Engineering Geology 51 (1998) 37–53

Effects of partial substitution of lime with ground granulated blast

furnace slag (GGBS) on the strength properties of lime-stabilised
sulphate-bearing clay soils
S. Wild a,*, J.M. Kinuthia a, G.I. Jones a, D.D. Higgins b
a Centre for Research in the Built Environment, University of Glamorgan, Pontypridd, Mid. Glamorgan CF37 1DL, UK
b Cementitious Slag Makers Association, Surrey, UK
Received 3 September 1997; accepted 26 June 1998

Abstract

Increasing global awareness of environmental pollution as well as increasing waste material disposal legislation is
providing impetus for material upgrading by stabilisation of in situ soil as an alternative to its export to land-fill and
replacement by imported granular fill. The use of ground granulated blast furnace slag (GGBS), an industrial
by-product, is well established as a binder in many cement applications where it provides enhanced durability,
including high resistance to chloride penetration, resistance to sulphate attack and protection against alkali silica
reaction (ASR). This paper reports on efforts to extend the use of GGBS to highway and other foundation layers by
determining the beneficial effect on strength of progressively substituting GGBS for lime in lime-stabilised clay soils,
particularly in the presence of gypsum. This paper describes the results of laboratory tests on lime-stabilised kaolinite
containing different levels of added gypsum and on lime-stabilised gypsum (selenite) bearing Kimmeridge Clay to
which, in both cases, the lime has progressively been substituted with GGBS. The tests determine the strength
development of compacted cylinders, moist cured in a humid environment at 30°C. The results illustrate that
substitution of lime with GGBS in stabilising gypsum-containing clays produces significant improvements in strength
development. The process has important practical applications, and the paper reports on-going pilot field trials aimed
at realising these applications. © 1998 Elsevier Science B.V. All rights reserved.

Keywords: Clay soils; Compressive strength; Highways; Lime; Slag

1. Introduction mechanical properties of the bulk material. These

From the extensive research work reported on
the pozzolanic activity between clay and lime, it is
apparent that the type and form of the reaction
products determine the physical, chemical and
* Corresponding author. Fax: +44 1443 482169;
e-mail: swild@glam.ac.uk
properties will initially depend on the nature of
the immediate cation exchange reaction which
occurs, while the much slower pozzolanic reactions
will determine the ultimate strength. Pozzolanic
reactions occur over long periods (i.e. years) and,
depending on the curing environment, eventually
cease altogether. Thus, the strength enhancing
reactions will also cease and deterioration may

0013-7952/98/$ – see front matter © 1998 Elsevier Science B.V. All rights reserved.
PII: S0 0 1 3 -7 9 5 2 ( 9 8 ) 0 0 03 9 - 8
38 S. Wild et al. / Engineering Geology 51 (1998) 37–53
even occur, depending on the chemical
environment.
In clay–lime systems in general, the cementing
agent is calcium silicate hydrate (C–S–H ) gel
(Croft, 1964; Sloane, 1965; Abdi, 1992; Wild et al.,
1993). The pore solutions of these systems contain
silicate and aluminate ion species which are pro-
duced by the dissolution of clay particles in the
highly alkaline environment provided by the dis-
solved lime (i.e. OH− ions in solution). Since
kaolinite is high in alumina, the gel in kaolin-
ite–lime systems also contains alumina. In addi-
tion, crystalline calcium–aluminate hydrate
(C–A–H ) phases (such as C AH and C AH )
3 6 4 13
and calcium–aluminate silicate hydrate
(C–A–S–H ) phases (such as C ASH ) are formed,
2 8
as observed by Croft (1964). In the absence of
excess alumina, only C–S–H phases would nor-
mally form, as in the case of Portland Cement
(PC ) and silica fume ( Wild et al., 1995).
Microstructural evidence of similar gels in PC and
PC–GGBS blends suggests that the gel forms on
the surface of the solid particles (Richardson
et al., 1994).
Introduction of slag into the clay–lime hydration
reaction will undoubtedly modify the clay–lime
reaction products. Currently, an indication of the
manner in which these products may be modified
by slag can only be obtained from previous findings
on activated slag systems, as virtually no previous
work has been carried out on clay–lime–slag sys-
tems. The slag clearly provides additional alumina,
calcia, silica and magnesia to the system, depend-
ing on the type and amount of slag (Regourd,
1980; Smolczyk, 1980). Since the principal reac-
tants introduced by slag are also present in the
clay–lime system as well as in PC–GGBS blends,
the strength of hydrated clay–lime–GGBS mix-
tures is likely to be governed by the same factors
observed in hydrated PC–GGBS blends. These
factors include the properties of the C–S–H gel,
such as its amount, porosity and permeability;
fineness of all reactants involved; efficiency of
mixing, temperature and curing time. Also, since
the slag introduces extra and more freely available
alumina and silica compared to the clay–lime
system alone (due to its high reactivity in the
presence of lime), the formation of the strength-
contributing silicates and aluminates is enhanced.
The lime in the clay–lime mix will provide the
required alkaline environment for slag activation
and hydration, whilst also producing modification
of the clay.
In the presence of sulphates in the soil, expansive
reactions are likely to occur. Researchers have
observed that the presence of SO2− ions in either
4
PC, GGBS or PC–GGBS blends results in the
formation of a potentially expansive system mainly
through the formation and subsequent growth of
ettringite (Mehta, 1973; Regourd, 1980). Similar
expansive behaviour has been reported in case
studies involving clay–lime–sulphate systems
(Mitchell, 1986; Hunter, 1988; Snedker and
Temporal, 1990).
Wild et al. (1993, 1996) have worked on clay–
lime systems with and without the presence of slag
and/or gypsum. In lime-treated kaolinite clay in
the presence of gypsum, they observed nucleation
and growth of ettringite crystals on the surface of
clay particles. They also observed that clay–lime–
slag–gypsum systems can be expansive [as the
principal reactants — Ca(OH ) , SiO , Al O ,
2 2 2 3
SO2− and H O — are the same as those in PC
4 2
and PC–slag blends), but more importantly the
presence of slag in clay–lime–gypsum systems
lowers the deleterious expansions normally associ-
ated with ettringite formation. The present paper
reports on the manner in which strength develop-
ment is modified when the lime in stabilised kaolin-
ite–lime mixes is partially replaced by GGBS, in a
system which contains gypsum levels of 0, 2, 4 and
6% by weight. It also reports on the effects of
partially replacing lime with GGBS on the strength
development of lime-stabilised Kimmeridge Clay,
the latter being a naturally occurring sulphate-
bearing clay soil of Jurassic age.
2. Materials
2.1. Kaolinite
Industrial kaolinite was used as a source of
kaolinite. It was supplied by ECC International
Ltd, St. Austell, Cornwall, under the commercial
trade name ‘‘Standard Porcelain’’, as a white
 S. Wild et al. / Engineering Geology 51 (1998) 37–53 39

