The role of ettringite in the deterioration of artificial lime stabilised soils: a microstructural study

The formation of ettringite has been defined as a major failure mechanism of lime stabilised cohesive soils. It can result in both disruptive volumetric changes and loss of mechanical strength. The mechanisms of its formation and the role it plays in deleterious processes are complex. This paper reports the dimensional and strength changes of a range of artificial lime stabilised cohesive soils subject to two swell test procedures: the UK linear California Bearing Ratio swell test (BS1924-2) and the European accelerated swell test (EN13286-49). The resulting microstructural composition was analysed using a combination of scanning electron microscopy and electron dispersive X-ray spectroscopy. The results are explained in terms of established theories of crystal formation and subsequent expansion mechanisms.


Introduction
Soil stabilisation can be defined as the enhancement of the engineering properties of a soil by blending in a chemical or granular additive (McNally, 1998). Engineering properties improved by the application of soil stabilisation techniques include increases in stiffness, strength and durability. Soil stabilisation allows in situ material to be used as either a temporary or permanent structure in engineering works. Applications include pavement foundations (capping and subbase), temporary working platforms, sea defences and slope stabilisation.
Under UK specifications (specifications for highway works, Highways Agency, 2009), if a pavement foundation is to be constructed on a weak subgrade (California Bearing Ratio (CBR) < 5%), a stabilised capping layer (with improved engineering performance) can be constructed, so that sufficient support is provided to the overlaying pavement. This incorporates the in situ material into the final works and also allows a reduction in the thickness of the overlaying layers (McNally, 1998). Economic and environmental benefits are realised through a reduction in primary material use, fewer lorry movements, avoidance of landfill tax (associated with the disposal of excavated material) and reductions in embodied energies (Britpave, 2010(Britpave, -2011, case studies 1-10).
However, under certain conditions stabilised soils can exhibit the phenomenon of what is often termed sulphate heave. This results from deleterious reactions that occur within the stabilised layer, resulting in disruptive volumetric changes and corresponding losses in strength. A number of often cited pavement failures (Cerato, Miller, & Elwood-Madden, 2011;Hunter, 1988;Kota, Hazlett, & Perrin, 1996;Mitchell, 1986;Perrin, 1992;Puppala, Hanchanloet, Jadeja, & Burkart, 1999;Rollings & Burkes, 1999;Snedker & Temporal, 1990) have been attributed to this deleterious This then enables the dissolved clay particles to react with free calcium ions in solution to form calcium silicate hydrates (CSH), calcium aluminate hydrates (CAH) and calcium aluminium silicate hydrates (CASH), usually described using generic formulas. These are groups of cementitious products that may be final products or metastable intermediates, the co-dependent reactivity of which contributes to the complexity and difficulty in the study of cementitious systems. For example, the hydrated minerals with cementitious properties: C 4 AH 13 , C 3 AH 11 , CAH 10 , C 3 S 2 H 3 and C 2 ASH 8 have all been identified in stabilised soils (Bell, 1996).
These cementitious products are equivalent to those formed during the hydration of cement pastes. The insoluble hydrates form a cementitious matrix which fills void spaces within the soil to encompass unreacted clay particles. Subsequent curing of this cementitious matrix results in material with increased engineering performance. This is illustrated in Figure 1.
So long as the soil remains alkali (pH > 10.4) the dissolution of clay minerals and subsequent pozzolanic reactions will continue. The strength of the soil will then also continue to increase with time for months or even years after mixing. The ultimate effect of lime stabilisation is dependent on the soil composition, lime content, clay mineralogy, curing temperature and soil pH (Mohamed, 2000).
Sulphate heave occurs when sulphate minerals already present in the soil (or transported there by mobile ground water) react with the lime and/or cement used in the stabilisation along with Figure 1. The development of a cementitious matrix in lime stabilised clay soil (adapted from Locat, Bérubé, & Choquette, 1990). alumina from dissolved clay particles to form expansive minerals called ettringite, monosulphate and thaumasite. When these minerals form after the stabilised layer has been mixed and compacted, they exert pressure on the soil, because they have a greater unit volume, than the reactants they were derived from (Little, Nair, & Herbert, 2010). This results in sulphate heave and causes significant damage to the entire pavement structure. This can manifest itself as both transverse and longitudinal ridges and cracks in the pavement surface, as well as discreet areas where the strength loss in the foundation is so severe that significant permanent deformation can occur (Hunter, 1988;Rollings & Burkes, 1999;Snedker & Temporal, 1990). As previously described, the addition of lime to a soil containing clay minerals increases the pH to >10.5 (Equation (2)) prompting the dissolution of aluminosilicates into the pore solution (Equation (3)): The aluminate ions -Al(OH) − 4 -react with sulphate ions from the dissolution of gypsum (Equation (4)) to form ettringite (in this case the trisulphate form) shown in Equation (5): Therefore the anionic sulphate (SO 2− 4 ), alumina (Al(OH) − 4 ) and cationic calcium (Ca 2+ ) ions can be considered the candidate ions (Little et al., 2010;Puppala, Intharasombat, & Vempati, 2005) for the formation of ettringite in cohesive soil and must be present for it to occur. The formation of ettringite during lime stabilisation of cohesive soils is shown in Figure 2.
