the mineral interfaces of CaW04, CaF2 and
CaC03 in the presence of multivalent cations (Ca2+, Mg2+, AP+) during the flotation process were
investigated. Rotation experiments, zeta potential and ESCA measurements were made for the
characterisation of the mineral surfaces and the utilisation of different kinds of zeolites in the
flotation systems was examined.

1. Introduction

The modification of the physicochemical properties of solid interfaces and the absorption on them
of surface-active agents is an important step in many beneficiation methods for raw materials and
intermediate products, as well as in modern environmental technology. The present work
investigated the flotation of calcium-containing minerals and primarily tried to explain the complex
mechanism of adsorption at the solid interfaces, which depends on the ion composition in the
flotation pulp. New ways of controlling the flotation of CaF2 and CaW04 minerals with high
CaC03 content were also examined.

2. Experimental

2.1. MINERALS

Pure scheelite, calcite and fluorite (purity> 98%) were used in the tests. The particle size range
was 63-100 ~m.

2.2. FLOTATION

Rotation tests were carried out in a modified Hallimond tube. Mineral particles were suspended in
a beaker and, after conditioning with the respective ions and the collector (sodium oleate 5x 10-5
mol/l, pH = 9), were transferred into the Hallimond tube and then floated.

419
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 419-425.
© 1992 Kluwer Academic Publishers.
420

2.3. ZETA P01ENTIAL MEASUREMENTS

The zeta potential was detennined in a stream-potential measuring device by means of a vibrating-
reed electrometer VA-752 (YEB Vakutronic, Dresden).

2.4. ESCA MEASUREMENTS

The Electron Spectroscopy Chemical Analysis (ESCA) measurements were carried out by means of
a photoelectron spectrometer (type ESCA-3, Vakuum Generators). The X-ray sources used are
AlKa (1486.6 eV) radiation.

2.5. ZEOLITES

Both synthetic and natural zeolites were used. The synthetic ones were: one called SASIL (of
Henkel KGaA) and one called ZEOSORB (CK Bilterfeld). Their partiele size was < 10 !-lm and
their specific surface (by B.E.T.) was found to be 2.28 m 2jg. The Si02fA1203 ratio of the zeolites
was 2.0 - 2.5.
The natural zeolites were from different deposits (Zl from Cuba and Z2 from Hungary), with
varying chemical composition. The particle size in this case was < 40 !-lffi and their specific surface
was 13.71 m 2jg and 22.88 m2jg respectively.

3. Results and Discussion

It has been established that flotation selectivity and recovery are strongly influenced by the presence
of multivalent ions such as calcium and aluminium ions. The addition of calcium ions leads to a
decrease in the recovery of scheelite and fluorite (Figure 1). The influence of Ca2+ ions on the
floatability of calcite is smaller. Mg2+ ions show a similar effect.

100 Recovery R [%]

80

20 .............. ,." .... " ... ,.

05W04

0
0 20 40 60 80 100
Oe concentration [mg/l]

Figure 1. Recovery as a function of calcium concentration.

 421

An explanation for the reduction of mass recovery is, on the one hand, the modification of
surface properties by chemical reactions and, on the other hand, the precipitation of carboxylates.
In this way, the collector - sodium oleate is usually used as collector [1] - is removed from the
system and its concentration is thus reduced.

100 Recovery R [%]

80

60

40 ~.,.m. . . OaW04
20 ...G...::c:
......"'"
..... ,."",.--_"""9'~~=====::::~

~~t:~::::~~~0a::F=2======::~~J
oLCa003
o 20 40 60 80 100
AI oonoentratlon [mg/l]

Figure 2. Recovery as a function of aluminium concentration.

The effect of Al ions on the floatability of CaF2, CaW04 and CaC03 was determined. Alumi·
nium ions in the pulp had a marked effect on the flotation of calcium-containing minerals (Figure
2); the recovery decreased rapidly to about 15%. It is assumed that the depressing effect is caused
by the adsorption of their hydroxo complexes in solution, the formation of an AI(OH)3 coating on
the mineral surface and the adsorption of oleic acid by the Al(OHh precipitates [2]. The adsorption
layers prevent the adsorption of oleic acid or cause the collector already adsorbed to be displaced
from the mineral surface.
In order to prove these assumptions, the influence of cations on the adsorption process on the
mineral surface was investigated by zeta potential and ESCA measurements. Using zeta potential
measurements (stream potential), it is possible to determine the electrokinetic properties of the
minerals [3].
In the pH range 8-10, which is of interest in these investigations, CaF2 and CaW04 have a
negative zeta potential, while CaC03 has a negative one in the pH range 10-12 (Figure 3). The
corresponding zeta potential values could be attributed to the adsorption and reaction of OH-,
C032- and HC03- ions on the mineral surfaces, which may finally lead to the formation of
Ca(OHh and CaC03, respectively, on these surfaces.
While the adsorption of protons and hydroxyl ions plays an important role for the zeta potential
of fluorite, the zeta potential behaviour of calcite is characterized by the preferential dissolution of
calcium in the presence of dissolved C02. By adding calcium ions, the surface charge of CaF2,
CaW04 and CaC03 is changed and it is possible that the zeta potential becomes positive.
Figure 4 shows the variation of the zeta potential values of scheelite in the presence of various
calcium ion concentrations as an example. The positive effect is caused by the high concentrations
of Ca 2+ ions on the mineral surface, which was previously negatively charged. These ions are
adsorbed at the mineral/solution interface and the measured zeta potential is changed to positive.
422

40 Zeta potential [mV)

D

130
20
10
Or---------------~------~------~

-10
-20
-30~~~--~~~--~~~--~~~~

123 4 5 e 7 8 9 ID n w ~
pH

Figure 3. Zeta potential as a function of pH.

10 Zeta potential [mY)

5 ........................................................"..........................- .... ".... - .... ".................. - ..........- ..

Or--4~~~~----------~~----~~2~

-5 ......S'...

-10
-15

D
-25~~--~~--~--~~--~~~~~
123 4 5 e 7 8 Q ID n
pH

Figure 4. Zeta potential as a function of calcium concentration and pH of CaW04. Calcium concentra. .
tions: I, IxlO..3 M; 2, lxlO..4 M; 3, lxl0..6 M; 4, no calcium ions added (orig.).

Figure 5 shows the influence of AI ions on the zeta potential of scheelite. Their positive effect on
the zeta potential is smaller than the effect of the bivalent Ca ions in the considered pH range. This
could be attributed to the formation of complex hydroxyl compounds as a function of pH. These
compounds are adsorbed on surfaces. The Al ions are also present as uncharged AI(OHh or so
called slime coating on the mineral surface.
Another method to determine the qualitative and quantitative chemical interaction of the minerals
in the presence of high Ca, Mg and Al ion concentrations is the Electron Spectroscopy for
Chemical Analysis (ESCA) [4]. The main feature of the ESCA technique is the possibility to obtain
information about the chemical state of the compounds in the surface layer, concerning for example
the degree of oxidation or the presence of charged groups. In mixtures, the minerals interact with
each other through the solution and that which is least soluble under the prevailing conditions tends
 423

to deposit on the surface of the other.

Depending on the composition of the solution, it is interesting to determine to which extent
chemical reactions take place on the mineral surface. In Table lone may see the change in chemical
composition of the mineral surface after treatment with solutions containing Ca, Mg and AI ions.

20 Zeta potential [mY)

10 .............................

-10 ..............•.................
orlQ.
-20 .

-30~~--~--L-~---L--~~---L--~~
1 2 3 456 7 8 9 10 11
pH

Figure 5. Zeta potential as a function of aluminium concentration and pH of CaW04. Calcium concen·
trations: I, IxlO- 3 M; 2, IxIO- 4 M; 3, IxlO-6 M; 4, no calcium ions added (orig.).

TABLE 1. Determination of minerals composition (%) using ESCA measurements

(S : standard ; Ca, M19, AI Add··
IUonal caUons
. ).
CaW04Probe
probe
CaW04S
Ca
9.5
Mg
-
AI
- 39.5 °
CaW04+Ca 9.1 - - 37.1
CaW04+Mg 6.S 2.8 - 34.1
CaW04+AI I.S - 8.9 S1.7

CaC03 probe
probe
CaC03S
Ca
11.1
Mg
-
AI
- °
49.2
CaC03+Mg 9.6 2.S - 4S.4
CaC03 +AI 1.7 - 9.9 S2.0

CaF2 probe
probe
CaF2S
CaF2 +Ca
Ca
IS.7
16.2
Mg
-
-
AI
-
-
°
11.7
9.5
CaF2+Mg 12.7 S.2 - 10.8
CaF2+ AI 13.3 - 1.9 10.1
424

CaC03 and CaW04 have a great effect on the chemical composition of alumina. In agreement
with the values for the bond energy, for instance for the AI-2p energy level (Table 2), it can be
shown that there are, on the one hand, adsorption of Al hydroxy-complexes present in the solution
and formation of Al(OH)3 coatings and on the other hand cation exchange on the mineral surfaces.

TABLE 2. Bond energy 0 f the AI-2p energy level *

I II
probe eV eV
CaW04 74.5 37% 74.6 63%
CaC03 73.8 60% 74.8 40%
CaF2 73.8 38% 75.1 62%
* Bond energy accordmg to ref. [4] : AI, 73.2; Alox' 75.6; A1203, 74.5; AI2(W04)3,
74.5 eV.

On CaF2, surface ion exchange takes place in a great extent in the presence of Mg ions. A Ca ion
present at the interface is replaced by a Mg ion and MgF2 is formed, for instance. It can be said that
in evaluating the results obtained by ESCA measurements, it is possible to determine which
compounds are built at the mineral surfaces under given conditions.

Recovery R [%] Recovery R [%1

100~~~~~-------------. 100~~~~~-------------.

80

ZEOSORB

20 ..... _......_.................. _................................... -._ ......

o~~--~~--~~--~~~ OL-~--~--~--~--~--~~

o 2 4 6 8 ro ~ u ~ o 2 4 6 8 ro ~ U
m [g] m [gJ

(a) (b)

Figure 6. Influence of SASIL, ZEOSORB, Zl and Zz zeolites on the recovery of scheelite in the presence
of: (a) Ca ions (2.5xlO- 3 M), (b) Al ions (lxlO· 3 M).
 425

Now there is a question: How is it possible to reduce the concentration of the multivalent ions in
the flotation pulp and therefore reduce their negative influence on recovery? This is possible with
insoluble inorganic ion-exchangers, such as zeolites [5]. These zeolites replace the multivalent ions
such as Ca and Al ions by the indifferent Na ions in the flotation pulp. It is assumed that the
application of zeolites in comparison with complex-fonning substances or other regulating reagents
brings advantages since the disturbing ions can be removed without changing the surface
characteristics of the ore to be floated. Aluminium silicates of the type A zeolite are distinguished
by a high exchange capacity especially compared to calcium ions. An ion exchange also takes place
with heavy metal ions as we found out.
Two different kinds of synthetic (ZEOSORB and SASIL), as well as natural zeolites (Z! and Z2)
were examined in order to detennine to what extent these conclusions can be applied to the flotation
of CaF2, CaW04 and CaC03. The flotation of pure scheelite ore is shown as an example in
Figure 6. The mineral was floated in the presence of a specified Ca ion concentration, with
addition of different zeolites. This led to an increase of recovery of scheelite from 30% to about
87% by adding SASIL and to 75% using Z!. The floatability of scheelite is also improved
considerably in presence of Al ions by adding zeolites. Thus the recovery of scheelite from about
15% increased to about 65% by adding ZEOSORB and to 58% by adding Z!.
Using zeolites, polyvalent cations can be removed from the flotation system as expected. This
is, in principle, a possibility for the control of polyvalent cations concentration in flotation systems.

4. References

1. Finkelstein, N.P. (1989) Review of interactions in flotation of sparingly soluble calcium

minerals with anionic collectors, Trans. IMM ru.
157-177.
2. Rakotonarivo, E., Bottero, 1.Y., and Cases, 1.M. (1983) Study of the adsorption of long
chain alkylsulfonates from aqueous solutions on aluminium hydroxide gels, ColI. & Surf. 2,
273-292.
3. Steiner, H.l. (1965) Elektrokinetische messungen in rahmen der flotationsforschung, Radex-
Rundschau fl, 733-758.
4. Wagner, C.D., Riggs, W.M., Davis, L.E., Moulder, J.F., and Muilenberg, G. E. (1979)
Handbook of X-ray Photoelectron spectroscopy, Perkin-Elmer Corporation.
5. Rybinski, W. von, Schwuger, M.I., Schulz, P.A., and Dobias, B. (1986) Einsatz von
Zeolith A fur erhohte Selektivitat und Ausbringung bei Flotationsprozessen, Schrifienreihe der
GDMB, Heft 48, pp. 69-92.
3.7. NON-CYANIDE TECHNOLOGY FOR THE FLOTATION OF
LEAD-ZINC ORES

V.PANAYOTOV
University of Mining & Geology
Darvenitsa, Sofia 1156
Bulgaria

EXTENDED SUMMARY. One of the most important initial studies in mineral processing is the
measure of floatability in order to determine the appropriate chemical reagents and also to
investigate the surface properties of the particles.
In the case of sulphide ores, the estimation of the type of the charge carrier (of n- or p- type) may
be useful. The latter is accomplished by measuring the contact potential difference, the thermo-
electromotive force and the specific electroconductivity. This information may then be processed to
obtain the energetic diagrams for each mineral. In this way, one is able to estimate the floatability of
the mineral and the appropriate treatment regime; the approximate concentration of collectors and
reagents for the real process may also be estimated from these diagrams.
The electrochemical treatment of mineral particles directly in the flotation pulp permits the
depression and/or activation of minerals without the use of collectors [1]; the apparatus constructed
is presented in the following picture. Results from an industrial-scale application of this treatment,
in the plant of Gorubso in Bulgaria, showed that this technology increased Ag recovery by 2%, Au
recovery by 1-2 g!ton in the Pb concentrate and copper recovery by approximately 2%. Recovery
of lead was approximately 92% and of zinc 84-85%.

REFERENCE

1. Panayotov, V. (1991) Design of a non-cyanide technology for floating lead-zinc ores. Energy
prerequisites, implementation, results, in Proceedings First Eur. Metals Conf. "Non-Ferrous
Metallurgy - Present and Future", Brussels, preprint.

427
P. Mavros and K. A. Matis (eds.i, Innovations in Flotation Technology, 427-428.
© 1992 Kluwer Academic Publishers.
428

Figure 1. Modification of a mechanical flotation cell for the electrochemical treatment of the pulp
directly in the cell.
4. Water and Wastewater Treatment
4.1. FLOTATION IN WATER TREATMENT

TH.F. ZABEL
W.R.C. pic.
Henley Road. Medmenham.
P.O. Box 16.
Marlow. Bucks. SL7 2HD.
U.K.

ABSTRACf. Dissolved-air flotation is a very effective solid-liquid separation process used in

water treatment. as an alternative to sedimentation, to lower the solids concentrations before
granular filtration. This enables filters to be operated more effectively and cost efficiently to
produce a filtered water quality suitable, after disinfection, for potable supply. Dissolved-air
flotation has been specifically developed for the treatment of algal laden and coloured waters. The
flocs produced by the treatment of these waters have a low settling velocity. particularly when the
water temperature is low and even when polyelectrolytes are added. Treating these waters, flotation
requires smaller tanks and produces better treated water quality, compared to sedimentation.
Dissolved-air flotation has become accepted as an alternative to sedimentation, in particular in the
Scandinavian countries and the United Kingdom. with more than 50 plants in operation or under
construction. The largest dissolved-air flotation plant currently being constructed is the 315 Ml/day
Frankley water treatment works in the UK.

1. Introduction

The application of dissolved-air flotation in water treatment has been discussed in detail previously
[1-3]. This paper provides a summary and update ofthese studies.

2. Types of Flotation Processes

Aotation can be described as a gravity separation process, in which gas bubbles attach to solid
particles to cause the apparent density of the bubble-solid agglomerates to be less than that of the
water thereby allowing the agglomerates to float to the surface. The different methods of producing
the gas bubbles give rise to different types of flotation processes which are electrolytic flotation.
dispersed-air flotation and dissolved-air flotation [1].

2.1. ELECfROLYTIC FLOTATION

The basis of electrolytic or electro-flotation is the generation of bubbles of hydrogen and oxygen in

431
P. Mavros and K. A. Matis (eds.). Innovations in Flotation Technology. 431-454.
© 1992 Kluwer Academic Publishers.
432

a dilute aqueous solution by passing a direct current between two electrodes [4]. Electrical power is
supplied to the electrodes at a low voltage potential of 5 to 10 VDC by means of a transformer-
rectifier. The energy required for electro-flotation depends largely on the conductivity of the liquid
and the distance between the elecrodes.
The process has the disadvantage that the electrodes need periodic replacement and fouling of the
electrodes is also a frequent problem. But as the bubble formation does not create turbulence. the
system is attractive for the removal of fragile flocs.
The bubble size generated in electro-flotation is very small and the surface loading is therefore
restricted to less than 4 m/h compared with up to 12 m/h for dissolved-air flotation.
The application of electro-flotation has been restricted mainly to sludge thickening and small
effluent treatment plants (10-20 m 3/h). The process has been reported to be suitable for very small
water treatment installations [5].

2.2. DISPERSED-AIR FLOTATION

Dispersed-air flotation has been discussed in the previous chapters. Both foam and froth dispersed-
air flotation are unsuitable for water treatment applications because the bubble size tends to be large
(>1 mm. compared to 20-1oo!Jl11 for dissolved-air flotation and electro-flotation) and either high
turbulence (froth flotation) which would break up the fragile flocs formed during the chemical
pretreatment. or undesirable chemicals (foam flotation) are required to produce the air bubbles
required for flotation.

2.3. DISSOLVED-AIRFLOTATION

In dissolved-air flotation the bubbles are produced by the reduction in pressure of a water stream
saturated with air. The three main types of dissolved-air flotation are vacuum flotation. micro-
flotation and pressure flotation.

2.3.1. Vacuum flotation. In vacuum flotation the water to be treated is saturated with air at
atmospheric pressure. A vacuum is then applied to the flotation tank releasing the air as small
bubbles. The amount of air available for flotation is limited by the achievable vacuum. This process
is being applied only in the paper industry for the recovery of fibres. Vacuum flotation has three
main disadvantages: it is a batch process. it requires sophisticated equipment to maintain the
vacuum. and the amount of air available is limited by the vacuum achievable. Therefore. it is being
replaced by pressure flotation.

2.3.2. Micro-flotation. In micro-flotation [6]. the entire volume of water is subjected to increased
pressure. This is achieved by passing the water down and up a shaft approximately 10 m deep. In
the down-flow section the water is aerated and the amount of air dissolved in the water increases
because of the increase in hydrostatic pressure. As the water rises in the up-flow section. the
hydrostatic pressure decreases and some of the dissolved air is released in the form of fine air
bubbles. The amount of air available for flotation is restricted by the depth of the shaft provided.
Application of this process has been restricted so far to a few small sewage and effluent treatment
plants.

2.3.3. Pressure flotation. Pressure flotation is currently the most widely used dissolved-air
flotation process. In pressure flotation air is dissolved in water under pressure.
 433

Three basic (pressure) dissolved-air flotation processes can be used: full-flow pressure flotation,
split-flow pressure flotation, and recycle-flow pressure flotation [2].
In full-flow pressure flotation, the entire influent is pressurised and aerated. In split-flow
pressure flotation, a fraction of the influent is subjected to pressurisation, while the remainder
flows directly to the flocculation-flotation tank. This flow scheme offers a cost saving over full-
flow pressure flotation, since the saturator and the feed pump only handle a portion of the total
flow and therefore smaller units are required. However, as less air is provided in the split-flow
system, this process has to operate at a higher pressure in order to provide the same amount of air.
For water treatment applications, requiring the removal of fragile floes, both systems are
unsuitable because any floc fonned prior to the air release device would be destroyed by the high
shear experienced during the pressure release. For this application recycle-flow pressure flotation
is the most appropriate system (Figure 1).

Level

Compressor

Air
pressure Saturator
regulator

Recycle
Raw In-line
Water
level
control

Coagulation Flotation treated

chemicals Flocculator Air Flotation tank water to rapid
injection gravity fi Iter
nozzle

Figure 1. Schematic diagram of a recycle dissolved-air flotation plant for water treatment [1].

In recycle-flow pressure flotation the whole influent flows directly to the flocculation tank or to
the flotation tank if separate flocculation is not required. Part of the clarified effluent is recycled,
pressurised and saturated with air. The pressurised recycle water is introduced to the flotation tank
through a pressure release device and mixed with the flocculated water. In the pressure release
device the pressure is reduced to atmospheric pressure so releasing the air in the fonn of fine
bubbles (20 to 100 11m in diameter). The air bubbles attach themselves to the floes and the
agglomerates float to the surface. The float can be removed from the surface and the clarified water
is taken from the bottom of the flotation tank. Because of the additional flow through the flotation
tank due to the recycle, the flotation tank has to be designed larger than for the full-flow system
especially if the recycle flow required is large, as for sewage sludge thickening.
434

3. Theory of Dissolved-Air Flotation

To achieve efficient clarification by dissolved-air flotation, the impurities present in the water must
be coagulated and flocculated effectively prior to the introduction of the micro-bubbles. Hoes and
bubbles must be well mixed to allow the formation of the bubble-floe agglomerates. These subse-
quently rise to the surface where they are removed.

3.1. MECHANISM OF FLOTAnON

The bubble-floc agglomerates may be formed by any of three distinct mechanisms

- entrapment of bubbles within a condensing network of flocs
- growth of bubbles from nuclei within flocs
- attachment of bubbles to flocs during collision
Work by Kitchener and Gochin (7) has shown that all three mechanisms can occur but that the
principal mechanism in dissolved-air flotation for potable water treatment is the attachment
mechanism. Their work has also indicated that the organic content of surface waters is generally
high enough to render the floc surface sufficiently hydrophobic for bubble attachment. Only in
very pure, organic "free" systems was the flotation efficiency significantly reduced.

3.2. SOLUBILITY OF AIR IN WATER

Over the temperature and pressure ranges used in dissolved air flotation (0 to 30°C and 200 to 800
kPa) both nitrogen and oxygen obey Henry's Law:

P=Hx

where P = partial pressure of the gas in the gas phase, x = mole fraction of the gas in the liquid
phase, and H = Henry's Law constant.
In a continuous saturation system the gas phase above the water does not have the same
composition as air, since oxygen is more soluble in water than nitrogen. In order that the quantities
of oxygen and nitrogen leaving the saturator in the pressurised water are the same as those entering
it in the compressed air, the nitrogen content of the gas in the saturator rises, creating a nitrogen-
rich atmosphere. This results in a reduction of about 9% in the mass of gas that can be dissolved.
In assessing the performance of a continuously operating saturation system, the 100% saturation
level should be taken as that achievable assuming a nitrogen-rich atmosphere (Figure 2).
The solubility of gas is also a function of temperature and dissolved solids concentration.
However, the range of solids concentration encountered in potable water treatment is small and this
effect can therefore be neglected when calculating the air available for flotation.
The total amount of gas dissolved in the pressurised water can be determined by the Van Slyke
method (9). However, for evaluating the amount of air available for flotation, it is only necessary
to measure the air precipitated by the pressure release device rather than the total amount of air
dissolved. A method for determining the amount of air available for flotation and the saturator
performance has been described in ref. [1). The unit, which may be connected in parallel to the
flotation plant, consists of the same air release device as used in the flotation plant (e.g. a nozzle) to
control the amount of recycle entering the unit and for releasing the pressure; a column to separate
the precipitated air from the water; a volumetric air meter to measure the amount of air released; and
a water outlet weir to control the air/liquid interface.
 435

160

140

~ 120
E
o
w
~ 100
w
-'
w
a:
D 80
w
>
-'
0
(J)
(J)
60
0
a:
~
40
Air atmosphere in saturator
Nitrogen-rich atmosphere in saturator

20 Temperature values in °c

o 100 200 300 400 500 600 700

PRESSURE (kPa)

Figure 2. Mass of gas dissolved in water as a function of pressure and temperature [8].

3.3. EFFECf OF BUBBLE SIZE

To achieve maximum agglomeration between gas and solid phases it is essential that the gas
bubbles rise under laminar flow conditions. This avoids shedding of the floes as can occur in the
turbulent regime. The maximum bubble diameter for laminar flow is 130 ~. For bubble sizes less
than 130 ~ Stokes' law can be applied to calculate the rise rate:

U _ g (0 - Og) d 2
t- 18!l

where Ut is the rise rate of the bubble, g the gravity constant, Q and Qg the density of the liquid and
the gas phase, respectively, d the bubble diameter and !l the dynamic viscosity of the liquid phase.
The maximum bubble diameter for laminar flow can be calculated from the above equation by
assuming that the limiting value of the Reynolds number for laminar flow is Re = 1. The
relationship between the rise rate and bubble diameter is given in Figure 3.
436

The specific gravity of flocs produced in water treatment is very similar to that of water.
Therefore only very small air bubbles are required to float the flocs to the surface. The smaller the
air bubbles which can be produced the larger the number of bubbles available per unit volume of
gas released. The presence of large numbers of bubbles increases the chance of bubble-floc
attachment in a very dilute floc suspension typical for water treatment application. However, the
smaller the bubble size the slower the rise rate of the bubble, which in turn increases the size of the
flotation tank required to allow the bubble to reach the surface. In practice the bubble size ranges
from 10 to 120!Ull with a mean size of approximately 40!Ull [8].

uQ)
UJ

E 0.1
~
<Ii
:c
Q)

.c
:::J
.c
~

'ffi
0.01
'0
Q)
UJ
.;::

'0
Q)
Ci5
c:
0.001

0.0001 L -_ _ _ _ _ _ _ _ _ _ _ _ _ _ ~

Bubble size, microns

Figure 3. Rise rate of air bubbles (calculated by Stoke's Law) as a function of bubble size [10].

4. Design of Flotation Plants

4.1. FLOTATIONTANKDESIGN

Both circular and rectangular flotation tanks are being used in water treatment applications.
 437

4.1.1. Circular tanks. Circular tanks are employed mainly for small flotation plants treating
wastewater or for sludge thickening applications, which require no preflocculation prior to
flotation.
For potable water treatment, a pre-flocculation stage is required, prior to flotation, to flocculate
the impurities present in water to larger agglomerates suitable for removal by flotation.
For circular tanks, the transfer of the flocculated water to the flotation tank without breaking up
the flocs creates problems, especially in larger plants as the flocculated water must be introduced
close to the bottom of the centre section of the flotation tank to achieve even distribution. Therefore
most of the larger flotation plants for water treatment applications employ rectangular tanks.
However, small circular package plants are being marketing in which one half of the tank is used
as the flotation stage and the other half as the flotation section (PTW Projects Ltd, personal
communication 1990). The use of circular flotation tanks has also been reported by Offringa and
van Vuuren [Ill.