Table 1 Table 3
Particle size distribution of ‘‘Standard Porcelain’’ ( ECC Engineering properties of ‘‘Standard Porcelain’’
International Ltd, 1987)
Consistency limits
Sieve size %
Liquid limit (%) 61
>53 mm 0.03 Plastic limit (%) 32
>10 mm 4 Plasticity index 29
<2 mm 70
Others

Table 2 Specific gravity 2.57

Chemical analysis and mineralogical composition of ‘‘Standard Maximum dry density (MDD) (Mg/m3) 1.50
Porcelain’’ ( ECC International Ltd, 1987) (Standard BS compaction)
Optimum moisture content (%) 26
Chemical analysis % pH 5.9
Natural moisture content (%) 2.5
SiO 48
2
Al O 37
2 3
Fe O 0.65 Table 4
2 3
TiO 0.02 Sulphate content of Kimmeridge Clay
2
CaO 0.07
MgO 0.30 Total sulphur 0.71%
KO 1.60 Sulphide (S2) 0.02%
2
Na O 0.10 Water soluble sulphate (SO ) 1.41%
2 3
Loss on ignition (LOI ) 12.5 Total sulphate (SO ) 1.73%a
3
Mineralogical composition a The total sulphate of 1.73% SO translates to 3.72% by weight
3
of gypsum (CaSO · 2H O).
4 2
Kaolinite 84
Micaceous material 13
Feldspar 1
other as a yellow grey (or brown) silty clay. The
Other minerals 2 material was sampled 1–3 m below ground level.
Borehole logs from this site report Kimmeridge
Clay as occurring at depths between 0.5 and 9 m
powder consisting of 84% kaolinite, 13% mica, 1% below ground level. The two types of clay men-
feldspar and 2% other minerals. Table 1 shows the tioned above have been identified as the same
particle size distribution of ‘‘Standard Porcelain’’, material, the change in colour being thought to
while Table 2 shows the chemical and mineralogi- result from localised weathering and oxidation
cal composition of the ‘‘as-supplied’’ material from (Littleton and Willavise, 1992). The clay also
ECC International Ltd. Table 3 shows the engi- contains pyrite [iron sulphide (FeS )] and fine
2
neering properties of Standard Porcelain. crystals of selenite (gypsum). The results of sul-
phate analysis, engineering properties and mineral
2.2. Kimmeridge Clay identification of the Kimmeridge Clay, which had
first been dried and crushed and fully homog-
Kimmeridge Clay was chosen for this study for enised, were carried out by Frodingham Cement
its proximity and similarity to Oxford Clay, which Company Ltd ( UK ) and are shown in Tables 4–6
was used in the construction of the M40 respectively. The presence of dickite, a type of
Oxford–Birmingham Motorway. It was obtained kaolinite, is significant as it makes the comparison
with the assistance and permission of Oxford City of test results with those of the model industrial
Council from a site near Oxford (an area of kaolinite (which also contains mica as a minor
farmland under development south east of phase) relevant. Thus, a study using kaolinite for
Oxford ). It occurred in two intermixed forms, one example might be expected to reflect the likely
as a firm to stiff dark (or blue) grey clay and the reactions in the Kimmeridge Clay and vice versa.
40 S. Wild et al. / Engineering Geology 51 (1998) 37–53

Table 5 Table 7
Engineering properties of Kimmeridge Clay Physical properties of lime (Buxton Lime Industries Ltd )

Consistency limits Trade (or product) ‘‘Limbux’’ hydrated lime

name
Liquid limit (%) 65 Chemical name Hydrated lime or calcium hydroxide
Plastic limit (%) 33 Physical form Dry white powder
Plasticity index 32 Melting/decomposition 580°C
temperature
Others Bulk density 480 kg/m3
Specific gravity 2.3 (water=1 at 4°C )
Maximum dry density (MDD) (Mg/m3) 1.57 Specific surface 300–1500 m2/kg
(Standard BS compaction) Solubility in water 1.76 g/l sat. sol. at 10°C
Optimum moisture content (%) 22 Solubility in Sugar solution
pH 9.6 other solvents
pH 12.4 (aqueous solution approx. 2 g/l )
Vapour pressure 0 at 20°C

Table 6 Table 8
Mineral phases in Kimmeridge Clay Chemical composition of lime (Buxton Lime Industries Ltd )
Mineral Chemical Compound Chemical Composition
formula formula (%)
Quartz SiO
2 Main (and hazardous) ingredient
Calcite CaCO
3 Hydrated lime Ca(OH ) 96.79
Ankerite Ca(Fe,Mg) (CO ) 2
32 (calcium hydroxide)
Dolomite CaMg(CO )
32
Gypsum (selenite CaSO · 2H O
4 2 Others
crystals) Calcite (calcium carbonate) CaCO 1.36
Muscovite (mica) ( K,Na)(Al,Mg,Fe) (Si Al )O (OH ) 3
2 3.1 0.9 10 2 Anhydrite (calcium sulphate) CaSO
4
0.06
Illite (mica) ( K,H O)Al Si AlO (OH )
3 2 3 10 2 Magnesia MgO 0.83
Dickite (kaolin) Al Si O (OH )
2 2 5 4 Ferric oxide Fe O
2 3
0.06
Anatase +Probable trace amount TiO
2 Alumina Al O
2 3
0.10
Silica SiO 0.46
2
Excess moisture HO 0.34
2

