Shortcomings in the estimation of clay fraction by hydrometer

Abstract

The estimation of clay fraction is important for predicting the engineering properties of a soil. SANS 3001 GR3 (SANS 2011) specifies a procedure for clay fraction determination using a hydrometer. It has long been suspected that there may be flaws in this approach. Some of the possible sources of error have been suggested, but little or no change has been made in the standard procedures for assessment of clay fraction in well over half a century. This paper deals with a microscopic examination of some typical South African clayey soils to assess the adequacy of dispersion and possible consequences for clay fraction determination in currently specified hydrometer procedures. Clays are examined both with and without dispersant, and with and without labelling of clay minerals using an exchangeable cation dye. © 2016, South African Institute of Civil Engineers. All rights reserved.

Full text

14 Stott P, eron E. S hortcoming s in the estima tion of clay frac tion by hydrometer.
J. S. Afr. I nst. Civ. Eng. 2016;5 8(2), Art. #1232, 11 page s. http://dx.doi.org /10.17159/2309-8775/2016/v58 n2a2
TECHNICAL PAPER
Journal of the South african
inStitution of civil engineering
Vol 58 No 2, June 2016, Pages 14–24, Paper 1232
PHILIP STOTT (Pr En g, MSAICE) received BSc
(Hons) and MSc degrees from Manchester
University. He le ctured at Ahmadu Bello
University in N igeria and the University of th e
Witwatersr and in Johannesburg, and has been
working as a consult ing engineer since 1984. He
received the Henry Adams Award from the
Institute of Structural Engineers, London.
Currently he is a DTech Candidate and a me mber of the Soil Mechanics
Research Group at the Cent ral University of Technology (CUT ), Bloemfontein.
He is also a member of t he Structural and Geotechn ical Divisions of SAICE
(South African Institution of Civil Engineering).
Contact details:
Departm ent of Civil Engineering
Central Universit y of Technolog y
Private Bag X2039
Bloemfontein 9300
South Africa
T: +27 82 253 4001
E: philip@Y10.co.za
DR ELIZABET H THERON (Pr Tech Eng) received
an NHDip and MTech in Civil Eng ineering from
the then Technikon Free State (renamed sin ce to
the Central Universi ty of Technology, or CUT for
short), Bloe mfontein, South Africa. She also has a
PhD in Geograp hy from the University of the
Free State, Bloemfontein. Sh e has lectured in
civil enginee ring at the forerunner of the CUT
since 1987, and has received the Pres tige Award of the Vice-Chan cellor
(Academic: Class G old). She is currently a senior lec turer at the CUT, and
project mana ger of, and researcher in, CUT’s Soil Mechan ics Research Group.
Contact details:
Departm ent of Civil Engineering
Central Universit y of Technolog y
Private Bag X20539
Bloemfontein 9300
South Africa
T: +27 51 507 3646
E: etheron@cut.ac. za
Keywo rds: hydrometer analysis, clay fraction, dispersion of clays,
de-occulation
INTRODUCTION
An estimation of the clay fraction of a soil
is required for a number of soil evaluations,
including common methods of assessing
heave potential relating to foundation
design. Van der Merwe’s method (Van der
Merwe 1964) uses the plasticity index (PI)
and clay fraction. Skempton’s “activity” is
defined as PI/clay fraction (Skempton 1953).
The method of estimating clay fraction
by hydrometer, as specified in the South
African standard SANS 3001 GR3 (SANS
2011), is very similar to that specified in
Britain, America, Australia and many other
countries. It is, however, somewhat dubious
in its efficiency. Savage suggested that the
hydrometer method may be doubtful due to
four factors (Savage 2007):
1. Stoke’s law assumes all particles to be
spherical, while clays are flaky.
2. De-flocculation of many clays is seldom
fully completed at the time of testing.
3. Clay particles are partially carried down
by the larger particles.
4. A relative density of 2.65 is assumed for
all particles, which may not be true.
Savage proposed a method of estimating clay
fraction indirectly by using Skempton’s activity
formula. Unfortunately there seems to be no
clear pattern of correlation bet ween hydrome-
ter results and Savage’s method. Savage did not
give examples, and the examination of samples
by the Central University of Technology (CUT)
Soil Mechanics Research Group revealed no
clear pattern of correlation (some values high-
er, some lower than the hydrometer). There
appears to be no way of telling which gives the
better estimate, or what the likely margins of
error may be. The method does not appear to
have found wide acceptance.
Progress has been made on Savage’s first
point, the question of non-sphericity of par-
ticles. It has been addressed by laser scatter-
ing techniques for particle suspensions (e.g.
Konert & Vandenberghe 1997; McCave et
al 1986; Ma et al 2000). This technique has
enabled an allowance to be made for particle
shape, and has generally led to a small but
significant increase in clay fraction estima-
tion. Such an allowance is not specified in
SANS3001GR3.
Savage’s fourth point seems to have
drawn little attention, since almost all
non-organic soil components have densities
reasonably close to 2.7, and the likely error
due to this factor is probably quite small. His
remaining two points concern dispersion,
and obviously merit attention.
Research currently being done on the
theoretical aspects of dispersion of clay par-
ticles suggests that the problem is far from
well understood (e.g. Robinet et al 2011),
and it remains very difficult to assess most
aspects of dispersion for any specific clay
and solute system. Experimental research
on de-flocculation/dispersion using non-
traditional de-flocculants currently appears
to be concentrated on ceramics (e.g. Al-Lami
2008). Such dispersants produce functional
groups acting as spacers between clay par-
ticles and may be too expensive for routine
soils testing at this stage of development.