4.1.2. Rectangulartanks. Rectangular flotation tanks offer advantages in terms of scale-up,

simpler design, easier introduction of the flocculated water, easier sludge removal and the smaller
area required. The flotation tanks are currently designed with a depth of approximately l.2 to l.5
m and upflow rates of between 8 and 12 m/h, depending on the type of water treated.
It is important that the flocculated water entering the flotation tank is uniformly distributed over
the whole width of the flotation tank to avoid hydraulic disturbances. Good mixing of the
flocculated water with the recycle stream should be provided, without creating too much turbulence
resulting in floc break-up. The flotation tank is usually equipped with a baffle directing the flocl
bubble agglomerates towards the surface of the flotation tank. The baffle is usually arranged at an
angle (approximately 60° to the horizontal) to reduce the velocity of the flocculated water in the
entrance section of the flotation tank, in order to avoid disturbing the sludge layer accumulating on
the surface of the flotation tank. Sufficient space should be provided between the top of the baffle
and the liquid surface in the flotation tank to ensure that the velocity is not increased which would
result in the disruption of the sludge layer. The liquid level in the flotation tank is controlled by
outlet weirs or by the outlet valve and care should be taken that the water is removed over the
whole width of the flotation tank to maintain stable hydraulic conditions in the unit. The residence
time in the flotation tank is between 5 and 15 min, depending on the loading (upflow rates) and
depth of the flotation tank.
A sludge beach for easier sludge removal and a sludge trough for collecting the sludge are
usually provided. To protect the flotation tank against wind, rain, and freezing of the sludge, a
protective roof over the flotation tank is essential for efficient operation of the flotation plant.
The size of the flotation tank is restricted by hydraulic and mechanical considerations and the
demand of the operator in terms of flexibility and number of treatment lines desired. flotation
plants with surface areas in excess of 80 m2 treating up to 18 Ml/d per flotation tank are operated
successfully.

4.2. AIR RELEASE DEVICES

Different types of pressure release devices are employed, ranging from proprietary nozzles and
needle valves to simple gate valves. To achieve effective air release, the pressure should be reduced
suddenly and highly turbulent conditions should exist in the device. The velocity of the recycle
stream leaving the pressure reduction device should be low enough to prevent floc breakup. As the
water passes through the pressure release devices at high velocities and as air is released, erosion
438

Liquid/air ir.n;=:=~=:;\

Recycle Recycle
pump Liquid out- pump Liquid out·
to flotation unit to flotation unit

INJECTION OF AIR INTO SUCTION LINE OF PACKEO SATURATOR

RECYCLE PUMP

-.J /' Inlet - from recirculation pump

Air feed -

pump suction pump discharge

SATURATOR WITH REAERATION PUMP AND EDUCTOR

Liqui~ir in Inlet - from recirculation pumJ:

- Air out

valve

Air air
discharge
Recycle
pump Outlet - to flotation unit Outlet - to flotation unit

ENTRAINING THE AIR WITH EOUCTOR UNPACK EO SATURATOR

Figure 4. Different air saturator systems [IJ.

 439

and cavitation can occur and hence the devices should be made from stainless steel. For larger
plants, a number of these devices usually at a spacing of approximately 0.3 m are used to obtain
good mixing and distribution between the flocculated water and the air bubbles.
An air injection nozzle has been developed and patented [12]. The nozzle consists of two orifice
plates to reduce the pressure and to create twbulence, and a shroud section to decrease the velocity
of the stream of recycled water before it is mixed with the flocculated water. The size of the first
orifice plate, which is the smaller, controls the amount of recycled water added to the flotation
tank. The second plate assists the air release and slows down the velocity of the recycle stream
before it enters the shroud section. More than 95% of the bubbles produced by the nozzle are in the
size range 10-120 1J.lll, with a mean size of about 40 1J.lll. A comparison of the proprietary nozzle
and a needle valve showed that the nozzle produced smaller air bubbles; however, both devices
achieved similar flotation treated water quality [8].

4.3. AIR SATURATION SYSTEM

As the air saturation system constitutes approximately 50% of the power costs of the flotation
process, it is important to optimise the design of the recycle system to minimise the operating
costs.
Various methods are employed for dissolving air under pressure in the recycle stream. These
include sparging the air into the water in a pressure vessel (saturator), trickling the water over a
packed bed, spraying the water into an unpacked saturator, entraining the air with ejectors, and
injecting the air into the suction pipe of the recycle pump (Figure 4).
Introducing the air to the recycled water, either on the suction side of the recycle pump or
through eductors before entering the saturation vessel, leads to substantially higher pumping costs
compared with using a separate compressed air supply.
Saturation levels of between 60 and 80% achieved by introducing the air on the suction side of
the pump could be increased to 90% by providing a turbine mixer in the saturation vessel [13].
Tests have shown that the packed saturator system has the lowest ratio of operating cost to
saturation level achieved [8,13]. A possible disadvantage of using packings is the danger of
blockage caused by biological growth. It is unlikely, however, that biological growth would pose
problems because the loading rates of saturators (between 1000 and 2000 m 3 m· 2 d· l ) are far
greater than typical biological filtration rates (less than 10 m 3 m· 2 d· l ) and this problem has not
been observed so far in plants treating potable water.
Extensive research has been carried out on optimising the design of packed saturators [8]. It has
been shown that the saturator can be operated over the range 300-2000 m 3m· 2d· 1 without any
decrease in saturation efficiency. Higher rates were not investigated, because of equipment
limitations. A packing depth of 0.8 m of 25 mm polypropylene pallrings, the least expensive
packing material investigated, was sufficient to achieve 100% saturation (Figure 5). This packing
depth, however, was substantially higher than the 0.3 m reported by Bratby and Marais [13]. The
type of packing material used was of minor importance provided it was in the same size range.
Casey and Naoum [14] found in their investigation on saturator performance that 0.8 m packing
of Raschig rings achieved only 80.6% saturation whereas packing of 25 mm and 38 mm spheres
produced 95% saturation. They used a semi-batch process and an air stripping system for
establishing the saturator efficiency rather than a continuously steady-state system and the actual air
release device used in full scale plants. In terms of operating costs, an unpacked saturator has to be
operated at a pressure of about 200 kPa above that of a packed saturator, in order to supply the
same amount of air to the flotation tank [15].
440

Water saturated with air under pressure is a very corrosive mixture and if mild steel is used for
the construction of the saturation vessel a corrosion-resistant lining should be provided. Plastic
pipes or stainless steel should be used for the connecting pipework between the saturator and the
flotation tank.

100 "
0 Water lemperature 12.SoC
Loading 1900 m 11m: Iday

SO "-
" Water temperature 4.8· 6.0',C
Loadmg 1900 m-'/m'/day

~ 60 Packing:
z
a
;:: Plastic pall rings
q:
a: Plate distributor
:> Pressure 415 kPa
>- 40
;;,

20

a
a 0.2 0.4 0.6 o.s 1.0 1.2
PACKING DEPTH 1m)

Figure 5. Effect of packing depth on air dissolution [8].

5. Factors Affecting Dissolved-Air Flotation Efficiency

5.1. COAGULATION

Raw surface waters used for a potable watcr supply contain colloidal and suspended solids, which
require the addition of coagulants such as aluminium sulphate or fenic sulphate for destabilisation
and subsequent flocculation. Similarly, if organic colour such as humic and fulvic acids are
present, coagulants are needed for the precipitation of colour and for flocculation. To achieve
efficicnt clarification, the pH of thc coagulated water has to be optimised, which might require the
addition of either acid or alkali. The optimum pH is dependent on the type of raw water being
treated and the coagulant being used. The optimum dose and pH are similar to that required for
sedimentation and can be detennined in the standard jar test apparatus. However, flotation does not
require the addition of polyelectrolytes as a coagulant aid. flotation is susceptible to over- or
underdosing, because of the short residence time in the plant (approximately 30 min).
Effective mixing of the raw water with the coagulant and the pH adjustment chemical is
important for the efficient operation of the flotation process. Most plant manufacturers offer flash
mixers for mixing the chemicals to the raw water. A flash mixer usually consists of a tank giving a
residence time between 0.5 and 1 min, which is agitated by a high speed mixer (usually 350 rpm).
Some fonn of in-line mixing using static mixers or other devices such as valves or orifice plates to
 441

crcate high turbulence is preferable to the use of flash mixers, as superior water quality tends to be
achieved.
If both a coagulant and a pH adjustment chemical are required, it is important that good mixing
of the first chemical with the raw water is completed before the second chemical is added. This is
particularly important in treating soft, coloured waters. The order of chemical addition appears to
be not important [16].

5.2. FLOCCULATION

Before the coagulated impurities can be removed successfully by flotation, flocculation into larger
agglomerates (flocs) is required. To achieve flocculation, agitation must be provided to promote
collisions between the primary particles and facilitate floc growth. This can be achieved by
mechanical agitation or by hydraulic means.
The flocculation time, degree of agitation and the means of providing agitation affect flotation
performance. The floc required for flotation is compact and relatively small compared with the
more fragile and larger flocs required for sedimentation.

5.2.1. Flocculation time. The flocculator usually consists of a tank subdivided into two or more
equal-sized compartments, each agitated by a slow moving paddle. The flocculation time required
differs with the type of water being treated and tests have shown that, for algal laden waters a
flocculation time of about 12 min is sufficient, whereas for highly coloured, soft water times of
about 20 min are required. Although some improvement can be achieved in treated water quality by
increasing the number of flocculation stages, this improvement can also be obtained by offering a
longer flocculation time. Therefore, most manufacturers offer flotation plants with a two-stage
flocculator and a flocculation time of about 20 min. Table 1 shows the improvement which was
obtained by increasing the flocculation time from 12 to 16 min, treating a highly coloured, soft
water [16]. This highljghl~ the importance of the llocculation stage for the efficient operation of the
flotation process.

TABLE 1 The effect of increaSing the flocculation time on the treated water quality
Flocculation Turbidity Residual coagulant
(min) (FfU) (mg/! Fe)
12 1.2 0.83
16 0.94 0.56

5.2.2. Degree of agitation. Besides the flocculation time, the degree of agitation is also very
important. Agitation is usually provided by a slow moving four-blade gate paddle, although
different designs arc being used. The degree of agitation is usually expressed as the mean velocity
gradient G calculated by

where P =total power input/volume of flocculator compartment and ~ =dynamic viscosity.

Tests have indicated that the optimum velocity gradient for flotation is about 70 s-I independent
of the type of raw water treated [16]. This compares with an optimum G value for horizontal
442

sedimentation of between lO and 50 S·I. Tapering the paddle speeds appears to give no
improvement in treated water quality except when treating turbid river water where tapering
provides some benefit. To avoid excess shear, which prevents adequate floc growth, the tip speed
of the paddles should not exceed 0.5 m S·I.

5.2.3. Hydraulic flocculation. An alternative approach to mechanical flocculation is the use of

hydraulic flocculation in which the energy required for flocculation is provided by the water
flowing through the flocculator. Tests have shown that half the flocculation time with a higher G
value (G= 150 S-I) were required for hydraulic flocculation, compared with mechanical flocculation
(G=70 s-l) (Figure 6).

FLOCCULATION TIME

/
• Hydraulic 4.6 min at 8.2 m)/h
& Hydraulic 57mJn at82m J /h
o Mechanical 10 min at 8 2 m 1 /h
::>
t; 0.8
>-
t-
O
~

-=i~~./
::>
I- 0.7
a:
w
t-
"s:
8
~=::---
------A
~ 0.6
a:
t-
------
A _ _- - A

O.5~--------y-------,---------,
10
FLOW RATE Im'/hl

Figure 6. Comparison of hydraulic and mechanical flocculation for flotation [17].

The difference in flocculation time and G value required is probably due to the more uniform
velocity distribution in the baffled tank hydraulic flocculator used, thus avoiding excess shear and
floc breakup.
The product of mean velocity gradient and flocculation time (G*t) is often used to express the
flocculation conditions required. It seems that for efficient flotation a G*t value of between 40 000
and 60 000 is required, independently of whether hydraulic or mechanical flocculation is
employed.

5.3. QUANTITY OF AIR REQUIRED FOR FLOTATION

The quantity of air supplied to the flotation tank can be varied by altering the saturator pressure
and/or the amount of recycle. If a fixed orifice is used for controlling the recycle an increase in
saturator pressure is associated with a small increase in recycle. Thus, different nozzle sizes require
different combinations of flow and pressure to deliver the same amount of air.
When flotation is used for the thickening of activated sludge, the volume of air required is
usually proportional to the mass of solids to be removed. Typical values of air/solids ratios in
 443

commercial practice for sludge thickening are 15 to 30 ml air/g solids [18]. For the treatment of
surface waters, therefore, low air/solids ratios should theoretically be required as the solids load is
much smaller. Experiments, however, have indicated that excess air is needed to ensure adequate
collisions between floc particles and bubbles and for bubble attachment and that the quantity of air
required depends mainly on the volume of water treated rather than on the solids content.
Approximately 8-10 g air/m 3 raw water, corresponding to air/solids ratios as high as 380 ml air/g
solids, are required for efficient solids/liquid separation, for a solids concentration in the raw water
of 20 mg/l (Figure 7).

Plant Flow rate 108 m3/h

o 3.18 mm nozzle
o 2.78 mm nozzle
I:>. 2.18 mm nozzle

Plant Flow rate 52.8 m3/h

s
~
X 2.18 mm nozzle--
LL
4
~
o m Turbidity
ED
a:
:::) 3
~
a:
w
~ 00
~ 2
w
~
«w
II:
~
Z
X-_~_~~~
o
________
-x------X _______ x ___ X_
~

o
~
~

g
LL
0 +-____
4
~------~----~----__------~----~----__------~
10

AIR ADDED (9 air/m 3 raw water)

Figure 7. Effect of air added on the performance of flotation [16].

The tests also showed that the treated water quality was only dependent upon the total quantity of
air introduced to the system and not upon the individual values of pressure or recycle (Figure 7).
Using a packed saturator at an operating pressure of between 350 and 420 kPa, with a recycle rate
of between 7 and 8%, was found to be adequate for optimum performance [2].

5.4. FLOAlEDSLUDGEREMOVAL

The sludge, also called float, accumulating on the flotation tank surface, can be removed by
flooding or mechanical scraping. Flooding involves raising the water level in the flotation tank
sufficiently for sludge and water to overflow into the sludge collection trough. Flooding has the
advantage oflow equipment cost and minimal effect on treated water quality, but at the expense of
high water wastage and low sludge solids content (less than 0.2%). Flooding can be operated
continuously or intermittently.
444

In continuous flooding, a small amount of water is allowed to flow continuously over the sludge
take-off weir, carrying the float with it. For the successful operation of continuous flooding, a
constant flow rate to the flotation plant is required to maintain a constant liquid level in the flotation
tank. If the liquid level drops for a period of time, the sludge tends to consolidate and a much
higher water flow over the weir is required to remove the sludge.
For intermittent flooding the sludge is allowed to accumulate on the surface of the flotation tank
for a specific time (usually several hours). For sludge removal, the flotation treated water outlet is
partially or completely closed for a period of time (5-10 min) and the water is allowed to flow over
the sludge take-off weir carrying the sludge with it.
However, if a sludge with a high solids content is required, a mechanical sludge removal
system is required. Different manufacturers of flotation equipment offer different types of removal
systems. Systems employed are:
I. Part or full-length scrapers, which use blades to scrape the surface of the flotation tank and
which push the sludge over the sludge beach into the sludge collection trough.
2. Beach scrapers, which are rolls with blades attached and scrape only over the sludge beach
and continuously thin out the sludge on the surface of the flotation tank by drawing the sludge
towards the sludge beach by natural flow.
3. Equipment which sucks sludge from the surface of the flotation tank; or
4. Rolls which move over the surface of the flotation tank and scoop the sludge from the tank
surface.
For water treattnent applications the part or full length scrapers and the beach scrapers are the
most widely used systems.
The sludges produced from the treattnent of different raw waters can vary considerably. Sludge
obtained from the treatment of algal laden waters or highly turbid river waters and accumulated
over a period of up to 24 h was stable and the accumulation of the sludge did not result in a
deterioration in treated water qUality. However, the sludge produced from the treatment of low
turbidity, highly coloured water tended to break up after accumulation periods of less than 30 min
resulting in a deterioration in treated water qUality. Removal of accumulated sludge by mechanical
scrapers usually results in a deterioration in treated water quality during the removal process and
for a short time after the completion of the scraping. To minimise the effect of the sludge removal
on treated water qUality, the sludge removal frequency, the scraper speed and the water level in the
flotation tank have to be optimised for each particular application.
For the removal of sludge produced from the treattnent of algal laden and turbid river waters the
sludge beach scraper was found to be an efficient sludge removal system, resulting in little
disturbance of the sludge thus minimising the degree of sludge break-up. It was found that for
optimum operation of the beach scraper, in terms of sludge solids concentration and treated water
quality, the scraper should be operated continuously. The water level in the flotation tank should be
adjusted close to the lower end of the beach and a thin continuous sludge layer (about 10 mm)
should be maintained on the surface of the flotation tank. Under these operating conditions sludge
solids concentrations of 3% were consistently produced with little effect on treated water quality
[16). These sludges were suitable for filter pressing, producing cake solids concentrations of
between 16 and 23% without polyelectrolyte addition. An advantage of the beach scraper is that
fewer mechanical parts are needed compared with a full-length scraper.
Flotation plants treating low alkalinity, coloured water require full-length scrapers operating
continuously as the sludge produced treating this type of water starts to break up after only 30 min.
The optimum speed in terms of treated water quality and sludge solids concentration was found to
be 30 m/h, producing a sludge of I % solids concentration. At lower scraper speeds the sludge
 445

remained sufficiently long on the surface for sludge break-up to occur. At higher speeds both the
sludge solids concentration and the treated water quality deteriorated [16].
Equipment cost for sludge removal can be considerable (as high as 10 to 20% of total plant cost
[19]) and therefore care should be taken in selecting the most appropriate and cost effective
removal system for a particular application.

6. Filtration of Flotation-Treated Water

Tests comparing the performance of rapid gravity sand filtcrs fed with flotation-trcated water and
upflow floc-blanket-clarified water have shown that these waters had similar filtering characte-
ristics, provided the water going on the filters had similar turbidities and residual coagulant
concentrations. The presence of air bubbles in the flotation-treated water had no influence on the
filter performance [16].

Flash Air injection Sludge

mixer F locculators through needle scraper

Coagulation
chemicals

H
Raw
Sludge
Backwash

water

Figure 8. Schematic diagram of a combined l1otation-filtration plant [I].

7. Combined Flotation and Filtration Plants

The combination of flotation and filtration was pioneered in Sweden (Figure 8), but experience
with this process has also been reported from other countries [5,20-22]. In this system a rapid
gravity sand or anthracite-sand filter was incorporated in the lower section of the flotation tank.
This arrangement has the advantage of providing an extremely compact plant. However, the
flotation rate of the plant is limited by the filtration rate. Thus in practice the plants can be operated
at a maximum surface loading of 6 to 7 m/h compared to a surface loading of 12 m/h for a separate
flotation unit. The tank depth of the plant tends to be deeper, approximately by 2.5 m, to
accommodate the filter bed and underdrain system compared with 1.5 m for a separate flotation
446

unit. In addition, the flow to the plant and any coagulant dosing have to be stopped periodically to
facilitate cleaning of the filter, which is backwashed in the nonnal way by air scour and water
wash.
The compactness of this system makes it particularly suitable for package plants. The reported
perfonnance of the plant in tenns of treated water quality is comparable with standard flotation
plants followed by rapid gravity filtration.

8 . The Ozone Flotation System

The combination of flotation with ozonation is being pioneered in France [23]. The ozonated air or
oxygen is introduced through porous plate diffusers, fonning bubbles in the size range 0.2 to 0.5
mm. It is claimed that the ozone assists in the removal of colour and algae. In addition, pressurised
recycle water is added above each diffuser plate; the small bubbles are needed for water
clarification. Thus, the process is a combination of dispersed-air and dissolved-air flotation. Table
2 gives the algae removal efficiencies reported for the process (Bourbigot, personal communication
1990) which appear to be lower than those achieved by dissolved-air flotation (> 90%).

T ABLE2 Al gal remov al bJy the ozone-fl otatlon process.

Operation conditions Algae species Mean removal
%
1987 Ozone dose 3 g OJ/m 3 Euglenn sp. 98
Coagulant dose" 20-25 g/m3 Melosira granulata 39
Upflow rate 16.7 m/h
1988 Ozone dose 2 g 03/m3 Melosira granulata 51
Coagulant dose" 20-25 g/m3 Synura urella 60
Upflow rate 16.7 m/h Dinobrion mallomonas 55
Staurastrum megacanthum 79
Scenedesmus armatus 87
Aphanizomenonflos aquae 42

9. Performance of Dissolved-Air Flotation Plants

Extensive studies on both pilot- and full-scale plants have been conducted on the perfonnance of
the dissolved-air flotation process with different raw waters. The types of raw waters investigated
include lowland, high alkalinity river water; soft, highly coloured water; and nutrient-rich, long-
tenn stored water with algae problems. Detailed perfonnance data have been given elsewhere
[1,3,5,16,24-26].
Summaries of the perfonnance of various flotation plants operating in Finland and the UK have
been prepared. Heinlinen, surveying 34 treatment plants [22], concluded that the Finnish plants
using dissolved-air flotation are working well with few operational problems, but that many of the
plants are overdesigned, resulting in unnecessarily high capital costs. Longhurst and Graham [27]
assessed the design and perfonnance of 12 dissolved-air flotation plants in the UK. A comparison
of the results achieved by 4 plants, for which quality data are provided, is given in Table 3.
 447

TABLE 3. Average water qu al'Ity pIrod ucedb)y fiour fl otatlon plants [27]
Works Raw Water Treated water
Colour TUlbidity Colour Turbidity
°Hazen NTU ~azen NTU
Range Average Range Average Range Average Range Average
A 6.22 14.7 0.5-6.4 2.3 0.5-8 2.8 0.2-3 1.0
B 12-27 20.0 1-9.5 5.3 0.3 1.5 0.6-4 2.3
C 2-60 19.5 0.6-3.6 2.4 4.0 0.6
D 22.5-24 59.2 0.5-29 4.0 4.5-35 19.4 0.5-8.5 2.7

9 .1. TREATMENT OF LOWLAND. HIGH ALKALINITY. HIGH TURBIDITY RIVER

WATER

Under optimum operating conditions. flotation reduced raw water turbidities of up to 100 NTU to
usually less than 3 NTU at the design upflow rate of 12 m/h. When the raw water turbidity
exceeded 60 NTU. the treated water quality could be improved significantly by reducing the flow
rate through the plant by about 10-20%, The colour was reduced from as much as 70° Hazen. and
residual coagulant concentrations before filtration were in the range 0.25-0.75 mg AI/!. An upflow
floc-blanket sedimentation plant. operated at an upflow rate of 2 m/h. produced similar treated
water quality to that of the flotation plant during the low-turbitidy periods but better quality (by 1 to
2 NTU) for raw water turbidities exceeding 100 NTU.
The selection of the correct coagulant dosage and coagulation pH is critical during flood
conditions. Because of the short residence time in the flotation plant. the change in raw water
quality has to be followed closely to maintain optimum coagulation conditions.
Drago [25] compared the performance of dissolved-air flotation with sedimentation treating high
turbidity water (Table 4). which shows that sedimentation produced better treated water quality
than flotation. However. the plants might not have been operated under optimum coagulation
conditions for the primary treatment process. as the aim of the investigation was to produce
optimum conditions for subsequent filtration. Although directly abstracted river water can be
treated successfully by flotation. sedimentation tends to be the more appropriate treatment process
for this application. especially if the raw water turbidity varies rapidly and high turbidity peaks
(>100 NTU) are experienced.

TABLE 4. Comparison of dissolved-air flotation and sedimentation

treatmg h'gh
I turb'd'
I Ity water.
Dissolved-air flotation Sedimentation
Raw water NTU 45-138 48-152
Treated water NTU 9.7-17.0 0.73-1.0

9,2. TREATMENT OF LOW TURBIDITY. SOFT. HIGHLY COLOURED UPLAND WATER

Table 5 shows a comparison of the water quality achieved by flotation. sedimentation and filtration
treating a low turbidity. soft. highly coloured water.
The flotation plant was operated at l2m/h upflow rate. whereas the floc-blanket sedimentation
plant could only be operated at less than 1m/h. even with the addition of polyelectrolyte. because
the floes produced by the coagulation of these waters are very light and have low settling
448

velocities. The quality of the water treated by the two processes was quite similar. Only the
residual coagulation concentration of the sedimentation-treated water was usually lower by about
0.2 mg Fell. By increasing the flocculation time from 12 to 16 min, however, the residual
coagulation concentration in the flotation treated water was reduced to that of the sedimentation-
treated water.

TABLE 5. Comparison of qualities achieved with flotation, sedimentation,

and fIltration [16]
Turbidity Dose Colour pH Iron Manganese Aluminium
NIU mgll Fe °Hazcn mg Fell mg Mn/l mgAl/l
Raw 3.2 45 6.2 0.70 0.11 0.23
Flot-TI 0.72 8.5 2 4.8 0.58 0.16 0.01
Flot-F2 0.19 9.0 0.01 0.02 0.01
Sed-T3 0.50 6.0+0.8 0 5.05 0.36 0.14 0.10
mg/l poly
electrolyte
Sed-f4 0.29 0 10.5 0.01 <0.02 0.10
• Improved flotation-treated water quality similar to that achieved with sedimentation was
obtained by increasing the flocculation time from 12 to 16 min.
1 - Flotation-treated
2 - Flotation-filtered
3 - Sedimentation-treated
4 - Sedimentation-filtered

Another advantage of flotation was that the plant produced consistently good treated water
quality, even at temperatures below 4 °C which occurred frequently during the winter months. At
these low temperatures, the floc blanket in the sedimentation tank frequently became unstable
resulting in a deterioration in treated water quality.