2.3. Lime Table 9

Particle size distribution of lime (Buxton Lime Industries Ltd )
Hydrated lime under the trade name ‘‘Limbux’’
was supplied by Buxton Lime Industries Ltd, Size (mm) % Passing
Buxton, Derbyshire; Tables 7 and 8 show its physi- 500 100.00
cal properties and chemical composition respec- 355 99.99
tively, while Table 9 shows its grading. 250 99.98
180 99.93
125 99.74
2.4. Slag 90 99.59
63 94.71
Ground granulated blast furnace slag (GGBS )
was supplied by Civil and Marine Slag Cement
Ltd, Llanwern, Newport; Table 10 shows its chemi- 3. Mix compositions and specimen preparation
cal composition and physical properties together
with a typical composition for Portland Cement For the purpose of sample preparation, it was
for comparison. found necessary to establish target dry density and
 S. Wild et al. / Engineering Geology 51 (1998) 37–53 41

Table 10 sions 50 mm diameter and 100 mm length, were

Chemical composition and physical properties of GGBS and thoroughly mixed in a variable speed Kenwood
Portland Cement (Civil and Marine Slag Cement Ltd )
Chef mixer for 2 min before slowly adding the
Oxide Composition (%) calculated amount of water. Intermittent hand
mixing with palette knives was necessary to achieve
GGBS Portland Cement a homogeneous mix. The material was compressed
CaO 41.99 63 immediately after mixing into cylinders to the
SiO 35.34 20 prescribed dry density and moisture content. A
2
Al O
2 3
11.59 6 steel mould fitted with a collar, so as to accommo-
MgO 8.04 4.21 date all the material required for one sample, was
Fe O 0.35 3
2 3 used. Compaction, which was carried out immedi-
MnO 0.45 0.03–1.11
S 1.18 — ately after mixing, was achieved using a hydraulic
2
SO 0.23 2.3 jack. The prefabricated mould, which was oiled to
3
Insoluble residue 0.3 0.5 reduce friction, ensured that no further compaction
Relative density 2.9 3.15 was subjected to the soil once the desired volume
Bulk density (kg/m3) 1200 1400
Colour Off-white Grey
had been attained. The compacted cylinder was
Glass content #90 — left in the mould under pressure for a short period
before release, in order to allow for relaxation.
The densities would therefore be similar to those
moisture content values. Proctor compaction tests obtained using the standard Proctor test since the
were conducted in order to establish values of the dry density and moisture content remain constant.
maximum dry density (MDD) and optimum mois- The cylinders were extruded using a steel plunger,
ture (OMC ) for the range of compositions in each trimmed, cleaned of releasing oil, weighed, mea-
system and the mean MDD and OMC values were sured and wrapped in several runs of cling film.
adopted as the target values for the preparation The cylinders were labelled and placed in polythene
of test specimens. Samples were therefore expected, bags before being placed on a platform in sealed
within experimental error, to be of the same den- plastic containers. Water was always maintained
sity, volume and moisture content for all the below the platform to ensure that there was no
material compositions in a given system. In the evaporation from the samples. The plastic contain-
kaolinite–lime–GGBS–gypsum system, a dry den- ers were then placed in an environmental chamber
sity value of 1.41 Mg/m3 and a moisture content capable of maintaining temperatures to ±1°C and
value of 30% were adopted for all mixes (measured humidity to ±2% relative humidity. The samples
MDDs and OMCs varied between 1.40 and were moist cured for 7 and 28 days at 30°C and
1.46 Mg/m3 and 29–31% respectively). Similarly 100% relative humidity. Table 11 shows the details
for the Kimmeridge Clay–lime–GGBS system, a of the mix compositions used.
dry density value of 1.41 Mg/m3 and a moisture
content value of 24% was adopted (measured
MDDs and OMCs varied between 1.36 and 4. Testing and analysis
1.43 Mg/m3 and 21 and 28% respectively).
Although this means that, within each series, speci- At the end of the prescribed moist curing
mens of different composition will deviate slightly periods, samples were removed from the environ-
from their MDD and OMC values, the minor mental chamber. Any condensation on the cling
variation in void space which this will produce film covering the cylinders was removed with paper
would not be expected to have any significant tissue, prior to their being weighed. Two cylinders
effect on physical properties. per mix proportion were subjected to unconfined
Using the adopted mean density and moisture compressive strength ( UCS ) tests and the mean
content values, dry materials, enough to produce strength determined. A third test was sometimes
three compacted cylindrical test samples of dimen- necessary, especially when the strength variation
42 S. Wild et al. / Engineering Geology 51 (1998) 37–53

Table 11
Details of mix compositions

Lime (L) Slag (S) Gypsum (G) Moisture Dry density Mixes selected for
content (wt%) content (wt%) contenta (wt% SO ) content (%) (Mg/m3) use in XRD analysis
3
Kaolinite–lime– 6 0 0 30 1.41 ⻬
slag–gypsum
5 1 0 30 1.41 ⻬
3 3 0 30 1.41 ⻬
2 4 0 30 1.41 ⻬
1 5 0 30 1.41 ⻬
0 6 0 30 1.41 ⻬
6 0 1 30 1.41 ×
5 1 1 30 1.41 ×
3 3 1 30 1.41 ×
2 4 1 30 1.41 ×
1 5 1 30 1.41 ×
0 6 1 30 1.41 ×
6 0 2 30 1.41 ×
5 1 2 30 1.41 ×
3 3 2 30 1.41 ×
2 4 2 30 1.41 ×
1 5 2 30 1.41 ×
0 6 2 30 1.41 ×
6 0 3 30 1.41 ⻬
5 1 3 30 1.41 ⻬
3 3 3 30 1.41 ⻬
2 4 3 30 1.41 ⻬
1 5 3 30 1.41 ⻬
0 6 3 30 1.41 ⻬