In the topochemical mechanism, crystal growth occurs at the solid-solution interface. This can occur when the rate of ettringite crystallisation is greater than the dissolution rate of the candidate ions. Considering the almost instantaneous hydration of the cement clinker phase of tricalcium aluminate -C 3 A -by the topochemical mechanism (Kirchheim, Ferna'ndez-Altable, Monteiro, Molin, & Casanova, 2009), the dissolving aluminate ions Al(OH) − 4 cannot migrate far into the pore solution, because the supersaturation of the liquid phase with respect to ettringite is so low. By this mechanism, damage occurs when the regions of crystal growth intersect and mutually exert pressure, forcing the particles apart (Cohen, 1983).
In the through-solution model, on hydration, the solid particles begin to dissolve and their corresponding ions move into aqueous solution. When the pore solution becomes supersaturated with respect to ettringite, then ettringite crystallises out. In this manner, it can form away from the source of alumina, generally in the bulk pore solution and is characterised by nucleation and crystal growth at many sites, frequently in void spaces throughout the material (Min & Tang, 1994). Because of this ability to migrate through the pore solution and crystallise in parts of the material which can accommodate crystal growth, the occurrence of ettringite in this case is not considered deleterious. No damage to the cementitious matrix may necessarily occur, due to the air-entrainment void space being able to accommodate the crystal growth. As such, the through-solution mechanism is considered less deleterious than formation by the topochemical mechanism (Kalousek & Benton, 1970;Soroka, 1979).
In the crystal growth theory, expansion occurs when reaction zones of ettringite formation intersect, continue growing and mutually exert pressure, this can result from crystals growing on the surface of other particles (topochemically) or in the pore solution (through-solution).
In the crystal swelling theory, expansion is caused by the swelling of relatively small ettringite crystals that are colloidal or gel size. Expansive stress results from the conversion of free energy of the reaction to work (Kalousek & Benton, 1970). Ogawa and Roy (1982) observed the formation of ettringite on the surface of C 4 A 3 particles. These grew radially out into the pore solution. Expansion began when these reaction zones intersected and mutually exerted pressure. Schwiete et al. (1966) studied the CaO-Al 2 O 3 -CaSO 4 -CaO-H 2 O system. They concluded that the crystallisation pressure was the result of topochemical AFt formation on the surface of C 3 A particles. Because the AFt crystals occupy a bigger volume than the C 3 A grains, they eventually peel off. It is this on-going process that eventually results in the crystallisation pressure. Nakamura et al. (1968) in their studies of the C4A3S-CH-CS system found that at high [OH − ], AFt formation is topochemical and expansion is caused by the intersection of reaction zones resulting from crystal growth. These crystals were comparatively fine. Conversely when the [OH − ] is low, AFt formed by through-solution mechanism precipitated large crystals in the pore solution.
In the crystal swelling theory, ettringite forms by the through-solution mechanism. The crystals are very small gel-like and colloidal in size (Mehta, 1973). The tiny crystals have a high specific surface area with unsatisfied surface charges. They can absorb ions and water molecules to decrease their surface energy, creating an electric double layer around the ettringite crystals in a similar manner that causes the swelling of particular clays (Little et al., 2010). Intersection of these layers produces the crystal swelling pressure (Min & Tang, 1994). Mehta and Hu (1978) found that an increase in the amount of absorbed water corresponded to an increase in volumetric expansion. Ouhadi and Yong (2008) showed that AFt can swell by the order of 50% and its fluid retention increase by as much as 400% in the soils studied in their research. They concluded that it is the crystal swelling mechanism that is responsible for ettringite-induced heave. Mehta (1973) stated that for colloidal ettringite to cause expansion, it must be in contact with outside water. Quoting Ouhadi and Yong (2008) 'the availability and flow of pore water is the single most important factor controlling lime-induced heave. Without an abundance of water AFt cannot form' and Little et al. (2010) 'the presence of external water is a decisive factor in causing deleterious reactions in stabilised soils'.
The correlation of ettringite formation and expansion of a cementitious material are thought to be extremely difficult. The precipitation of ettringite in the existing pore structure of a material is thought to not contribute to expansion. The inherent variability of soils results in difficulty in charactering and thereby quantifying the degree of expansion. Overall ettringite and monosulphate formation is dependent on many factors. However, the available alumina content and an excess of water are considered critical factors.