9.3. TREATMENT OF NUTRIENT RICH, LONG-TERM STORED WATER WITH ALGAE

PROBLEMS

Algae tend to float and are therefore difficult to remove by sedimentation. Table 6 shows the
removal efficiency of flotation and sedimentation for different algae species. At times the algae
counts in the flotation treated water were lower than those in the sedimentation filtered water,
Figure 9.
Efficient flocculation is essential for effective algae removal. Only 10 to 20% algae removal was
obtained when the flotation plant was operated without coagulant addition.
Figure 10 shows a comparison of the algae removal rates achieved by three different coagulants
at their optimum pH for minimum coagulant residuals. Aluminium sulphate gave the best removal.
For polyaluminium chloride (PACI), an equivalent dosage in terms of aluminium was required to
achieve an algae removal similar to that of aluminium sulphate. The poorest treated water quality
was obtained with chlorinated ferrous sulphate. Tcsts havc shown that algae removal is improved
by lowcring the pH [17]. The poorer algae removal achieved with chlorinated ferrous sulphate
might have been a result of the higher coagulation pH required for minimum coagulant residual in
the treated water (8.3-8.7 for chlorinated ferrous sulphate, compared with 6.8-7.2 for aluminium
sulphate).
 449

TABLE 6, Comparison of the algal removal efficiency of flotation! and sediment-

ation2 using chlorinated ferrous sulphate as coaguJ1ant [16]
Algal type Raw water Sedimentation-treated Rotation-treated
water water
(cells/ml) (cells/ml) (cells/ml)
Aphanizomenon 179000 23000 2800
MicrocystiS" 102 000 24000 2000
Stephanodiscus 53000 21900 9100
Chlorella 23000 3600 2200
* Aluminium sulphate was used as the coagulant
! upflow rate 12 m/h
2 upflow rate 2 m/h

200 Aphanizomenon M icrocystis

102000
179 000 100

150 u u u
j!l j!l ~

~'" '"~
Q)
E .:::
~ uQ)
----'!! +-' :;::: U
j!l
u
~
a; C +-' C C

M
u 0
'';::; '"~
+-' M
OJ
U
0
+-'
0
+-' ~'"
j!l
:;:::
~ ~ ~
~
I
100 . ~. c c I
'j!l c c c c
0
.", Q) 0
'';::;
0
50
,'":l: <l! OJ 0 0
:l: .. .S E E +-' +-'

0'" 0'" 0'"

X X
~ .:l: .. uQ) ~
:l: u
OJ
u
OJ
c ~ <f) ;;::: c .~
<f) <f) ;;::: ;;:::
:l :l
0 0
u u

9'" 50 9'" 24000

« «
23000

58
0 0

Figure 9. Removal of algae by flotation, sedimentation and filtration [16].

Edzwald and Wingler [28] compared dissolved-air flotation with conventional treatment and
concluded that DAF produced lower treated water turbidities and better algal removal. The
removals by DAF were 99 to 99.9% compared to less than 90-99% by sedimentation treating a raw
water with a very high algal count of 5 x 1OS algal cells/ml.
450

TABLE 7. Example of operating costs for a 22.7 MUd flotation plant with 95%
load factor [26]
Annual costs Water costs
(£OOOs) (p/m3)
1. Fixed charges from the capital for
a) Civils of £180000
(for concrete. flash mixing.
flocculating and flotation tanks.
plus foundations for saturators).
Civils - 30 year plant life
10% interest rate
10.6% amortisation rate 19.1 0.24

(b) Mechanicals of £160000

(for all mechanical plant for
flash mixing. flocculation.
chemical mixing and dosing.
manual control and nominal
coverage of the flotation tanks).
Mechanicals - 20 year plant life
10% interest rate
11.75% amortisation rate 18.8 0.24

Note that these costs are exclusive

of raw water and high lift pumps.
sludge disposal facilities and all
buildings.

2. Power for saturation at 400 kPa

and with 7% recycle. plus flash
mixing. flocculating. chemical
mixing and dosing and sludge
removal. Estimate 22kWh/Ml
@3.2p/kWh 5.51 0.Q7
3. Chemicals - average dose
7mg/litre as Fe and sodium
hydroxide for pH adjustment 21.9 0.28
4. Labour - operator. quality control
1 man year with overheads 8.0 0.10
5. Maintenance - fitter. 1/4 man year
and materials 6.0 0.08
Total annual costs £79310
Total water costs 1.01 p/mj
 451

Raw Water Algal Count Raw Water

Coagulant
(cells/m) - Microcystis Turbidity (FTU)

Aluminium sulphate o 40000 3.5 - 4.1

Aluminium sulphate • 120000 5.5 - 5.9
PAC • 120000 5.5 - 5.9

.
Chlorinated ferrous sulphate • 48000 3.6

100

,- ---• .-
,
~ ~
...J 90 ..A--
« /
>
0 /. "
~
L.I.I
a: 80 /",.'"
...J
« ",'7'"
<!l
...J /'.
« 70
/
0 1.0 2.0 3.0 4.0 5.0 6.0 mgjl AI
0 2.0 4.0 6.0 8.0 10.0 12.0 mg/I Fe
COAGULANT DOSE

Figure 10. Comparison of effectiveness of three coagulants on algal removal by flotation [16].

10. Cost of Flotation

In their survey of operational dissolved-air flotation plants in the UK, Longhurst and Graham [27]
produced capital cost information based on 8 plants. For comparison, all costs were converted to
1984 prices.

Cost (£000) = 7.7 x Area - 342

where Area = total surface area of flotation tank including the riser section (m 2). A similar cost
function was calculated based on plant capacity:

Cost (£000) = 24 x Capacity + 15.9

where Capacity = treatment works design capacity (MUd). These two equations give slightly
different estimates of capital costs, because the equations were derived from data referring to a
452

different sample of plants and the large margin of error in the cost data.
An alternative cost function (corrected to Q4 1981 prices), based on an earlier report on the cost
of water treatment plants [29], is given by Zabel [2].

Capital cost (mOO) = 23 x Area°.52

where Area = total surface area of flotation tank (m 2). This equation was derived from a very small
database but indicates some economy of scale.
Estimated operating costs for a 22.7 MIld flotation plant are given in Table 7. The costs are
based on 1979 prices and do not include the cost for raw water and high lift pumping. It is
interesting to note that, although the energy cost for flotation are significant, the chemical costs are
substantially higher.
A cost comparison of flotation with floc blanket clarification, has shown that the overall costs of
the two processes is simillar. Flotation tends to have lower capital costs but higher operating costs
than sedimentation, thus a low plant utilisation would favour flotation [30].

11. Conclusions

Dissolved-air flotation is an effective means for solid/liquid separation and has been successfully
applied for the clarification of surface waters for potable supply. It can offer several process
advantages over sedimentation, including better water quality, rapid start-up and shut down, high
rate operation and thicker sludge. It is particularly effective for the removal of algae and the
treatment of highly-coloured, soft waters. The compactness of the plant makes it particularly
suitable for colder climates which requires housing of the plant. The rapid start-up has advantages
for use in touristic areas with high variation in demand.

12. References

1. Zabel, T.F., and Melbourne, J.D. (1980) Flotation, in W.M. Lewis (ed.) Developments in
Water Treatment 1, Applied Science Publishers Ltd., London, Chapter 6.
2. Zabel. T. F. (1984) Flotation in water treatment, in K.J. Ives (ed.) The Scientific Basis of
Flotation, Martinus Nijhoft, The Hague.
3. Gregory, R. and Zabel, T. F. (1990) Sedimentation and flotation, in F. W. Pontius (ed.)
Water Quality and Treatment. A Handbook of Community Water Supplies, A.W.W.A.,
McGraw Hill, Fourth Edition, Chapter 7.
4. Barett, F. (1975) Electroflotation - development and application, Water Polution Control
11(1),59.
5. Krofta, M., and Wang, L.K. (1984) Development of innovative flotation-filtration systems
for water treatment Part A, Firstfull-scale sand float plant in U.S.A., Proceedings Water
Reuse Symposium III. Future of Water Reuse, San Diego.
6. Hemming, M. L., Cottrell, W. R. T., and Oldfelt, S. (1977) Experience in the treatment of
domestic sewage by the micro-flotation process, Proceedings of the Water Research Centre
Conference on Flotation for Water and Waste Treatment, Water Research Centre,
Medmenham, UK, Paper 2.
7. Kitchener, J.A., and Gochin, K.J. (1981) The mechanism of dissolved-air flotation for
 453

potable water: basic analysis and a proposal, Water Research 15,585.

8. Rees, A.J., Rodman, D.J., and Zabel, T.F. (1980) Evaluation of dissolved-air flotation
saturator performance, Technical Report TR 143, Water Research Centre, Medmenham, UK.
9. Packham, RF., and Richards, W.N. (1975) Water Research Centre, Report TM 106.
10. Bennett, G.F. (1988) The removal of oi/from wastewater by air flotation: a review, CRC
Reviews in Environmental Control ,lli, Issue 3.
11. Offringa, G., and van Vuuren, L.RJ. (1987) Treatment of eutrophic waters, Paper presented
at Symposium Hartbeespoort Dam 'Quo Vadis', CSIRO Conference Centre Petrovia,
September 1987.
12. Brit. Pat. Spec. Nos. 1.444.026 and 1.444.027.
13. Bratby, J., and Marais, G.v.R. (1977) Flotation, in D.B. Purchas (ed.) Solid/liquid
separation equipment scale up, Uplands Press Ltd., Croydon, London, chapter 5.
14. Casey, T.J. and Naoum, I.E. (1986) Air saturators for use in dissolved air flotation process,
Water Supply (Amsterdam) 1" pp. 69-82.
15. Zabel, T.F., and Hyde, RA. (1977) Factors influencing dissolved-airflotation as applied to
water clarification, Proceedings of Water Research Centre Conference on Flotation for Water
and Waste Treatment, Water Research Centre, Medmenham, UK, Paper 8.
16. Rees, A.J., Rodman, D.J., and Zabel, T.F. (1979) Water clarification by flotation-5,
Technical Report TR 114, Water Research Centre, Medmenham, UK.
17. Rodman, D.J. (1982) Investigation into hydraulic flocculation with special emphasis on algal
removal, MSc thesis, Water Research Centre, Stevenage, UK.
18. Maddock, J.L. (1977) Research experience in the thickening of activated sludge by dissolved-
air flotation, Proceedings of the Conference on Flotation for Water and Waste Treatment,
Water Research Centre, Medmenham, UK, Paper 5.
19. Rowlands, H. S. (1977) Proc. Soc. Chern. Ind. Symp. on Principles Practice of Flotation.
20. Van Vuuren, L.R, de Wet, F.l., and Cillie, F.J. (1984) Dissolved-air flotation-fi/tration
studies in South Africa, A.W.W.A. Research Foundation News, Denver, Colorado, pp. 10-
12.
21. Vosloo, P.B.V., Williams, P.G., and Rademan, R.G. (1986) Pilot and full-scale
investigations on the use of combined dissolved-air flotation and filtration (DAF F) for water
treatment, Water Pollution Control 85, 114-121.
22. Heinanen, 1. (1988) Use of dissolved-air flotation in potable water treatment in Finland, Aqua
Fennica,1.8., 113-123.
23. Bourbigot, M.M., Faivre, M., and Langlais, B. (1988) Experimental plant demonstrates
advantage of combined ozone andflotationfor pretreatment of coloured algae laden waters,
I.O.A. Conference Report - Ozone in Water Quality Management - Olympia Conference
Centre, London, 19-20 October 1988.
24. Krofta, M., and Wang, L.K. (1985) Application of dissolved-air flotation to the Lennox
Massachusetts Water Supply, Water purification by flotation, 1. New Engl. Water Works
Assoc. 22, p. 249.
25. Drago, J.A. (1984) Clarification of Sacramento Delta Water in large scale dissolved-air
flotation pilot plant, Proceedings A.W.W.A., Annual Conference, Dallas, U.S.A.
26. Wilkinson, P.D., Bolas, P.M., and Adkins, M.F. (1979) Institution of Water Engineers and
Scientists, South Eastern Section, Summer Meeting Paper.
27. Longhurst, S.A., and Graham, H.1.D. (1987) Dissolved-air flotation for potable water
treatment: a survey of operational units in Great Britain, The Public Health Engineer 11(6),
April 1987.
454

28. Edzwald, I.F., and Wingler, B.J. (1990) Chemical and physical aspects of dissolved-air
flotationfor the removal of algae, Aqua.32. 24-35.
29. Water Research Centre (1988) Cost information for water supply and sewage disposal.
Technical Report TR 61. Water Research Centre, Medmenham, UK.
30. Gregory. R. (1977) A cost comparison between dissolved air flotation and alternative
clarification processes, Proceedings of the Water Research Centre Conference on Flotation for
Water and Waste Treatment. Water Research Centre. Medmenham, UK, Paper 11.
4.2. MECHANISM OF SELECTIVE ION FLOTATION

WL. WALKOWIAK
Institute of Inorganic Chemistry and Metallurgy of Rare Elements
Technical University ofWrociaw
Wybrzeze WyspiansJdego 27
50-370 Wroclaw
Poland

ABSTRACf. It has been determined that ion-water and cation-surface active anion interactions
govern the ion flotation selectivity, when a sublate was not formed in the bulk solution. The
flotation selectivity orders of anions were correlated with the experimentally-determined viscosity
coefficients, whereas the flotation selectivity orders of cations were correlated with their ionic
potential values. It has been found that, if sublate precipitation takes place in the bulk solution, the
surface charge of sublate was the key factor influencing the floatability of an ion. Basic properties
of sublates, e.g. the solubility and electrokinetic potential in aqueous solution have been
determined. Ion flotation kinetics, using cobalt(II) and cobalt(III) ions as an example, is also
considered in this study.

1. Introduction

Ion flotation, as one of the adsorptive bubble separation techniques, involves the removal of
surface inactive ions from aqueous solutions by the introduction of a surfactant and the subsequent
passage of gas bubbles through the solution. As a result of the process, a solid known as the
sublate appears on the gas/aqueous solution surface. Higher concentrations of floated ions and
surfactant may lead to precipitation of sublate in the bulk solution. If the species is a cation or an
anion (generally named colligend), which forms a soluble complex or an ion pair with the opposite
charged surfactant ion, the process is named foam fractionation. Foam fractionation is often
confused with ion flotation and sometimes will occur simultaneously.
Another possibility to remove the surface inactive ions from aqueous solution is to precipitate
those ions and then float this precipitate into the foam phase. This process is termed precipitate
flotation of the first kind and requires the use of a surfactant.
These foam separation processes have been applied to concentrate and separate effectively a
variety of constituents from dilute aqueous solutions. Ion flotation has been introduced by Sebba
[1] and more recently a few monographs were written [2-4], as well as some reviews [5-7].
The objective of this study is to investigate experimentally the mechanism and selectivity as well
as the kinetics of ion flotation of transition metal cations and anions from dilute aqueous solutions,
using cationic and anionic surfactants. Basic properties of sublates, e.g. the solubility and
electrokinetic potential in aqueous solutions, are also presented.

455
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 455-473.
© 1992 Kluwer Academic Publishers.
456

2. Sublate Solubility

Since the presence of a sublate in the bulk solution detennines the ion flotation process, the
sublate's solubility has been detennined by measurements of the aqueous solution's turbidity. Two
systems were measured, e.g. transition metal cations (Ag+, Mn2+, C02+, Zn2+, Cd2+, C~+' and
Fe3 +) with anionic surfactants (sodium dodecylobenzenesulphonate - DBSNa and sodium
dodecylsulphonate - DSNa), as well as transition metal anions (Ag(CNh-, Au(CNh-, Au(CNk,
Zn(CN)42-, Pt(CN)4 2., Cd(CN4)2-, Hg(CN)i-, PdC42-, PtCI62-, and Co(CN)6 3-) with cationic
surfactants (cetyl-trimethyl-ammonium chloride - CTMACl, cetylpyridinium chloride - CPCI, and
cetyl-dimethyl-benzyl-ammonium chloride - CDMBACl).

4.00
6.0

3.75
III 5.0 III
~
~
~ ~
I
I 3.50
4.0

3.25
3.0

50
MOLECULAR

Figure 1. Sublate solubility vs. the molal mass of ions studied for metal cations with sodium
dodecylobenzenesulphonate-DBSNa and metal complex anions with cetyl-dimethyl-benzyl-ammonium
chloride - CDMBACI.

In the case of metal cation sublates, it was found that the Ag+ cation did not fonn sublates with
the anionic surfactants studied. For other sublates, the solubility was found to be influenced mainly
by the size and charge of metal cations. The lowest sublate solubilities are for those sublates, which
are fonned by metal cations with the highest ionic potential values. The ionic radii and ionic
potentials of the metal cations studied are presented in Table 1. In the case of sublates of metal
complex anions, the lowest values of solubilities are for trivalent complex anions, like Co(CN)6 3-
and the highest for monovalent complex anions, like Ag(CNh- and Au(CNh-. Sublate solubilities
for anions with the same charge are decreasing with the increase in ion size.
 457

TABLE 1 Ionic radii and ionic potentials of studied metal cations

Cation Ionic radii [A] Ionic potentials [A-']
cr3+ (>-76 9_95
Fe 3+ 0.79 3.80
Mif+ 0.96 2.08
C02+ 0.88 2.27
Zn2+ 0.89 2.25
Ag+ 1.29 0.78
Cd2+ 1.09 1.83
In3+ 0.93 3.23
* Taken from ref. [8].

Generally, sublate solubility was found to depend upon the molal mass of the ions which form
the sublates (Figure I). On the other hand, the sublate solubility is influenced by the size of the
surfactant molecule. For quaternary ammonium salts, the sublate solubility decreases with the
number of carbon atoms in the surfactant molecule:

no. of
calbon atoms 19 22 25

Analogically, the solubility of sublates, which are precipitated by sulphonates, is also decreasing
with the number of calbon atoms in the surfactant molecule:

no. of
carbon atoms: 12 18

3. Selectivity of Ion Flotation in the Absence of Sublate in the Bulk Solution

Knowing the sublate solubilities as well as the initial concentrations of the floated ions and the
surfactants used, it is possible to calculate whether the sublate can be precipitated in the bulk
solution or not. For all ion flotation experiments presented in this chapter, no sublate was formed
in the bulk solution.

3 .1. FLOTATION OF TRANSITION METAL CATrONS

The ion flotation selectivity of anionic surfactants for inorganic cations has been studied in several
investigations [9-13]. Walling et aI. [9 J have found that the order of increased selectivity of the
foam separation process is: H+ < Na+ < K+ < NlLt and Ca 2+ < Mg2+. Wace and coworkers [10]
found values of the selective adsorption coefficient in foam separation:

as r 2+,Cs+ = 58,
aCa2+,Cs+ = 38.5,
aCe3+,Cs+ = 3.3
458

1.0

0.5

\J
.......
\J
1.0 Ir----_

0.5

o
TIME. min

Figure 2. Flotation curves for competitive ion flotation of 5xlO-5 M metal cations by 4xlO- 4M sodium
dodccylsulphonate-DSNa (A) and sodium dodccylobenzenesulphonate-DBSNa (8) at pH 2.2.

The low value of aCe3+,Cs+ was explained by the possibility of the high hydration nunber of
Ce(III) cations, which reduces the effect of the high valency. Rubin and Jome [11] have compared
their experimental data on the separation of sr2+ and uol+ ions with the proposed theory based
on the diffuse layer model, with the restriction of a different closest approach to the surface for
each ion. Huang and Talbot [12] have investigated the continuous flow from fractionation ofPb2+,
Cd 2+ and Cu 2 + cations with sodium dodecylsulphonate. They have found that the order of
decreased selectivity is : Cu2+ < Cd 2+ < Pb2+, which is the order of decreasing ionic radii. Their
results are in good agreement with a theory based on the Gouy-Chapman model of the double layer
[11]. Kubota and Hayashi [l3] have carried out experiments of continuous-flow foam fractionation
of Na+, Cd2 + and Cr 3+ cations with sodium dodecylbenzenesulphonate. They found that the
 459

selectivity order, i.e. Na+ < Cd 2+ < cr3+ is in accordance with the Gouy-Chapman theory of the
diffuse layer. Recently, Grieves and his coworkers [14] have investigated the continuous-flow
foam fractionation of each of the colligends Li+, K+, Rb+ and Cs+ versus Na+. The selectivity
coefficients were determined as 0.92, 1.10, 1.51 and 1.65, respectively. These coefficients were
correlated with the absolute partial molal entropy of the cations studied in an aqueous solution,
which increases as the degree of cation hydration decreases, for a series of monoatomic cations of
like charge.
The objective of this study was to establish the selectivity of anionic surfactants, i.e. sodium
dodecylsulphonate and sodium dodecylbenzenesulphonate, for transition metal cations with
different ionic radius and electrical charge; results are interpreted in terms of the ionic potential of
metal cations.
In the first series of experiments, the removal of two metal cations with different electrical
charge, i.e. C02+ and Fe 3+, from multi-ion aqueous solutions by DBSNa was investigated and
results are shown in Figure 2. Removal of C02+ cations from the dilute aqueous solution is lower
than the removal of the Fe 3+ cations. During the first 30 min of flotation, only Fe 3+ cations are
practically removed from the aqueous multi-ion system, when DBSNa is used as an anionic
surfactant.

1_0 l~~======== Mn 2+
Zn 2+

u - - - - - - - Co2+

"u
0.5

o 50 100
TIME, min

Figure 3. Flotation curves for competitive ion flotation of lxlO,5M metal cations by 2.5xlO-4M sodium
dodecylobenzenesulphonate-DBSNa at pH 2.2.

In Figure 3 the removal of five metal cations i.e. cr3+, Fe 3+, Co2+, Zn2+ and Mn2+ is shown.
The removal of divalent cations, i.e. Mn 2+, Zn2+ and Co 2+, is much lower than the removal of
trivalent cations, i.e. Fe3+ and cr3+.
A final series of experiments was conducted to determine the removal by flotation of three metal
cations, having the same 4d lO electron configuration but different electrical charge, i.e. Ag+, Cd 2+
and In3 +. Results are shown in Figure 4; the removal of metal cations by flotation is seen to be
460

Ag+
1.0 r===========Cd2+

u
"u
0.5

- - - - - - - In3+

o 25
TIME. min

Figure 4. Flotation curves for competitive ion flotation of lx10- 5M metal cations by lxlO-4 M sodium
dodecylobenzenesulphonate-DBSNa at pH 3.0.

highest for trivalent cations and lowest for monovalent ones. Also, In 3 + cations may be removed
from solution by flotation, while Ag+ and Cd 2+ cations remain in the aqueous solution.

1,0 ~:- _ _ _ __
Ag(CN ):2

u
Au(CNI 2
"u

0.5

min

Figure S. Flotation curves for the competitive ion flotation of LOx 1O-5M cyanide complex anions by
1.0xlO-4M cetyI-trimethyI-ammonium chloride-CTMACI at pH 11.0 [16].

3.2. BATCH FLOTATION OF ANIONS

First, the study of batch ion flotation involved the selective separation of Au(CNh' and Ag(CN)z
 461

complex anions with dodecyl-dimethyl-bcnzylammonium bromide from a two-component aqueous

solution [15). The kinetic curves for the anions studied show that the AU(CN)2- anion has much
greater affinity to the surfactant used over the Ag(CN)f anion.
The next research concerned competitive ion flotation of Ag(CNh-, Au(CNh·, Au(CN)4-
anions by cetyl-trimethyl-ammonium chloride; results are presented in Figure 5. The selectivity
sequence is as follows: Au(CN)4- > Au(CNh- > Ag(CNh-.
A series of batch foam separation experiments were conducted to determine the selectivity
sequence for the cyanide complex anions of Zn(II), Cd(II), Hg(II) and Au(III) with the cationic
surfactant cetyl-trimethyl-ammonium chloride [17). The effect of nitrates on the ion flotation of the
cyanide complexes studied, for solutions containing equimolar concentrations of all four metals,
yielded the selectivity sequence:

Also, the selectivity of chloride anion complexes of palladium(II), platinum(II) and gold(III) were
studied using cetyl-trimethyl-ammonium chloride as a cationic surfactant. flotation curves for
competitive ion flotations are presented in Figure 6. The selectivity sequence deduced from these
curves is as follows:

1.0

u
'-
u

0.5

o 50 100
TIME, min

Figure 6. Flotation curves for competitive ion flotation of 3.0x IO- 6 M chloride complex anions by
2.0xIO-5M cetyl-trimethyl-ammonium chloridc-CTMACl at HCl concentration equal to O.OIM.
462

3.3. CONTINUOUS FLOTATION OF ANIONS

Steady state, single equilibrium stage experiments can be conducted in an ion flotation unit shown
previously [18].
An original approach to selectivity problems of anions was presented by Grieves [19]. A cationic
surfactant, concentrated from a dilute aqueous solution at aqueous solution/air bubble interfaces has
been modeled as a soluble ion exchanger.
Consider the continuous flow with the feed stream containing an ion X- with a concentration Ci,
and the quaternary ammonium surfactant Surf+Y- with a concentration ej of Surf+ and bi of yo.
When the system reaches an equilibrium, the residual stream contains a concentration c,. of X-, a
concentration er of surfactant, and a concentration b r of Y -_ An assumption can be made that the
foam consists of entrained bulk liquid, containing the surface excess of surfactant having a surface
concentration r e plus the fixed and diffuse layers of counterions of surface concentrations r c and
rh, respectively.
When each bubble goes through the bulk solution, an exchange reaction may occur:

(1)

and thus the selectivity coefficient can be defined by:

(2)

It was shown that the liquid height in the column had no influence on K' [19]. This indicates that
either the surface ion exchange reaction is rapid and reaches equilibrium at a short distance above
the gas sparger or no surface ion exchange occurs and then the selectivity is determined by ion pair
formation in the bulk solution:

K' = (Surf-X)r (Y-)r

(3)
(Surf-Y)r (X-)r

A possible cause for an ion exchange reaction at the surface can be the attraction of the less
hydrated anion to the surface, in preference to the more hydrated anion_ The concentration of the
colligend in the surface phase, r c, is proportional to (Ci - c,.):

(4)

where D is the average bubble diameter, L is feed flow rate and A is the gas flow rate_ A similar
expression can be written for rb. The ratio of rcfrb in eqn. 2 can be replaced by (Ci - er)/ (bi - b,.).
It leads to the simple and direct experimental determination of K':

(5)

For divalcnt anions, an exchange reaction can be expresscd as:

 463

(6)

and the selectivity coefficient is consequently:

(7)

An alternative exchange reaction is also possible. For divalent anions and a quaternary
ammonium surfactant, when all the X 2- anions are instantly paired as (Surf-X)", the surface ion
exchange reaction is defined as:

(8)

For this, the selectivity coefficient is determined according to eqn. 5. For divalent anions, the
expression for the selectivity coefficient given by eqn. 5 fits the data rather well, considerably
better than a second-order relation between f JCr and fblbr. It indicates that the (Surf-X)' ion is
exchanged with counterion Y- [20].
Steady state, single equilibrium experiments of cyanide complex anions with quaternary
ammonium salts were conducted. Values of K' for each colligend are given in Table 2.