Kimmeridge Clay– 5 0 1.73 24 1.41 ×

lime–slaga
3 2 1.73 24 1.41 ×
2 3 1.73 24 1.41 ×
1 4 1.73 24 1.41 ×
0 5 1.73 24 1.41 ×
6 0 1.73 24 1.41 ×
4 2 1.73 24 1.41 ×
2 4 1.73 24 1.41 ×
1 5 1.73 24 1.41 ×
0 6 1.73 24 1.41 ×
8 0 1.73 24 1.41 ×
5 3 1.73 24 1.41 ×
3 5 1.73 24 1.41 ×
2 6 1.73 24 1.41 ×
0 8 1.73 24 1.41 ×
10 0 1.73 24 1.41 ×
6 4 1.73 24 1.41 ×
4 6 1.73 24 1.41 ×
2 8 1.73 24 1.41 ×
0 10 1.73 24 1.41 ×

a It should be noted that the gypsum in the kaolinite–lime–slag system was added, while in Kimmeridge Clay it is inherent in the
clay as selenite crystals.
 S. Wild et al. / Engineering Geology 51 (1998) 37–53
in the first two tests differed by more than 10% of
the mean strength (normally a variation of ±10%
of the mean was observed ). Before testing, the end
surfaces of samples to be tested were gently
abraded to ensure a flat surface and good contact
with the testing rig platens. A special self-levelling
device was used to ensure uniaxial load applica-
tion. An M30K JJ Instruments testing machine
capable of loading up to 30 kN was used to apply
the load at a compression rate of 1 mm/min. After
testing, a small quantity of material was taken
from the interior of the tested sample for moisture
content determination to establish the moisture
condition at the end of each curing period. This
was found to be within ±5% of the initial moisture
content of the samples, and therefore any loss or
gain of moisture by the samples during curing was
negligible (i.e. a moisture content of 30% would
normally vary within 30±1.5%).
X-ray diffraction ( XRD) analysis was per-
formed on selected mixes from the kaolinite–
lime–GGBS–gypsum system. The fewer mineral
phases in the industrial kaolinite used compared
to those in the Kimmeridge Clay (see Section 2)
made identification of new products simpler and
was thus more effective in contributing to a clearer
understanding of the mechanisms involved in the
interaction between the clay, lime, slag and
gypsum. The selected mixes are indicated in
Table 11. All X-ray diffraction runs were per-
formed using fixed settings for the X-ray equip-
ment and a standard amount of powder specimen.
The work was carried out at the Department
of Mechanical, Materials and Manufacturing
Engineering, University of Newcastle, Newcastle
upon Tyne, using a Philips diffractometer PW1965
and generator PW1730, a graphite monochroma-
tor and Cu Ka radiation of wavelength l=
1.54179 Å.
In addition to strength determination and X-ray
analysis, linear expansion measurements were also
carried out on these specimens. The dimensional
changes exhibited by the specimens during 7 days
of moist curing, which are the curing conditions
prior to strength testing, were found to be relatively
small (+2.1% to −0.55%). Also, specimens which
exhibited the greatest swelling were those of high
lime–low slag contents, which were also those
43
observed subsequently to contain the greatest
amount of ettringite. In fact, specimens with low
lime–high slag contents tended to shrink during
moist curing rather than expand. Swelling of these
specimens during both moist curing and saturation
is addressed in a further paper ( Wild et al., in
press).
5. Results
5.1. Kaolinite system
Fig. 1(a–c) illustrates the effects (on the 7 and
28 day UCS ) of progressive substitution of lime
with GGBS in kaolinite–6% by weight lime (con-
trol ) samples with and without added gypsum.
The levels of gypsum used (2, 4 and 6% by weight,
based upon the clay and stabiliser combined) are
equivalent to approximately 1, 2 and 3% by mass
of SO .
3
When no gypsum is present, for the 7 day curing
period [Fig. 1(a–c)] partial substitution of lime
with slag has very little effect on strength, produc-
ing only a slight increase in strength with increasing
substitution. However, after 28 days the level of
substitution has a significant effect on strength,
producing an accelerating increase in strength with
increasing substitution of lime with slag until zero
lime is reached, when strength falls sharply. This
indicates that by 28 days the hydration of the slag
is at a much more advanced stage with regard to
its contribution to strength than is the pozzolanic
reaction between the lime and the kaolinite.
When gypsum is present, for the 7 day curing
period there is increasing strength enhancement,
relative to the sulphate free material, with increas-
ing substitution of lime with slag. This suggests
that the presence of gypsum not only accelerates
the cementitious reactions with lime and kaolinite,
but also even more effectively accelerates the
hydration of the slag. After 28 days curing, there
is further strength enhancement by the gypsum,
but in this case the strength enhancement is much
greater at low substitution levels and the enhance-
ment becomes less as the lime is substituted with
slag. Thus, at a 5:1 slag to lime ratio there is
virtually no difference in the 28 day strengths of
44 S. Wild et al. / Engineering Geology 51 (1998) 37–53
Fig. 1. Unconfined compressive strength vs. slag ( lime) content
for kaolinite–lime–GGBS cylinders with and without: (a) 1; (b)
2; (c) 3% SO equivalent of gypsum and moist cured for 7 and
3
28 days at 30°C and 100% relative humidity.
specimens with gypsum and without gypsum, irre-
spective of the SO level. This indicates that the
3
accelerating effect of the gypsum on the slag hydra-
tion is complete. However, at 0:6 slag to lime ratio
there is, at 28 days, substantial strength enhance-
ment when gypsum is present, particularly at the
higher SO levels, which suggests that the effect of
3
gypsum on the cementitious reaction between lime
and kaolinite is more pronounced at this later
stage. It should also be noted that, if only slag is
present and no lime (6:0), then strengths are
extremely low, there is no strength development,
and gypsum has no influence whatsoever on
strength (point B on graphs in Fig. 1). This sug-
gests that gypsum will only accelerate slag hydra-
tion if the slag is activated with lime, and will not
activate the slag on its own.
Fig. 2 presents the same 7 and 28 day strength
data, but in this case plotted against gypsum
content (as per cent SO ) rather than slag ( lime)
3
content. This again emphasises the increasing
strength enhancing effect of the sulphate additions
with increasing substitution of lime with slag, again
excepting 0% lime. In addition, it shows that there
is an optimum sulphate content of 1% SO for
3
maximum 7 day strength, which is principally a
result of acceleration of the slag hydration reac-
tions. For the composition 6L0S, sulphate content
has little beneficial effect on strength at 7 days.
After 28 days, however, sulphate content has a
very marked effect on the strength of this composi-
tion, which increases considerably in going from
0% SO to 2% SO , emphasising as previously
3 3
suggested that the reaction product resulting from
the clay–lime–gypsum reaction contributes sig-
nificantly to strength at these later periods.
Conversely, specimens containing a small amount
of lime and a large amount of slag (1L5S ) show
negligible variation in strength with increase in
sulphate level (at 28 days) simply because the
initial accelerating effect of the sulphate on slag
hydration is complete and the hydration level of
specimens without sulphate has reached that of
specimens with sulphate.
The way in which different levels of sulphate
influence the rate of strength development is clearly
illustrated in Fig. 3(a–d). When no sulphate is
present [Fig. 3(a)], strength development of high
lime–low slag material (e.g. 6L0S) is very limited,
emphasising the very slow pozzolanic clay–lime
reaction. In contrast, strength development of the
low lime–high slag material (e.g. 1L5S) is substan-
tial, thus confirming the much more rapid activated
 S. Wild et al. / Engineering Geology 51 (1998) 37–53 45