Methodology
Clay that had undergone an industrial manufacturing process was selected for a number of reasons; the typical spatial inhomogeneities of natural soils make studying their behaviour under stabilisation difficult. Organic, sulphide/sulphate and other minor constituents, typical of a natural soil, have been removed in the processed material, so that it is of known composition. This allows the physicochemical properties of a particular soil mixture to be attributed to the imposed mixture composition and/or test condition. Again, if a natural soil were used in the research, other effects such as expansion/relaxation (a risk with consolidated clays), oxidation of pyritic deposits and action of autotrophic bacteria mean that changes in the characteristics of natural clay are likely, over time. This would again make control of mixture composition difficult. To make the soil easier to work with and more closely reflect the composition of in situ soils, quartz sand was also incorporated into the soil mixtures.

Materials
The kaolin was supplied by Sibelco, UK, as powdered China clay under the trade name Puraflo S. The montmorillonite as calcium bentonite and quartz sand (Chelford 14/25) was supplied by RS Minerals. The lime used was calcium oxide supplied by URS/Scott Wilson. Compositional analysis by X-ray fluorescence (XRF) is presented in Table 1. The gypsum was supplied by Saint Gobain. It was supplied as crushed gypsum stone direct from the quarrying operation in Lincolnshire. This was prepared by drying at room temperature to remove surface water, then sieved to remove trace impurities of mudstone. The gypsum was passed through a jaw crusher, and then sieved again. The material passing the 425 μm sieve was retained for use in the study. Confirmation that the material was the dihydrate was obtained by thermogravimetric analysis and purity by quantitative X-ray diffraction (QXRD). The purity of the lime was also determined by QXRD.

Classification
Material classification was undertaken in accordance with the current highway agency guidance on soil stabilisation for capping and subbase. The classification tests included in the programme are taken from Table 3/1 soil tests for suitability and design (Highways Agency, 2007aAgency, , 2007b. The soils used in the investigation are defined by the following classes (Figure 3/1 Highways Agency, 2007a): • Class 7E -cohesive material improved with lime for lime stabilisation.
Compositional analysis by XRF of the materials used in the study is presented in Table 1. Classification data and other physicochemical property data are presented in Table 2.

