The 1998 BGS touring lecture was given by Professor Antonio Gens of the Technical University of Catalunya, Barcelona. In Nottingham the East Midlands Geotechnical Group was the host. Gens confessed that he had once avoided unsaturated soils, considering it an alien subject not amenable to study with the usual tools of soil mechanics. He suggested that many people viewed unsaturated soils in this manner and a main aim of his lecture was to bring unsaturated soils into a more unified framework within mainstream soil mechanics.
The importance of unsaturated soils was underlined. One third of the earth surface is in arid or semi arid climates, where the water table is very low and the upper part of the ground is unsaturated. In addition, compacted soils are, at least initially, always unsaturated. Flow through unsaturated soils is also of interest in geoenvironmental engineering, such as in the study of the spread of contaminants through compacted clay liners or in the vadose zone.
An idealised representation of an unsaturated soil is shown in Figure 1. The soil consists of particles in contact, which enclose pores containing air and water. A meniscus is formed by the water which has a negative pore water pressure with respect to the air. Differences between the air and water pressure create capillary forces between the soil particles. These intergranular forces are normal to the contact between the particles, so that the overall result is stabilising.
Total suction of the soil controls water flow between two points at the same height and under the same air pressure. It is composed of matric suction, caused by the capillary forces, and osmotic suction, caused by differences in concentration of the pore water solution. In most cases matric suction controls mechanical behaviour, with osmotic suction only significant in some clay-rich materials. The relationship between matric suction and degree of saturation is described by the characteristic curve (Figure 2) which provides a form of 'identity card' for the soil. The shape of the characteristic curve is determined by the pore structure of the material, with uniformly graded soils exhibiting flat curve sections, and well graded materials having more smooth curves.
Gens gave a brief overview of the mechanical behaviour of unsaturated soils focusing on shear strength, collapse and swelling behaviour. Shear strength is enhanced by suction, due to the stabilising effect of intergranular forces. Laboratory tests have shown that as suction increases, apparent cohesion increases but often with little change to the friction angle. This high shear strength is responsible for the high slope stability sometimes exhibited by unsaturated soils. If the suction conditions change in these slopes due to saturation of the soil, then slope failure can occur. This phenomenon is observed in slopes all over the world, a classic example being Hong Kong.
Another characteristic feature of unsaturated soils is collapse. Collapse is a compression that occurs quickly when the soil increases its water content. The intergranular forces produced by suction create a metastable structure with a high void ratio. On saturation the suction disappears so the intergranular forces also disappear, causing compression of the soil. If saturation is fast, there is a rapid reduction in volume, which can cause significant problems. This phenomenon was illustrated with a stress/strain curve of compacted Boom clay tested under suction control (Figure 3).
As a sample under suction is loaded, the curves move outside of the saturation line. If the sample is saturated at any point, it returns to the saturation line through a volumetric strain corresponding to collapse. However, if the saturation occurs under a reduced vertical stress, a slight swelling may occur, caused by elastic recovery at the particle contacts. Both of these effects can occur in the same soil, depending on the applied stress level. Truly expansive or swelling clays have a mineralogical composition that allows water into the inner structure of the clay. This can generate large strains in expansive clays. Swelling is in fact highly dependent on density, with dense soils producing the greatest swelling. In terms of hydraulic behaviour, the most important parameter is permeability. In unsaturated soils, permeability is very much dependent on suction, where moderate variations in suction can change permeability by orders of magnitude.
Development of unsaturated soil mechanics
There has been interest in unsaturated soils since the first conference on soil mechanics at Harvard in 1936, where there were a number of papers covering their hydraulic and mechanical behaviour. Between the early 1950s and mid 1960s, fundamental research was undertaken, and the significance of suction was recognised and properly defined. During this time an attempt was made to establish an expression for an effective stress that would describe all soil behaviour. This was not successful, mainly because it was found to be impossible to account for the phenomenon of collapse.
Failure to unify soils within a common theory led to separation of the study of saturated and unsaturated soils from the mid 1960s. During this period fundamental advances were made in the study of saturated soils. Unsaturated soils began to be regarded as problematic or strange soils and the subject became fragmented, with separate conferences held on expansive, collapsing, arid and residual soils. However, from the late 1970s attempts began to unify the subject and to incorporate it back into the soil mechanics mainstream.