Work on de-flocculation/dispersion relevant
to soil mechanics continues to use methods
and dispersants which have been in use for
many years (e.g. Rodriguez et al 2011; Rolfe
et al 1960). Attempts to assess the magnitude
of error likely to be involved in incomplete
dispersion continue to use the hydrometer
itself as the instrument of investigation
Shortcomings in the
estimation of clay
fraction by hydrometer
P Stott, E Theron
The estimation of clay fraction is important for predicting the engineering properties of a
soil. SANS 3001 GR3 (SANS 2011) specifies a procedure for clay fraction determination using
a hydrometer. It has long been suspected that there may be flaws in this approach. Some of
the possible sources of error have been suggested, but little or no change has been made in
the standard procedures for assessment of clay fraction in well over half a century. This paper
deals with a microscopic examination of some typical South African clayey soils to assess the
adequacy of dispersion and possible consequences for clay fraction determination in currently
specified hydrometer procedures. Clays are examined both with and without dispersant, and
with and without labelling of clay minerals using an exchangeable cation dye.
Note to readers:
This paper contains 18 supporting photogr aphs which are discussed
in detail on page s 20–22. However, due to space constraints, the
photographs have b een spread evenly throughout the p aper. We
trust that this will not inconvenience our readers .

Journal o f the South Africa n Institution of Ci vil Engineerin g • Volum e 58 Number 2 June 2016 15
(Nettleship et al 1997; Rodriguez et al 2011).
This paper is primarily concerned with
Savage’s second point, the dispersion of clay
particles. His third point, clay being carried
down with larger particles, follows from this
as a matter of course.
THEORETICAL BACKGROUND
TO DISPERSION OF CLAYS
The behaviour of dispersants is complex and
appears to be still imperfectly understood.
This outline synthesises information from
Das (2008), Zschimmer and Schwartz (2014),
Nettleship et al (1997) and Robinet et al (2 011).
Clay particles carry charges which leave
their inner structure negatively charged and
tend to leave their outer edges positively
charged. When active clay soils are mixed
with water, two things tend to happen.
Firstly, water molecules, which are polar
(their atomic structure leaves one side posi-
tively charged and the other side negatively
charged, while remaining neutral as a whole),
surround cations (positively charged metal
ions) in the soil. When coated with water the
cations become mobile. They are strongly
attracted by the negative charges in the inte-
rior of some types of clay minerals and pen-
etrate between the tetrahedral and octahe-
dral sheets of these clays, forcing the sheets
apart. This is the reason why some clays can
increase in volume powerfully when wet-
ted. Secondly, the positively charged outer
edges of the clay particles attract negatively
charged ions which form a diffuse layer
whose concentration diminishes with dis-
tance from the clay surface. Multi-valent ions
provide multiple electro-negativity and rela-
tively few of them need to congregate around
a clay particle to balance the positive charge
on the surface of the clay. The resulting field
surrounding the clay has marked peaks at the
ions and troughs between them. This allows
adjacent clay particles to maintain mutual
electrical attraction by fitting troughs on one
particle to peaks on another.
In order to assess clays by their rate of
precipitation, as in the pipette and hydrom-
eter methods, it is necessary to disperse
the particles of clay into the water through
which they precipitate. Mechanical agitation
is essential for this, but is not sufficient on its
own. Chemical dispersion is needed to break
the bonds of electrical attraction holding
assemblages of clay particles together.
Dispersants work in three ways. The
first is to replace multi-valent ions at the
clay surface by mono-valent ions. When
an individual clay particle is surrounded
by sufficient mono-valent ions to render it
electro-neutral, the field surrounding it is
relatively uniform; clay particles in such a
state cannot attract each other by fitting
electrostatic peaks to troughs. The second
way is by reacting with multi-valent ions
to form chemical complexes, making them
unavailable for attraction to clay surfaces.
The third manner is by forming functional
groups which act as spacers between the clay
particles, effectively preventing them from
approaching each other.
The combined action of clay particles,
cations and dispersing agents is complex.
Above a certain concentration of dispersant
the diffused double layer starts to become
thinner, repulsion between the particles
Photograph 1 Olive-grey residual clay not treated with dispersant but mechanically stirred as
specified in SANS 3001 GR3
Comparison with the 30 micron by 2 micron rectangle suggests that there are particles of
about 2 microns attached to several of the silt particles. Some clay-size particles appear to
be dispersed in suspension. Note the faint pinkish cloudy patches covering a considerable
part of the field of view (most of which is not within the lens’s range of sharp focus).
Photograph 2 Olive-grey residual clay, not treated with dispersant, after addition of 3 mg
methylene blue per 1 g of soil
Many of the particles of 2 microns and a little larger adhering to the silt particles show
a faint blue outline. Much of the pinkish cloudy area has taken in methylene blue and
appears to be composed of extremely small clay particles. Little of the field of view is in
focus, but some deeply-stained individual particles smaller than 1 micron are discernible.

Journal o f the South Africa n Institution of Ci vil Engineerin g • Volum e 58 Number 2 June 201616
reduces, and at higher concentration turns to
attraction, allowing flocculation to occur.
Sodium hexamataphosphate is one of the
most popular dispersing agents. It is specified
in the standards for assessing clay fraction in
Britain, America, Canada, Australia, Japan
and other countries. It provides mono-valent
sodium ions to coat the clay surface, as well
as phosphate groups to form complexes with
multi-valent cations. Sodium carbonate may
be added to increase alkalinit y; this has been
found to improve the dispersive efficiency
slightly in some circumstances (Rolfe et al
1960) and to extend the useful life of the
dispersant (Nettleship et al 1997).