Mixture design
Twelve soil mixtures were used in the study. The lime content is determined by the ICL test + 2.5% in accordance with the recommendations in HA74/07 (Highways Agency, 2007aAgency, , 2007b. The amount of lime required (ICL) was determined in accordance with the procedure set out in BS1924-2 (British Standards Institute [BSI], 1990). The results of which are presented in Table 3 and are rounded to the nearest 0.5%.
The soils used in the study comprised unsulphated kaolin and montmorillonite clay (K and M), sulphated clay (K5S and M5S), lime stabilised clay (K4L and M6L) and sulphated lime    stabilised clay with a low, medium and high sulphated content (K4L0.5S, K4L1.5S, K4L5S and ditto montmorillonite). These are shown in along with other compositional data ( Table 4). The soils were dry blended in a 40 L rotary drum mixer. Deionised water was added to the soils optimum moisture content + 2, left to condition for 1 h, after which the lime was added with further mixing. The soil was then left to condition in a sealed bag for 24 ± 1 h at 20 • C prior to production of the test specimens.

Swell tests
Two swell tests were used in the study. The UK linear CBR swell test (BS1924-2, BSI, 1990) and the European accelerated volumetric swell test (EN13286-49, CEN, 2007). The conditions of each test are presented in Table 5 along with the pass/fail criteria associated with the relevant test. If using BS1924-2 (BSI, 1990), then a minimum average strength after immersion, as well as maximum individual and average linear swell values are defined for a soil to be considered suitable for use. Using EN13286-49 only maximum volumetric swell values are taken. The average volumetric expansion (G v ) is calculated using Equation (6): where G v is the volumetric swell, V 0 the initial volume, V 1 the volume after immersion and V 2 the volume of the confinement wrapping.

Analytical techniques
To investigate the microstructure of the soils subject to the two swell tests, a combination of scanning electron microscopy (SEM) and energy dispersive X-ray spectroscopy (EDX) was used: • SEM: samples were coated with platinum using a Polaron SC7640 sputter coater. Analysis was undertaken using a Philips FEI XL30 SEM under high vacuum at a pressure of about 3.5 × 10 −6 mbar, using an accelerating voltage of 5-20 kV, working distance of 10-20 mm and a spot size of 2-5. The image was composed from the detection of secondary electron emissions. Micrographs were analysed using ImageJ software (Rasband, 1997(Rasband, -2012. • EDX: undertaken to determine the elemental composition using point analysis. This was achieved using an accelerating voltage of 20 kV, a working distance of 15 mm and a spot size of approximately 5. The INCA EDX software programme [Oxford Instruments], was used for data and image processing. Samples were taken from the top 5 mm of the CBR moulds after they had undergone testing in accordance with the UK linear swell test procedure and from the side of the European accelerated swell test specimens using a palette knife. This material was crumbled into glass dishes. These were then freeze-dried, by first freezing at −80 • C overnight, then vacuum drying. This procedure halted the hydration reactions thus preserving the phase assemblage and what original fine microstructure remained (petrographic analysis is difficult to perform on low strength materials due to the cutting and polishing required to produce a surface suitable for analysis). Samples for SEM and EDX were reserved as is.

Physical property testing results
The following section reports results of the physical property testing and results of the kaolin and montmorillonite soils subject to the two swell tests: UK linear CBR swell test (BS1924-2, BSI, 1990) and the European accelerated swell test (EN13286-49, CEN, 2007). The results of linear swell and strength are reported for the 28-day test.