Milestones in the development of saturated soil mechanics include the theory of effective stress, Terzaghi's consolidation analysis, the critical state behaviour framework and the understanding of the structural effects of natural soils. Parallels can be found in the development of unsaturated soil mechanics. The abandonment of a single effective stress approach has allowed the features of unsaturated soils to be more readily understood. Modern analysis is usually based on two independent stress variables, net stress (excess of total stress over air pressure), and matric suction (Fredlund and Morgenstern, 1977). Recently the particular significance of microstructure in unsaturated soils has also been recognised (Delage et al, 1996).
Two approaches for describing the mechanical behaviour of unsaturated soils were discussed: state surfaces and elasto-plastic models. The state surface approach is based on the relationship of a characteristic, such as void ratio, with both matric suction and net stresses. This is both practical and popular but has some drawbacks - it is not very generalised and not readily consistent with modern saturated soil frameworks.
The Barcelona basic model (Alonso et al, 1990) was given as an example of the use of the elasto-plastic model approach. Gens described how unsaturated soil behaviour could be incorporated into an elasto-plastic framework. Figure 4 shows a yield curve for an unsaturated soil. Saturation at any point on the yield curve creates a collapse represented by a translation of the yield curve. A similar displacement of the yield curve can also be created by loading. Saturation of the sample at a small stress within the elastic domain generates a small swelling corresponding to the dual swelling/collapse behaviour typical of unsaturated soils. Another feature of unsaturated soils in terms of a critical state model is that as suction increases, cohesion increases and so the elastic domain expands.
It was suggested that elasto-plastic models for unsaturated soils exhibit consistency with saturated soil mechanics in a 3-D space (Figure 5). Saturation at any point on the yield curve causes a collapse onto a saturated soil model and elasto-plastic unsaturated soil models can therefore be viewed as an extension of classical saturated soil mechanics. A number of elasto- plastic models for unsaturated soils have been developed in recent years. Most of them share common features: adoption of two independent stress variables, use of loading/collapse yield surfaces and consistency with saturated soil models (Gens, 1996).
It was suggested that with a few modifications, a natural extension of Terzaghi's consolidation analysis could also be applied to unsaturated soils. Unsaturated soil has the capability to store water through changes in the degree of saturation. These changes are represented by the characteristic curve, which is basic in the description of flow characteristics through an unsaturated soil.
Gens questioned specifications which allow compaction above 95% of standard proctor dry density (95% relative compaction) on the dry side of optimum (Figure 6). This can be dangerous because soils compacted on the dry side exhibit a granular structure with large pores and so are prone to collapse. This was illustrated by a collapse on part of the Girona to Lleida highway north of Barcelona. The road was constructed on residual granite fill. A serious collapse occurred after heavy rain, with settlements of 200mm to 300mm. Longitudinal cracking occurred along a 10km stretch and voids were created under pavement slabs. Other problems included erosion, shallow slides and foundation instability, and both vertical and horizontal strains were apparent. Investigation of core samples showed relative compaction above 95% which was within specifications but had not prevented serious settlement problems.
The problems associated with compacting to a high density were illustrated by a Turkish motorway. The road experienced longitudinal cracking up to 300mm wide through the sub-base to the fill. Cracking was concentrated along the hard shoulder and fast lane and was generated by swelling in the top layers of the fill. Tests showed the compaction density was extremely high, in some cases exceeding modified proctor dry density, and compaction water content was low. The investigation revealed that swelling had been caused by water flowing into the fill from the hard shoulder and central reservation. Gens stressed that the soil was a low plasticity silty clay which was not particularly expansive. Swelling can occur in unsaturated soils when active minerals are not present, if the density is high enough.
There are strong grounds for specifying a maximum compaction density as well as a minimum, Gens said. These two case histories corroborated Trenter and Charles' (1996) recommendation that for sensitive projects the specification should stipulate compaction between the 0% and 5% air voids line, but with minimum and maximum dry densities (Figure 7). However, Gens warned that compaction on the wet side does not constitute complete 'life insurance'. Recently, he conducted experiments where a sample compacted on the dry side was compared to one compacted on the wet side but transferred to the dry side with suction control (Figure 8). On saturation, the sample initially compacted on the wet side exhibited significant collapse, equal to about half that of the dry side one (Figure 9). Gens said this showed that if soils dry out after compaction there is potential for collapse.