TABLE 2. Selectivity coefficients for cyanide complex anions from the continuous
··t
Iioam separation WIh catIoruc surt:actants.
95% Correlation
Anion Surfactant K' confidence coefficient r Reference
limits for K' for K'
Ag(CN)i cetyl- 2.71 0.189 0.86 [18]
Au(CNh- triethyl- 10.8 0.391 0.98 [18]
ammonium
iodide
AU(CN)4' cetyl- 28.6 1.11 0.96 [21]
Zn(CN)42. trimethyl- 8.86 0.642 0.83 [22]
Cd(CN)42 - ammonium 21.8 1~17 0.91 [22]
Hg(CN)42- iodide 25.1 1.33 0.91 [22]
Pt(CN)4 2- 6.16 0.41 0.95 [21]

3.4. CRITERIUM FOR SELECTIVE FLOTATION OF CATIONS AND ANIONS

According to the ion-exchange model for flotation of ionic species from aqueous solution, two
main kinds of interaction, which govern the process, can be distinguished. Namely, the
counterion-absorbed surfactant ion and counterion-water interactions.
The first interaction is essentially of an electrostatic nature and thus depends upon the ionic size
and ionic charge [11]. The ion-water interaction is much more complex - the structure of water and
its changes caused by dissolved ions plays an important role. Substantial dissimilarity of cations
and anions behaviour in aqueous solution means that the anions are generally lower-hydrated than
the cations. Small ionic species and those of large ionic charge were found to be 'water structure
makers', because of their high electric field which polarizes, immobilizes and contracts water
464

molecules at the intermediate vicinity to an ion [23]. There is also an ordering of water molecules at
long distance range. As a result, negative values of hydration entropy and of ionic partial molal
volume are observed. Simultaneously, lower mobility of water molecules in the hydration sphere
over these far bulkier water molecules results in an increase of viscosity at the intermediate vicinity
of an ion and thus in possitive values of viscosity coefficient, B.
On the other hand, large monovalent anions were found as 'water structure breakers' [23].
Dipole-dipole repulsive interaction between water molecules in the hydration sphere of such anions
can allow for polarization, immobilization and electrostriction of water molecules only at the first
hydration layer. Outside this layer, the structure of the surrounding water is highly disordered.
Consequently, neighbouring water molecules are more mobile than bulk water and this results in
possitive values of the ion hydration entropy and of ionic partial molal volume and also in
negative values of the viscosity coefficient for large monovalent anions in aqueous solution.
These observations have been confirmed by the studies of monolayers formed at the air/water
interface by dodecyl-pyridinium halides. Parreira [24] explains the expansion observed in the
surface isotherms in the presence of electrolytes, like KCl, KBr and KI, by the formation of ion-
pairs. The tendency to form ion-pairs between the surfactant ions and the halide counter-ions in the
order I' > Br- > Cl- can be explained by the water structure-enforced mechanism, which was
proposed by Diamond [24].
My attempt to calculate the selectivity criterion for ion flotation from an aqueous solution by an
ionogenic surfactant is based upon the discussion of experimental results of batch and continuous
ion flotation, in connection with a thermodynamic analysis of the adsorbed surface active-
counterion and ion-water interactions. The ion flotation results illustrated in Figures 2, 3 and 4
show that trivalent metal cations have the highest affinity to anionic surfactants, while the
monovalent metal cation has the lowest affinity to surfactants. The sequences of growing affinity of
metal cations to anionic surfactants are the same as the sequences of ionic potential values:

ionic potential 2.08 2.25 2.27 3.80 3.95 0.78 1.83 3.23

Presented data confirm the selective foam fractionation model proposed by Rubin and lome
[11], which is based on the Gouy-Chapman theory of the diffuse double layer, with the restriction
that the closest approach to the surface is determined by the size of the hydrated ions.
Quite a different situation occurs in the case of anionic flotation, where monovalent large ions
have the highest affinity to a cationic surfactant. It suggests that ion-water interactions govern the
selectivity of the ion flotation process in this case. Since only very few thermodynamic data for
complex anions are available, the characteristics of ion-water interaction were based on
experimentally evaluated values of viscosity coefficients for adequate aqueous electrolytes. At
electrolyte concentrations lower than O.IM, the relative viscosity of the solution (lJr) is given by the
Jones-Dole equation [25]:

(9)

where lJ and TJs are the viscosities of solution and water, respectively.
Values of the B coefficients were experimentally determined for potassium salts of several
cyanide complexes at 25°C. A negative value of the viscosity coefficient was found only for
 465

KAu(CN)4 salt, whereas this coefficient was positive for the other salts studied. There is a
correlation between the foam separation selectivity sequence and the values of an adequate viscosity
coefficient for the studied electrolytes:

Au(CN)z- > Ag(CN)z-

B: 0.020 0.041

and

B: -0.039 0.0057 0.0078 0.0086

This reinforces the previously fonnulatcd selectivity criterion for quaternary ammonium surfactants
towards oxyanions, where higher affinity to the surfactant was detennined for less hydrated anions
and the ion-water interaction was found to govern the affinity of an anion to the cationic surfactant.

4. Selectivity of Ion Flotation in the Presence of Sublate in the Bulk Solution

In aqueous solutions containing di- or trivalent complex anions, like Zn(CN)4 2 - or CO(CN)6 3-, and
a cationic surfactant such as cetylpyridinium chloride, at concentrations of the components higher
than lxlO- 4 M, sublates are precipitated in the bulk solution. The presence of the sublate
determines the run of ion flotation and a characteristic band on flotation curve is observed [26-28].

I
1.0

50 2 -
4
2 C(-
u
'- 3 CW
u
4 Br-
0.5 5 N03

0.00
150
TIME, min
Figure 7. Flotation curves for ion flotation of I.Ox 1O-4M Co(CN)6 3 - by 3.5xIO- 4M CPCI (concentration
of anions O.OO2M).
466

The main factor influencing a sublate flotation is the presence of electrolytes in the bulk solution.
When there is no sublate in the bulk solution, the influence of electrolytes on the rate and efficiency
of ion flotation is disadvantageous [28]. A quite reverse effect of electrolytes is observed when the
sublate is precipitated in the bulk solution. This is confirmed in Figure 7, in which kinetic curves of
CO(CN)6 3- anions with CPCl in the presence of various anions at a concentration of O.OO2M are
shown. The accelerating influence of electrolytes on the flotation of Co(CN)6 3- is seen to increase
in the following sequence:

To explain the results observed, electrokinetic potentials of aqueous suspensions of the sublates
studied were measured in identical conditions as in ion flotation experiments. The values of
electrokinetic potentials for these sublate suspensions versus partial molal volumes of anions in
aqueous solutions are pre&i:nted in Figure 8. It may be seen that anions with the lowest values of
electrokinetic potentials of sublate suspensions are those, which accelerate most the flotation rates.
The order of accelerating influence is the same like the order of increasing values of partial molal
volumes. The larger, less hydrated anions, such as N03 - or Br- accelerate the flotation rate of the
sublate much more than smaller, more hydrated anions, such as soi-, Cl- or CN-.

Figure 8. Electrokinetic potentials of aqueous suspension vs. partial molal volumes of anions for
(CPhCo(CN)6 sublate precipitated in the presence of a concentration of O.OO2M of anions.

In the next series of experiments, flotation of a synthesized (CP)3CO(CN)6 sublate was carried
out. The (CPhCo(CN)6 sublate was precipitated out in a column, separated from solution by
filtration, washed with distilled water and finally dried at ambient temperature. The aqueous
 467

suspensions of sublate were prepared and floated in a flotation column; electrokinetic potentials
were measured as well. Both flotation curves and electrokinetic potential values for (CP}JCo(CN)6
sublate are illustrated in Figure 9. After 10 min, the flotation run is practically over; this indicates
that the synthesized sublate is floated much faster than the sublate precipitated inside the flotation
column. The reason for this is the high hydrophobicity of the surface, which is confirmed by the
values of the electrokinetic potential for aqueous suspensions of the sublate.

1.0

u 1 . subia te
~ pot. • mY
+ 4.9
...
"- 2. sublate + 1.0·1Q-5 M CPCI +13.5
u 3. sublate + 3.5 '10- 6M K3Co (CN)6 0.0

0.5

~-- __ 2
-----1
~~----~------.3
o 5
TIME. min

Figure 9. Flotation curves for synthesized (CPhCo(CN)6 sublate in the presence of K3CO(CN)6 and
CPCl, with values of the electrokinetic potential of this sublate.

To better explain the behaviour of sublate flotation, the electrokinetic potentials of aqueous
suspensions of (CP)3CO(CN)6 and (CP)2Zn(CN)4 sublates were measured as a function of
complex anion concentration (Figure 10). The values of electrokinetic potential for the aqueous
suspension of sublates, at Co(CN)6 3- and Zn(CN)42- concentrations which result from sublate
solubility in water, are somewhat higher than zero. On the other hand, the values of the
electrokinetic potential are equal to zero, when the values of the negative logarithms of the anion
concentrations (-log[Co(CN)63 -] and -log[Zn(CN)4 2 -]) are 5.30 and 4.00, respectively. It means
that the tendency of cyanide complex anions to transfer from the sublate surface into solution is
stronger than the similar tendency of cetyl-pyridinium cations. It indicates that in the inner
Helmholtz plane of electrical double layer, cetyl-pyridinium cations are more adsOlbed than cyanide
complex anions. This results from a stronger polarity of the large cetylpyridinium cations.
The selective flotation of ions in the absence of sublate in the bulk solution, which was presented
in chapter 3, shows that there is a possibility of separating ions from multi-ion systems. Another
possibility of selective ion flotation exists, when there is a difference between sublate solubility of
floated ions. An example of such a separation is illustrated in Figure 11. It may be seen that
Au(CNk anions are floated faster, since they do not form a sublate in the bulk solution. Quite a
different situation exists, when the CN- concentration is much higher, i.e. 0.25 M (Figure liB);
468

4.0 5.5
>
E
+60 ~
N

....0
Q.
+40
(.)
....
1.1.1
z +20
~
0
....0:
(.)
0
1.1.1
..J
1.1.1
-20

6 7 8
-log [Co (CN )~-I

Figure 10. Electrokinetic potentials of aqueous suspensions of (CPbCo(CN)6 and (CPhZn(CN)4

sublates vs. cyanide complex anions concentmlion.

1.0 A

0.5

o 50 o
TIME. min

Figure 11. Flotation curves for competitive ion notation of 5.0x 10- 5 M cyanide complex anions by
2.0xlO- 4 M CTMACI at a cyanide concentration of (A) 1.0xlO-4 M and (8) 0.25 M.
 469

the accelerating influence of cyanide anions on the flotation of Pt(CN)42- sublate causes a much
faster flotation of these anions.

5. Ion Flotation Kinetics

The first kinetic studies of ion flotation were carried out by Rubin and his coworkers [29-30]. For
homogeneous systems a first-order reversible equation can be applied. In terms of flotation
removal. R. the integrated form of the rate equation is :

0.434 k Ci
log (M - R) =10gM - M t (9)

where M is the maximal flotation removal. t is the time, k is the rate constant and Ci is the
concentration of ions at time O.
For insoluble systems. i.e. when the sublate is precipitated in the bulk solution. the equation is
exponential :

log (M - R) =10gB - m log t (to)

where m and B are constants. Their product. m.B. is the rate constant in the differential form of the
equation.
According to Shakir and Samy (31), the colloidal sublates. which are formed between the
dissolved thorium species and anionic surfactants, are removed with rates described by a first-order
equation [9]. According to Golman (3) and Skrylev and Amonov [32]. the results of ion flotation
from homogeneous systems were found to fit into the zero-order equation:

C=Ci-kt (11)

On the other hand. Skrylev [32] has found that the rate of removal of colloidal sublates is described
by a second-order equation:

lIc - l/ci = kt (12)

The brief review presented here shows that the kinetics of ion flotation is not so far clearly
established. In this study. the kinetics of ion flotation of cobalt(lI) and cobalt(lII) ions have been
investigated as an example. A Turbo-Pascal program (for IBM PC microcomputers) has been
developed for treating the flotation curves. The program calculates the value of the flotation order.
n. for which differences between real and theoretical curves are minimum. The rate constant values
are evaluated from the following equations :

for n = I.

k=-ln£..:....'1 '" ~ (13)

ci - cr Ci t
470

and forn* 1

Cj - cr 1
k = cj(n - I)t [(c _ cr)n-l

The measure of the error is the relative standard deviation of the concentration:

st.dev. = Cj _ c r * 100% (15)

where N is number of points on the flotation curve.

Cj
M = 0.628
k = 0.0262 min- 1
n = 1.11
st.dev. =2.05 Z

time

Figure 12. Kinetic curve for the ion flotation of 5x 10- 5 M Co2 + by 2xlO- 4 M DBSNa with results of
the kinetic calculations.

In the first series of experiments, the C0 2+ cations were floated using sodium dodecyl-benzene
sulphonate as an anionic surfactant. In this case, no sublate was precipitated in the bulk solution.
An example of the flotation curves is shown in Figure 12; values of final flotation removal,
flotation rate constant, flotation order, and standard deviation arc presented. It may be seen that the
flotation order is close to one (equal, in fact, to 1.11). For this order, the removal was found to fit
reasonably into the kinetic equation with a standard deviation of 2.05%. The first-order equation
gives also a quite reasonable fit to the data, wi th a standard deviation of 2.70%.
A quite different situation occured when Co(CN)6 3 - anions were floated by cetyl-pyridinium
chloride. In this case, the sublate was precipitated in the bulk solution and, as it may be seen in
Figure 13, its presence has a substancial influence on the flotation curve. The kinetic curve is not
._- ~ ~--- -~. h' .h~ AM. r"'.<{)n~hJv. since the standard deviation is 10.5%. Dividing this
 471

Cj
M = 0.684
k = 0.382 min- 1
n =0.669
st.dev. =10.5~

Part 2

Cj

o
Pad 1 M = 0.504
k =3.30 min-1
n = 0.0000
C r 1 - - - - - - - - - - - 1 st. dev. =3.70 l f - - - -

M = 0.574
Cj k = 0.120 min- 1
n = 1.19
st. dev. =2.60 Z'

r TINE

Figure 13. Kinetic curve for the ion flotation of 5xlO-5 M CO(CN)6 3 - by 2.5xlO- 4 M CPCI with results
of the kinetic calculations.
472

curve into two parts leads to a better correlation of the theoretical model with the data. The first part
of the flotation curve follows a zero-order equation with a standard deviation of 3.7%, while the
order of the second part is close to one (equal to 1.19), with a standard deviation of 2.6%.

6. References

1. Sebba, F. (1962) Ion Flotation, Elsevier, Amsterdam.

2. Lemlich, R. (ed.) (1972) Adsorptive Bubble Separation Techniques, Academic Press, New
York.
3. Golman, A.M. (1982) Ion Flotation, Nedra, Moscow.
4. Wilson, D.l., and Clarke, A.N. (1983) Topics in Foam Flotation, Marcel Dekker, New
York.
5. Grieves, R.B. (1975) Foam separation: A review, Chern. Eng. I. 2, 93-106.
6. Mizuike, A., and Hirade, M. (1982) Separation and preconcentration of trace substances -
III. Flotation as a preconcentration technique, Pure & Appl. Chem.~, 1555-1563.
7. Caballero, M., Cela, R., and Perez-Bustamante, I.A. (1990) Analytical applications of some
flotation techniques - a review, Talanta 37, 275-300.
8. Burgess, 1. (1989) Ions in Solution: Basic Principles of Chemical Interactions, I. Wiley and
Sons, New York.
9. Walling, Ch., Ruff, E.E., and Thornton, J.L. (1957) The adsorption of cations by anions
foams, I. Phys. Chern . .6l.486-489.
10. Wace, P.F., Alder, P.I., and Banfield, D.L. (1969) Foam separation process design, Chern.
Eng. Progress Symp.Series 65, 19-28.
11. Rubin, E., and lome, I. (1969) Ion fractionation by foams, Sep. Sci. :!,313-320.
12. Huang, R.C.H., and Talbot, F.D. (1973) The removal of copper, cadmium, and lead from
dilute aqueous solutions usillgfoamfractiollation, Can. I. DIem. Eng. .2l, 709-713.
13. Kubota, K., and Kayashi, S. (1977) The removal of sodium, cadmium and chromium ions
from dilute aqueous solutions using foam fractionation, ibid.~, 286-292.
14. Grieves, R.B., Burton, K., and Craigmyle, J.A. (1987) Experimental foam fractionation
selectivity coefficients for the alkali (Group I A) metals, Sep. Sci. & Techno!. 22, 1597-
1608.
15. Charewicz, W., and Gcndolla, T. (1972) Ion flotation of cyanide complexes of gold(1) ,
Chemia Stosowana lQ, 383-391.
16. Charewicz, W., Walkowiak, W., and Gendolla, T. (1976) unpublished data.
17. Walkowiak, W., and Grieves, R.B. (1976) Foamfractionation of cyanide complex anions of
Zn(II), Cd(II), Hg(l/) and A u(lII) , J. Inorg. Nucl. Chern. 38, 1351-1356.
18. Walkowiak, W., and Rudnik Z. (1978) Selectivity coefficients for Ag(CNJ2' and Au(CNJ2'
from continuous foam fractionation with a quaternary ammonium surfactant, Sep. Sci. &
Techno!. .1.3,127-139.
19. Grieves, R.B., Bhattacharyya, D., and The, P.J.W. (1973) A cationic surfactant as a soluble
ion exchanger in the continuous foam fractionation of nitrate, iodide, and bromide, Can. J.
Chern. Eng. i1. 173-177.
20. Grieves, R.B., Drahuskuk, R.L., Walkowiak, W., and Bhattacharyya, D. (1976) Selectivity
coefficients for divalent oxyanions from the continuous foam fractionation of quaternary
ammonium surfactant, Scpo Sci. 11, 241-253.
21. Gendolla, T., and Charcwicz, W. (1981) unpublished data.
 473

22. Gendolla. T .• and Charewicz. W. (1979) Note: Selectivity coefficients for Zn(CN)/·.
Cd(CN)/- and Hg(CN)/- from continuous foam fractionation with quaternary ammonium
surfactant. Sep.Sci. & Techno!. H. 659-662.
23. Kavanau. J.L. (1964) Water and Solute-Water Interactions. Holden-Day Inc .• London.
24. Parreira. H.C. (1969) Counter-ion effects in soluble monolayers. J. CoIl. Interphase Sci. 12.
235-241.
25. Jonnes. G .• and Dole. M. (1929) The viscosity of aqueous solutions of strong electrolytes
with special reference to barium chloride. 1. Am. Chern. Soc. ~. 2950-2964.
26. Walkowiak. W .• and Bartecki. A. (1973) Flotation of complex anions of the VIII subgroup.
I. Flotation of PdCli- ions. Nukleonika U. 133-141.
27. Walkowiak. W .• and Bartecki. A. (1973) Flotation of complex anions of the VIII subgroup.
II. Flotation of PtCl,:- ions. ibid. U. 209-218.
28. Walkowiak. W. (1980) 10njZotation and solvent sublation of cobal:-cyanide complexes. J.
Chern. Tech. Biotechnol..:ill. 611-619.
29. Rubin. A.J. (1968) Removal of trace metals by foam separation processes. J. Amer. Water
Association 60. 822-845.
30. Rubin. A.J .• Johnson. J.D .. and Lamb. J.C. (1966) Comparison of variables in ion and
precipitatejZotation. Ind. Eng. Chern .• Process Design Develop.~. 368-375.
31. Shakir. K.. and Samy. S. (1979) Kinetics studies on the foam separation of thorium( IV) with
sodium lauryl sulphate. ColI. Polymer Science ill. 420-426.
32. Skrylev. I.D .. and Amonov. K.B. (1973) Kinetics of ion flotation. Z. Prikl. Chimii 46.
819-824.
33. Skrylev. L.D .• Legaczenko. LA .. Menczuk.W.W .• and Kondratieva. S.S. (1981) Kinetics
ofuraniumjZotation removalfrom diluted carboxylic solutions. ibid. 54.2137-2140.
4.3. FLOTATION TECHNIQUES IN WASTE WATER TREATMENT

A.I. ZOUBOULIS, K.A. MATIS and G.A. STALIDIS

Laboratory o/General and bwrganic Chemical Technology
Department o/Chemistry. Aristotle University
Thessaloniki 54006
Greece

ABSTRACT. In today's world of environmental strain, waste water treatment is becoming a more
or less conventional application of flotation as, for instance, in the oil, food or chemical industries,
sewage treatment, etc. In this paper, different flotation methods (ion flotation, precipitate flotation,
adsorbing colloid flotation), and techniques (dispersed-air, electrolytic and dissolved-air flotation)
are rewiewed and several applications of these for waste water treatment are presented, noting the
need for further chemical engineering development and knowledge of this process. Selected bench-
scale experimental results are also given. specifically for the removal of dyes, phosphates and toxic
metals, including chromium, silver and copper, the latter on pyrite fines. The processing of dilute
aqueous metal solutions for precious metals recovery is stressed. possibly as an alternative to
hydrometallurgy .

1. Introduction

1.1. HISTORY OF ROTATION

Flotation is a rather unique separation process, combining three separate phases. gas/liquid/solid.
It originated in minerals processing (known as froth flotation). a field which presents special
peculiarities compared with the application of flotation to wa~te water treatment.
A systematic application of flotation in minerals beneficiation had already started the previous
century. i.e. in 1860 by Haynes on sulphides, to separate them from gangue with the use of oil.
The first apparatus of froth flotation was recorded, in 1905, in Broken Hill of Australia. where air
bubbles were produced by intense stirring of the pulp [1]. Quite early, applications of this method
have also been reported in water and wastewater treatment [2].
The scope of this paper is to deal with the application of flotation in waste water treatment. a
field covered by Hahn in the previous similar Advanced Study Institute [3]. The application of
flotation to water treatment is dealt in this volume by Zabel.

1.2. BUBBLE GENERATION TECHNIQUES

According to the technique used for the generation of gas bubbles (usually air). which are the
transport medium of this separation process, two broad categories of flotation are generally

475
P. Mavros and K. A. Matis (eds.). Innovations in Flotation Technology. 475-497.
Ii:) 1992 Kluwer Academic Publishers.
476

Re% Zeolite
100 i~-
?V-- .
80
--" ~ J
1
•
i -..

60
-" --7 !
j
40
!

1
20

o I !
o 0.6 1.6 2 2.6 3
[Dodeoylamlne), Mx10E(6)

Figure 1. Zeolites removal by flotation; comparison of (I) aqueous and (2) alcoholic solutions of the
surfactant. Reprinted with permission from ref. [9]; copyrighl1990, Gordon-Breech.

distinguished:
- dispersed-air flotation, where electrolytic and froth flotation belong; the latter is applied mainly in
the mineral processing field, with rather large air bubbles being generated, with a mean diameter
of 0.5-1 mm; and
- dissolved-air flotation - in which vacuum (and pressure) flotation are also classified - finds
applications mainly in water and waste water treatment; smaller bubbles of diameter 0.05-0.1
rnrn are generated.
Today, a number of companies construct and sell suitable flotation units of various sizes, such
as Wemco, Oow, Krofta, Water Management, I.e.I., Simon Hartley, P.O. Process Engineering,
etc.

1.3. SURFACTANTS

flotation is, in principle, a separation process based on surfactants [4]; the most applicable of them
are usually termed collectors. The chemical equilibria and the effect of pH value on the form of
surfactant in flotation aqueous solutions should always be carefully examined and kept in mind for
each system, as for instance that of dodecylamine, a rather common cationic surfactant [5]. The
ionisation of this amine is a function of pH and the respective species may (or may not) help in
flotation.

1.4. EFFECT OF PARTICLE SIZE

The significance of particle size and its consequences in the flotation process is well known [6].
Usually, the coarser size is that of interest in mineral processing, for obvious economic reasons.
Nevertheless, it is the very fine (or ultrafine) particles, having diameters lower than 20 !ffi1, which
create most problems in the flotation of minerals [7]. A solution to this problem has been proposed
 477

Re% x/m
100 100
2! .. ,
.~
80 80
I •

i
60
60

40
~ 40

20
----,~ 20

o ~Vi 0
o 50 100 150 200 250 300
Aclntol FA-l, (mg/I)

Figure 2. Adsorption of a commercial fatty acid (Acintol) on dolomite fines at natural pH. (1) adsorbed
quantity Cads/m (ppm/g); (2) removal of fatty acid (%). Reprinted with permission from ref. [11];
copyright 1989, Gordon-Breech.

by using finer bubbles. produced by dissolved-air or electrolytic flotation [8J. In wastewater

treatment. the removal of turbidity - consisting mainly of fine particles as clays, oxides, etc. - using
flotation, has also been examined; Figure 1 presents some of the results obtained in the case of
zeolites, in which toxic metals have been previously ion-exchanged [9,10].
The adsorption of fatty acids (typical anionic collectors) on fines has also been studied [II]. as
shown in Figure 2. for the case of dolomite. a widely found salt-type mineral, with good results.
Subsequently, the addition of fine mineral particles (diameter < 45 ~m) - which are usually
considered as solid industrial wastes!byproducts, produced during mineral processing operations -
was recommended as a means of removing residual flotation reagents by adsorption; i.e. as a
purification method of the final effluent, if this is not recycled in the processing circuit.