UK linear CBR swell test
The results of the swell testing for both clay types after 28 days immersion using BS1924-2 (BSI, 1990) are presented in Figures 3 and 4.
The unstabilised soils showed the greatest increase in linear swell. During the first three days, little swell was observed. On flooding the top of the specimen, as per the test procedure at three days immersion, the soil exhibited rapid swelling. K5S swelled significantly more than the unstabilised kaolin without sulphate from the outset of immersion. They reached a plateau at different time intervals with linear swells of 5.7 mm (ca. 13 days) and 9.1 mm (26 days), respectively (K and K5S). The stabilised clay exhibited the greatest degree of dimensional stability swelling by only 0.6 mm at 6 days and remained stable thereafter. The sulphated soils all exhibited swelling. K4L0.5S swelled to approximately 2.0 mm in 4 days immersion and again remained stable till the end of the test. Both K4L1.5S and K4L5S exhibited a proportional increase over time. Interestingly, although the overall proportion of swell was larger at the highest sulphate content (K4L5S), the rate of increase was approximately the same for both soils. At 28 days, both  were continuing to swell, indicating on-going changes in composition within the cementitious matrix.
Montmorillonite soils subject to 28-day linear swell testing are shown in Figure 4. The unstabilised clays once again exhibit the greatest degree of linear swell (M -24.5 mm and M5S -16.5 mm). Contrary to the unstabilised kaolin clays at 28 days, it is the unsulphated clay that swelled the most. Immediately on immersion, soil mixture M exhibited a relatively large swell response reaching in excess of 8 mm in the first 24 h, in comparison to M5S, which took over 4 days to achieve the same degree of linear expansion. The stabilised soil M6L exhibited negligible swell (maximum 0.02 mm) over the duration of the test. The low sulphate soil (M6L0.5S) exhibited a swell response similar to that of the equivalent kaolin (K4L0.5S). A total linear swell of 2.1 mm was observed, which reached a maximum after approximately 4 days immersion. The medium and high sulphate clays (M6L1.5S and M6L3S) exhibited a similar swell response, both in the rate of swell and the total dimensional increase. The rate of swell at 28 days of soil M6L3S remained constant indicating that expansive reactions were continuing, whereas the results indicate that the medium sulphate clay (M6L1.5S) was starting to plateau at around 20 days, as seen by a flattening of the swell curve and a lower total swell at the end of the immersion (7.3 mm compared with 8.1 mm). Figure 5 shows the water contents of the soil taken from both the top and bottom of the kaolin CBR specimens after 28 days immersion (approximately 300 g taken from the top 20 mm down from each soil face). The equivalent montmorillonites are shown in Figure 6.
Common to both is a proportional increase in water content of the top of the specimen with sulphate content. The bottom of the specimens remained broadly similar through the whole range of soil mixtures. Again (and as expected) the unstabilised montmorillonite soils exhibited larger water contents than the kaolin equivalents. The stabilised mixtures (K4L and M6L) had similar water contents. The difference between the top and bottom increased with sulphate content and linear swell. Table 6 presents the soil strength of the kaolin mixtures after 28 days immersion. The unstabilised clays had CBR values of <1% tested at both the top and bottom of the specimens. The low sulphate mixture K4L0.5S had a higher strength than the control (K4L), measuring 7% higher on the top and 35% higher on the bottom. Both the medium and high sulphate mixtures had strengths lower than the control, with K4L5S approaching that of the unstabilised soil measured on the top  of the CBR specimen. Again the bottom faces of the sulphated kaolin soils developed a 'crust' of material that gave an artificially high strength as measured by the CBR test. The pH of all the specimens remained sufficiently alkaline despite the extended duration of immersion. The results of the 28-day soaked CBR testing of the montmorillonite soils are given in Table 7. The unstabilised soils again had CBR values of <1%. The control soil (M6L) achieved a relatively high strength; double that of the kaolin (K4L), 42% higher on the top and 33% higher on the bottom than the equivalent soil tested at 7 days. All the stabilised soils exhibited higher strengths on the bottom compared with the unconfined top of the specimen. Increasing the amount of sulphate, resulted in a proportional decrease in soil strength in the order M6L > M6L0.5S > M6L1.5S > M6L3S. Again the pH of the soil remained alkaline after 28 days immersion.   Figure 7 shows the plots of the average volumetric expansion (G v ) based on testing of three identical specimens of the kaolin soils over the standard immersion period. The unstabilised clays disintegrated almost immediately. The stabilised clay swelled by 8.1%, which was reached 2-3 days after immersion. The degree of volumetric swell exhibited by the sulphated clays was proportional to sulphate content and followed the series K4L < K4L0.5S < K4L1.5S < K4L5S. The low sulphate soil K4L0.5S swelled to completion within the standard immersion period (168 h/7 days). K4L1.5S appeared to be reaching a plateau at 7 days immersion, while K4L5S continued to remain volumetrically unstable and was continuing to exhibit a positive swell response.