A mixture of sodium hexametaphosphate
and sodium carbonate is specified as the
dispersant for the hydrometer procedure of
SANS 3001 GR3.
BACKGROUND AND AIMS
OF THE STUDY
Investigations are being undertaken by the
CUT Soil Mechanics Research Group seek-
ing solutions to the problem of large num-
bers of failures in government subsidy houses
due to heaving foundations. It appears that
in some of the failures investigated, the geo-
technical investigation had given misleading
indications of clay fraction. In one case
hydrometer analysis indicated less than 10%
clay on a site where notable shrinkage cracks
in the ground surface suggested at least 20%
clay content. Since hydrometer analysis is
normal for almost all construction projects
in South Africa, such shortcomings in the
method may be relevant for a wide range of
situations. The aim of this study was to gain
an insight into the reliability of the clay frac-
tion indicated by the hydrometer for a range
of clays typical of those found in construc-
tion projects in South Africa.
It is common practice among researchers
to examine clays with an electron microscope.
This has the advantage of very high magni-
fication. Preparation involves samples being
treated by techniques such as drying and coat-
ing with gold (Nettleship et al 1997). This does
not replicate conditions in the hydrometer. A
series of exploratory tests were conducted at
the geotechnical research laboratory of CUT
to examine the possibility of using an optical
microscope/digital camera combination to
investigate the efficiency of the dispersion
of clays using method GR3. The results sug-
gested that dispersion was not always satisfac-
tory. Many clay particles appeared to remain
as conglomerations, while others remained
adhered to silt and sand particles.
The procedures used by soil science
laboratories differ somewhat from those for
engineering materials. Previous cooperation
with the Soil Science Department of the
University of the Free State had sometimes
found higher clay fractions indicated by
their procedures. It was arranged for six
samples to be prepared by the UFS soil sci-
ence laboratory using their normal method.
The dispersant is 50 g per litre sodium
hexametaphosphate solution (the amount
applied depends on soil type), sonification in
a dismembrator is specified for clay soils, and
mechanical dispersion is of shorter duration
but at a higher speed than SANS 3001 GR3.
The samples were examined to determine
whether dispersion by this treatment was
visibly more effective than SANS 30 01 GR3
for these six soils.
Photograph 3 Olive-grey residual clay not treated with dispersant after addition of 8 mg
methylene blue per 1 g of soil
None of the pinkish cloud remains, it appears to have taken in dye and can be deduced
to be an aggregation of very small clay particles. Almost all of the silt particles appear to
be completely covered with clay. The assembly moves as a flexible unit in the currents
caused by evaporation around the edge of the cover slip. No individual clay particles,
which are not part of an aggregation, can be seen.
Photograph 4 Olive-grey residual clay treated with dispersant and stirred as specified in
SANS3001GR3
Many clay-size particles are dispersed but some remain attached to silt particles or form
associations of three or more particles not attached to silt.

Journal o f the South Africa n Institution of Ci vil Engineerin g • Volum e 58 Number 2 June 2016 17
EQUIPMENT, MATERIALS AND THEIR
USAGE IN THE INVESTIGATION
Microscope and camera
An optical microscope with objectives
of 10x, 40x, 60x and 100x was equipped
witha digital camera (resolution 9 mega
pixels). Magnification resulting from the
combinedeffects of the microscope’s
lenses andthe camera was assessed
by measurements on a 100 lines/mm
diffractiongrating.
A drop of sample prepared for hydrome-
ter analysis was placed on a microscope slide,
covered with a cover-slip and photographed
at various magnifications.
Photographs were taken at various loca-
tions on the slide. Most of the photographs
in this investigation were taken using the
microscope’s 40x objective since more pow-
erful lenses give a very small depth of focus.
Magnification
The combined optical and digital magnifica-
tion can be defined in different ways. The
computer screen that was used to view the
images showed lines spaced at 10microns
on the diffraction grating, spaced at 40 mm
on the screen when using the 40x objective.
This implies a magnification of 4000times.
Alternatively, the 10-micron spacing on the
diffraction grating corresponds to 150 pixels
on the photographs produced by the
camera using the same objective. The most
convenient way of indicating magnification
is by incorporating a reference object of
known size. All of the photographs in this
article have a rectangle superimposed to
indicate the scale. The length and breadth
of each rectangle represent 30 microns and
2microns respectively.
Variations in procedure
Samples were also prepared employing vari-
ations to the normal procedures in order to
examine the inf luence of time of soaking in
dispersant, time of agitation, concentration
of dispersant and volume of dispersant used.
Method GR3 specifies only minimum times
of soaking and agitation. All of the times
involved are within these specifications,
and this part of the investigation served
only to verify whether this aspect of the
specification is adequate. Examination of the
concentration of dispersant was prompted
by the finding of a difference in hydrometer
yield for certain clays using the Japanese and
American standards, which specify different
concentrations of sodium hexametaphos-
phate (Mishra et al 2011).