2. Flotation Methods and Techniques

2.1. ION FLOTATION

The method generally involves the removal of surface-inactive ions from aqueous solutions by
adding the proper surfactant; as a result. a scum appears on the surface of the solution. consisting
of solids and containing the surfactant usually as a chemical constituent. aftcr the passage of gas
bubbles through the solution.
Figures 3 and 4 present some of the results. obtained from applying the method to the recovery
of germanium. a valuable metal. The flotation mechanism could be described by the following
equations :
478

100 RdQe RdQe

80 /
:r
i ~ l00r--TI/--'-·~=::::ti_;:::::::::::in

eo ......
V
:-~(·I.·..··.I. . ,. . . . . . . . . . .:. . . . . . . . . . .;. . .
/

if
60

_I
1

1. . . . .
40

, i ,,1' +............................ ·..···,.···························+.··1

V /
20
:
i
o o.---~l-----+----~
o 234 o 2 4 e
[pyrogalloll MxlOE( 4) [Dodeoylamlnel. MxlOE(4)

Figure 3. Ion flotation of germanium from dilute solutions. Effect of pyrogallol concentration, keeping
constant dodecylamine at 2.2x 10-4 M (A), and of dodecylamine concentration, keeping constant pyrogallol
at 3.3xlQ-4 M (B). Reprinted with permission from ref. [12]; copyright 1987, Elsevier.

where L represents the ligand-activator and S+ denotes the cation of the surfactant (in the Ge
experiments. pyrogallol was used as the ligand-activator and dodecylamine as the surfactant). The
selection of surfactant type and concentration is of great importance, as is also shown in Figure 4.
It should be noted that the forementioned is not a typical example of ion flotation, since the use of a
suitable activator was necessary, for the transformation of molecular (uncharged) germanium
compound into an anion.
Sebba, who established ion flotation back in 1959, proposed the use of the so-called aphrons
(dispersions of gases in liquids) for these separations [14). As it will be seen later. the main
advantage of ion flotation seems to be its high selectivity, i.e. between different metals in their
dilute aqueous solutions. However, ion flotation has mainly remained in the laboratory or pilot
scale, and only during the last few years applications of this method in full industrial scale have
been reported in the literature [15).

2.2. PRECIPITATE FLOTATION

It is an evolution of the previous method, which involves initially the formation of some "kind" of
precipitate of the ionic species from dilute aqueous dispersions and its subsequent removal by
 479

Re% Ge
100

BO

50

40

20

0
OB 010 012 014 015
Number of Oar bon atoms

Figure 4. Effect of aliphatic primary amines (surfactants), having different number of carbon atoms in
their chain, on germanium flotation. Reprinted with permission from ref. [13J; copyright 1988, Marcel
Dekker.

RaW, Zn

•
100 .,---..,..----,,..--...,---,..--~----,

60+·· ........· .. ,· .. ··············1··········,,/.·····;········\:,·······..··!·..·················,···················l

60
80 +1/· 'i'ii.i.,.II ............... .
.. i:j.li.,.,. . H.iH ......i .. ,.... i.. i

40+················+··············/···,,····················;·· .. ··················;·\·················f·····............ 1

70 •.......;.... i;,···;··j··HH·······
20 + . . . . . . . . . .;. . . ./.............;.....................;. . . . . . . . . . . . . . . . . . . .".····,·· . · · . · · · · . ··1

o+-~~--~--~--~---+--~
e
"
8 9 10 12 10 100 1000
pH Iln1 mg/l

Figure 5. Removal of zinc by flotation; (A) effect of pH and (B) effect of initial zinc concentration.
Reprinted with permission from ref. [17]; copyright 1987, Technical Chamber of Greece.
480

transfer on air bubbles to the surface of the solution. According to Pinfold [16], there are three
categories (termed kinds) of precipitate flotation.
Figure 5(A) shows an example of precipitatc flotation of zinc (as a hydroxide), occuring
preferably at the pH range of 9-9.5; in Figure 5(B), the effect of varying initial concentrations on
zinc removal is also illustrated.

2.3. ADSORBING COLLOID FLOTATION

The method involves the removal of a solute by adsorption on or coprecipitation with a carrier floc,
which is then floated, usually with the help of a suitable surfactant [18]. Figure 6 shows results
obtained with As(V); in this case, the presence of surfactants was not necessary for the efficient
flotation and removal of arsenic. However, As(III) was also removed, using the same method; in
this case, the further application of sodium oleate was found to be necessary [20].

Ra% As(V)
100~--~4P~--~~========~~====~

BO

60

40

20 -~-----~-~-------~---+---------------.-~---------~.~

O+-------~-------+------~--------+-------~~

o 40 BO 120 160 200

Fe(lll), (mg/l)

Figure 6. Adsorbing colloid flotation of As(V) by ferric hydroxo-compJexes at pH:5.0 ± 0.1. Initial
As(V) concentration: I, 10; 2, 50; 3, 100 mg/1. Reprinted with pennission from ref. [19]; copyright
1989, Marcel Dekker.

2.4. DISSOLVED-AIR FLOTATION

This bubble generation technique has bcen used for the selective separation of copper by ion
flotation, in presence of As(V), presented in Figure 7. A short-chain xanthate has been applied as
collector for the copper ions, at pH 3.0-5.0.
A statistical approach to the (dissolved-air) flotation of copper and zinc, precipitated as
sulphides, by cationic SUrfactants at a pH of about 2 has also been reported [21]. The experiments
were designed by a full factorial experiment and the results treated statistically in order to optimise
the process.
 ---
481

Re%
100 ~u T -
i
80
!
,
:
60 ,
I
. :
40 , ,
, I
20
ASI
o 1 ~

2 3 4 5 6
pH

Figure 7. Selective flotation of copper from arsenic ions by potassium ethyl-xanthate (10% excess of
stoichiometric). Initial Cu and As(V) concentrations equal to 50 mg/I. Reprinted with permission from
ref. [19]; copyright 1989, Marcel Dekker.

2.5. ELECfROLYTICFLOTATION

This bubble-generation technique has been investigated for the removal of metals from their dilute
solutions; chromium(VI) ions were effectively removed [22], as seen in Figure 8. Dissolved-air
and dispersed-air flotation techniques were also tested and the results compared [23J.

Re% Cr(VI)
100

80

60

40

20

0
0 100 200 300 400
Current Density (A/m2)

Figure 8. Effect of current density on Cr(VI) removal, using different collectors (in tap water): (1)
dodecylamine, (2) cetyl-trimethyl-ammonium bromide. Reprinted with permission from ref. [22];
copyright 1989, Sci. and Techn. Letters.
482

It should be pointed out that toxic inorganics, even in trace amounts, have to be removed in high
degree from industrial discharges, because they could cause problems in the biological waste water
treatment plants. This is a promising field for the application of flotation as a pretreatment method.
Further on, other conventional methods are not as effective; for example, sedimentation and
flocculation remove less than 80% of them.
In oil industries, for the primary separation in waste water treatment, the American Petroleum
Institute gravity separator is often used; its function is mainly to separate the free oil from waste
waters, but the efficiency is not more than 70-80%. For this reason, electrolytic flotation has been
applied for the destruction of the remaining oil emulsions, as a secondary (polishing) treatment
step, with high efficiency [24].

2.6. PARAMETERS OF THE PROCESS

Many parameters affect the process; the most important are [25] ; reagents concentration, pH value
ofthe solution, ionic strength, gas flowrate, flotation time and feed concentration.

Re'll Pb
100
-
---
~
80

60 -----K ~
V y
'-.
V

40

20
r--7 I
V
o
o 1 2 3 4

Reagent Concentration (Mx10E(4))

Figure 9. Effect of variable reagent concentration on the dispersed-air flotation of Pb (lxlO-4M):

(I) variation of collector (dodecy\amine) concentration; (2) variation of activator-ligand (pyrogallol)
concentration. The concentration of the other reagent was kept constant each time and equal to 2x 10- 4 M.
Reprinted with permission from ref. [25]; copyright 1990, Gordon-Breech.

An example for the effect of reagents concentration is given for the case of lead removal, as
shown in Figure 9, where a similar mechanism with the germanium removal has been investigated.
On the other hand, one of the most critical parameters, as presented illustratively for Cr(VI) in
Figure lOA, is the pH value, while the effect of ionic strength is presented in Figure lOB [17].
The gas flow rate I flotation time parameter has been shown for the case of germanium in Figure
11; it can be noticed that an increase in retention time was needed to obtain the same flotation
recoveries, when using lower flowrates. Example of the feed concentration effect has already been
presented (Figure 5B).
 483

Rd
_~Q~(V~I)__~~__~__~__~-.
100,....

BO+···········,···············-········;··············......-... .; ..• -........ ~.+- ...... -.-..;..~ Bna··--·· .. ····iC......................; •.....•..... _ .... ···.;················...···.;..·· ..·•· ..·······················1

60+-·--·········-·····:····"·.. ,-·-·--·,···-····..·~·,,+,-,-·'"-,,--~······,·········,,·,I

40+'"····~·"··,,·····:,············,,-·,·,,;,,····,·········;·,,--····· ... -;,···~····........+··..·l

20+·····..···· ....·i....................... , ..........................·.. ,··,··\ .. •• .. ••... ·1

0+----+----~--~---4----~
3 6 7 9 11 0.03 0,06 0,09 0,12 0.16

pH [Salt] M

Figure 10. Removal of Cr(VI) ion by dispersed-air flotation. (A) effect of pH value; (8) effect of ionic
strength: I, NaCl; 2, NH4CI; 3, (NH4h S04; 4, (NH4hP04. Reprinted with permission from ref. [20).

Re% Ge
100

80

60

40

20

0
0 100 200 300
Retention Time, (8)

Figure 11. Flotation recovery of germanium ions versus time under various gas (nitrogen) flowrates;
1,100; 2,150; 3,200; 4, 300 cm 3/min. Reprinted with permission from ref. [20).
484

2.7. CONTINUOUS FLOW OPERATION - SELECTIVE SEPARATION

In laboratory flotation experiments continuous flow has proved to give comparable results with
batch tests. Figure 12 is an example for the system of coppcr-zinc/dodecylamine flotation from a
solution, where the metals were precipitated previously as sulphides [26]. It has been further
found out, that selectivity can be achieved by the careful selection of conditions. The ratio of
surfactant concentration to metal ion (to be removed) was equal to cp = 0.1-0.2, much lower than
the stoichiometric amount.

Ad Ael
eli ........................... •
• ••
100 100 ..............

•
80 •····..···..........·......·i·· ..·.... ....................,.........................._. ,................................ 80 ,..... ,,,.,".......... ,.._,,...... ..····.. "1- .....-.-'''-.......--".~ ..

flO .. _ ...................... i ............................... j ...........................+. 60 ........................... i ............. . . . . . +. . . . 'H .... M._._.._,....

_~

40 ............................., ............................... i.............-....-...... ~............ 40

20 20 ............. '''.'''................ .. .... " . -".~

o+--k~~~~==*=~==~
Zn'
0
Zn. + •
o 20 40 00 eo 0 20 40 00 eo
Time (min) Time (min)

Figure 12. Selective separation of the mixture Cu(Zn, using continuous flow dispersed-air flotation. Zn
concentration: 250 mg/I; Cu concentration: (A) 50 mg/!, (B) 250 mg/l. Reprinted with permission from
ref. [26]; copyright 1989, Marcel Dekker.

One of the important advantages of flotation, illustrated also in Figure 13, is the high selectivity.
Lead was reversely separated from germanium by dodecylamine as collector, after precipitation as
a hydroxide at pH 10 [27].

3. Application to Wastes

3 .1. SEWAGE TREATMENT

Flotation (dissolved .. air) was found to thicken reliably and effectively activated sludge up to 4%
w/v, under a wide range of conditions (varying S.V.I.'s) [28]. It was found to thicken the sludge
to 4% w/v faster than gravity would thicken it to 2.5% w/v sludge solids concentration, obtaining a
 485

Re%
100~--------------~------------------------~
•
_. _._. __. . ._._.______~"__._,_._. __,__. _._j __._"____.,_.,J. ____._. ____.___,_~.--.
80
f i l ; ;
60 -------------~----------f_----l.--------L-------!-------
: ! ' i
40 ._,.__....._._-_. . . . . . ._- . .--..--.-,-..~.--,- . .----.---~--. . .-.--.. ---.--~. - . .-..-.. ---.--.---.-.. ~.-.-., . .--.

20 ------------~--Ge
-------~ --------------T--====r
O~------~-----+------~------+------+--~
o 2 3 4 5
(Pb] (Mx10E(4))

Figure 13. Effect of initial lead content on separation of Pb/Ge with dispersed-air flotation;
[Ge]: Ix 10- 4 M, pH : 10.0-10.3, collector ratio <p: 0.5. Reprinted with permission from ref. [27];
copyright 1990, Marcel Dekker.

clear effluent in a shorter period of time. In addition, the two products - the float and the effluent -
both contained large amounts of ai r, remaining aerobic for long periods of time.
An excellent design example was given by Bratby and Marais [29] for dissolved-air flotation
applied to activated sludge, with the following procedure:
- the air/solids ratio is calculated first, based on the influent waste water solids concentration. The
optimum value for depth float above water level is a function of the desired float solids
concentration. The solids loading rate is then calculated and the area of unit required for
thickening is found by a material balance; finally, the area required for clarification is given.
- For a given air/solids ratio, there is a limiting value of down-flow rate and this relationship is
found experimentally for each case. Details for inlet arrangements are also useful. A suitable
saturator pressure is usually chosen from locally available pumps and compressors.
The application of (electrolytic) flotation to sewage was first reported by Matov, using cells with
consumable aluminium anodes [30]. According to our knowledge, this researcher was the founder
of electrolytic flotation.
Some technical problems were reported with a circular basin [31]; in this casc, however, an
existing thickener was converted to a flotation unit. The dry solids concentration achieved was
promising. Finally, the economic aspects of the process, including basic principles, have been
discussed in the past [3].

3.2. OIL AND CHEMICAL INDUSTRY

Due to gravity reasons, oil removal should be perhaps the main application of flotation. The
advantages of electrolytic flotation in this sense have already been analysed [24,32). The current
density (100-300 A/m2), the conductivity of the solution, the depth of the tank and the initial
concentration of oil emulsion have been examined as operating variables. Electrodes of different
configuration and material were also tested. Preliminary information on flotation characteristics
486

[011], (mg/I)
80.-~~~--~--------,---------,---------.

60 +---~~__t - - - - - - + - - - - - + . - - - -

40

20

o+---------+---------+---------+-------~
o 40 80 120 160
Time, (min)

Figure 14. Pilot-plant scale removal of emulsified oil dispersion using electrolytic flotation. Effect of
depth (height) : 1, 1 m; 2, 1.5 m. Reprinted from ref. [32].

obtained from batch experiments were tested in a continuous unit. The kinetics and scale-up of the
process were then studied. with a surface area ratio of 0.017 to 0.36 m2. Typical results are shown
in Figure 14.
Gravity separation and dissolved-air flotation were also found to produce an effluent. that met
federal regulations for oily discharges. at one of the largest onshore installations in Alaska [33]. It
is noted. that the A.P.I. Design Manual [34] gave no prediction of the effluent quality from the
gravity separator, but only provided a separator size for specified gravity and flow rate. Such a unit
does not break or separate emulsions. nor it is designed for such purposes. An onboard
electrolytic flotation system was investigated. by Lockheed. for bilge and ballast water mixed with
seawater to increase conductivity. thus lowering the voltage applied [35].
flotation was also examined, among other methods. for heavy metals removal from wastewaters
[36.37]; obvious sources are metal processing, metal finishing and plating industries. Advantages
of each treatment scheme, with respective comparison, were discussed.
A successful effort to reduce the cost of biological treatment applying flotation as sludge
thickener has also been presented [38]. A case history was analysed from a major integrated
refinery and chemical plant of Exxon.

3.3. FOOD INDUSlRY

The feasibility of dissolved-air flotation in the treatment of a palm oil mill effluent was reported, but
relatively high operating costs were suggested [39]. Dissolved-air flotation was also applied to
slaughterhouse [40], poultry processing (41), seafood processing [42], soap and food industrial
wastes [43].
By-product recovery (Le. proteins. fats, oil, etc.) seems to be a potentially profitable aspect of
flotation. as it is shown by the following example. In a Swedish abattoir, the installation was based
upon chemical precipitation and dissolved-air flotation, followed by sludge treatment [44]. With an
 487

Re% Dye
100 ,
...
80 ,....--
~ • -'-
i

1~
i
-"- L i:
l"
60

,~ /
IT
:

I.
40
i
20 ,

o
V V i
o 30 60 90 120 150
[polyelectrolyte]. mg/l

Figure 15. Dissolved-air flotation of a dye effluent: effect of the addcd polyelectrolyte concentration on
dye removals in presence of (I) 0.88 gil CaCl2 and (2) 0.06 gil FC2(S04h.

Re% Dye
100
i
80 ,
• V </

60
, / V
40 r - i
i
'Kiil!,:
' ~

! ; [ ii
! i ; V"",.,
20

I
0
0,01 0.1
[Flocculant], gil

Figure 16. Dye removal as a function of flocculant addition: (1) Fe2(S04h. (2)CaCI2' Concentration
of polyelectrolyte (Zetag) : I. 50 mgll; 2, 100 mgll.
488

initial capital cost of £95,200, the net saving and contribution towards capital was £11,841 per
annum.

4. Other Applications

4.1. REMOVAL OF DYES

The application of flotation to dyes and paint effluents is quite old, but still active [45-47]. Recent
work in our laboratory applied the dissolved·air flotation technique (l L cell, 10% recycle) and was
tested for Irgalan dye; results are presented in Figures 15 and 16. The process involved coagulation
and flotation steps, i.e. it was similar to adsorbing colloid flotation. Two flocculants have been
tried, calcium chloride (at pH 12) and ferric sulphate (at pH 4), which hydrolyse to hydroxide.
Zetag, a cationic polyelectrolyte, kindly supplied by Allied Colloids (UK), has also been added.
The obtained recoveries were over 80%, even without the presence of a surfactant. Details of the
experimental apparatus and procedure can be found elsewhere [25].

4.2. SILVER RECOVERY

The recovery of silver by ion flotation has attracted the researchers' interest for many years [48],
with possible application to the treatment of photographic wastes and in particular the hypo
solutions used in photographic fixing baths, in which the thiosulphate is already present. Silver and
thiosulphate can interact in aqueous solutions to form three complex anions [48] :

Ag+ + S20:3 2. ? Ag(S20:3Y, kl =7.14x108

Ag(S203Y + S20:3 2- ? Ag(S20 3h3-, k2 = 4.01xlQ4
Ag(S203)3- + S2032- ? Ag(S20 3h5-, k3 = 6.47

For [S203]/[Ag] ratios greater than 1.0, the system is soluble; below 1.0, Ag2S particles are
formed, particularly for ratios below 0.7 :

Some of the results are presented in Figure 17, showing the influence of pH on dispersed-air
flotation of Ag, using dodecylamine as surfactant to float the thiosulphate-silver complex; flotation
time was 5 min and the experiments were conducted in a 100 cm 3 cell. The use of thiosulphate in
excess, i.e more than the calculated stoichiometric amount, and up to a pH value of around 9.5, has
been found to give the worst recoveries; while using less thiosulphate, the flotation pH from
around 3.5 up to 12 did not substantially affect the flotation of silver.

4.3. SEPARATION OF CR(VI) FROM CR(III) : A SPECIATION APPLICATION

Aotation may also be used for speciation studies. Other speciation methods commonly used are
electrodialysis, anodic stripping voltammetry, ion chromatography, etc. The need for simple,
precise, relatively cheap and selective speciation methods is a well known necessity, either for
 489

Re"
100 .-~.

80

60

40

20

0
2 4 6 8 10 12
pH

Figure 17. Effect of pH on ion flotation of Ag (50 mg/l) from solutions using dodecylamine (3.48x
10-4 M). Thiosulphate concentrations: 1, 1.16xlO-4 M; 2, 2.32xlO- 4 M; 2, 3.48 xlO-4 M.

environmental pollution research and control, or for waste water engineering, where according to
the existing species, a different removal method has to be followed.
As an example of speciation studies, the selective separation of Cr(III)/Cr(VI) waS investigated;
the Case of chromium constitutes, because of its aqueous chemistry, a special heavy/toxic metal,
forming anions as well as cations. Chromium has an unusual ability, too, to co-exist in two
soluble and stable oxidation states, having quite different geochemical and toxicological properties.
Previous work on chromium has been already cited [20,22,23].
Sodium dodecylsulphate (l x I 0-4 M in 1% ethanolic solution) was used as collector, in the pH
range 7.5-10.5; radioactive tracers were employed for the analytical determination of the species.
Figure 18 presents some of the results, showing hexavalent chromium (as anionic chromate) to
remain in solution, while 90% of the trivalent cation was floated (see also ref. [49] for more
details).
An interesting example of the influence of the ionic strength on the Cr(VI)/Cr(III) separation is
presented in Figure 19, where Na3P04 waS used. The pH value WaS measured after the salt
addition, without any other modification. The presence of the salt may alter the pH of the solution
outside the optimum range.

4.4. COPPER-ADSORBING FLOTATION

Copper ion removal from dilute aqueous solutions has been investigated by various procedures,
i.e. ion flotation by O-ethyl-dithiocarbonate [19] and precipitate flotation as sulphide [21, 26].
Copper (50 mg/!) adsorption on pyrite fines (of size < 45 !-lm) followed by flotation is reported
here. This can be considered as an extension of the previous work, where mineral slimes have
been used forthe removal offatty acids by adsorption and flotation [II].
490

Re% Or
100
i /- ~---0N
80 ,
60
I J ,
,

1 ~ i

-------
40
.- ,

. ~
20
2 ..L
..... I
o
2 4 6 8 10 12
pH

Figure 18. Ion flotation of a Cr(I1I)/Cr(VI) mixture (S mg/l each); (1) Removal of Cr(lll), (2) Removal
ofCr(VI).

pH

-
13

11

9
I

, 7
.......
:--...., .

i i

5
:
T 3
1.000E-04 1.000E-03 0.01
[Na3P041, M

Figure 19. Effect of Na3P04 concentration on Cr(VI)/Cr(lII) separation; (1) Removal of Cr(In);
(2) Removal of Cr(VI); (3) pH.

The principles of pyrite flotation in mixed sulphide ores processing are well known [50); today it
is believed that the mechanism of flotation is of electrochemical nature, due to dixanthogen fonned
on the pyrite surface. The role of copper sulphate in the mineral flotation of pyrite has also been
discussed [51); it has been found that this improves the recovery and grade of pyrite, when used in
appropriate dosages.
 491

Re%
100

80

50

40

20

0
2 4 5 8 10 12
pH

Figure 20. Adsorption of copper on pyrite and flotation of the fines using 20 mg/l of potassium ethyl
xanthate (KEX); (1) copper removal (%); (2) fines removal (%).

Re%
100

80

50

40

20

0
2 4 5 8 10 12
pH

Figure 21. Adsorption of copper on pyrite fines and collectorless dissolved-air flotation of the fines
(notation as in Figure 20).

The dissolved-air technique has been applied in this case (initial volume 1 L, 10% recycle) to
pyrite fines from the Stratoni area, in Chalkidiki (Northern Greece); the initial pyrite and copper
concentrations were 0.5 g/I and 50 mg/l, respectively. The time allowed for adsorption and
conditioning by the collector was 10 min each. Initially, potassium ethyl xanthate was tested as a
492

Re%
100
1 t. ......

80
.1 ..-

J--~2 .
rz
60 y~

40 ~", -iii--
20 V
o ~
4 6 8 10 12
pH

Figure 22. Adsorption of copper on pyrite fines and flotation of the fines by cetyl-trimethyl-ammonium
bromide (concentration of CTMABr 10 mg/l in 0.5% ethanol solution - notation as in Figure 20) .

ro~·······"····""""""+·""~-·:""-"i'"''''''''''''''''''''''''''''''''''''''+ ............................; ....... - .......................- ...." ............ ~ -li . . . . . . . . . . . ·. . . . t. . ·_·. ·.·+. . . . ·.·...·. . ~
j'.. .................................... +. . . . . . . . . . . . . . . . . . . .+ . . -1

-10 . . . . . . . . . . . . . . . . . . ...l.

10+ . _. . . . . . . . . . . . . . . . . . . !'o.:

-1 Ii ............................. f . . . . . . . . . . . . . . ·-j . . . . ·. . ·. . . . . ·........+. . . . . . . . . . . . . . . . _·. .+. ·. . · . 1

•
j;,'", · ...... • .. • ........ • ..............

-20+---t---I----t----t--+-.....
0.2 0.4 0.8 0.8 o 0.2 0.4 o.e O.B
[Floooulantl (mole/me) [Floooulantl (mole/me)

Figure 23. Effect of added flocculant during flotation of phosphates (0.18 g/m3 polyelectrolyte); I, total
solids (mg/l); 2, total phosphates (mg/l); 3, effluent turbidity (TEIF); 4, dissolved phosphates (mg/l); 5,
l;,-potential (m V).
 493

collector. The influence of pH on recoveries is shown in Figures 20 and 21. In the latter, the
ability of (dissolved-air) flotation to float fines without any collector was illustrated, giving better
removals on both copper and pyrite, in comparison with the above case. Another collector, cetyl-
trimethyl-ammonium bromide (CfMABr) was also tested giving similar results (Figure 22).

4.5. PHOSPHATES REMOVAL

The removal of phosphates from surface waters is generally necessary to avoid problems, like
eutrophication, in mainland lakes - used for recreational purposes, particularly near the big cities,
such as Berlin. Common treatment methods for the removal of phosphates are either physico-
chemical or biological.
Experimental work was conducted at a pilot-plant scale induced-air flotation unit, constructed
by Kleindienst (Germany). A commercial coagulant, based on AI(III), was applied together with
Sedipur, a cationic polyelectrolyte; a concentration of 0.2 glm 3 of the latter was found to be
sufficient. No other surfactant was added. Measurements of!;'-potential were carried out in parallel.
Some of the results are shown in Figures 23 and 24. Rotation conditions were as follows: influent
raw water flow rate 2.2 m3/h, the pressure in the system was kept at 6.5 bar, with the help of a
centrifugal pump (used also for the partial recycle of the clarified water, at a rate of 0.5 m3/h) and
the corresponding air flow rate was 0.02 m3/h [52].

1\ =-==+ -

0.1+-----+-----+-----4-----~----~----~~
i
o 0.2 0.4 0.8 0.8 1.2
[Polyelectrolyte) g/m3

Figure 24. Effect of added polyelectrolyte on the flotation of phosphates (0.2 mM flocculant); notation
as in Figure 23.