Fig. 2. Unconfined compressive strength vs. gypsum content (%SO ) for kaolinite–lime–GGBS cylinders with additions of varying
3
amounts of gypsum, moist cured for 7 and 28 days at 30°C and 100% relative humidity.

slag hydration reaction. When sulphate is intro-
duced [see Fig. 3(b)], the 7 day strengths of low
lime–high slag compositions increase substantially,
whereas those of high lime–low slag compositions
show much smaller increases, thus emphasising the
much greater accelerating effect of sulphate on
activated slag hydration compared with its influ-
ence on the clay–lime reaction. The effect of sul-
phate on the activity of kaolinite–lime only mixes
is much more pronounced after 28 days [see
Fig. 3(c)]. Here the strength increase between 7
and 28 days for the high lime–low slag composi-
tions is substantial, particularly when compared
to that where no sulphate is present [Fig. 3(a)].
This strength increase is attributed to formation
of a cementitious reaction product which is not
formed in the absence of sulphate. It is suggested
that this product is ettringite.
X-ray analysis was used principally for identifi-
cation of ettringite, lime and gypsum phases for
46 S. Wild et al. / Engineering Geology 51 (1998) 37–53

Fig. 3. Unconfined compressive strength vs. curing time for kaolinite–lime–GGBS cylinders with additions of varying amounts of
gypsum, moist cured for 7 and 28 days at 30°C and 100% relative humidity.

specimens cured at 1 and 4 weeks. The intensities
of these peaks relative to the kaolinite peaks give
an indication of the amounts present, although
this could be distorted by preferred orientation.
After 1 and 4 weeks of moist curing of kaolinite–
6% by weight lime mixes, only X-ray diffraction
peaks due to mica, kaolinite and lime were appar-
ent [Fig. 6(a,b)], which indicates that either little
reaction had occurred or that the products of
reaction were amorphous. The latter is undoubt-
edly the case, because cementation does indeed
occur and the C–A–S–H gel formed due to reaction
of lime with kaolinite is not crystalline (Croft,
1964; Sloane, 1965; Wild et al., 1993, 1996). The
 S. Wild et al. / Engineering Geology 51 (1998) 37–53 47

crystalline products of the reaction which have
been reported develop, if at all, at a very slow rate
(Croft, 1964; McCallister and Petry, 1992). When
gypsum was present, the lime and gypsum peaks
present after 1 week of moist curing were not
present after 4 weeks of moist curing [Fig. 6(c,d )],
and new peaks identified as those of ettringite were
also visible after 4 weeks of moist curing. These
ettringite peaks were present for mixes with little
or no substitution of lime with GGBS (i.e. mixes
of composition 6L0S06G and 5L1S6G). At higher
lime substitution (i.e. mixes of composition
2L4S6G), no ettringite peaks were identifiable and
peaks of residual gypsum appeared [Fig. 6(f )].
5.2. Kimmeridge Clay system
Figs. 4 and 5 illustrate the effects (on the 7 and
28 day UCS ) of progressive substitution of lime
with GGBS in Kimmeridge Clay–lime cylinders at
various magnitudes of total stabiliser (TS ) content
(i.e. lime+GGBS=5, 6, 8 and 10%). In contrast
to the kaolinite–lime–gypsum system, where the
SO level was artificially varied, in Kimmeridge
3
Clay the sulphate content is fixed at its as-sampled
level of 1.73% SO . Thus, for purposes of compari-
3
son with the kaolinite–lime–gypsum system,
Fig. 1(b) is the nearest equivalent (2% SO ). Also
3
in the case of the Kimmeridge Clay, the total
stabiliser content was varied from 5 to 10% in
order to establish, from a practical engineering
aspect, what an appropriate level of stabiliser
would be. There are, with regard to strength
development, clear similarities in behaviour
between the two clay systems [swelling of these
specimens is addressed in a further paper ( Wild
et al., in press)]. Comparison of Fig. 4(b) with
Fig. 1(b) shows parallel behaviour in the relation-
ship between 7 day and 28 day strength, in that at
low slag to lime ratios (6L0S) the 28 day strength