European accelerated volumetric swell test
The average volumetric swell response of the montmorillonite based soils is shown in Figure 8. The unstabilised clays swelled rapidly, almost reaching their maximum swell within 5 h of immersion. Unlike their kaolin equivalents, the clay remained cohesive enough such that it could still be measured and the specimen volume calculated up to the end of the immersion period. The stabilised soil (M6L) exhibited a much reduced swell response of 1.4% within the first 5 h, after which the clay remained volumetrically stable; until the end of the test. The swell of the sulphated stabilised clays followed the series M6L0.5S < M6L1.5S < M6L5S. After approximately 3 days immersion, the mixture M6L0.5S almost reached a plateau. During the remainder of the immersion period, the soil continued to swell, but only by a relatively small amount, approximately 0.13%. The medium and high sulphate soils continued to swell, although again, at a relatively reduced rate, having largely swelled to completion by approximately 5 days immersion. Tables 8 and 9 show the final strength and swell values of the soils subject to the two swell tests presented in relation to the pass/fail criteria of the test. That is, the strength and swell values measured after a period of immersion as defined in the appropriate standard.  It is possible that relatively large variances in swell of the unstabilised clays are a result of small differences in the permeability of the soils, introduced during specimen manufacture. This could result in differences in the rate of water uptake of the clay and the associated swell response, rather than fundamental differences in material behaviour. The unstabilised clays for both soils exhibited characteristically high swelling and extremely low strengths after immersion. As would be expected, the montmorillonite swelled significantly more than the kaolin and is characteristic of 2:1 layer structure clays. Some variation in the ultimate linear swell values was found for the sulphated clays over their unsulphated counterparts, while the volumetric swells were similar. This could be due to variation in permeability of the top layer of the soil in the CBR moulds resulting in varying swell rates. Although given the large values, from a suitability point of view the soils would definitely be unsuitable.

Discussion
The effect of lime stabilisation is apparent both in terms of a reduction in linear and volumetric swell and an increase in CBR strength. This can be attributed to the development of a cementitious matrix resulting from the pozzolanic reactions forming CSH, CASH and CAH under the localised alkaline conditions within the soil specimens (Bell, 1996). The increase in strength was much more pronounced for the montmorillonite soil than for the kaolin. The slower rate of strength gain of lime stabilised kaolin soils has been reported by Bell (1996), who found that ultimate strengths tended to be higher (in the region of 30-40%) over the long-term.
The addition of increasing amounts of sulphate to the soils resulted in an incremental decrease in soil strength and an increase in both linear and volumetric swell. Given the relative stability of Figure 10. High sulphate kaolin soil (K4L5S) UK linear CBR swell test. the control soil (M6L and K4L) this can be attributed to deleterious sulphate reactions within the soil matrix. Expansive sulphate reactions, in particular the formation of ettringite would account for the increase in heave as well as the loss of strength resulting from damage to the cementitious matrix (Little et al., 2010).
In terms of the relative severity of the tests used and their associated suitability criteria, the European accelerated volumetric swell test (EN13286-49, CEN, 2007) appears to be more onerous than the UK linear swell test (BS1924-2, BSI, 1990) on account of both the control soil (K4L) Figure 11. Ettringite found in the high sulphate kaolin soil (K4L5S) UK linear CBR swell test.
For both tests, the degree of reaction can be considered a function of the rate of swell. Considering the results of the kaolin-based soils; in BS1924-2, at 28 days, both the medium (1.5S) and high sulphate (5S) were continuing to swell indicating on-going deleterious reactions. The low sulphate soil (0.5S) had reached a plateau relatively early. In EN13286-49 (CEN, 2007), the low and medium sulphate soils had reached a plateau relatively early, while the high sulphate soil was continuing to expand, although in the montmorillonite soil, it had started to level off.