Long-term experiments, presented in Figure 25, were conducted for comparison of flotation
with sedimentation. Better results were obtained with flotation, in terms of turbidity, proving the
applicability of flotation at a larger scale. A general comparison is also presented in the same figure
between the turbidity of raw water before the sedimentation (but after the addition of chemicals -
flocculant and polyelectrolyte) with that following sedimentation, flotation, and rapid sand
filtration. The results were promiSing.
494

Turbidity, (TE/F)
50.---,---,---,---,---,---,---,---,---~--,

40+---+---+-+-~·--+-~~--~

30

20

10

o 2 4 6 8 10 12 14 16 18 20
Time (Hours)

Figure 25. Efficiency of flotation for phosphates removal, through comparison of turbidities between
flotation and sedimentation; 1, raw water; 2, before sedimentation; 3, after sedimentation; 4, after flotation;
5, after filtration.

5. Conclusions

From the discussion and results given above, it is made clear that the flotation process is quite
established in the field of waste water treatment. Certainly, it deserves more attention in different
new application fields, as well a~ for materials and by-products recovery; see for example the paper
of Dobias et al. in this volume. Due to the aforementioned advantages, further and closer
investigation is warranted.

6. Acknowledgements

Thanks are due to Messrs. N. Lazaridis and K. Kydros from our laboratory, as well as to Dr. L.
Goetz (lspra) and Prof. A. Grohmann (Berlin), for their help in many ways.

7. References

1. Crabtree, E.H., and Vincent, J.D. (1962) The early days of froth flotation, in D.W.
Fuerstenau (ed.) Froth Flotation - 50th Anniversary Volume, A.I.M.M.E., N. York, pp. 39-
54.
2. Hansen, C.A., and Gotaas, H.B. (1943) Sewage treatment by flotation, Sew. Works J. li,
242-252.
3. Hahn, H.H. (1982) Wastewater Treatment, in K.J. Ives (ed.) The Scientific Basis of
Flotation, Proceedings NATO ASI, Cambridge, U.K., 5-16 July, preprint.
 495

4. Scamehorn, J.F., and Harwell, J.H. (1987) An overview of surfactant-based separation

processes, in J. F. Scamehorn (ed.) Surfactant Science Series 26, Marcel Dekker, U.S.A.,
pp. 169-185.
5. Leja, J. (1982) Surface Chemistry of Froth Flotation, Plenum Press,N. York.
6. Schulze, H.J. (1984) Physicochemical Elementary Processes in Flotation, Elsevier,
Amsterdam.
7. Sivamohan, R (1990) The problem of recovering very fine particles in mineral processing - a
Review, Int. J. Miner. Process. 28., 247-288.
8. Matis, K.A., and GaUios, G.P. (1986) Dissolved-air and electrolytic flotation, in B.A. Wills
and RW. Barley (cds.) Mineral Processing at a Crossroads, M. Nijhoff, Dordrecht,
pp.37-70.
9. Zouboulis A.I., Matis, K.A., and Zamboulis, D. (1991) Foam Flotation for Fine Particles
Removalfrom Water: the example of Zeolites, Toxicol. & Environm. Chem.ll, (in press).
10. Zouboulis, A.I., Zamboulis, D., and Matis, K.A. (1991) Foam flotation of zeolites -
application on zinc ion removal, Sep. Sci. & Techno!.~, 355-365.
11. Matis K.A., Gallios G.P., and Zouboulis A.1. (1989) Adsorption of fatty acids on fine
mineral particles as a purification method of industrial wastewaters, Toxic. & Envir. Chern.
20, 425-435.
12. Matis K.A., Papadoyiannis LN., and Zouboulis A.1. (1987) Separation of germanium and
arsenic ions from effluents by flotation techniques, Int. J. Miner. Process.ll, 83-92.
13. Matis K.A.. Stalidis G.A., and Zouboulis A.I. (1988) Flotation of germanium from dilute
solutions. Separ. Sci. & Tech. n. 347-62.
14. Sebba, F. (1985) Separations using aphrons, Separ. & Purif. Methods H, 127-148.
15. Golman, A.M. (1988) Principles of theory and industrial application of ion flotation, in
K.S.E. Fossberg (ed.) Proceedings XVI Intern. Mineral Process. Congress, Elsevier.
Amsterdam, pp. 1701-1712.
16. Pinfold, T.A. (1972) Precipitate flotation, in R Lemlich (ed.) Adsorptive Bubble Separation
Techniques. Academic Press, N. York, pp. 75-89.
17. Zouboulis. A.I.. Matis. K.A., and Spathis P.K. (1987) Removal of zinc from solutions by
precipitateflotation. Techn. Chron. f:1, 5-27 (in Greek with extended english summary).
18. Clarke, A.N., and Wilson. DJ. (1983) Foam Flotation - Theory and Applications. Marcel
Dekker, N. York.
19. Stalidis G.A.. Matis K.A .• and Lazaridis N.K. (1989) Selective separation of Cu, Zn, and
Asfrom solution by flotation techniques. Sep. Sci. & Technol. 24, 97-109.
20. Stalidis, G.A., Matis. K.A., and Zouboulis, A.I. (1986) Flotation techniques for the
separation of trace pollutants, Chim. Chron. (New Ser.) 15, 133-46.
21. Stalidis, G.A., Matis, K.A., and Lazaridis. N.K. (1988) A statistical approach to precipitate
flotation ofcuSlznS, Int. J. Min. Process. ~ 203-216.
22. Zouboulis. A.I .• and Matis, K.A. (1989) Electrolytic flotation of chromium from dilute
solutions, Envir. Techno!. Let. lQ. 601-612.
23. Zouboulis. A.I., and Matis, K.A. (1989) lonenflotation von Chromo in D. Uhlig (cd.)
Fortschritte in Theorie und Praxis der Aufbereitungstechnik, F.I.A., Freiberg, pp. 418-424.
24. Matis, K.A.. and Backhurst, J.R. (1984) Laboratory studies of electrolytic flotation as a
separation technique, in J. Gregory (cd.) Solid-Uquid Separation. E. Horwood. Chichester,
pp.29-40.
25. Zouboulis, A.I., Matis. K.A., and Stalidis, G.A. (1990) Parameters influencingflotation in
the removal of metal ions, Intern. J. of Envir. Studies ~ 183-196.
496

26. Stalidis, G.A., Lazaridis, N.K., Matis, KA., and Papadoyiannis, LN. (1989) Continuous
precipitate flotation ofCuS/ZnS, Sep. Sci. & Technol., 24,1033-1046.
27. Zouboulis A.I., Matis, K.A., and Lazaridis, N.K (1990) Separation of germanium from
lead in solutions by flotation, ibid. 25,463-476.
28. Komline, T.R. (1976) Sludge thickening by dissolved-air flotation in the USA, in
Proceedings Wat. Res. Centre Conf., Flotationfor Water and Waste Treatment, Felixstowe,
U.K.
29. Bratby, J., and Marais, G.Y.R. (1977) Flotation, in D.B. Purchas (ed.) Solid-Liquid
Separation Equipment Scale-up, Uplands Press, Croydon, U.K, pp. 155-198.
30. Matoy, B.M. (1966) Consumable anode electroflotation cell, Elektron. Obrab. Mater. 1, 94-
96 (Transl. from Russian).
31. Preiss, W. (1988) Thickening of surplus sludge by means offlotation, Wat. Sci. Tech. 20,
217-222.
32. Matis, KA. (1982) Electrolyticflotation in industrial effluent treatment, Chim. Chron. (New
Ser.) 11, 159-171.
33. Lysyj, I. (1982) Treatment of oily wastewaters from onshore operations, J. W.P.C.F. 54,
309-315.
34. Amer. Petr. Inst. (1969) Oil-water separator process design, in Manual on Disposal of
Refining Wastes - Volume on Liquid Wastes, chapt. 5.
35. McKenna, Q.H., Helber, H., Carrell, L.M., and Tobias, R.F. (1973) Electrochemical
Flotation Conceptfor Removing Oilfrom Water, U.S. Coast Guard - Dept. Transp., Proj.
No. 4101.
36. Zouboulis, A.I., and Matis, K.A. (1988) Separation of heavy metals from effluents by
flotation, in H.H. Hahn and R. Klute, (cds.) Pretreatment in Chemical Water and Wastewater
Treatment, Springer-Verlag, Berlin, pp. 159-166.
37. Peters, R.W., Ku, Y., and Bhattacharyya, D. (1985) Evaluation of recent treatment
techniques fur removal of heavy metals from induJtrial wastewaters, A. 1. Ch. E. Symp. Scr.
.Bl(No. 243), 165-203.
38. Wozniak, J., and Collen, T.L. (1984) Case history - Optimisation of dissolved-air flotation
sludge thickener, reduces sludge treatment costs by eliminating additional dewatering
process, in M. La Crega and D. Long (eds.) Toxic and Hazardous Wastes, 16th Mid-Atlantic
Indust. Waste Conf., pp. 4 11-417.
39. Ng, W.J., Goh, A.c., and Tay, J.H. (1988) Palm oil mill effluent treatment - Liquid/solid
separation with dissolved-air flotation, BioI. Wastes 25, 257-268.
40. Cooper, R.N., and Denmead, C.F. (1979) Chemical treatment of slaughterhouse wastes with
protein recovery, J. W. P. C. F. n, 1017-1023.
41. Reed, S. W., and Woodard, F.E. (1976) Dissolved-air flotation of poultry processing waste,
ibid. 48,108-119.
42. Krofta, M., Wang, L.K., and Pollman, C.D. (1989) Treatment of seafood processing
wastewater by dissolved-air flotation, carbon adsorption andfree chlorination, in I.M. Bell
(cd.) Proceedings 43th Purdue Industrial Waste Conf., Lewis, Michigan, pp. 535-550.
43. EI-Gohary, F.A., and Abo-Elela, S.I. (1987) Management of waste waters from soap and
food industries: a case study, Sci. of Total Enyir. 66, 203-212.
44. Hopwood, A.P. (1975) The recovery of proteins and fats from waste water during effluent
treatment by physico-chemical processes, Proceedings I.Chem. E. Symp. Application of
Chemical Engineering to the Treatment of Sewage and Industrial LiqUid Effluents, York,
U.K, Symp. Ser. No. 44, preprint.
 497

45. Winget, R.L. (1964) Color removal and B.O.D. reduction in kraft effluents by foam
separation, Techn. Bull. no. 17, Nat'l. Council for Stream Improvment, Inc., N. York.
46 Backhurst, J.R., and Matis, K.A. (1981) Electrolyticjlotation in effluent treatment, J. Chern.
Tech. Biotech. n, 431-434.
47. Cheng, M.H., and Huang, S.D. (1988) Solvent sublation and adsorbing colloidjlotation of
Direct Dye, J. Colloid Interf. Sci. 126,346-354.
48. Bhattacharyya, D., and Grieves, R.B. (1972) Foam separation of complex anions: silver
thiosulphate, A.I.Ch.EJ.l8., 2()()-205.
49. Zouboulis, A.I., and Goetz, L. (1990) Ion flotation as a toolfor speciation studies - Selective
separation in the system Cr(IlI)/Cr(Vl), Toxic. & Envir. Chern. 21 (in press).
50. Fuerstenau, M.e., Miller, J.D., and Kuhn, M.e. (1985) Chemistry of Flotation, A.I.M.E.,
N. York.
51. 0' Connor, e.T., Botha, e., Walls, M.J., and Dunne, R.C (1988) The role of copper
sulphate inpyritejlotation, Min. Engng.l, 203-212.
52. Zouboulis, A.I., and Grohmann, A. (1989) Removal of phosphates from surface waters
throughjlotation,2nd Intern. Symp. Metals Separation, Speciation and Recovery, Rome,
preprint.
4.4. FLOTATION OF PIGMENTS AND INKS FROM WASTE PAPER

B. DOBIAS, W. KLAR and K. SCHWINGER

Institute of Physical and Macromolecular Chemistry,
University of Regensburg,
D-8400 Regensburg
Germany

ABSTRACf. Waste paper processing gains more and more importance for environmental and
technological reasons. Herewith in connection, the deinking flotation plays an important role. In
this paper the basic terms of the deinking process are presented.

1. Introduction

A general flowsheet of fibre recycling is outlined in Figure 1. The diagram shows the paper
circulation from the manufacturing (from primary fibre materials) through treating (e.g. printing)
and use - up to the stage when it becomes waste paper and then on to processing and the
manufacture of new paper. In this circulation, the paper gets altered, while moving through the
individual system areas, e.g. inks and binders are added to paper during its processing. If waste
paper is sent for deinking (Le. removing ink) by a magazine publisher, it already contains 35-50%
glues. Paper use and waste paper collection lead to further soiling.
Only graphic papers, i.e. writing and printing paper, can be recycled. The role of waste paper
processing is to collect assorted papers and to remove all impurities using special sorting
aggregates. However, these machines cannot handle particles smaller than a certain size or soluble
particles. These substances markedly disturb paper manufacturing. Such impurities are usually
added to paper during an earlier phase of its recycling, when it is manufactured and treated in order
to create the product properties which are required.
As can be seen (Figure I), the waste paper recycling is not true recycling but rather
"downcycling". The reason is that during processing, secondary fiber from waste paper get
irreversibly damaged and without addition of primary fibers, Le. cellulose and wood substances,
the so called recycling collapse would follow. This is due to today's possibilities ofink removal by
washing or flotation deinking that result in an incomplete purification and thus a bad quality of the
new paper.
Figure 2 contains data of the production of different sorts of new paper and the percentage of
waste paper in different new papers. As can be seen, it is almost impossible to increase the use of
waste paper as a raw material for the production of wrapping paper and corrugated board. On the
other hand, if one considers the small proportion of waste paper used for manufacturing graphic
papers, it is obvious that in the first place the quality of recycled waste paper must be markedly
improved for this product group, if we want to guarantee a higher use of secondary fibres.

499
 500

INKS, GLUES

0-100%

100-0%

WASTE WATER

PAPER
PROCESSING

SOLID WASTE LONG-LIVING PRODUCTS

Figure 1. Principle of recycling fibre material (1).

For hygienic and special papers, the use of waste paper can be increased, but the demand is
much smaller than in other product groups.

2. Printed Paper Composition

The composition of printed paper, Le. writing and printing paper, will be discussed in the next
section.
For newspaper and illustrated magazine paper, fibre raw materials are used, such as various
cellulose and wood materials, further deinked semiproduct and kaolin and calcium carbonate as
fillers. In addition, chemical auxiliary materials, such as starch and optical brighteners and inks are
required, according to the scope of use.
 501

2.1. eOMPOSmON AND MANUFAcruRE OF FIBRE MATERIALS

Fibre materials are composed mainly from cellulose. Besides this. hemicelluloses. lignine - a
mixture of linked phenols - and accompanying materials such as oils. resins. fats. wax and
minerals occur in the paper.
Hemicelluloses are macromolecular carbohydrates with an average degree of polymerisation of
less than 200. which is substantially smaller from the average degree of polymerisation in
cellulose. where it is less than 450.

3%
68_wrapping
0,47 Mlo t
f:~4%
special
0,49 Mlo t

graphlo papers

~
3,86 MID t

oorrugated
1.57 Mia t
boxboard
1,33 Mia t
-- 46%
newsprint
0,72 Mlo t
sanitary
0,61 Mia \
39%

Figure 2. Production of new paper and perentage waste paper portion for the various papers [2].

In the last years a number of manufacturing processes for wood have been developed that lead to
a higher compactness and allow for a partial substitution of cellulose. Today. the most widely used
processes of wood material production are the stone grinding. pressurised grinding. thermomecha-
nical grinding and the chemi-thermomechanical grinding. The list is not complete but it shows the
many possibilities of the wood material production.
The difference between cellulose and the wood materials is in the way of manufacturing. Wood
materials are split mechanically and finally. according to their scope of use. are treated by grinding
and steaming at temperatures of 110 to 130 °e. as is the case with thermomechanical grinding. or
they are additionally impregnated by chemicals. as is the chemi·thermomechanical grinding. On the
other hand. cellulose is produced by a chemical treatment. Boiling with calcium bisulphite and
sulphonic acid or with sodium sulphide and sodium hydroxide yields the sulphite or sulphate
cellulose. Boiling removed lignine. a component of wood fibres. A detailed explanation of the
individual manufacturing processes will not be given here.
502

2.2. FILLING PIGMENTS

Of the filling pigments that are used in paper production, calcium carbonate causes, unlike kaolin,
great problems in the deinking process. Table 1 shows a number of paper filling materials and their
physical properties. However, kaolin and calcium caroonate are the main raw materials used.

TABLE 1. Ph SICa! propertIes 0 f paper fillers.

Natural Chemical Crystalline Mean particle SpecifIC Spec. surface
comoosition fonn size weight aream2/K
kaolin A4[(OH)sSi 4O IO)1 hexagonal 0.3-5 2.8 4-7
scales
talc Mg3[(OHhSi4O I01 monokline 0.2-5 2.7 5-20
lamellae
montmorillo- A4[(OH)sSi 4O IO)1 hexagonal 0.1-2 2.8 30-50
nite .nH20 scales 300-400 *
calcium CaC03 onhorombic 0.5-5 2.7 3-4
carbonate trigonal
barvte BaS04 rhombic 2-5 4.5 0.7-2
calcium CaS04 orthorombic 1-5 2.8 0.7-2
sulphate
(gypsum
anhydrite)
synthetic varies mostly 0.1-0.5 2.1 30-150
precipitated amorphous
silicates
precipitated CaC03 hexagonal 0.2-0.5 2.7 3-10
calcium or rhombic
carbonate
aluminium Al(OH)3 hexagonal 0.5 2.4 2-4
hydroxide
titanium Ti0 2 tetragonal 0.2-0.5 4.2 8-11
dioxide
barium BaS04 orthorombic 0.5-1 4.4 3-5
sulphate
* The actIve fonn of montrnonllomte Increases ItS surface area to the second value In an aqueous
suspension, due to swelling

2.3. PRINTING INKS AND PRINTING PROCESSES

Printing inks are mainly composed of the following components: color giving materials, binders
and auxiliary materials.
Color giving components can be basically classified as soluble dyes, and organic and inorganic
pigments.
Dissolved dyes used in inks to obtain a better quality for printing in other product areas (e.g. in
wrapping) cause additional difficulties during deinking since substances which are water soluble
can be neither floated nor washed out and they cannot always be suppressed during bleaching.
 503

Soot is the most important of the pigments used in inks; this pigment is an industrial product
with defined properties. The size of the primary particles is in the range from 10 to 50 [lm. The
primary particles get structurised to larger aggregates during the manufacture of ink pigments.
Medium particle size, degree of structurisation, specific surface area, wetting ability, and oil
demand are some parameters that determine the properties of soot during ink manufacture and that
have later an importance for deinking.
Binders are necessary for the attachment of pigments on the paper surface and to one another.
Auxiliary products modify the printability and print properties. Binders and auxiliary products
are chiefly natural products but in most inks synthetic compounds are also contained and their
proportion is high in new products. An exact composition of inks cannot be given since it
constitutes a secret formula of each ink manufacturer.
According to the printing process, one can distinguish rotogravure, reliefprinting, offset, and
flexo print inks. Table 2 shows the thickness of the ink films and the ink amount for print on both
sides for different printing processes.

TABLE 2 Thickness of the ink film and the applied ink amount in printing on both sides
Printing process Ink film thickness, [lITI kg dry ink per 1 ton of paper
relief printing 4-6 15-25
offset printing 1-3 10-20
rotogravure 3-12 20-50

The rotogravure inks used for illustrated magazines represent the main part of all inks. They
have a low viscosity. Volatile organic compounds with a low boiling point and high vapour
pressure (e.g. toluene) are used as solvents. A typical composition of rotary rotogravure inks is
shown in Table 3.

TABLE 3 Composition of rotogravure inks (in %)

Reference [4) [5)
Pigment (soot) 6-8 6-10
Binder (resin) 20-30 20-30
Complementary substances 5-10 1-4
Solvents (toluene) 50-70 60-70

Films of rotogravure inks come relatively easily off the paper surface and the inks can be
removed from the suspension, especially because many rotogravure products are printed on coated
papers where an adhesion of colour films on paper fibres is not possible [6).

TABLE4 ComposltIOn 0 f convention. al rerIe f pnntlng

. . I·nks C o.
III ~)
Reference [4) [5) [7) [7)
Colour black no data black varicoloured
Pigments 10-15 9-15 10-14 6-10
Mineral oils up to 85 80-89 80-88 50-75
Binders 0-10 no data 0-3 10-20
Solvent - no data 0-3 2-15
Auxiliary materials (wax etc.) 0-4 2-5 2-5 6-10
504

Relief printing inks are mainly used for newspapers. The main binders in conventional relief
printing inks are highly viscous mineral oils with high boiling points. The inks come off the fibres
relatively easily and can be removed from the suspension. The composition has been given by
various authors (see Table 4).
Offset inks are oily and highly viscous. They are classified as heatset- and coldset-offset inks
and are of a similar composition. They differ by the boiling points of mineral oils contained. In
Table 5, the composition of offset inks is given. The main part of the mineral oil mixture used for
heatset inks containing heatset oils of a narrow boiling range of 230 to 290°C. The dryers used in
the heatset process accelerate the solidification of the ink layer on the paper, because of expulsion
of 70-85% of the specific heatset oils. Therefore, the heatset ink films contain substantially less
mineral oils than other offset inks.

TABLE 5 Composition of offset inks (in %)

Reference [8] [8] [5]
Roll-offset Roll-offset Heatset
heatset coldset
Pigments 15-25 20-25 15-20
Mineral oils 25-40 ca. 60 30-40
Hard resin 25-35 8-10 -
Soft resin/drying oils 5-15 0-12 3040
Complemetary substances 5-10 1-5 5-20

Contrary to rotogravure. relief printing and offset inks, flexoprint inks have an aqueous base.
The basic difference between water-based and oil-based inks is that, during ink production, the
pigments are mixcd with oil in one case and with water in the other one.
The pigments can be wetted both with oil or water. By oil adsorption the inks get a hydrophobic
character, while pigments of an water-based ink remain hydrophilic in aqueous solution. The
induced hydrophobicity, or hydrophilicity, of the particles influences their behaviour in the later
de inking process. Table 6 presents the composition of newspaper flexoinks.

TABLE 6. Com position ofnewspapcr flcxo inks [7].

black varicoloured
Pigment 12-16 12-20
Solvent (water) 65-77 65-80
Binders: resins 9-12 5-8
amines 1-3 1-3
Complementary substances (wax. dispersion auxil. 2-10 2-10
material. defoamer)

Easily hydrolysable synthetic dispersions are used as binders. mainly acrylates. that make the
presence of amines necessary. Therefore, and because of cellulose contained in fibre materials.
cartxJxy groups are present that become important in deinking.
 505

COARSE
PURIFICA nON 2

GLUE, INK, FILLER

FOILS, GLUE, WOOD

AGGLOMERATES

dilution

drainage

Figure 3. Overview of the waste paper recycling process.

3. The Waste Pape~ Recycling Process

Recycling of waste paper consists of four basic processes: suspension, purification, fractiona-
ting and improvement. Figure 3 shows a simplified process scheme. In the first step, waste paper
506

gets suspended by use of primary and secondary solvents in water. Coarse purification is
performed with hydrocyclones. At this stage, foils, glues and dirt are removed. In the next step,
(Fine Purification I), a flotation or washing equipment is indicated. If deinking is performed by
flotation, sodium hydroxide, fatty acids, hydrogen peroxide and complexing agents are added to
the suspension. Larger ink particles adhere to finely dispersed air bubbles due to a hydrophobic
interaction and get removed together with the foam.

TABLE 7. Advantages and disadvantages of the wash or flotation deinking.

Wash deinking Rotation deinking
- low capital expenditure - larger yield
- cheaper reagents - smaller filler and fine material loss
- worse AP quality - small waste water amount
- simpler circulating water purification

- lower yield - high capital expenditure

- higher filler and fine material loss - expensive reagents
- complicated circulating water - only partial filler reduction
purification - better AP quality

Wash deinking is a physical process involving washing out of ink particles up to 10 !lm,
whereas larger fibres are held back in sieves. In the Fine Purification II equipment, further
impurities such as foils, glues and fibre aggregates get separated by punched or slotted sheet
metals.
Fractionation separates fibres in a long fibre and a short fibre fraction. Milling and bleaching
follow after dewatering. This increases the strength, binding ability and whiteness of the recycled
waste paper.
Table 7 compares the flotation and washing processes with each other. Because of the advanta-
ges of the flotation process, deinking is chiefly performed by this method in Europe.

4. The Flotation Deinking Process

Flotation deinking is determined mainly by the following two processes:

1. ink removal from suspended fibre material or waste paper in pulp;
2. adsorption of the detergent on ink particles and their attachment on gas bubbles.
The removal of ink films from fibre substances occurs because of friction, impact, and shear
forces acting during dissolution. The prerequisite of the mechanical separation is a reduction of
binding forces between the ink and the fibre surface by chemical means. Therefore, the following
compounds are used as deinking auxiliary substances:
- sodium hydroxide
- detergents
- complexing agents
- hydrogen peroxide
The added sodium hydroxide has the following functions:
1. it reacts with the used fatty acid to a sodium salt and an interfacially active compound is
formed:
 507

NaOH + R-COO -> R-COONa + H20

2. the surface tension of the aqueous suspension gets reduced and this leads to a better
wettability of the hydrophobic ink particles;
3. both the fibres and ink particles get negatively charged. Therefore a pH range between 8 and
10 is used. Because of repulsive forces of the negatively charged partcles, the separation of
fibres and colour pigments is facilitated and a new agglomeration is hindered.
As fatty acids, mainly oleic and stearic acid are used. As mentioned above, fatty acids together
with sodium hydroxide cause a reduction of the surface tension of water. Besides it, oleates and
stearates react with Ca ions in water and with the dissolved calcium carbonate that is present in
paper as filler and forms precipitates of Ca soaps.
Hydrogen peroxide is used as a bleaching agent. The presence of sodium hydroxide leads to the
yellowing of wood fibres in paper. The bleaching effect is based on a heterolysis process of the
peroxide in a hydronium and a perhydroxy ion. The latter acts as a bleaching reagent through
destruction of chromophore groups present in lignine. a wood fibre component causing the
yellowing:

Because most inks contain heavy metals that act as catalysts and bring about homolysis of H202
in hydroxyl radicals and finally decomposition to water and oxygen, the bleaching effect of H202
gets reduced:

2H202 -> 40H·-> 2H20+02

Therefore, complexing agents are used in the deinking process. They have the role of removing
heavy metal traces from the reaction mixture. As examples, let us name diethylenetriamine acetate
(DTPA) and ethylenediaminetetraacetate (EDTA). They bind heavy metals (like Fe 3+) but also Ca
ions present in tap water get complexed.
The removing process is presented schematically in Figure 4. In the first step, hydoxyl ions
charge fibres and ink particles negatively. In the second step. the hydrophobic part of the detergent
attach to ink: particles - supported by a negative repulsion between fibres and colour pigment - and
the ink particles come off the fibre [9]. Dispersion brings about a stabilisation of colour pigments.
The hydrophobised colour particles adhere onto finely dispersed air bubbles streaming through
the suspension and get removed as foam. The dispersed colour particles are sometimes so small
that they cannot be floated directly. In the method employed in Germany, calcium salts of fatty
acids are used which form larger, floatable aggregates with the colour particles.
Experiments with various fatty acids have been performed and it has been found that a certain
calcium ion concentration is necessary to ensure flotation of ink particles. This is about 2. 7x 10.3
mol/l (Figure 5), corresponding to a water hardness of 15° dH. At lower concentrations, the ink
flotation is incomplete. It is assumed that a calcium bridge is being formed between the negatively
charged colour particles and the anionic detergent (Figure 6). so that the hydrophobic colour
particles adhere to air bubbles.
Very high Ca ion concentrations occur in paper mills, due to :
- the use of CaC03 as a filler in paper manufacturing.
- the use of water containing Ca ions and
 508

water
ink + OW water OH·

j .I C ;;dS
initial state fibre loading c1 fibre and ink

+ detergent

.. dIll! :;£.9 -:?

adsorption of detergent

hydrophobic
o
hydrophaic

come-off of ink and

stabilization

Figure 4. Schematic prcsentation of the come-off process.

o 10 20 30 40 50 60 70 80 90 100
water hardinessodH

Figure 5. Effect of water hardness on dcinking.