Methylene blue
In addition, samples were treated with
methylene blue (MB), with the aim of
labelling clay particles for positive identi-
fication. Methylene blue (C16H18 N3SCl) is
an effective indicator of clay, as it readily
exchanges places with cations in the clay
mineral structure, the amount depending
on the cation exchange capacity (CEC) and
specific surface area (SSA) of the clay min-
erals (Turoz & Tosun 2011). Active clays like
montmorillonite have high CEC and SSA,
and readily take in methylene blue. When
MB is available in large concentrations,
Photograph 5 Olive-grey residual clay treated with dispersant and methylene blue
The small agglomerations of clay-size particles show little, if any, staining, suggesting very
low CEC. The silt particle at bottom centre appears to be coated with very small, high CEC
particles which are very darkly stained. The deeply stained agglomeration in the centre
is about 50 microns in length and 25 microns in width. It appears to be made of very
small, high CEC particles, and it seems to engulf several silt and clay-size particles. Such
agglomerations were not uncommon in this sample, but probably not common enough
to ensure a meaningless clay fraction from hydrometer analysis.
Photograph
6 Red-brown soil from Limpopo after mechanical stirring without addition of
dispersing agent
Many of the silt-sized particles appear to have clay-sized particles adhering to them. A
faint cloudy pinkish haze, as noted in Photograph 1, is again evident. Most of the silt
particles appear to be clustered together in loose associations.

Journal o f the South Africa n Institution of Ci vil Engineerin g • Volum e 58 Number 2 June 201618
such clays rapidly become totally opaque
and appear in photographs as dark blobs in
which no structure can be seen. Inactive
clay minerals like kaolinite have low CEC
and SSA and show little colouring until
high CEC/SSA fractions present are already
deeply stained. Progressive addition of small
amounts of dye can therefore give an indi-
cation of the types of clay mineral present
in a sample, and can also help to establish
whether the clay-size particles which can
be seen adhering to silt and sand particles
are, in fact, composed of clay minerals. Any
additive to the soil solution which affects
the cation balance will inevitably influence
the effectiveness of the dispersant. Only
small quantities of methylene blue were
therefore added to the dispersed samples. It
could be expected that silt and sand would
not be coloured, and high CEC / high SSA
clays (e.g. montmorillonite) would be col-
oured after adding very little dye, whereas
low CEC / low SSA clays (e.g. kaolinite)
would be coloured only after the addition of
considerably more dye.
THEORETICAL CONSIDERATIONS
AND STRENGTHS / WEAKNESSES
OF THE METHOD EMPLOYED
Soil mechanics and soil science generally
consider all particles of 2 microns and
smaller to be clay-size particles, and those
from 2 microns to 60 microns (or some
other arbitrary figure of this order) to be
of silt-size. But particles and agglomera-
tions of clay minerals typically range from
about 0.1 micron to slightly more than 2
microns; non-clay particles typically range
from about 1 micron upwards (Robinet
et al 2011). Some clays, e.g. kaolinite and
haloysite, may have particles considerably
larger than 2 microns, as can be seen in
electron micrographs by Bühmann and
Kirsten (1991). There is thus a range where
size classification may not correspond with
mineral classification. Certain important
aspects of soil behaviour (e.g. volume
change) depend on clay mineral content,
while hydrometer and pipette analyses
attempt to establish only particle sizes, not
mineral content. Many of the individual
particles observed were in this ambigu-
ous range of 1 to 2 microns, raising the
question of whether they are clay particles
which need to be dispersed, or silt particles
which do not flocculate and should not
needdispersion.
The magnifications possible with the
optical microscope and camera combina-
tion used in this investigation are probably
sufficient to distinguish most of the range
typical for clay through silt to sand, but not
adequate to measure the smallest particles
in this range. Since all samples remained in
aqueous suspension, all of the smaller indi-
vidual particles were subject to Brownian
motion. At the highest magnification (100x
objective – 37.5 pixels per micron, 10000x
magnification on the computer screen),
particles at the lower end of the clay-size
range could be distinguished in many of
the samples, but their Brownian motion
hindered observation or measurement since
they suddenly appear in the focal plane, and
disappear as they move away from the focal
plane. Photographing them was not very
successful, possibly because the exposure
time of the camera/computer combination
was too long. Many small particles were vis-
ible and could be photographed where they
formed part of large agglomerations or were
attached to silt or sand particles.
Photograph 7 Red-brown Limpopo soil as in Photograph 6 after stirring without dispersant
The large silt/sand particle appears to have lost part of its coating of clay-size particles.
The pinkish haze and loose groupings of silt particles are again evident.
Photograph 8 Red-brown soil from Limpopo after addition of 3 mg methylene blue per gram of soil
The pinkish cloud appears to have absorbed methylene blue, revealing itself to be an
extensive agglomeration of small clay particles, as in the case of the olive-grey clay in
Photographs 1, 2 and 3. Most of the silt grains which appear to be covered with clay-size
particles have scattered spots of dark blue stain suggesting a few high CEC clay particles
among many low CEC particles.

Journal o f the South Africa n Institution of Ci vil Engineerin g • Volum e 58 Number 2 June 2016 19
GENERAL CONSIDERATIONS
The following considerations in terms of the
microscopic investigation should be noted:
1. Those samples extracted for microscopic
investigation at the UFS laboratory were
taken by pipette after a settling time
calculated to give only silt- and clay-sized
particles at the depth of extraction (larger
particles having settled below this level).
The largest particle sizes measured were
of the order 50 microns, suggesting that
the sample was, indeed, restricted to clay
and silt-sized particles. Samples prepared
in the CUT laboratory were taken imme-
diately after agitation, and some samples
contained particles considerably larger
than 50 microns, allowing examination of
sand grains as well as silt.