Embankment dams and the application of unsaturated flow-deformation analysis were then discussed. Gens suggested that unsaturated-flow deformation analysis could reproduce the observed settlement of embankment dams. Saturated analysis of embankment dams predicts heave during impounding due to hydraulic uplift, which is fundamentally wrong. However he conceded that although saturated analysis could be modified to reproduce the observed settlements, unsaturated flow deformation analysis includes collapse, and therefore often provides a preferable 'natural' approach.
The Limonero dam near Malaga in southern Spain was analysed. This included a comparison of the effects of compacting the dam core on the wet and dry side of optimum, and a study of the influence of the soil microfabric. Results showed less collapse and a greater dissipation of pore pressures with a wet core. The dry core showed little pore pressure dissipation, and a sharp wetting front. The Beliche dam in Portugal was also discussed. This suffered a sudden temporary impoundment during construction. A clear collapse behaviour was identified, which could be separated from construction settlements. Again, settlement, collapse and piezometer results of the embankment dam could be successfully modelled (Pagano, 1998).
The final case history concerned expansive ground at the Asco nuclear power station project, south of Barcelona. The site consisted of a large bowl-shaped excavation in mudstone, which began to experience heave. Although the mudstone was a swelling soil, heave was initially attributed to elastic rebound.
Extensometers installed on the site revealed that swelling was occurring in the top 10m of the mudstone. Piezometers in this zone either showed a dry reading, with suction higher than 100kPa, or a standing water table of 32m, with no intermediate readings. This implied that fissures were channelling water to the swelling rock. Maximum swelling occurred where the mudstone was thickest. Material characterisation of the mudstone showed insitu suction to be very high at around 20Mpa and laboratory tests showed the swelling period to be very long at over 300 days. A conceptual model of the microstructure was developed, incorporating swelling in a two phase process. First, water penetrates the macrostructure of the pores and then activates the microstructure of swelling minerals (Figure 10). Water exchange between the two structural levels accounts for long term swelling behaviour. The model was found to provide a remarkably consistent framework for reproducing all field observations and laboratory results and was successfully used to match observed permeability and heave and to predict behaviour over the life of the power station (predicted heave was found to be not critical).
To date, the predicted values have compared well with observations. Gens said that the important point here was that the model reproduced actual physical processes behind the observed behaviour of the soil. This allowed confident predictions to be made.
Gens concluded by emphasising that consideration of both suction and microstructural effects is essential to achieve an appropriate understanding of the behaviour of unsaturated soils. Increased use of suction control laboratory equipment and insitu suction measurement will allow further development of this understanding.
Future advances will include incorporation of microstructural effects within a quantifiable framework, along with chemical and temperature effects (Gens, 1995).
In the last few years major effort had been made to incorporate unsaturated soils into the soil mechanics mainstream and that trend was set to continue, he said.
Alonso E, Gens A and Josa A, 1990. A constitutive model for partially saturated soils. Geotechnique, Vol 40, pp405-430.
Brooks RM and Corey AT, 1966. Hydraulic properties of porous media. Hydrol Paper 3, Colorado State Univ, F Collins.
Delage P, Audiguier M, Cui YF and Howat MD, 1996. Microstructure of a compacted silt, Canadian Geotechnical Jnl, 33, pp150-158.
Fredlund DG and Morgenstern NR, 1977. Stress state variables for unsaturated soils. ASCE Jnl Geotech Eng Div, Vol 103 (GT5), pp447-466.
Gens A, 1996. Constitutive modelling: Application to compacted soils. Proc 1st Int Conf on Unsaturated Soils, Paris, 3, pp1179-1200, Balkema.
Gens A, 1995. Constitutive laws, Modern issues in non-saturated soils (Eds Gens A, Jouanna P and Schrefler B). Springer Verlag, pp129-158.
Pagano L, 1998. Interpretation of mechanical behaviour of earth dams by numerical analysis. Prediction and Performance in Geotechnical Engineering, Hevelius Edizioni, Napoli, pp89-150.
Trenter NA and Charles JA, 1996. Model specification for engineered fills for building purposes. Geotechnical Engineering, Proceedings of the Institution of Civil Engineers, Vol 119, 4, pp219-230.