Swelling and collapse of compacted soils to be used as earth dam cores

The mechanical behaviour of fine-grained soil materials to be used as impermeable cores for earth dams has been extensively studied by numerous researchers. The required properties of these materials have also been very well described by standards and specifications. Yet, more than often it is required to assess their swelling/collapse potential, especially at various vertical stresses, as a means to estimate their volume changes upon inundation which is going to be caused by filling the dam reservoir. In the paper, experimental results of five different soils are presented. The soils tested ranged from non-plastic silty sands with clay to medium plasticity clayey silts that were compacted in conditions dry, at and wet of optimum moisture content as derived from standard compaction energy Proctor tests, then subjected to one-dimensional loading conditions and then inundated. The vertical stress was up to 7.6 MPa. The experimental results are categorized according to initial moisture content relative to Proctor optimum and indicate expected magnitudes of strains due to inundation for various grain-size distributions and plasticities.


Introduction
Earth and rockfill dams have become increasingly more attractive relative to other types of dams over the past decades.More than often, clayey materials for the construction of the clay core in these types of dams may be hard to obtain in the necessary quantities at close proximity to the dam under construction, pose various concerns regarding their mechanical behaviour and constitute the subject of extensive experimental programmes before construction and quality control during construction.The required properties of these materials are generally very well described by specifications.These describe in detail the tests needed during geotechnical investigation and quality control and the standards that should be followed [1].Yet, in addition to usual classification and mechanical properties tests, it is often required to assess the swelling/collapse potential of these soils, especially at various vertical stresses, as a means to estimate their volume changes upon inundation which is going to be caused by filling the dam.Given the fact that earth dams can be very high, the range of vertical stress that needs to be applied can be quite large.
Experimental results from five different soils examined for use in earth dam clay cores are presented in the paper along with results from a sedimentary clayey soil for comparison.The soils tested, ranged from nonplastic silty sands with clay, to medium plasticity clayey silts and were compacted in conditions dry, at and wet of optimum moisture content as derived from standard compaction energy Proctor tests.They were then subjected to various vertical stresses under onedimensional loading conditions and then inundated.The vertical stress was up to 7.6 MPa.The experimental results are categorized according to initial moisture content relative to Proctor test optimum moisture content and indicate expected magnitudes of strains due to inundation for various grain-size distributions, plasticity indices and geological origins.

The soils tested
The soils tested by this stage of the ongoing research are Livadero clay, Skiros clay, Skiros weathered Serpentinite, and two Platy river non-plastic sandy silt to silty sand.For comparison, results from Maroussi clay, a sedimentary clayey silt with sand found in abundance in the north of Athens are also presented.Livadero clay is a sandy silt with clay found in the region of Drama close to the city of Livadero in northern Greece and has been examined for use in the construction of the clay core of an earth dam.Skiros clay is a clayey silt with sand used for the construction of the Ferakambos river dam on the island of Skiros in the Aegean Sea.In the same dam a locally found weathered Serpentinite found in abundance has been used for the construction of the shoulders of the same dam.Given its low permeability and that the clay found in the vicinity eventually was not found in the necessary quantities, the same material (in its more plastic and fine-grained phase) has been used for the construction of the top 5m of the clay core (the dam is 27m high).After all a preliminary study for the dam (before finalisation of the design study to the design that was adopted for construction) had proposed the construction of a uniform dam made totally with weathered Serpentinite and a major central drain.Details of the mechanical behaviour of the Skiros weathered Serpentini- te have already been presented [2,3].Another dam for irrigation purposes on the Platy river in the Rethimno region of southern Crete has been examined.The 55m high dam will be an earth dam with clay core built from clays found in the reservoir basin and shoulders built from coarse-grained alluvial deposits found also in the reservoir basin.As part of the geotechnical investigation and in search of clay deposits for use in the construction of the clay core, two non-plastic materials found in abundance in the reservoir basin were also examined.One was a sandy silt with clay (M1) and the other a silty sand with clay (M4B).Index properties of the soils tested are summarised in Table 1.
Standard energy Proctor compaction tests have been performed on all the soils tested.The compaction characteristics of the soils are presented in Table 2 and their dry unit weight-moisture content compaction curves are presented in Fig. 1.Depending on the shape of the compaction curves in Fig. 1, moisture contents for preparation of samples in conditions dry and wet of optimum moisture content were decided so as samples with dry unit weight 95% of the maximum dry unit weight could be prepared.These moisture content values are also shown in Table 2.