- because of closed water circuits, in order to reduce waste water discharge.

Added soaps and Ca ions form heavily soluble Ca soaps that deposit on fibres and paper mill
machines. However, since increasing the calcium concentration enhances fibre flotation and a
minimum Ca ion concentration is necessary to achieve a satisfactory colour pigment flotation, the
problem is being solved by the use of non ionic detergents. These are relatively insensitive toward
 509

Ca-ion-bridge interface
air bubble-Na-alkylcarboxylate
==--..;=+Ca+ -Oz C( CH z).CH J
\/\/Vco,Na,
air bubble _ _.......
hydropnObiC 0 hydropnillc

(RCO,),Ca . ,
interface
sorld-Ca(alkylcarooxylate) 2

Figure 6. Model presentation of the flotation process.

~_~~~~==:v-Dispersion unit

Multifunction rotor

Figure 7. The Voith flotation cell

Ca ions. Further, when using anionic detergents, NaOH must be added in order to convert them in
surface active compounds; this results in high pH values and acidification during waste water
processing is then necessary. These disadvantages can be avoided by the use of nonionic
detergents.
510

5. Concluding Remarks

Finally, attention is paid to some of the flotation equipment available. There are certainly various
flotation cells in use, but some only will be mentioned, related to the deinking process.
In the 60's, the Voith Company developed a quadratic flotation cell with a mechanical
distribution of air (Figure 7). It works according to the following principle: the suspension comes
in the cell to a rotor, that sits on the cell bottom and is surrounded with a punched coat driven from
above. The cell works as self-sucking. The rotor provides for fine air dispersion in the suspension,
for contact of ink particles with bubbles, and for the distribution of the material over the cell
bottom. The pulp is removed by an overflow, the foam with a wiper.

Figure 8. Tube flotation cell

Foam
Separation Bubble Separation

I I Separation Zone IMixing IAeration

Zone Zone

Figure 9. Beloit Pressurised Deinking Module (PDM) flotation cell.

 511

To improve the hydrodynamics of the cell, the tube flotation cell was introduced in 1978. TIle
aeration system and the mechanical fine air dispersing were taken from the old system. In 1982,
the tube design was improved. Only the form of the cell remained unchanged. Aeration is carried
out with an injector which WoIXs self-sucking and allows for an aeration of up to 30%. The foam
removal is automatic, without the aid of a paddle (see Figure 8).
A recent novelty (Figure 9) in the area of flotation cell design is the pressurised deinking module
(PDM). It is the first cell working under pressure in the separation zone. Besides a good separation
of ink particles and fibres, the following advantages have been reported by the manufacturer:
- it WOIXs without secondary cells;
- it has a simpler function than ordinary cells
- the maintenance effort is reduced
- optimisation of aeration, mixing and separation zones
- minimum danger of clogging
- the small diameter of the separation zone minimised the total space demand
- finally, there is no vacuum; therefore, a system for removing foam is no longer necessary. The
foam is transported by ascending air automatically to a reject opening in the upper part of the
separation chamber.

6 • References

1. Kocurek, M.J., Hamilton, F., and Leopold, B. (1987) Secondary Fibers and Non-Wood
Pulping, Pulp and Paper Manufacture, Joint Textbook Committee of the Paper Industry .2, pp.
221-223.
2. Grossman, H. (1989) Zur Entwicklung der Altpapiernutzung und der Papiertechnik. PTS-
Lehrgang-Grundlagen der Altpapiemutzung und Papiertechnik 4, pp. 1-3.
3. Autorenkollektiv (1987) Lehrbuch der Papier- und Kanonerzeugung, VEB Fachbuchverlag,
Leipzig, pp. 24-36.
4. Berndt, W. (1976) Druc/ifarben und Deinken, Wochenblatt fUr Papierfabr. 104(3),95-98.
5. Bassemir, W.R. (1979) The chemical nature of modern printing inks and deinking, Tappi
n2.(7), 25-26.
6. Lausch, H., and Ortner, H. (1966) Uber den Einfluss der Druc/ifarbenzusammensetzung auf
den Deinking-Prozess. Wochenblatt fUr Papierfabr. 21(5), 129-136.
7. Wasilewski, O. (1987) Composition and Chemistry of Novel Inks used in the Newspapaer
Industry and Deinking, Proceedings Tappi Pulping Conference (Washington D.C.), vol. 1,
pp.25-28.
8. Bremer, H. (1988) Druc/ifarbenfur den Heatset-RollenojJJsetdruck, Deutscher Drucker
M(15), 14-15.
9. Hanecker, E., and Grossman, H. (1989) Wirkungsweise von Deinking-Chemikillien, PTS-
Lehrgang, Grundlagen der Papierfabrikation 2, pp. 1-8.
10. Ortner, H., and Schweiss, P. (1982) Neue Flotations-Deinkingverfahren und Maschinen,
Wochenblatt fUr Papierfabr.llO(10), 343-347.
11. Pfalzer, L., und Schweiss, P. (1988) Die Voith-Multi-Injectorzelle- Funktionsweise und
Anwendung in einer modernen Deinkinganlage, Wochenblatt ftir Papierfabr. 112(8),
303-307.
4.5. A CASE STUDY OF FLOTATION APPLIED TO AN EDIBLE OIL
FACTORY WASTEWATER

A. TURKMAN and F. SENGUL

Department of Environmental Engineering
Dokuz Eylal University
Bornova.Izmir
Turkey

N.OZEL
Ayvalik Komili Oil Industry
Ayvalik. Balikesir
Turkey

ABSTRACf. Rotation is applied to many edible oil wastewater treatment plants in Turkey. In this
study, the wastewaters of an edible oil industry near Izmir have been treated by flotation in
laboratory conditions and various kinds of polymers have been applied in order to increase the
flotation efficiency.

1. Introduction

The contribution of vegetable oil in total edible oil production of Turkey is significant. The most
important ones are : sunflower seed oil, cotton seed oil, sesam oil and corn oil.
The industry which has been studied in this publication is an oil factory which refines olive oil,
corn oil, sunflower seed oil (and soap), depending on the seasonal supply of the related raw
materials. The industry - which was established in 1982 - produces armuaUy 200,000 tons of
distilled oleic acid, 1,000 - 1,200 tons of glycerin and 6,000 tons of stearic acid. The plant refines
about 20,000 - 25,000 tons of oil per year.
The wastewater treatment plant receives about 1,200 - 1,500 m3 of wastewater/day. It consists
of physical, chemical and biological treatment units. In Figure I, it may be seen that the treatment
plant includes screens, pumping station, equalisation tank, oil and grease removal, chemical
treatment and biological treatment units.
At the outlet of the equalisation tank the pH value is adjusted to 4.0 and the oil, which is freed by
an emulsion cracking operation, is scummed at the surface. Following this, chemical coagulation is
applied. Chemical coagulation is done by adding lime and some polyelectrolyte, rapid mixing, then
slow mixing and finally two settling tanks. Following the pH control tank, a two-step activated
sludge treatment is applied. The final sedimentation tank of the biological unit is followed by
chlorination tanks. The effluent is discharged to a surface water stream which goes into Edremit
Bay, after flowing for about 3-4 kilometers.

513
P. Mavros and K. A. Matis (eds.). Innovations in Flotation Technology. 513-519.
© 1992 Kluwer Academic Publisiu!rs.
 V>

WASTEWATER -I:>

LIME SLOW
ALUM MIXING

'---____-iISETTLING I -I
TANKS

TO RIVER

Figure 1. Flowsheet of the edible oil wastewater processing plant.

 515

The scum obtained at the flotation unit is evaluated as a side product and used in soap
production. Chcmical sludge obtained from coagulation is sent away after dewatering at press
filters. Biological sludge is thickened in a sludge thickener and then sent to sludge drying beds.
The aim of this study is to investigate the possibility of increasing the efficiency of a flotation
unit. The results of the experimental study will give an idea about the new alternatives that can be
applied in the plant.

20 Oil and Grease Removal from Wastewater by Flotation

Control of oil and grease is a stringent requirement of municipal authorities responsible for
permitting the connection of industrial wastewaters into the sewer system or direct discharge into
receiving waters. According to the Turkish Water Pollution Control Regulation [1], established in
1988, the discharge limits into the sewerage system and receiving waters are given in Table 1.

TABLE 1 D·lSCh arge rlmlts

. fioredObl
1 e 01·1·IIIdustry wastwaters [1)
Parameter Standard for discharging into Discharge into
receivins water the sewer system
2 hrs composite 24 hrs composite
COD,mg/l 250 230 4000
Oil and Grease, mg/l 60 40 250
pH 6-9 6-9 6.5-10
Suspended solids, mg/l 500

Oil industries have to treat their wastes to a great extent in order to meet the discharge standards
given in Table 1. Organic matter and oil and grease are important parameters that have to be
considered in treatment. In order to meet the above limits, grease removal, flotation and coagulation
are applied as unit operations. Coagulation and flotation are often accompanied by polyelectrolyte
application. There are many successful separation examples of industrial wastewaters by poly-
electrolyte application [2).
Free oil and grease which is not emulsified presents no serious problem with respect to its
removal from water because it will tend to float and agglomerate. The first-stage treatment can be
utilised to separate free floatable oil and grease from water. The typical process for oil and grease
removal, particularly in situations where emulsification exists, is a gravity separator followed by a
second-stage unit using one of the several methods - physical, chemical, electrical and biological-
for breaking the oil emulsion.Emulsions can be broken by chemicals that will neutralise stabilised
electrical charges, or precipitate emulsifying agents. Common chemicals that have been utilised to
break emulsions or coagulate the colloidal particles include alum, ferrous sulphate, ferric sulphate
or chloride, sodium hydroxide, calcium chloride, lime, soda, sulphuric acid, sodium sulphate and
polyelectrolytes [3,4]. After the addition of chemicals, freed oil and grease are concentrated and
removed by a suitable physical process such as flotationo

3 0 Experimental

In the experimental part of this study a laboratory flotation cell consisting of an 1 liter beaker and an
516

air supply system has been used. The analysis of the composite wastewater sample taken from the
outlet of the plant gave the results indicated in Table 2.

TABLE 2 Anal ytic

. al results 0 fthe wastewater ffik,en fro m the plant ou tlet.
Parameter 24 hrs composite
(5.3.91 - 6.3.91) (6.3.91 - 7.3.91)
BODs 2400mg/l 1250mg/l
Total nitrogen 26 mg/l 40mg/l
Total phosphorus 0.20mg/l 5.6 mg/l
Oil and grease 3200mg/l 2100mg/l

The effect of various polyelectrolytes on the flotation efficiency was investigated. The experi-
mental procedure was conducted with water samples being taken from the outlet of the equalisation
tank. Since the study aimed at increasing the efficiency of the oil and grease removal unit and
obtaining a constant water quality with respect to oil before the biological unit, composite
wastewater samples were prepared to determine the general characteristics of wastewater. Each
sample was treated by five different ways: natural or "free" flotation, dispersed-air flotation,
dispersed-air flotation with the addition of a cationic polyelectrolyte (K 5065 or K 50) and
dispersed-air flotation with the addition of an anionic polyelectrolyte (A 8687). The polyelectrolytes
used were polyacrylamide-based. The volume of scum, which contained oil, grease and suspended
matter, as well as the remaining water was recorded. As an example, if 240 mIll floated oil was
measured, it meant that 760 ml/l of treated wastewater remained at the bottom of the cell.

4. Results and Discussion

The experimental results for the flotation of sunflower seed oil and com oil, as well as mixtures of
them with olive oil wastewater, for various time periods, are given in Table 3 and in Figures 2
and 3. Hotation in most cases took place for about 5-10 min. Mixtures of com oil and olive oil
wastewaters were floated in about 1-7 min.
When the floatable material was allowed to float on its own (with and without any air supply),
the efficiency was rather low. Addition of polyelectrolytes was investigated as a means of
increasing efficiency: it was found to shorten flotation time to approximately 2 min.
The effect of the type of polyelectrolyte added on flotation efficiency depended upon the type of
wastewater treated, but in general the anionic polyelectrolyte (A 8687) gave best results for sun-
flower seed oil wastewaters or mixtures containing high ratios of olive oil wastewater, cationic
polyelectrolytes (K 5065 or K 50) gave more clear water as compared to anionic ones. Figure 2
illustrates the effect of polyelectrolyte addition on sunflower seed oil flotation efficiency: in almost
all experiments, efficiency decreased in the following order: anionic polyelectrolyte (A 8687),
cationic polyelectrolyte (K 5065), air flotation without any addition and "free" flotation. The two
cationic polyelectrolytes gave similar results. Figure 3 illustrates similar results, obtained with
mixtures of com oil with olive oil wastewaters.
In some cases, the application of polyelectrolyte brought no improvement and their use was not
justified. This is obvious in Table 3 for the mixture of com oil and olive oil wastewaters; in such
cases, flotation may be performed without any polyelectrolyte addition. In a few cases, the
difference between dispersed-air flotation and "free" flotation was very small, too. Thus, the opera-
 517

Floatable matter [mIlO 350 Floetabl~ matter Cml/il .

350 ,• ,1
I ~ I
300 ---------~----.- ---.-+---------~--------.- 300

250 250

200 200

150
Free
-+- Air 100 -+- Air

0 2
--
-*- K5065

4
AB687

6 B
50

0 2 4
--
-*- K5065

6
A8687

B 10
Time [minI Time [min]

Figure 2. Flotation of sunflower seed oil wastewaters (left) and of sunflower seed and olive oil mixture
wastewaters (right).

Floatable matter [ml/il

250

200

Free
-+- Air

0 2 3
--
-*- K5065

4 5
AB687

6 7 B
Time [mInI

Figure 3. Flotation of com and olive oil mixture wastewaters.

518

TABLE 3. Experimental results of flotation of sunflower seed wastewater and

mixtures with olive oil or corn oil wastwaters with the addition of polyelectrolytes
Type of wastewater TIme Free Air Cationic Anionic
sample (min) floated oil floated polyelectrolytes polyele-
(normal) K5065 K50 ctrolyte
A8687
floatable matter (ml/l)
Sunflower seed oil 1 240 320 320 324 324
wastewater (5.1.91) 2 250 325 325 325 325
5 250 325 325 325 325
Sunflower seed + olive 1 25 70 150 175 175
oil wastewaters 2 50 150 225 240 240
(10.1.91) 3 100 200 250 260 255
5 200 300 320 320 320
10 250 325 325 325 325
Sunflower seed+ olive 1 250 325 325 325 325
oil wastewaters The effluent of the equalisation basin was highly clear, so
(17.1.91) the oil removal was carried out in a short time
Com oil + oilve oil 1 320 320 320 320 320
wastewaters (7.1.91) 2 320 325 325 325 325
3 325 325 325 325 325
5 325 325 325 325 325
Com oil + oilve oil 1 240 323 323 323 323
wastewaters (18.1.91) 2 250 325 325 325 325
Com oil + oilve oil 1 75 200 225 225 275
wastewaters (19.1.91) 2 200 275 280 275 285
3 225 280 285 285 290
5 230 285 285 285 300

tion of a flotation unit in treatment plants of edible oil factories necds continous and close control to
save chemicals and perform efficient treatment.
In the wastewater treatment plant, a certain amount of oil removal is achieved by applying acid
cracking at pH 3-7. Since a constant quality of wastewater entering into the biological unit was not
obtained, it was necessary to increase the flotation efficiency in the plant. This study showed that
application of polyelectrolytes will be useful in reaching this goal.

5. References

1. Turkish Official Gazette (1988) Turkish State Water Pollution Control Regulation, September
4, Number 19919, Ankara.
2. Tiirkman, A. (1991) Polymer Application Examples in Industrial Wastewater Treatment, in A.
Tiirkman and o. Uslu (eds.), New Developments in Industrial Wastewater Treatment,
Kluwer, Dordrecht.
3. Sengiil, F. (1990) A Case Study on Sunflower Seed Oil Industries, Waste Characterisation,
Classification and Treatment, Water Science Tech. 22(9), 241-248.
 519

4. Eckenfe1der, W.W. (1980) Principles of Water Quality Management, CBI Inc., Boston.
5. A WWA Standard Methods (1981) Standard Methods for the Examination of Water and
Wastewater, 15th Edition, AWW NAmerican Public Health Association, Washington, D.C.
4.6. THE NEED FOR ADVANCED TREATMENT OF STORM-WATER
RUN-OFF FROM SEPARATE SEWER SYSTEMS

R. PFEIFER and H.H. HAHN

I nstitut fur Siedlungswasserwirtschaft,
Universitat Karlsruhe,
Am Fasanengarlen,
D-W7500 Karlsruhe,
FRG

ABSTRACT'. A wide spectrum of anthropogenic pollutants exists in surface run-off resulting from
precipitation. These are mainly organic micropollutants and heavy metals, which are transported in
association with finer suspended solids. Conventional treatment methods for storm-water run-off
using sedimentation alone arc not sufficient for reducing the input of particulate materials with
adsorbed contaminants to natural waters. Chemical precipitation/coagulation is effective in produ-
cing separable aggregates and phase separation using flotation or cyclone and vortex separators can
then be applied to remove them from storm-water.

1. Introduction

During precipitation events, a great number of solids loaded with pollutants are introduced to
natural waters. Depending on hydraulic conditions, particles may be transported or deposited as
sediments which may be resuspended and transported later. It is probable that there are higher
concentrations of these pollutants in areas with increased sediment deposits [I). The quality of
natural waters in urban collection areas is also affected by emissions from communal public sewage
plants. Both plant effluents and storm-water run-off must be considered in assessing the pollutant
load to a natural water. Advanced treatment technologies are already applied to reduce pollutant
loads from conventional sewage. Thus it may be necessary to treat storm-water run-off to a greater
extent than is currently practiced to achieve further reductions in pollutant loading of natural waters.
Discharges from paved areas, especially those from areas of high traffic, may present a
significant loading of pollutants. Increased concentrations of heavy metals, such as lead, cadmium
and zinc have been observed, as well as significant organic pollutants such as 3,4-benzopyrene and
3,4-bcnzofluoranthene [2-4). Colloidal particles appear to be the primary vector of transport for
these pollutants. Due to the fine particle size (d < 80 (.lIII), these contaminated suspended solids are
not removed to a great extent by sedimentation within practical periods of time. Improvements in
solid/liquid separation in the treatment of storm-water run-off can be realised through the use of
chemical pretreatment for particle removal in subsequent unit processes.

521
P. Mavros and K. A. Matis (eds.), Innovations in Flotation Technology, 521-525.
© 1992 Kluwer Academic Publishers.
522

2. Physicochemical treatment methods

Where sewage and storm-water run-off are collected by separate systems, surface run-offs are
typically discharged into natural water by storm-water sewers. In addition these discharges may
include flows from misconnections and extraneous waters, which exist even during dry weather
periods. Surface water sewers are either discharged directly to the receiving waters or to interme-
diate storm-water sedimentation tanks, which remove some ofthe larger particulate materials and
light liquids.
Characteristics of the discharge events, which depend upon the collection area, must be
considered in the design of plants for the treatment of storm-water run-off. This applies not only
for physicochemical plants but also for the traditional storm-water sedimentation tanks. Load
profiles for a hypothetical planning situation are given in Table 1. Treatment plant characteristics
that may be required for a given load profile are also indicated.

cummulative frequency ,%, separation treatment

method
100 %
SST
<:: :J

cyclone-SST
<:=:::::::::===
VS - SST

50 %
PIC - SST
<==--~-
PIC - VS

flotation
0%
<:=====
particle diameter [log d)
separable
solid fractions

Figure 1. Treaunent methods and estimated separation efficiencies (SST: storm-water sedimentation tank;
VS: vortex separator; PIC: precipitation/coagulation stage).

Besides these operational criteria, which must provide a given level of reliability, the solid/liquid
separation efficiency obtainable by a specific treatment configuration must be determined.
Treatment configurations are arranged in Figure 1 by their estimated separation efficiencies. As
separation efficiency decreases, removals shift to particles oflarger sizes.
 523

TABLE 1: Loads of a treatment plant (load profile) and the resulting construction
demands (plant profile).
Given load profile Required plant profile
* time/duration of the discharge is * automatic operation
unknown
* large load fluctuations (quantity and * adaptable operation (by controlling)
quality)
* load peak at the beginning of event * short response times
* long periods of stagnation (dry * permanent stand by
weather periods)
* flexible utilisation of capacity * compensation without
performance damage

One possible means of improving the removal of poorly settling particles in systems using
sedimentation tanks is the addition of cyclone or vortex separators (Figure 1). This has already
been realised using storm-water tanks in combined systems [5]. While cyclone separators require a
sedimentation tank, vortex separators may be used in conjunction with them or alone [6]. Both
cyclone and vortex separators work under atmospheric pressure. The influent water is rotated and
the solids are separated by centrifugal and gravity forces, acting in a twisted stream.
Alternatively, chemical precipitation/coagulation might be used to improve the efficiency of
particle removal in storm-water sedimentation tanks. Chemical pretreatment is also likely to
improve the separation efficiency of vortex separators (Figure 1) (7). The chemical pretreatment
helps in forming aggregating small colloidal materials into larger, more easily removed floes. The
characteristics of these flocs can be tailored to the needs of the solid/liquid separation process.
However, the use of chemical precipitation/coagulation in conjunction with flotation is a
particularly promising treatment configuration.

3. Application of flotation

The use of precipitation/coagulation with flotation in the treatment of storm-water run-off is being
investigated in a recently initiated pilot study [7]. The pilot (Figure 2) treats water at an outlet of the
Karlsruhe storm-water sewer system; the first results are expected in autumn 1991. The aim of this
work is to develop treatment and control strategies for storm-water treatment. The plant includes
monitoring and sampling equipment, which records quality and quantity aspects of the storm-water
run-off. In addition to the dissolved-air flotation unit, the usage of a pneumatic flotation unit and a
vortex separator was proposed.
The application of the precipitation/coagulation-flotation process requires automatic data
acquisition and plant operation. This is accomplished using two computers, which communicate
with each other and undertake different fields of activities in measuring and controlling the pilot
unit (Figure 3). Both computer process operation and feed forward process control have to respond
quickly to precipitation events. The pilot unit and research efforts have been designed to allow for
simultaneous evaluation of the treatment process and development of a process control system as
well as for parallel improvement of both processes. These efforts can then be combined as the
project evolves.
flotation is a treatment method, which can be integrated into the presented plant profIle and also
524

in-flow (drum) screen flotation

storage tank precip.l coagul. out-flow

Figure 2. Schematic presentation of the pilot plant.