Compositional analysis
Compositional analysis of selected soils was undertaken and the results reported in the following sections.

Scanning electron microscopy-electron dispersive X-ray spectroscopy
SEM-EDX shows micrographs taken from the control soils (K4L and M6L) after testing to EN13286-49. In the kaolin soil (left-hand image), the typical morphology of cementitious product (Kim & Kim, 2011) can be seen on the surface of the clay particle. In the right-hand image, smectite lamellae typical of this type can be seen as found by Yilmaz and Civelekoglu (2009). The typical acicular crystal appearance of ettringite was not found in the control specimens (Figure 9). Figure 10 shows the microstructure of K4L5S after the UK linear swell test. The spherical structures at EDX1 were identified as hydrated lime particles due to strong Ca and O emissions, as observed by Peethamparan, Olek, and Lovell (2008).
The two micrographs of Figure 11 show the characteristic needle-like morphology of ettringite. The crystals are small (<3 μm long and <0.1 μm wide), randomly orientated, and appear to have formed uniformly throughout the material.
The equivalent montmorillonite soil is shown in Figure 12. No residue lime was detected. Ettringite (AFt) was found throughout the soil, confirmed by EDX (point EDX2) showing strong emissions of Ca, Al, O and S. Adsorption of Si into the structure of AFt may account for the strong Si emission, as well as background from the surrounding clay material. The ettringite has a different morphology to the high sulphate kaolin. The crystals are shorter, but are much greater in width (4 μm × 0.4 μm). These are shown at greater magnification in Figure 13.
The low sulphate soils of kaolin and montmorillonite are shown in Figures 14 and 15.
In the kaolin soil, the characteristic plate-like crystal morphology of monosulphate was present, confirmed by EDX analysis (EDX2). Again strong emissions of Ca, Al, O and S were recorded. These are similar to those found by Stutzman (2006) in their petrographic studies of concrete. In Figure 15, the area highlighted appears to be the remnants of what once was a void in the structure of the montmorillonite soil. Several nucleation points and radial crystal growths of AFt were found.   Figure 16 shows the high sulphate kaolin after the European accelerated swell test. Ettringite (AFt) was found throughout the soil with morphology very similar to that found in the BS1924-2 test.
In addition, spherical formations high in Ca and Al were found (EDX 3) as shown in Figure 17. These are much the same as those found by Tosun and Baradan (2010) in their studies of heatcured mortars. They suggest that this is a Al-rich species that acts as a seed of what they term 'ball ettringite' that forms in the voids and narrow spaces in cement paste. Figure 18 shows the equivalent montmorillonite soil. EDX analysis (EDX 4) confirms the needle-like crystals to be ettringite due to strong elemental emissions of Ca, S, Al and O. The bottom micrograph shows what appears to be AFt growing within a void of the microstructure (circled). The crystals in the void space are much thicker (>1 μm) than those found distributed throughout the material. Figure 19 shows the analysis of the low sulphate kaolin after the EN13286-49 test. The characteristic plate-like morphology of monosulphate was found throughout the material. The EDX analysis (EDX 5) confirms that these crystals are those of AFm due to high emissions of Ca, Al, O and S, rather than what can be another plate-like phase -CSH (Stark & Möser, 2002). The EDX spectrum of AFt and AFm is similar as they are comprising the same elements. The high sulphate form, AFt, was also found although compared with the other soils, this was much less abundant ( Figure 20). Figure 21 shows the equivalent montmorillonite soil. AFt was found (circled), although its occurrence was much less than the high sulphate soil. Its morphology was also similar Figure 16. Acicular ettringite crystals found in the high sulphate kaolin (K4L5S) after the European accelerated swell test, EN13286-49 (CEN, 2007). to the kaolin having undergone the UK linear CBR swell test. The crystals were relatively long (>15 μm) yet still very thin. Their width was difficult to measure but was probably <0.5 μm.
EDX6 shows the emission spectrum of small cubic structures high in Ca, Si and O. This is possibly a CSH cementing product similar to that found by Peethamparan et al. (2008) (Figure 22).