Experimental method
Two methods were employed for samples' preparation.One was dynamic compaction in the Proctor test mould using an automatic compactor and then trimming the specimens to be compressed and inundated from these larger samples, and the other dynamic compaction by hand in the oedometer cells by controlling dry soil mass, water mass and the height of the sample by gradual compaction until the required dry unit weight is achieved.Maroussi clay and Skiros weathered Serpentinite samples were prepared using the former method, and samples of all the other soils using the latter.
A review of available methods for preparation of compacted samples for swelling/collapse tests and their comparison is beyond the scope of this paper; still the two methods used constitute widely used and well accepted methods for preparation of compacted samples.If one had to point to a difference standing out is that the preparation of large size samples in the Proctor test moulds allows the measurement of suction of each sample before trimming the specimens to be tested.Therefore, although conventional oedometers may be used for the swelling/collapse tests, still the magnitude of suction in each moisture condition relative to Proctor test optimum moisture content is known and can be associated with the observed strains due to inundation other than just the moisture content.As already reported [3] the Skiros weathered Serpentinite exhibited matric suction of 12 kPa and total suction of 240 kPa for samples compacted wet of optimum, 30 kPa and 240 kPa respectively for samples compacted at optimum, and finally 70 kPa and 380 kPa respectively for samples compacted dry of optimum moisture content.Matric suction in this particular case was measured using a Soilmoisture Equipment Corp. 2100F laboratory tensio- meter and total suction using a Decagon Devices, Inc. chilled mirror hygrometer.Matric suction was measured by installing the tensiometer in the large size sample before trimming the specimen for the swelling/collapse test and total suction on trimmings from locations immediately above and below the trimmed samples.Similar measurements on compacted samples of Maroussi clay [4] yielded 60 and 130 kPa for samples compacted wet of optimum, 80 and 330 kPa for samples compacted at optimum, and a total suction of 2500 kPa for samples compacted dry of optimum.In this last case the tensiometer could not measure the matric suction of the samples.These measurements of suction and the insight they offer in the interpretation of observed swelling/collapse upon inundation come at the cost of larger inhomogeneity of specimens finally trimmed for actual swelling/collapse testing, and at a cost and effort that cannot be routinely applied in geotechnical practice.