/~~-
( rain gauge )
senso~

~~=.~.~~_:"
. / pH. IUrbldlt~

monlm,in. and

~~~~
~-+- sampling station

( water level
sewer
/ jt samplers
------------------ Sl, S2, ...

V
-------- ..,,--
~tment stati~
~ \.
~
--
CU~G'
- V
Unit part n

Figure 3. Organisation of monitoring and controlling of the pilot plant.

allows for a high separation efficiency. In the seventies. flotation plants were utilised in the USA
for the treatment of combined sewer overflows. This experience indicated that the high process
efficiency and the compactness of the treatment unit were the determining factors [8.9].

4. Conclusions

Solid/liquid separation units such as cyclone or vortex separators are required in addition to storm-
 525

water sedimentation tanks to achieve high removals of suspended solids. Using a chemical
precipitation/coagulation step, suspended substances can be aggregated to larger sizes and separa-
ted using flotation or other unit processes. High separation efficiency may require process
automation.
While problems associated with storm-water run-off can be dealt with using end-of-pipe
technologies, there is also a need to initiate pollution-reduction activities such as reducing the
paving of areas and separating storm -water run-off based as pollutant load.

5. References

1. Hahn, H.H. (1990) Niederschlagsbedingte Schmutzstoj]belastung der GewiisserBeitrag der

verschiedenen Belastungspunkte, in Schriftenreihe Bd. 58, Institut fUr Siedlungswasser-
wirtschaft, Universitat Karlsruhe.
2. Dannecker, W., Au, M., and Stechmann, H. (1988) Bilanzierung des Eintrags von
Schadelementen aus der Luft und durch den Kraftfahrzeugverkehr, Abschlussbericht, Institut
fUr Anorganische und Angewandte Chemie, Univesital Hamburg.
3. Chebbo, G., Musquere, P., and Bachoc, A. (1990) Solids transfered into sewers,
Proceedings 5th Int. Conf. Urban Storm Drainage (Osaka), pp. 885-890.
4. Hahn, H.H., Augustin, A., Pfeifer, R, and Xanthopoulos, C. (1990) Niederschlagsbe-
dingte Schmutzbelastung der Gewiisser aus stiidtischen befestigten Fliichen, BMFI'-
Verbundprojekt, Zwischenbericht 1989. Institut fUr Siedlungswasserwirtschaft, Universitat
Karlsruhe.
5. Geiger, W.F., and Jonasson, A. (1991) Zyklone zur Leistungssteigerung von Regenbe-cken,
BMFI'-Zwischenbericht 1990. Institut fUr Siedlungswasserwirtschaft, Universitat!
Gesamthochschule Essen.
6. Brombach, H. (1991) personal communication, Umwelt- und Fluidtechnik GmbH (UFf),
Bad Mergentheim.
7. Hahn, H.H., and Pfeifer, R (1991) Chemisch-physikalische Behandlung von Nieder-
schlagsabflQssen aus Trennsystemen stiidtischer Gebiete, BMFf-Zwischenbericht 1990.
Institut fUr Siedlungswasserwirtschaft, Universitiit Karlsruhe.
8. Gupta, K., and Agnew, RW. (1973) Combined sewer overflows. Screening and dissolved
air flotation. Water & Sewage Works.8.. 74-78.
9. Field, R., and Fan, C. (1981) Recycling urban stormwater for profit, Water/Engineering &
Management~, 30-33.
LIST OF CONTRIBUTORS

Prof. I.M. Cases ENSGA - Institut Poly technique de Lorraine, B.P. 40, 54501
Vandoeuvre Cedex, France
Dr. M.W. Chudacek MD Research Co. PTY Ltd., P.O. Box 22, North Ryde, NSW 2113,
Australia
Prof. C. Ek Universite de Liege, Faculte des Sciences Appliquees, 2, rue Annand
Stevart, 4000 Liege, Belgium
Prof. I.A. Finch McGill University, Dept. of Mining & Metallurgical Engineering, FD
Adams Building, 3450 University Street, Montreal, PQ H3A 2A7,
Canada
Prof. W. Forsling Lulea University of Technology, Dept. of Inorganic Chemistry,
95187 Luiea, Sweden
Dr. St. Gaidarjiev Mineral Processing Department, Mining and Geological Institute,
Sofia, Bulgaria
Dr. G.P. Gallios Department of Chemistry, Aristotle University, 540 06 Thessaloniki,
Greece
Prof. J. Gregory University College, Dept. of Civil & Municipal Engineering, Gower
Street, London WClE 6BT, UK
Dr. U.A. Ipekoglu Dokuz Eyliil University, Dept. of Mining Engineering, 35100
Bomova/ Izmir, Turkey
Ms. W. Klar Institute of Physical and Macromolecular Chemistry, University of
Regensburg, P.O. Box 397, 8400 Regensburg, Gennany
Dr. M.R.M. Leite University of Porto, Dept. of Mineral Engineering, Rua dos Bragas,
4099 Porto, Portugal
Mr. D. Malliaris IGME - Mineral Processing Dept., 70 Messogion Street, 11527
Athens, Greece
Dr. K.A. Matis Department of Chemistry, Aristotle University, 54006 Thessaloniki,
Greece
Dr. P. Mavros Department of Chemistry, Aristotle University, 54006 Thessaloniki,
Greece
Dr. D.J. McKee Julius Kruttschnitt Mineral Research Centre, Isles Rd.,
Indooroopilly, Queensland 4068, Australia
Dr. V. Panayotov Dept. of Mineral Processing, Higher Institute of Mining & Geology,
Darvenitsa, 1156 Sofia, Bulgaria
Dr. W. Petruk CANMET, 555 Booth Street, Ottawa KIA 001, Canada
Mr. R. Pfeifer Institut fUr Siedlungwasserwirtschaft, Universtitat Karlsruhe,
Postfach 6980, D-7500 Karlsruhe I, Gennany
Prof. R.I. Pugh Institute for Surface Chemistry, Box 5607, 11486 Stockholm,
Sweden
Ms. S. Raatz Forschungsinstitut fiir Aufbereitung, Chemnitzer Strasse 40, 9200
Freiberg, Gennany

527
528

Dr. B. Radoev Dept. Phys. Chern., University of Sofia, Anton Ivanov I, 1126
Sofia, Bulgaria
Dr. K.H. Rao Division of Mineral Processing, Lulea University of Technology,
95187 Lulea, Sweden
Dr. H.J. Schulze Forschungsinstitut fur Aufbereitung, Chemnitzer Strasse 40, 9200
Freiberg, Gennany
Prof. St. Stoev Higher Institute of Mining and Geology, Dept. of Mineral
Processing, 1156 Sofia, Bulgaria
Dr. A. TIirlanan Dokuz EylUl Univerity, Dept. of Environmental Engineering, 35100
Bomova - Izmir, Turkey
Prof. W. Walkowiak Technical University ofWroclaw, Wybrzeze S., Wyspianskiego 27,
50-370 Wroclaw, Poland
Dr. T.F. Zabel WRc Medmenham, PO Box 16, Marlow, Bucks. SL7 2HD, UK
Dr. A.1. Zouboulis Department of Chemistry, Aristotle University, 540 06 Thessaloniki,
Greece

LIST OF PARTICIPANTS

Prof. A.A. Abramov Mineral Proc. Dept., Higher Institute of Mining, Leninsky prosP. 6,
Moscow, USSR
Mr. B. Aksani Hacettepe University, Dept. of Mining Engineering, Beytepe,
Ankara, Turkey
Dr.M. Akser Dupont Co., Inc., Chambers Works, Jackson Lab., Deepwater, New
Jersey 08023, USA
Dr. G. Anastasakis Dept. of Mineral Ores, Technical University of Crete, 127, El.
Venizelou Street, 73133 Chania, Greece
Dr. Th. Angelidis Department of Chemistry, Aristotle University, 540 06 Thessaloniki,
Greece
Mr. D. Aslanidis MTM - K.U. Leuven, De Croylaan 2,3001 Heverlee, Belgium
Dr. G. Atesok lTV Maden Fakiiltesi, Cevher Hazirlama Anabilim, Dali, Maslak,
80626 Istanbul, Turkey
Dr. P. Bevilacqua University of Trieste, Istituto di Miniere e Geofisica Applicata, P.le
Europa I, 34127 Trieste, Italy
Prof. J .M. Cases ENSGA - Institut Polytechnique de Lorraine, B.P. 40, 54501
Vandoeuvre Cedex, France
Dr. M.W. Chudacek MD Research Co. PTY Ltd., P.O. Box 22, North Ryde, NSW 2113,
Australia
Mr. R. Cruz Moreira University of Porto - ISEP, R. de Araujo 1413,4465 S. Mamede de
Infesta, Portugal
Ms. L.A.Cunha Faculty of Sciences & Technology, New University, Dept. Cil!ncia
dos Materiais, Quinta da Torre 2825, Monte de Caparica, Portugal
Ms. A. Daniilidou Department of Chemistry, Aristotle University, 540 06 Thessaloniki,
Greece
Prof. C. Ek Universit~ de Li~ge, Faculw des Sciences Appliqu~s, 2, rue Annand
St~vart, 4000 Li~ge, Belgium
 529

Mr. z. Ekmek~ Hacettepe University, Dept. of Mining Engineering, Beytepe, 06532

Ankara. Turkey
Mr. L. Ergiin Hacettepe University. Dept. of Mining Engineering. Beytepe. 06532
Ankara. Turkey
Dr. S. Ersayin Hacettepe University. Dept. of Mining Engineering. Beytepe. 06532
Ankara. Turkey
Mr. P. Fecko Mining University. Dept. of Mineral Dressing. 70833 Ostrava-
Poruba. Czechoslovakia
Prof. J.A. Finch McGill University. Dept. of Mining & Metallurgical Engineering. FD
Adams Building. 3450 University Street, Montreal, PQ H3A 2A7.
Canada
Prof. W. Forsling Lulea University of Technology. Dept. ofInorganic Chemistry.
95187 Lulea. Sweden
Dr. St. GaidaIjiev Mineral Processing Department. Mining and Geological Institute.
Sofia. Bulgaria
Dr. G.P. Gallios Department of Chemistry. Aristotle University. 54006 Thessaloniki.
Greece
Prof. J. Gregory University College. Dept. of Civil & Municipal Engineering. Gower
Street. London WCIE 6BT. UK
Ms. M. Grossou-Valta IGME - Mineral Processing Dept.. 70 Messogion Street. 11527
Athens. Greece
Dr. S. Hall University of Nottingham. Dept. of Mining Engineering. University
Park. Nottingham NG7 2RD. UK
Mr. M. Harrison M.I.M. Technology Marketing Ltd. Botany Road. Northfleet. Kent
DAl19BG. UK
Dr. B. lpekoglu Turldsh Atomic Energy Authority. <;ekmece Nuclear Research &
Training Centre. PO Box 1. Havaalani. Istanbul. Turkey
Dr. U.A. lpekoglu Dokuz Eyliil University. Dept. of Mining Engineering. 35100
Bomova/ Izmir. Turkey
Ms. N. Karaoglu Dokuz Eyliil Univerity. Dept. of Geology. Engineering &
Architectural Faculy. 35100 Bomova/Izmir. Turkey
Ms. W. Klar Institute of Physical and Macromolecular Chemistry. University of
Regensburg. P.O. Box 397. 8400 Regensburg. Germany
Mr. K. Kydros Department of Chemistry. Aristotle University. 540 06 Thessaloniki.
Greece
Mr. N.K. Lazaridis Department of Chemistry. Aristotle University. 540 06 Thessaloniki.
Greece
Ms. S.Y. Lee University College. Dept. of Civil & Municipal Engineering, Gower
Street. London WCIE 6BT. UK
Dr. M.R.M. Leite University of Porto. Dept. of Mineral Engineering. Rua dos Bragas.
4099 Porto. Portugal
Mr.C.Lucion Universit~ Catholique de Louvain. Laboratoire de Traitement des
Minerais. Place Ste Barbe 2. 1348 Louvain-la-Neuve. Belgium
Mr. D. Malliaris IGME - Mineral Processing Dept., 70 Messogion Street. 115 27
Athens. Greece
Prof. A.P. Mathews Kansas State University. Dept. of Civil Engineering. Seaton Hall.
Manhattan. Kansas 66506. USA
530

Dr. K.A. Matis Department of Chemistry. Aristotle University. 540 06 Thessaloniki.

Greece (ASI Director)
Dr. P. Mavros Department of Chemistry. Aristotle University. 540 06 Thessaloniki.
Greece
Dr. D.J. McKee Julius Kruttschnitt Mineral Research Centre. Isles Rd .•
Indooroopilly. Queensland 4068. Australia
Dr. H. Mordogan Faculty of Engineering - Architecture. Dokuz Eylill University.
Bornova - Izmir. Turkey
Ms.O.Ozcan Chern. Eng. Dept.. Istanbul Technical University. Maslak. Istanbul.
Turkey
Dr. V. Panayotov Dept. of Mineral Processing. Higher Institute of Mining & Geology.
Darvenitsa. 1156 Sofia. Bulgaria
Dr. R. Pascoe Cambome School of Mines. Pool. Redruth. Cornwall TR15 3SE.
UK
Dr. M. Patronis IGME - Mineral Processing Dept.. 70 Messogion Street. 115 27
Athens. Greece
Ms. M.J. Pereira Technical University of Lisbon. CVRM - 1ST. Av. Rovisco Rais.
1096 Lisboa Codex. Portugal
Dr. W. Petruk CANMET. 555 Booth Street. Ottawa KIA 001. Canada
Mr. R. Pfeifer Institut fUr Siedlungwasserwirtschaft. Universtitlit Karlsruhe.
Postfach 6980. D-75oo Karlsruhe 1. Gennany
Prof. R.J. Pugh Institute for Surface Chemistry. Box 5607.11486 StOCkholm.
Sweden
Ms. S. Raatz Forschungsinstitut fUr Aufbereitung. Chemnitzer Strasse 40. 9200
Freiberg. Gennany
Ms H. Raclavska Mining University. Dept. of Geology and Mineralogy. 70833
Ostrava - Poruba. Czechoslovakia
Dr. B. Radoev Dept. Phys. Chern .• University of Sofia. Anton Ivanov 1. 1126
Sofia. Bulgaria
Dr. K.H. Rao Division of Mineral ProceSSing. Lulea University of Technology.
95187 Lulea. Sweden
Mr. K. Repanas Dept. Chern. Eng.• The University. Newcastle-upon-Tyne NEI
7RU. UK
Dr. J.A.C. Rocha e Silva University of East Asia. P.O. Box 3001. Macau (via Hong-Kong)
Mr. J. Saario Kernira OY. Box 44, 02271 Espoo. Finland
Dr.M. Save BRGM. Mineral Processing Dept.• B.P. 6009. 45060 Orleans-Ia-
Source. France
Dr. H.J. Schulze Forschungsinstitut fUr Aufbereitung. Chemnitzer Strasse 40. 9200
Freiberg. Gennany
Prof. F. Smith Laurentian University. Ramsey Lake Road, Sudbury. Ontario P3E
2C6.Canada
Prof. G.A. Stalidis Department of Chemistry. Aristotle University, 540 06 Thessaloniki,
Greece
Prof. St. Stoev Higher Institute of Mining and Geology. Dept. of Mineral
Processing. 1156 Sofia. Bulgaria
Dr. A. Tiirkman Dokuz Eylii1 Univerity. Dept. of Environmental Engineering. 35100
Bornova - Izmir. Turkey
 531

Dr. T.S. Uyar Mechanical & Energy Engineering Department, Marmara Scientific &
Industrial Research Centre, P.O. Box 21, 41470 Gebze, Kocaeli,
Turkey
Dr. T. Varmakis Department of Chemistry, The University of Ioannina. 450 00
Ioannina, Greeee
Prof. W. Walkowiak Technical University ofWroclaw, Wybrzeze S., Wyspianskiego 27.
50-370 Wroclaw, Poland
Dr. J.B. Yianatos University of Santa Maria, Dept. of Chemical Engineering, Casilla
llO-V, Valparaiso, Chile
Dr. T.F. Zabel WRc Medmenham, PO Box 16, Marlow, Bucks. SL7 2HD. UK
Mr. P. Zarbos Department of Chemistry. Aristotle University. 54006 Thessaloniki.
Greece
Dr. A.I. Zouboulis Department of Chemistry. Aristotle University, 540 06 Thessaloniki,
Greece
532

1 Mr. K. Kydros 14 Dr. T .S. Uyar 27 Dr. HJ. Schulze 40 Dr. W. Petruk
2 Dr. S. Ersayin 15 Dr. OJ. McKee 28 Ms. S. Raatz 41 Mr. C. Lucion
3 Mr. N.K. Lazaridis 16 Prof. W.Walkowiak 29 Dr. P. Bevilacqua 42 Mr. RC. Moreira
4 Dr. U. lpekoglu 17 Dr. G.P. Gallios 30 Mr. J. Saario 43 Dr. J.A.C. Rocha e Sil,
5 Prof. 1. Gregory 18 Dr. A.l . Zouboulis 31 Prof. A.P. Mathews 44 Ms. W. Klar
6 Dr. G. Atesok 19 Dr. S. Hall 32 Dr. P. Mavros 45 Ms. O. Ozcan
7 Mr. D. Malliaris 20 Ms. N. Karaoglu 33 Dr. B. lpekoglu 46 Dr. K.A. Matis
8 Dr. M.RM. Leite 21 Dr. M. Akser 34 Dr. St. Gaidarjiev 47 Prof. F. Smith
9 Dr. T. Aggelidis 22 Dr. M. Patron is 35 Ms. L.A. Cunha 48 Ms. M.J. Pereira
10 Mr. D. Aslanidis 23 Ms. S.Y. Lee 36 Dr. M. Save 49 Dr. M.W. Chudacek
11 Mr. L. Ergiin 24 Dr. H. Mordogan 37 Dr. J.B. Yianatos 50 Dr. T. Vai"makis
12 Mr. R Pfeifer 25 Dr. A. Tiirkman 38 Prof. A.A. Abramov 51 Prof. RJ . Pugh
13 Ms. A. Daniilidou 26 Mr. B. Aksani 39 Dr. V. Panayotov 52 Mr. J. Ekmekci
INDEX

abstraction 43,335 calcite 331,357

acceleration, additional 178 capillary
acoustic wave 383 - force, attachment 175
admicelles 26,332 - (Laplace) suction 2
adsorption cassiterite 143
- electrostatic 335 cavities, air 221
- energy distribution 37 cellulose 500
- selective 331 chalcocite 74,78,310
ageing effect 317 - rims 313
agitation intensity 292,441 chalcopyrite 63,132,201,284,309
aggregate stability 177 - anhedral 309
aggregation number 26 - disease 135,310
air characteristic frequencies 103
- injection nozzle 439 charge
- insonification 389 - carrier427
- precipitation 214 - neutralisation 121
- release devices 437 chemisorption 331
algae 446 chitosan 122
a1ginates 122 chlorite schist 140
Alizarin Red S 268 chromites 415
alum 121 chromium speciation 489
amphiphilic substances 26 circuit stabilising 237
antimony 311 circulating load 251
apatite 264,331,357 clausthalite 139
aphrons478 clay 313
API gravity separator 482 coagulation 101,440
arsenic311 coal
axial dispersion coefficient 227 - activated 289
- brown289
backmixing 229 - coking 391
ballotini spheres 298 collection zone 214
baryte 309,348,384,413 collectors
base metal ores 132 - amines 357
binder 499 - fatty acids 331
bismuth 311 - kerosene 372
bleaching 507 - sodium myristate 345
bond, adhesive 2 - sulphydryl 84
bornite 284,310 - xanthoformiates 200
bridging, polymer 109 colligend 455
Brownian motion 10 1,112 collision
bubble efficiency 113
- column 229 - factor 166
- concentration 296 - radius 174
entrapment 434 comminution
- growth434 - perigranular 149
rising velocity 177 ,296 - time 152
- vibration 387 - transgranular 149
bypassing 224 complexation 263

533
534

contact angle 176,367 disengagement, bubble 214

continuous stirred tank reactors (CSTR) 184 dodecylammonium chloride 16
control dolomite 357
- adaptive 258 double layer, electric 3,105,270
- algorithm 257 dyes 488,502
- automatic, flotation 235
- objective 237 EDTA341
- proportional-integral (PI) 258 elasticity, surface 3
strategy 238 electrochemical effect 303
copper, porphyry 241,307 electrode
counterions 105 mineral 66
covellite 79,284,310 - noble metal 65,77
critical - poisoning 71
- angle 175 - reference 65
- micelle concentration 8,27 - saturated calomel 60,62
- thickness 174 - sensing 62
crocoite 284 - standard hydrogen 59
custelite 284 electroflotation 301,431
cyanide 283,312 electrophoresis 105
electrostatic patch 111
dead space 224 electrum 139,284,309
Debye-Hiickell06 Elexan 73
deinking 499 energy dissipation 3, 176
density Euler beta function 158
- joint probability 151 expert systems 259
- marginal size distribution 151
depressants fahlerz 309
- chromate 320 feldspar 412
- cyanide 320 fibre recycling 499
- dextrin 322 film
starch 322 aqueous 1
- sulphites 320 black
- sulphoxy compounds 320 - common 5
depression - Newton 5
- galena 320 - calcium sulphite, Bayraktar flot.
- pyrite 321 - drainage 2
deprotonation 264 rupture 2,174
design, flotation plant 436 - soap 2
diagrams - stability 1
- distribution 270 - thickness 3
- predominance area 270 - thin 4
dielectric constants 103 - aqueous 1
dimensionless number - common 5
- flow 215 filtration 445
- Froude 215 floatability 291
- Peclet 227 - correlations 291
- power215 - criterion 189
- Reynolds 215,435 - factor 166
- Weber215 - inherent 84
disjoining pressure 3 spectra 292
Direc delta function 154 flocculants
Dirichlet-Voronoi polyhedra 162 - inorganic salts 120
dixanthogen 64 - polymer 110,122
DLVO theory 6, 111 flocculation 101,433,441
dioctadecyldimethylammonium bromide 14 - hydraulic 442
 535

kinetics 115 galena 47,64,132,264,284,309

orthokinetic 116 gamma function 226
- perikinetic 115 gas
- rapid 112 - solubility 434
- sweep 121 - sparger 214,385
floes 101,433 geology 409
flooding 221 Gibbs 3
flotation goethite 211,283
- activation 267 gold 140,211,283,311,411
- cell granite 414
Bahr291 grind
- Benoit51O - minimum 127
- Davcra231 - optimum 127
- Jameson 291 gypsum 357
- Leeds 367
- Voith 510 Hallimond tube 303,360,419
- collectorless 84 Hamaker
- column 12,176,213,229 - constant 103
deinking 506 - function 4
depression 267 hematite. 350
- dispersed-air 431 hemimicelle theory 25
- dissolved-air 431 Henry's law 434
- efficiency 314 humic acid 440
- electrochemical 78 hydroeyclone 213
- electrolytic 301,431,481 hydrodynamics 224
- froth 303,475 hydrogoethite 284
- grease 515
- ion 455 IGME409
- kinetics 183,295,469 image anal ysis 131
- machines 393 induction
micro- 432 - gas 220
nitrogen 87 - time 173
- precipitate 455,480 ink 211,499
- pressure 432 insonification 384,402
vacuum 432 interactions
flow pattern 217 - electrical 104
fluorite 43,331,357,419 galvanic 80
fluoroapatite 265 - hydrophobic 107
foam fractionation 455 - steric 107
formamidine disulphide 284 - van der Waals 102
freibergite 132 interferometry 6,13
froth Jrgalan dye 488
- collapse 2 iso-gradelines 160
- drainage 188 iso-size lines 161
frother isotherm
- adsorption I Frunkin-Fowler 30
- MmC200 - Langmuir 30
- performance 200 - Temkin 31
- pine oil 200,404
- polypropylene glycol200 kaolin (mler) 502
fulvic acid 440 kaolinite 40
function kinetic models 194
- breakage 155 Krafft point 27
- transition 154
lamellae 2
536

Langmuir-Blodgeu 14 - volcanogenic 323

lapis lazuli 211 ozonation 446
Laser-Doppler Anemometty 223
layer paint effluent 488
diffuse 104 particle-bubble equilibria 291
- double, electric 104,333 particle popUlation 149
- Stem 105 particulate system 149
leaching patch
- electrum 285 - electrostatic III
- gold 284 - hydrophobic 21
- silver 284 pentIandite 92,201
Lewis acid 263 period, rapid mix 121
liberation 126,293,313 permittivity 105
- degree 150 phosphates 357,412,493
lignine 501 phylittes 310
lime 93 pigments 211,499
Lifshitz theory 21,103 PIRANA 73
Liouville equation 156 Plateau border 2
loading 221 plumbojarosite 284
locked-cycle test 20 1 point, zero charge 363
lyotropic liquid crystals 26 Poisson-Boltzmann equation 14
Poisson-Voronoi texture 162
magnesite 357 polyethylene oxide 17
manganese 412 potash salt 179
Marangoni effect 3 potential
marcasite 139,309 - electrochemical 77
Markovian property 154 - electrokinetic 467
materials balancing 125 - mineral60
mercury 311 - mixed 59
miargirite 284 - pulp 60,77
mica 13, 106 - redox 322
Microflot 291 - rest 81
mineral - solution 60
- activity 188 - Stemlayer4
- exploration 409 surface 265
- salt-type 357 zeta 4,105,269,333,360,420
mixer, vibration 402 potentiometric titration 269
mixing 437 potentiostats 94
modifier 358 Pourbaix diagram 73
Calcon 375 precipitation
- pyrophosphate 375 - bulk331
sodium fluorosilicate 372 - surface 42,331
sodium hexametaphosphate 370 probability
sodium silicate 370 - adhesion 171,184,212
molybdenite 202 - aggregation 184
collision 171,212
Nemst equation 61 - detachment 212
non-ideal flow 224 particlelbubble stability 171
three-phase-contact 171
on-stream analysis (OSA) 235 programs
ore - CALFLOT 202
- beneficiation 211 - FITEQL 269
- metamorphosed 127,309 - MATBAL 143
- microscopy 313 - SOLGASWATER 270,334
- unmetamorphosed 309 projection path. bubble 172
 537

protonation 264 starch 14,122

pyrene 26,40 stator 215
pyrite 48,78,132,187,251,307 Stem
- euhedral 309 - equation 30
- subhedral 310 - layer 105,332
pyroclastic rocks 308 - potential 4
pyrolusite 284 stibnite 414
pyromorfite 284 stimulus 225
pyrrhotite 63,74,91,187,251 Stoke's law 435
sublate 455
quartz 132,179,187,264,276,284,309 sulphide
minerals 57
rate - ore type
- attachment 296 - black 309
- detachment 296 - Cyprus 310
recycle 433 - Keiko 309
recycling collapse 499 - Kuroko309
Rehbinder 171 - Oko 309
repulsion pyritic 310
- hydrationl06 - siliceous 308
residence time distribution 218,224 - yellow 309
rhyolite 309 - oxidation 316
rotor 215 surface
- condensation 332
sapphire 14 - elasticity 3
scheelite 345,357,419 - hydration 264
Schulze-Hardy rule 114 - light scattering 3
sea-floor fracture 308 loading 432
selenium 139 modification 263
self tuning 257 precipitation 42,331
sensitisation 109 speciation 263
separation viscosity 3
- efficiency 166 suspension 219
- magnetic 284 SYRAN0410
sepiolite 373
set point 237 talc 372,390
settling tannins 14
- differential 118 tennantite 312
- velocity, particle 178 tetrahedrite 132,309
shear thickness, critical 173
- rate 215 thiourea 283
- stress 215 tin 143
silicic acid 374 ttansition function 154
silver 283,310,488 Traube's law 27
- metallurgy 140 Tromp curve 152,190
size turbidity 447
- distribution 126 turbulence 199,213,222,395,432
sliding time 173 Turing 149
solubility product 333
sorption 285 ultrafines 313,359
speciation 488 uranium 412
sphalerite 74,132,292,309
- activation 134 value, intrinsic 311
stabilisation steric 107 vamadite 284
stability ratio 113 velocity
538

- fluctuating 223
- turbulent 223
vibrations, mechanical 383
vibro-
acoustic 383
attrition 393
desorption 392
emulsification 400
mill 385
viscosity, surface, 3
volcanism 309
volcanogenic ores 126,132
voltammetry, cyclic 62
voltammogram 62
vortex separator 523

washability curve 164

wastewater flotation 513
water
- glass 392
- potable 437
- storm 521
- wash 212
- waste 437,475,513

zeolites 420
zero-time factor 198
zircon 199