2. The cover slip over the sample was
supported by the largest particles, and
consequently a depth of about 50 microns
was filled with suspension for the UFS
samples, and up to about 100 microns for
the CUT samples. Depth of sharp focus
at high magnification is far smaller than
this and consequently photographs neces-
sarily had most of their field out of focus.
3. Since clay sizes range from 2 microns
downwards, the concentration of suspen-
sion specified in the hydrometer method
allows too many clay particles in a depth
of 50 microns for convenient optical
differentiation. This made dilution of
the hydrometer samples necessary. The
majority of samples were diluted with
three times their own volume of de-ion-
ised water. This dilution was arbitrarily
chosen and was considered adequate for
this purely qualitative investigation.
4. The gap of approximately 50 to 100
microns bet ween slide and cover slip
allows evaporation of the suspension’s
water around the edges. It may be pos-
sible to seal around the edge of the cover
slip and prevent evaporation, but it was
found that the movement of water caused
by evaporation was helpful in distinguish-
ing between clay-size particles which
were dispersed and free-f loating, and
those which were attached to silt or sand
particles or formed agglomerations with
other clay particles. This consideration
results in a very limited time available for
the observation of each slide.
5. When samples dry out they conglomer-
ate, making it difficult to draw conclu-
sions about the behaviour of the clay in
conditions relevant to the pipette and
hydrometer tests. Only obser vations in
suspension conditions were considered in
this investigation.
SAMPLES USED IN THE
INVESTIGATION
Samples of six widely different clay soils
(from the Free State, Northern Cape,
Western Cape and Limpopo) were mechani-
cally agitated, as in procedure GR3, but
without first soaking in dispersant. Samples
of the same soils were prepared with both
dispersant and mechanical agitation at
CUT, as per SANS 3001 GR3, and at the Soil
Science Laboratory of UFS using standard
soil science procedures. From the six soils,
two were selected as showing typical features
and illustrating the general effectiveness
of the investigation’s procedures. One soil
appeared to show fair dispersion, the other
Photograph 9 Red-brown Limpopo soil mechanically stirred (without treatment with dispersant)
and subsequent addition of 6 mg of methylene blue per gram of soil
The dense mass of deeply stained clay particles appears to almost completely engulf a silt
particle covered with barely stained clay-size particles.
Photograph 10 Red-brown Limpopo soil not treated by dispersant with 6 mg/g methylene blue added
The large blue structure is more than 150 microns by 50 microns in size. Within this
structure silt particles can be distinguished. Much of the agglomeration appears to
consist of clay particles of about 1 micron or smaller. The agglomeration slightly above
and left of centre appears to consist almost entirely of a different species of clay particles
of about 2 microns which are very lightly stained. Very few soil particles are visible which
are not part of an agglomeration.

Journal o f the South Africa n Institution of Ci vil Engineerin g • Volum e 58 Number 2 June 201620
very inadequate dispersion. The first of these
samples is shown in Photographs 1 to 5, the
second in Photographs 6 to 13. They provide
a reference frame and show widely different
clays with and without dispersant, both
with and without methylene blue. Features
of some of the other soils are shown in the
remainder of the photographs.
OBSERVATIONS
Photograph 1 is of an olive-grey residual clay
from a proposed housing development in
Bloemfontein. Tests at the UFS Soil Science
Laboratory gave LL 72, PI 26 and clay frac-
tion 41% (by both hydrometer and pipette
methods). Photograph 2 shows the same
sample after the addition of 3 mg of MB per
gram of soil. Photograph 3 shows the same
sample after the addition of a further 5 mg/g
of MB dye.
It appears that a large number of very
small clay particles bind considerable
numbers of various kinds of particles into
associations. Currents caused by evaporation
of the suspension’s water show that these
associations are flexible, but strongly tied
together and move as a unit.
Photograph 4 shows the same soil after
treatment with dispersant as specified
in SANS 3001 GR3. Comparison with
Photograph 1 shows a very large increase in
clay-size particles dispersed throughout the
water. There are, however, some clay-size
particles adhering to silt particles, and a
number of small agglomerations of clay-size
particles with no visible silt core.
Photograph 5 shows this same dispersant-
treated sample after the addition of meth-
ylene blue. While this sample shows good
dispersion compared to the untreated state,
it is apparent that dispersion is not complete.
The majority of dispersed particles appear to
be about 2 microns in size, but small, deeply
stained particles of less than 1micron can
also be seen, and it is difficult to discern
whether they are free or attached to larger
clay-size particles. The agglomerations of
clay-size particles, the clay-coated silt parti-
cles and the agglomeration of very fine clay
will probably not precipitate at a rate which
will ensure their contribution to the clay
fraction being recorded by the hydrometer.
Photograph 6 is of a red-brown soil from
the Limpopo Province, which has a history of
giving variable results in soil tests and caus-
ing difficulties in construction. Treatment
was only mechanical stirring of the raw soil
without dispersant. Commercial laboratory
results for samples sent by CUT as part of a
parallel testing programme ranged bet ween
17% and 56% for clay fraction, and between
31 and 43 for PI.
As in the case of the olive-grey
Bloemfontein soil there are shadow y pinkish
bands associated with the distinctly visible
particles. There are also many clay-size
particles adhering to most of the silt-size
particles.
In Photograph 7 a large grain of silt
appears to be mostly covered with clay-size
particles. Part of the grain, however, is com-
pletely clean and free from clay coating. It is
possible that it was struck by one of the pad-
dles of the mechanical stirrer and some of
the coating was torn away. The coating of the
lower right area seems to have come loose
from the large particle, but remains attached
to the clay coating above.