Fig. 4. Unconfined compressive strength vs. slag ( lime) content for Kimmeridge Clay–lime–GGBS cylinders containing a total
stabiliser ( TS) content ( lime+GGBS ) of: (a) 5; (b) 6; (c) 8; (d) 10% by weight, and moist cured for 7 and 28 days at 30°C and
100% relative humidity.
48 S. Wild et al. / Engineering Geology 51 (1998) 37–53
Fig. 5. Unconfined compressive strength vs. slag ( lime) content
for Kimmeridge Clay–lime–GGBS cylinders moist cured for 7
and 28 days at 30°C and 100% relative humidity [ TS=total
stabiliser ( lime+GGBS) content].
is much greater than the 7 day strength and at
high slag to lime ratios (1L5S) the 7 day and 28
day strengths move closer to each other. Also, in
both cases the 7 day strength initially increases as
the slag to lime ratio increases, reaches a maximum
and then decreases, although the maximum occurs
earlier with Kimmeridge Clay. The 28 day strength
developed at high slag–low lime compositions is
attributed principally to the hydration of the slag
which is activated by lime and initially accelerated
by the presence of sulphate.
At both 7 days and 28 days of moist curing the
initial gradients of the strength versus binder com-
position curves [Fig. 4(a–d)] increase systemati-
cally with increase in total binder content. Also
the peak in strength versus composition curves
tends to move to higher slag/lime ratios as the
binder content increases. This suggests that at low
binder contents and high slag/lime ratios there is
insufficient total lime available in the system to
fully activate the slag. Fig. 5(a,b) suggests an
optimum slag/lime ratio of about 1.5 is required
to give maximum strength for the higher stabiliser
contents.
Another factor observed with Kimmeridge Clay,
not present with kaolinite, is that compositions
containing slag but no lime do show small strength
gains with time (points B and B∞ on the curves in
Fig. 4). This suggests that the Kimmeridge Clay
contains some component(s) which can produce
partial and limited activation of the slag. However,
this was not observed to occur in the kaolinite–
lime–gypsum system, which suggests that the com-
ponent producing this effect is something other
than gypsum alone.
6. Discussion
Enhanced strength development in clay–lime
and clay–lime–slag mixtures in the presence of
gypsum has been confirmed by many researchers
( Hasaba et al., 1982; Gupta and Seehra, 1989;
Abdi, 1992; Wild et al., 1993, 1996). In the current
work, it was noted that the rate of strength devel-
opment was generally higher when gypsum was
present. Inclusion of gypsum in the clay–lime
system accelerates the strength development by a
second mechanism in addition to C–S–H gel for-
mation, that is ettringite formation. This mode of
strength enhancement results from the growth and
development of crystalline ettringite [Fig. 6(d)].
The enhanced strength development, particularly
during the 7 to 28 day moist curing period, con-
firms the well-established accelerating effect that
gypsum has on the clay–lime reactions (Abdi,
1992; Wild et al., 1993, 1996), particularly at later
stages. The acceleration is known to be due to
 S. Wild et al. / Engineering Geology 51 (1998) 37–53 49

Fig. 6. X-ray diffraction traces of: (a,b) moist cured kaolinite–lime specimens without gypsum at 1 week and 4 weeks of moist curing
respectively; (c–f ) moist cured kaolinite–lime–GGBS specimens with 6% by weight of gypsum, at either 1 week and/or 4 weeks of
moist curing. (Note: Only peaks due to individual compounds are marked.)