Swelling/collapse strains due to inundation
The swelling/collapse strains upon inundation for each of the soils tested are plotted against vertical stress in Fig. 2 (a) The strains observed are really impressive ranging from 7.3% swelling for 1 kPa vertical stress (practically free swelling) to 10.7% collapse for 500 kPa of vertical stress which is also the maximum collapse strain observed with values for higher stresses decreasing linearly with the logarithm of increasing vertical stress.Another important observation is that these very high values of strain due to inundation correspond to a soil and a moisture condition that yields a value of swelling pressure of just 75 to 80 kPa (defined as the intersection of the strain-log stress line with the zero strain axis); a value that would not cause serious concern in everyday geotechnical practice although experimental results indicate that strains due to inundation above and below this value are remarkably high to the point of being called alarming.To add to this point, strains due to inundation for this soil compacted at optimum moisture content indicate a value of swelling pressure in the order of 700 kPa but strains below that value as low as 60 kPa and above that value as high as 2000 kPa are really small (between -0.5% and +0.02%) and only at very low stresses and very high stresses exhibit very high values (-5% for 1 kPa and +4% for 7.6 MPa).In this case therefore, a truly alarming value of swelling pressure (700 kPa) is not associated with similarly significant strains due to inundation in the range of usually applied vertical stresses in engineering works.Finally, strains due to inundation for this soil compacted wet of optimum moisture content indicate much smaller strain values especially for collapse and also for swelling, although the variation there is much higher (yet values for vey low values of vertical stress may reach even 1.2% swelling).Swelling pressure in this case is again between 70 and 80 kPa.The behaviour of this material upon inundation justified the investigation of its mineralogy.The XRD method was employed indicating a total of 65% clayey minerals with the rest being equal proportions of calcitic and silica minerals.Clayey minerals included 15% illite, 11% montmorillonite, 10%  kaolinite, 10% Muscovite, 8% Halloysite, 5% Chlorite and 4% Serpentinite.A mineralogy like that is not commonly found, yet the observations on the behaviour of Maroussi clay upon inundation shed considerable light into the relation (literally the lack of it) between the magnitude of swelling pressure and the magnitude of strains due to inundation.Similar observations can be made for Livadero clay (Fig. 3), although not as pronounced as with Maroussi clay.Dry of optimum the soil exhibited the highest swelling strains at low stresses and the highest collapse strains at higher stresses, again with the maximum exhibited at 250 kPa with considerable decrease with stress after that value.Swelling pressure was 65-70 kPa.At optimum the soil exhibited a similar behaviour with smaller strains in both ranges, an increase of swelling pressure to 150 kPa and similar decrease of collapse strains with vertical stress after a maximum exhibited at 250 kPa.Finally, for wet of optimum, the strains observed were the smallest (maintaining considerable values for swelling at low stresses,) with the interesting observation that swelling pressure climbed to 270 kPa.
Slightly different observations can be made for Skiros clay (Fig. 4).Dry of optimum the soil exhibited the highest swelling strains at low stresses and the highest collapse strains at higher stresses, with the maximum exhibited at 500 kPa with considerable decrease with stress after that value.Swelling pressure was 100-110 kPa.At optimum the soil exhibited a continuous increase of collapse strain up to the maximum of 4000 kPa of vertical stress.Finally, for wet of optimum, the strains observed were the smallest (maintaining considerable values for collapse at high stresses) with the interesting observation that collapse strains started decreasing after 2000 kPa.Skiros weathered Serpentinite was the only one of the soils tested that did not exhibit a maximum of collapse strain at any moisture conditions (Fig. 5).For moisture content 2.5% dry of optimum, from 250 kPa of vertical stress the collapse strain started to increase rapidly up to 4000 kPa where a tendency for stabilization of the collapse strain with vertical stress was observed.Maximum collapse strain reached 6.4%.Zero deformation upon wetting was observed at 125 kPa.For lower stresses swelling was observed but its maximum value was very small (-0.2%).For optimum moisture content a similar behaviour was observed although the increase in collapse strain is less rapid and the maximum collapse strain smaller (it reached a maximum value of 2.7%).The rapid increase started between 500 and 750 kPa and lasted up to 4000 kPa where stabilization of the collapse strain occurred again.Zero deformation upon wetting was observed at 225 kPa.For lower stresses, swelling was observed but its maximum value was only -0.34%.Finally for wet of optimum, collapse strain observed even at vertical stress as high as 7600 kPa was practically negligible (max.0.07%).In this particular case of so low collapse strains, it may be argued that there was essentially no strain due to wetting as the deformation observed after wetting for 24hrs was only secondary compression deformation of the sample.Zero deformation upon wetting was observed at 500 kPa (a value already reported and discussed [2]).For lower stresses swelling was observed but its maximum value was only -0.09% for the minimum applied stress.The Platy river sandy silt (M1, Fig. 6) and silty sand (M4B, Fig. 7), although non-plastic, they did exhibit for conditions dry of optimum a swelling pressure of 40 and 25 kPa respectively.In conditions dry of optimum, M1  exhibited a rapid increase of collapse strains between 125 and 250 kPa and no signs of stabilization of collapse after that stress, while M4B did not exhibit a rapid increase and collapse stabilized after 1000 kPa.In conditions at optimum, both soils exhibited a maximum of collapse strain of 0.7% at 500 kPa and a subsequent decrease of collapse strain values.Finally, for conditions wet of optimum, both soils exhibited no swelling with the collapse strains being the smallest and with a linear increase in their values with the logarithm of the vertical stress.Finally in Fig. 8 swelling/collapse strain is plotted against vertical stress (in logarithmic scale) for all soils at the same moisture conditions; dry (Fig. 8a), at optimum (Fig. 8b) and wet (Fig. 8c).All soils in conditions dry of optimum exhibit a maximum of collapse strain or stabilisation.Something similar cannot be observed for conditions at optimum moisture content, and finally for conditions wet of optimum usually a linear increase of collapse strains is observed although stabilisation has been observed too (Skiros clay).As found by other researchers too [5], both swelling and collapse may be reduced practically to extinction by compacting the soil wet of optimum.