Photograph 8 shows the same sample
after addition of 3 mg/g of methylene blue. A
few small blue spots are visible on the larger
silt-sized particles, but the majority of par-
ticles of about 2 microns remain unstained.
A clearly visible cloud of very small blue-
stained particles has largely replaced the
faint pink cloud, suggesting that the cloud
consists of very small and possibly translu-
cent clay particles with high CEC/SSA.
Photographs 9 and 10 show the same
sample after the addition of a further 3 mg/g
Photograph 11 Red-brown Limpopo soil after treatment with dispersant at the UFS Soil Science
Laboratory
Many clay-size particles are well dispersed, but a substantial number remain attached to
silt particles. There are also cloudy pinkish areas similar to those in Photograph 1.
Photograph 12 Same sample as Photograph 11 after addition of methylene blue

Journal o f the South Africa n Institution of Ci vil Engineerin g • Volum e 58 Number 2 June 2016 21
of methylene blue. Photograph 9 shows a
grain of silt, apparently covered with low
CEC clay, almost completely enveloped by
high CEC clay particles. A number of small
clay particles appear to be dispersed into the
surrounding water, but the majority are asso-
ciated with aggregations. In Photograph 10 a
large agglomeration of clay appears to contain
several small silt particles of varioustypes.
Photograph 11 shows a sample of the
same soil treated with dispersant. Many
small clay particles are dispersed, and, com-
pared to Photographs 6 and 7, the groupings
of silt particles have largely disappeared. Yet
many of the slit particles remain totally or
partially coated with clay-size particles, and
many agglomerations of two or more clay-
size particles can be seen.
Photographs 12 and 13 show the same
soil after addition of methylene blue.
Compared to Photographs 9 and 10, where
no dispersant was used, dispersion is clearly
improved, but most of the silt particles are
seen to be covered with clay, and the faint
pinkish clouds again appear to be revealing
themselves as very fine clay particles which
are not well dispersed and may settle in the
hydrometer as silt-size aggregates rather than
as individual clay particles.
Photograph 13 shows an exceptionally
large agglomeration of small clay particles
engulfing several silt particles and numer-
ous 1 to 2 micron clay particles against a
background of well-dispersed clay particles.
Nettleship et al (1997) came to the conclusion
that their anomalous observations of settle-
ment in the hydrometer might be explained
by agglomeration taking place while particles
were settling during the test. It seems more
likely that this could be the case here than
that such an agglomeration could have sur-
vived 15 minutes of stirring at 1 500 rpm after
prolonged soaking in dispersant.
The amount of clay which is obviously
not dispersed in Photographs 12 and 13 sug-
gests that it is very unlikely that the hydro-
meter will give a reliable estimate of the true
clay fraction of this soil.
Photograph 14 shows a soil from a
housing project in the Northern Cape.
Hydrometer analysis had shown the clay
fraction for almost all of the samples from
this site to be very low. This led to a low
value of Van der Merwe’s predicted heave
being accepted for design. Heave damage did,
however, occur on the project.
A considerable fraction of the clay-size
particles visible in the photograph are attached
to silt particles of various sizes. It is not clear
whether all the agglomerations of clay-size
particles have a silt core, but it appears that
much of the clay in this sample has not been
dispersed. Hydrometer analysis could underes-
timate the clay content quite drastically. This
might explain the unexpected damage which
occurred at the housing project.
Another sample, from less than 100m
away, was prepared according to SANS3001
GR3 at the CUT laboratory. A house
had become structurally unsound due to
heave while still under construction a few
metres from where this sample was taken.
Photographs 15 and 16 are of this sample. In
Photograph 15 many of the visible clay-sized
particles are attached to silt particles. It
appears that the large congregation of clay-
sized particles in Photograph 16 surrounds a
root hair or similar thin, thread-like structure.
Considerably more clay is undispersed
than is dispersed. This would suggest the
likelihood of a misleading estimate of clay
fraction by the hydrometer method.
Photograph 17 shows a low-activit y
kaolinitic soil described as “light yellow
silty clay” from the Western Cape after
Photograph 13 Same sample as Photograph 12
The dispersion and agitation procedures have produced many well-dispersed clay
particles, but the large agglomeration shown here is far from dispersed. This sand-size
grouping of silt and clay is unlikely to settle at the rate expected of clay, nor are the
numerous smaller aggregations.
Photograph 14 Soil from a subsidy housing project in the Northern Cape after treatment with
dispersant and mechanical agitation at the UFS Soil Science Laboratory
The majority of silt particles remain covered with clay-size particles.

Journal o f the South Africa n Institution of Ci vil Engineerin g • Volum e 58 Number 2 June 201622
dispersion and addition of methylene blue.
It has LL 34, PI 13 and LS 4.6 suggesting
that its heave potential is very low. All of the
clay particles are about 2 micron or slightly
larger in size, which is consistent with the
clay being kaolinite. This is the only sample
tested which showed no small, high CEC clay
particles; all of the other samples showed
a range of sizes and probable clay mineral
types. Many agglomerations of clay particles
are evident. Since it is unlikely that these will
settle at the rate expected of individual clay
particles, the clay fraction determination is
again likely to be unreliable. The description
“silty clay” seems inappropriate, since little,
if any, silt is evident. The agglomerations of
clay are, however, of silt size and it could be
that they had settled in the hydrometer at
the rate expected of silt-size particles and
had been incorrectly assessed as silt.