rapid removal of aluminium from solution, by
formation of ettringite.
When slag is present in clay–lime mixes, a major
factor determining the degree of formation of the
cementitious hydration products is the hydration
properties of the slag. The activation of slag by
lime (and also in some cases by components in the
clay soil ) promotes slag hydration and hence also
the formation of the cementitious calcium silicate,
aluminate and alumino-silicate hydrates.
In the presence of gypsum, the sulphate further
promotes activation of the slag through lime–gyp-
sum activation (Daimon, 1980; Smolczyk, 1980;
Wu et al., 1990; Gollop and Taylor, 1996). It
accelerates and enhances slag hydration such that
strength is enhanced significantly by 7 days. Added
50 S. Wild et al. / Engineering Geology 51 (1998) 37–53
gypsum activates slag only in the presence of lime
(and/or any other alkali). Inclusion of gypsum in
the clay–lime–GGBS system at high lime–low slag
compositions also accelerates the strength develop-
ment due to the development and growth of
crystalline ettringite. This enhanced strength devel-
opment, particularly at low slag to lime ratios and
high SO contents ( Fig. 3), is particularly apparent
3
after 28 days of moist curing, which suggests that
gypsum’s accelerating effect on the lime–kaolinite
reaction is slower than on the slag hydration
reaction.
From Fig. 2(a,b), the optimal SO concentration
3
for maximum strength development appears to
depend on the curing period. Obviously at pro-
longed curing more sulphate will be consumed,
resulting in an increase in optimum SO concen-
3
tration and strength. However, there will clearly
be a limit to the amount of sulphate which can be
consumed, which will depend on the original lime
content. This limit, for the lime content employed
in the current work (6% by weight for the kaolinite
system), and for 28 days of moist curing, would
appear to be approximately 2% SO .
3
XRD analysis results of clay–lime–GGBS–
gypsum mixes show that the favourable conditions
for crystalline ettringite formation prevailing in
the kaolinite–lime–gypsum system are reduced
significantly by the introduction of slag. In
PC–GGBS systems, it has been reported that the
higher amounts of C–S–H gel that are produced
readily absorb Al O (Gollop and Taylor, 1996),
2 3
and possibly CaO, in a manner that these compo-
nents are not readily available for further reaction.
In the kaolinite–lime–slag–gypsum system at low
lime–high slag ratios where larger amounts of
C–S–H gel will be produced, it might be expected
that less ettringite will be formed as more CaO
and Al O are bound in the C–S–H gel. Thus in
2 3
this case the pozzolanic reactions produce pre-
dominantly non-crystalline products and the
strength enhancement is principally via C–S–H gel
formation, the ettringite decreasing with increasing
slag/lime ratio. Thus, the most significant effect
that replacement of lime with slag has on the final
product is the reduction or complete elimination
of crystalline ettringite. Consequently, the strength
enhancement in clay–lime–slag–gypsum mixes
must result from hydration of the slag and the
formation of more C–S–H gel, and not from
ettringite formation. Therefore, the increase in
strength as lime is substituted with slag is predomi-
nantly due to the increased formation of the
cementitious silicate, aluminate and alumino-sili-
cates (mainly of calcium). This will then lead to
reduced porosity of the hydrated system (Bijen,
1996) as the gel develops within the pore spaces
and closes off the capillary pores. Strength will
increase with increasing slag level as long as suffi-
cient lime is present to enable alkali–slag activa-
tion. Therefore at low binder contents and high
slag/lime ratios, where insufficient lime is available
to fully activate the slag, strength decreases with
increasing slag level. For example for Kimmeridge
Clay, when the total stabiliser ( lime+slag) does
not provide enough lime for both cation exchange
and for pozzolanic activity to take place in addition
to activation of the slag, any strength increase of
clay–lime–slag mixes with increase in slag content
appears to be limited to short curing periods of 7
days [i.e. TS=5% in Figs. 4(a) and 5 (a)]. During
this period, it is suggested that some proportion
of the calcium ions are used for the cation exchange
while at the same time the short-lived alkalinity
only manages to activate the slag to a low level.
Over longer periods (28 days), those samples with
unsubstituted lime (i.e. low slag/lime ratios) will
develop the highest strengths, the others being
gradually denied hydration as the proportion of
lime decreases, as shown in Fig. 5(b).
Unlike in the case of kaolinite, there is significant
strength development at total lime replacement
with slag [points B and B∞ in Fig. 4(a–d)]. The 28
day strength at total lime replacement generally
increases with increasing stabiliser content,
whereas at 7 days there is little change in strength.
There is no doubt therefore that in this case the
slag is slowly activated. The activation must be
due to one or more of the various mineral phases
in Kimmeridge Clay. Two possible contenders
would be dolomite and calcite, both of which are
sparingly soluble (dolomite less so than calcite)
and generate an alkaline environment in solution
(the pH of Kimmeridge Clay in solution was
observed as 9.6 compared to 5.9 for kaolinite).
Thus, there is the possibility that some of the
 S. Wild et al. / Engineering Geology 51 (1998) 37–53
phases shown in Table 6 help in activating the slag.
More research is definitely required here to estab-
lish the mechanism or the factors involved in this
form of slag activation.
7. Practical applications
Results from the current work indicate that high
(7 day and 28 day) strengths not previously achiev-
able either with clay–lime or with clay–lime–gyp-
sum compositions can been achieved by partially
substituting the lime with ground granulated blast
furnace slag (GGBS ) in the stabilisation process.
Further, results reported by Wild et al. (1996, in
pr) and Higgins et al. (in press) indicate that the
compaction properties of both lime-stabilised
kaolinite and lime-stabilised Kimmeridge Clay are
not significantly affected by the substitution of
lime with GGBS, both in the absence and in the
presence of gypsum, and that both the material
plasticity and expansion potential are significantly
reduced. These observations result in very signifi-
cant and potentially viable commercial and indu-
strial applications, particularly for use in lime
stabilisation of sulphate-bearing clay soils. At pre-
sent, there are no specific guidelines on the use of
sulphate-bearing clay soils, and the most common
practice has been to avoid stabilisation of these
soils, thus incurring heavy economic penalties in
cutting, transporting and spoiling. Environmental
factors, such as increasing global awareness of
pollution, and increasing legislation through taxes
on waste disposal methods, such as dumping,
encourage material upgrading by stabilisation of
in situ soil as an alternative to its export to land-
fill and replacement by imported granular fill.
Thus, the lime stabilisation of sulphate-bearing
soils will result in global material resource savings
and better control of environmental degradation
by waste material dumping.
Curing at 30°C is an accelerated test, as in
practice stabilisation occurs at ambient temper-
atures (normally 10–25°C ). Work carried out by
the authors ( Wild et al., 1998) on the shear
strength of slag/lime-stabilised Kimmeridge Clay
at different temperatures (10–30°C ) has shown
that increase in temperature accelerates the rate of
strength gain. In particular, in this temperature
51
range, the clay–lime reaction is more sensitive to
temperature than the slag hydration reaction,
which further supports the case for using lime-