Discussion
The deterioration of the soils tested in this study can be linked to the formation of ettringite and monosulphate. Analysis of the control specimens (K4L and M6L) found them to be both dimensionally stable and retained their strength after testing. Ettringite was found in the high sulphate soils of both clay types (K4L5S and M6L5S), although its morphology was dependant on the clay and the conditions of the swell test. Ettringite in the kaolin soil was relatively difficult to detect, due to the small size of the crystals, despite relatively large and on-going dimensional increases in the specimens at the time of analysis. It is suggested that this is evidence of topochemical colloidal AFt formation and the expansion being associated with a crystal swelling mechanism (Schwiete et al., 1966). Some researchers have reported that kaolin clays in an alkaline environment produce a pore solution with a higher concentration of alumina compared with montmorillonite (Mitchell & Dermatas, 1990). This coupled with a high hydroxide ion (OH − ) concentration would promote such a mechanism (Min & Tang, 1994). Conversely, the AFt formed large stubby crystals throughout the montmorillonite soils suggestive of a through-solution mechanism (Ogawa & Roy, 1982).  The low sulphate soil of the kaolin formed monosulphate (AFm) and smaller amounts of AFt. It is suggested that the lower concentration of gypsum, coupled with the high alumina concentration in the pore solution, resulted in conditions that were more thermodynamically favourable to the formation of monosulphate, which is known to form in what is termed a low sulphate environment (Mitchell & Dermatas, 1990). This was surprising as monosulphate is thought only to be metastable at room temperature and very susceptible to drying (Damidot & Glasser, 1992;Mitchell & Dermatas, 1990). The montmorillonite (M6L0.5S, Figure 15) showed evidence of AFt formation in the void structure of the soil, further supporting a through-solution mechanism of formation.  In the European accelerated swell test the high sulphate kaolin soil exhibited significant on-going volumetric swell at the time of analysis (Figure 7). SEM-EDX analysis again showed evidence of colloidal AFt formation, but also Al-rich structures ( Figure 16) that act as seed points for later formation of the more typical needle-like AFt crystals (Tosun & Baradan, 2010). This 'early-age' ettringite is consistent with the recorded on-going increase in volumetric swell. In the equivalent montmorillonite soil, the long and thin crystals of AFt associated with a topochemical formation mechanism, as well as short, thicker crystals in void structures (through-solution) were found. It is possible that the higher immersion temperature increased the solubility of candidate ions and localised variations in conditions allowed the co-formation of ettringite by both mechanisms. The low sulphate kaolin soil (K4L0.5S) formed comparatively more AFm than the same soil tested under BS1924-2. After some initial expansion on immersion, it remained dimensionally stable. This supports the accepted theory that monosulphate does not contribute to expansion (Mitchell & Dermatas, 1990). The low sulphate montmorillonite (M6L0.5S) yielded AFt only (see Figure 21). Its morphology was similar to that formed in the UK linear CBR swell test of the same soil (Figure 15).

Conclusions
Based on the results of this research the following conclusions can be drawn: • Lime stabilised cohesive soils with increasing amounts of sulphate exhibit a proportional degree of dimensional instability and loss of strength, when subject to swell testing. This can be attributed to the formation of the hydrous calcium sulphoaluminate species ettringite (AFt). • For formation of monosulphate (AFm) was limited to those soils that were low in sulphate (0.5%). Aside from the controls, these also exhibited the lowest degree of expansion. • Relating the physical property results to suitability criteria of the two swell tests, the European accelerated swell test is considered more onerous than the UK linear CBR swell test at 28 days. • The mechanisms of formation and expansion were not the same for each soil and were found to be affected by clay type, sulphate content and swell test conditions. • In the soils tested evidence was found to support both swelling mechanisms (topochemical formation/crystal swelling and through-solution/crystal growth). • Ettringite can exhibit a range of morphologies dependant on the physicochemical environment in which it forms.