ADDITIONAL TESTS
Although it appears that one or two of the
samples tested showed fair dispersion (as in
Photographs 4 and 5), none showed unques-
tionably satisfactory dispersion (i.e. little or
no clay-sized material forming agglomera-
tions or associations with other particles).
Some showed very poor dispersion (as in
Photographs 12 to 16). Tests at the CUT
geotechnical research laboratory were carried
out to assess the effect of concentration of
dispersant, volume of dispersant, length of
time of submersion in dispersant and time
of mechanical agitation. The GR3 procedure
includes only minimum times for soaking
and agitation, so this served only to check the
adequacy of this aspect of the specification.
SANS 3001 GR3 calls for a minimum of
16 hours submersion in the dispersing agent.
Various periods from 16 hours to 2 weeks
were tested. No visible improvement in dis-
persion was observed.
SANS 3001 GR3 calls for a minimum
mechanical stirring time of 15 minutes at
1500rpm. Various periods from 15 min-
utes to 24 hours were tried with no visible
improvement in dispersion observed. This
tends to confirm that the specified minimum
times are adequate.
No visible improvement in dispersion was
observed by doubling the quantity of disper-
sant used to treat the samples or by increasing
the concentration of dispersant from 40g/l
to 60 g/l. This was not unexpected, since
the UFS samples used 50 g/l and showed no
visibly significant improvement in dispersion
from the GR3 samples.
Samples taken from suspensions permit-
ting little time for settlement, allowed assess-
ment of the dispersion of clay particles from
sand-size particles. Dispersion appeared to
be no better than from silt, as can be seen in
Photograph 18.
With particles as large as this, the depth
of the suspension between slide and cover
slip is so great that very little of the sus-
pended material is in focus. It appears that
treatment with dispersant and subsequent
mechanical agitation had failed to dislodge
clay particles from the sand grain.
DISCUSSION
All of the clays tested showed some lack of
dispersion. Every sample showed instances
Photograph 15 Sample of soil from the Northern Cape after soaking in dispersant and
mechanical agitation as specified by SANS 3001 GR3
This sample was taken a few metres from where a house became structurally unsound
due to heave and was demolished before the roof was installed. Soil apparently identical
to this from a nearby test pit was assessed by hydrometer as containing only 6% clay.
Many apparently clean silt particles can be seen, as can many dispersed particles smaller
than 2 microns. The larger particles are all covered with clay-size particles.
Photograph 16 Same sample as Photograph 15 after treatment with dispersant and mechanical
stirring
A large number of clay-size particles form an agglomeration around a thin, thread-like
structure – possibly a root-hair – while towards the top right many similar size particles
appear to be well dispersed.

Journal o f the South Africa n Institution of Ci vil Engineerin g • Volum e 58 Number 2 June 2016 23
of clay remaining attached to larger parti-
cles or forming agglomerations with other
clayparticles.
In some cases the lack of dispersion was
fairly small, but in some cases a substantial
fraction of the clay particles appeared to be
undispersed. This suggests that it will not be
reasonable to look for some universal factor
by which hydrometer results could be cor-
rected. It appears that predictions based on
clay fraction determined by the procedure
of SANS 3001 GR3 may be very unreli-
able for some soils. Since the SANS 3001
procedure is quite similar to that of many
other countries, it is likely that this problem
may be widespread. The samples prepared at
the UFS Soil Science Laboratory, using soil
science procedures with some differences
to those of SANS 3001 GR3, showed visibly
similar results to those prepared at the CUT
laboratory using the GR3 procedure.
It might be reasonable to consider specify-
ing different de-flocculants for different types
of soil. Rodriguez et al (2011) found that lith-
ium hydroxide is very efficient for high-CEC
soils, but is not effective for dispersing low-
CEC electropositive soils. Rolfe et al (19 60)
found considerable difference in the clay
yield given by a number of dispersants across
different types of clay in hydrometer tests.
Perhaps it is not surprising that the single
dispersant specified for all South African soils
appears to be reasonably adequate for some
soils and completely inadequate for others.
Changing dispersant may be futile, however,
since most of the soils tested showed mixtures
of clay ranging from small, high CEC particles
(much of it probably montmorillonite) to
large, low CEC particles (much of it probably
kaolinite). If a dispersant is not efficient for
several t ypes of clay, it will not give reliable
results for thesesoils.
The methods of quantitatively assessing
the efficiency of de-flocculants for geotech-
nical and soil-science purposes (Rodriguez
et al 2011; Rolfe et al 1960) take hydrometer
yield as the standard of comparison. There
appears to have been no attempt to assess
how much of the clay remains undispersed.
The use of even the most efficient dispersant
for any particular clay may therefore give
poor results.
There is also the question of mechanical
agitation. Rodriguez et al (2011) noted that
horizontal mechanical shaking in helicoidal
motion, with the addition of coarse sand as
an abrasive, is more effective for dispersion
than the conventional method. They did
note, however, that its efficiency is not the
same for all soils.
CONCLUSION
It appears that Savage’s suspicion that “de-
flocculation of many clays is seldom fully
completed at the time of testing” is well
founded. None of the clays tested reached
good dispersion, even when all aspects of
the dispersion procedure were extended
substantially. Clay coating of large (silt/sand)
particles was obser ved to some extent in all
samples which contained silt and sand par-
ticles – in some cases to a very considerable
extent. Such particles will probably settle
at the rate of silt/sand particles and their
clay coating will not be assessed with the
clayfraction.