activated slag stabilisation rather than lime
stabilisation.
Based on these findings, the Cementitious Slag
Makers Association (CSMA), in collaboration
with the Transport Research Laboratory ( TRL)
and other relevant industries such as the Buxton
Lime Industries (BLI ) ( UK ), have carried out a
pioneering highway pavement pilot field trial in
the United Kingdom using GGBS on the A421
Tingewick bypass, west of Buckingham. The trial
uses in situ soil stabilisation of sulphate-bearing
boulder clay using quicklime–Portland Cement–
GGBS optional compositions. The clay–lime–
GGBS option resulted in higher soaked and
unsoaked strength laboratory test results than
clay–lime–PC at equivalent stabiliser contents.
8. Conclusions and recommendations
The following principal points may be deduced
from the results, with regard to the influence of
GGBS and gypsum on kaolinite and Kimmeridge
Clay when stabilised with lime.
(1) Partial substitution of lime with GGBS gives
improved 7 day and 28 day strengths for both
kaolinite and Kimmeridge Clay. The maxi-
mum level of lime substitution is different for
the two clay types. In the case of kaolinite,
the effects are more pronounced in the pres-
ence of gypsum.
(2) The most significant strength enhancement of
kaolinite stabilised with lime/GGBS over the
first 28 days was either for (i) high lime–low
slag mixes with gypsum, due to the contribu-
tion of gypsum to the longer-term kaolinite–
lime–gypsum reaction, or (ii) low lime–high
slag without gypsum due to the lime-activated
slag hydration.
(3) The greatest short-term strength enhancement
was for low lime–high slag mixes with gypsum,
due to the accelerating effect of gypsum on
the lime-activated slag hydration.
(4) Slag alone has no effect on both 7 day and 28
day strength of kaolinite while it does provide
52 S. Wild et al. / Engineering Geology 51 (1998) 37–53
limited but significant strength enhancement
in the case of Kimmeridge Clay.
(5) Substitution of lime by slag in stabilised
Kimmeridge Clay provides a maximum in 28
day strength at a specific replacement level
which depends on the total stabiliser content
(i.e. 2S4L at TS 6% by weight, 5S3L at TS
8% and 6S4L at TS 10%). In the case of
stabilised kaolinite, where only one total stabi-
liser content was investigated, the 28 day
strength continues to increase gradually up to
a replacement level of at least 5S1L, although
at 6S0L the strength falls sharply.
It is clear from the observations in this study
that the amount of lime added, relative to the slag,
must be sufficient to activate the slag. Also, the
total stabiliser content (slag+lime) must be at the
necessary level to provide the required bearing
capacity and strength. Further work indicates that
the lime content must also, however, not be so
great that excess lime is freely available over an
extended period because the lime could react with
sulphates and cause swelling. Any expansive reac-
tions involving sulphates must be exhausted during
the initial curing period prior to any possible
inundation by water. Work on the expansive
behaviour of this material has been reported in a
further paper ( Wild et al., in press).
Acknowledgment
The authors would like to thank Civil and
Marine Slag Cement Ltd for supply of the slag.
The authors would also like to thank Buxton Lime
Industries and ECC International for supplying
respectively the lime and the kaolinite. In addition,
the authors would like to thank Professor Richard
Neale, Head of School of the Built Environment
and the technical staff in the School for provision
of facilities and assistance.
References
Abdi, M.R., 1992. Effects of calcium sulphate on lime stabilised
kaolinite. Ph.D. Thesis, University of Glamorgan,
Pontypridd.
Bijen, J.G., 1996. Blast Furnace Slag Cement, Association of
the Netherlands Cement Industry ( VNC ).
Croft, J.B., 1964. The pozzolanic reactivities of some New
South Wales flyashes and their application to soil stabiliza-
tion. In: Proc. ARRB, Australia, 1964, Vol. 2, Part 2 (Paper
120), pp. 1144–1167.
Daimon, M., 1980. Mechanism and kinetics of slag cement
hydration. In: 7th Int. Congress on the Chemistry of Cement,
Paris, France, pp. III-2/1–III-2/9.
Gollop, R.S., Taylor, H.F.W., 1996. Microstructural and micro-
analytical studies of sulfate attack. V: Comparison of
different slag blends. Cement Concrete Res. 26, 1029–1044.
Gupta, S., Seehra, S.S., 1989. Studies on lime-granulated blastf-
urnace slag as an alternative binder to cement. Highway Res.
Bull. 38, 81–97.
Hasaba, S., Kawamura, M., Torii, K., 1982. Reaction products
and strength characteristics in the stabilised soil using desul-
furization by-product and blastfurnace slag. Trans. Jpn. Soc.
Civ. Eng. 14, 251–253.
Higgins, D., Kinuthia, J.M., Wild, S., in press. Soil stabilisation
using lime-activated ground granulated blast-furnace slag [to
be presented and published in Proc. Sixth CANMET/ACI
Int. Conf. on Fly Ash, Silica Fume, Slag and Natural Pozzo-
lans in Concrete, Bangkok, Thailand, 31 May–5 June 1998].
Hunter, D., 1988. Lime-induced heave in sulphate bearing clay
soils. ASCE J. Geotech. Eng. 114, 150–167.
Littleton, I., Willavise, G., 1992. Some observations on the use
of lime to stabilise Kimmeridge Clay. Buxton Lime Indu-
stries Report, Ref. No. 91.
McCallister, L.D., Petry, T.M., 1992. Leach tests on lime-
treated clays. Geotech. Test. J. 15 (2), 106–114.
Mehta, P.K., 1973. Effect of lime on hydration of pastes con-
taining gypsum and calcium aluminates or calcium sulpho-
aluminate. J. Am. Ceram. Soc. 56, 315–319.
Mitchell, J.K., 1986. Practical problems from surprising soil
behaviour. The twentieth Karl Terzaghi Lecture. ASCE
J. Geotech. Eng. 112, 274–279.
Regourd, M., 1980. Structure and behaviour of slag Portland
cement hydrates. In: 7th Int. Congress on the Chemistry of
Cement, Paris, France, pp. III-2/10–III-2/26.
Richardson, I.G., Brough, A.R., Groves, G.W., Dobson, C.M.,
1994. The characterisation of hardened alkali-activated
blast-furnace slag pastes and the nature of the calcium sili-
cate hydrate (C–S–H ). Cement Concrete Res. 24, 813–829.
Sloane, R.L., 1965. Early reactions in the kaolinite–hydrated
lime–water system. In: Proc. 6th Int. Conf. on Soil Mechan-
ics and Foundation Engineering, Montreal, pp. 121–125.
Smolczyk, H.G., 1980. Slag structure and identification of slags.
In: Proc. 7th Int. Conf. on the Chemistry of Cement, Paris,
France, pp. III-1/3–III-1/17.
Snedker, E.A., Temporal, J., 1990. M40 Motorway Banbury IV
Contract — Lime Stabilisation. Highway and Transporta-
tion, December.
Wild, S., Abdi, M.R., Leng Ward, G., 1993. Sulphate expansion
of lime-stabilised kaolinite. Part II. Reaction products and
expansion. Clay Miner. 28, 569–583.
 S. Wild et al. / Engineering Geology 51 (1998) 37–53
Wild, S., Sabir, B.B., Khatib, J.M., 1995. Factors influencing
development of concrete containing silica fume. Cement
Concrete Res. 25, 1567–1580.
Wild, S., Kinuthia, J.M., Robinson, R.B., Humphreys, I., 1996.
Effects of ground granulated blast-furnace slag (GGBS ) on
the strength and swelling properties of lime stabilised kaolin-
ite in the presence of sulphates. Clay Miner. 31, 423–433.
Wild, S., Tasong, W.A., Veith, G., 1998. Utilisation of waste
materials in the stabilisation of land. Progress Report No. 2
for the Progress Meeting with Industrial Partners, Higher
53
Education Funding Council for Wales (HEFCW ) Technol-
ogy Foresight Initiative Project by the Construction Materi-
als Research Unit, University of Glamorgan, April.
Wild, S., Kinuthia, J.M., Jones, G.I., Higgins, D.D., in press.
Suppression of swelling associated with ettringite formation
in lime-stabilised sulphate bearing clay soils by partial substi-
tution of lime with ground granulated blast-furnace slag
(GGBS). J. Eng. Geol.
Wu, X., Jiang, W., Roy, D.M., 1990. Early activation of slag
cement. Cement Concrete Res. 20, 961–974.