Conclusions
Swelling/collapse of a total of six soils was investigated experimentally.Five of these soils have been used for or are intended (or have been examined) for the construction of clay cores of actual dams.All soils in conditions dry of optimum exhibit a maximum of collapse strain or stabilisation of collapse.Something similar cannot be observed for conditions at optimum moisture content, and finally for conditions wet of optimum usually a linear increase of collapse strains is observed although stabilisation has been observed too.There seems to be no direct relation between the magnitude of swelling pressure and the magnitude of strains due to inundation whether that is swelling strains or collapse strains.

Figure 2 .
Figure 2. Strain due to inundation versus vertical stress for Maroussi clay at each of the three moisture contents tested for vertical stress in a) linear scale, and b) logarithmic scale.

Figure 3 .
Figure 3. Strain due to inundation versus vertical stress for Livadero clay at each of the three moisture contents tested for vertical stress in a) linear scale, and b) logarithmic scale.to7. Diagrams denoted (a) present results with the vertical stress axis in linear scale, and diagrams denoted (b) present results with the vertical strain axis in logarithmic scale.Curves for uniform moisture conditions are presented in each figure, i.e. curves for samples compacted dry of optimum, at optimum, and wet of optimum moisture content.Swelling/collapse strains upon inundation for Maroussi clay are plotted in Figure2.Tests were performed both at very low vertical stress (just 1 kPa) and at very high vertical stress (7.6 MPa).The strains observed are really impressive ranging from 7.3% swelling for 1 kPa vertical stress (practically free swelling) to 10.7% collapse for 500 kPa of vertical stress which is also the maximum collapse strain observed with values for higher stresses decreasing linearly with the logarithm of increasing vertical stress.Another important observation is that these very high values of strain due to inundation correspond to a soil and a moisture condition that yields a value of swelling pressure of just 75 to 80 kPa (defined as the intersection of the strain-log stress line with the zero strain axis); a value that would not cause serious concern in everyday geotechnical practice although experimental results indicate that strains due to inundation above and below this value are remarkably

Figure 4 .
Figure 4. Strain due to inundation versus vertical stress for Skiros clay at each of the three moisture contents tested for vertical stress in a) linear scale, and b) logarithmic scale.

Figure 5 .
Figure 5. Strain due to inundation versus vertical stress for Skiros weathered Serpentinite at each of the three moisture contents tested for vertical stress in a) linear scale, and b) logarithmic scale.

Figure 6 .
Figure 6.Strain due to inundation versus vertical stress for Platy River Sandy Silt (M1) at each of the three moisture contents tested for vertical stress in a) linear scale, and b) logarithmic scale.

Figure 7 .
Figure 7. Strain due to inundation versus vertical stress for Platy River Silty Sand (M4B) at each of the three moisture contents tested for vertical stress in a) linear scale, and b) logarithmic scale.

Figure 8 .
Figure 8. Strain due to wetting versus vertical stress for all soils tested at a) dry of optimum, b) optimum, and c) wet of optimum moisture content.

Table 1 .
Index properties of the soils tested.

Table 2 .
Index properties of the soils tested.