It may be advisable to consider the
hydrometer unreliable for any critical
Photograph 17 Western Cape yellow kaolinitic soil after addition of methylene blue following
treatment with dispersant and mechanical stirring to the specification of
SANS3001GR3
Almost all of the particles are about 2 microns or slightly larger. Many form groups of
from a few to many particles. Many appear to be properly dispersed, but many are not.
Photograph 18 Dark-brown sandy clay from a road project at Thaba Nchu in the central Free
State after preparation for hydrometer analysis by soaking in dispersant and
mechanical agitation according to SANS 3001 GR3
This sand grain is of the order of 200 microns long and 100 microns wide. It appears
to be completely covered with clay-size particles. Many clay particles around it are
dispersed, although most are out of focus and cannot be clearly seen.

Journal o f the South Africa n Institution of Ci vil Engineerin g • Volum e 58 Number 2 June 201624
analysis such as heave prediction. Work
in progress at the CUT soil mechanics
research group is attempting to find more
reliable methods of assessing clay frac-
tion, but this is at a very early stage and is
therefore unlikely to be able to give reliable
quantitative results for at least two years.
The quest for better methods of assessing
clay content should perhaps become a
priority on a wider scale.
ACKNOWLEDGEMENTS
The authors wish to express their thanks
and appreciation to Prof L van Rensburg and
MrsYDessels of the Free State University,
and Prof SW Jacobsz of the University of
Pretoria, for their assistance and encour-
agement, as well as the National Research
Foundation (NRF) for its financial sup-
port of the CUT soil mechanics research
group’swork.
REFERENCES
Al-Lam i, H S 2008. Evaluat ion of some water-soluble
dispersants for d ifferent types of kaolin powders.
Al-Taqani, 21: 95–100. Available at: ht tp://www.ia sj.
net.
Bühma nn, C & Kirsten W F A 1991. The mineralogy of
five weathering prof iles, de veloped from Archaea n
gran ite in the ea stern Tran svaal, Republic of Sout h
Africa. South African Journal of Plant and Soil,
3:126–152. DOI: 10.1080/02571862.1991.10634823.
Das, B M 20 08. Advanced Soil Me chanics, 3rd ed. New
York: Taylor and Francis.
Konert, M & Vandenberghe, J 1997. Comparison of laser
grai n size analysis w ith pipet te and sieve a nalysis: A
solution for t he underest imation of the clay fraction.
Sedimentology, 44(3): 523–535.
Ma, Z, Merk us, H G, de Smet, J G A E, Heffels, C &
Scarlett, B 2000. New developments in particle
charac terization by laser diff raction: Size and shape.
Powder Technology, 111(1–2): 66–78.
McCave, I N, Br yant, R J, C ook, H F & Coughanowr,
C A 1986. Eva luation of a la ser-dif fract ion-size
analyser for use w ith natu ral sed iments. Journal of
Sedimentary Petrolog y, 56(4): 561–564.
Mishra, A K, Rid lo, M O & Higashi, T 2011.
Comparison of the clay fract ion of Aria ke clay
determined by Japanese and ASTM standa rds.
Marine Georesources and Geotechnology, 29(2):
110–116. DOI: 10.1080/1064119X .2012. 525138.
29:110-116, 2 011.
Nettleship, I, Cisko, L & Vallejo, L E 1997. Aggregation
of clay in the hydrometer test. Canadian
Geotechnical Journal, 34: 621–626.
Robinet, J C , Coelho, D & A ltmann, S 2011. State of
the art on c ation dif fusion in clay mineral systems.
Karlsr uhe, Germ any: European Com mission
(Community Resea rch), CatClay Report 1/5/2011.
Rodrig ues, C, de Ol iveira, V A , da Silveira, P M &
Santos, G G 2011. Chemica l dispersants and pre-
treatments to determ ine clay in soils with dif ferent
mineralogy. Journal of the Brazili an Society of Soil
Science, 35(5): 1589–1596. Ava ilable at: http://www.
redalyc.org.
Rolfe, B N, Mi ller, R F & McQueen, I S 1960. Dispersion
characteristics of montmorillonite, kaolinite, and
illite cl ays in water s of varying quality, and their
control with phosph ate dispersants . Washing ton DC:
Geological Sur vey and Colorado State University,
Professional Paper 334-G.
SANS (South A frican National Standard) 2011. SANS
3001: 2011, Edition 1.1. Civil Engineering Test Methods,
Part GR3. Pretoria: SA BS Standards Div ision.
Savage, P F 20 07. Evaluation of possible swelling
potential of soil. Proceedings, 26th South Afr ican
Transpor t Conference (SATC 2007), Pretoria, 9–12
Ju ly 2 007.
Skempton, A W 1953. The colloida l “activity” of clays.
Proceeding s, 3rd International C onference on Soil
Mechan ics, Zur ich, 1, pp 57–60.
Turoz, M & Tosun, H 2011. The use of methylene blue
test for pred ictin g swell pa rameters of natural clay
soils. Scient ific Research and Essays , 8: 178 0–1792.
Van der Merwe, D H 1964. The pre diction of heave
from the plasticit y index and percentage clay
fract ion in soils. The Civil Engineer in South Af rica,
June : 103 –107.
Zschi mmer & Schwa rz GmbH & Co 2014. Mechanisms
of action of deflocculates and di spersants in
ceramic bodies. Ceramics. Available at: htt p://www.
zschimmer-schwarz.com/en/simon/zschimmer-
schwarz/media/site/downloads/fachinfo/eng/
Fach/E_WirkmechVerfluess.pdf.