Studying the stress-suction coupling in soils using an oedometer equipped with a high capacity tensiometer

Abstract

In the context of research into deep nuclear waste disposal, various works have concerned the hydromechanical behavior of Boom clay, a stiff plastic clay extracted in the SCK-CEN Underground Research Laboratory near the Mol City (Belgium), at a depth of 223 m. Due to some amount of smectite minerals in the clay fraction, Boom clay exhibits swelling properties when hydrated under low stresses. To investigate some aspects of the hydromechanical behavior of Boom clay, oedometer compression tests were carried out on samples of Boom clay close to saturation and submitted to an initial suction. During oedometer compression, the changes in suction with increased vertical stress are monitored by means of a high capacity tensiometer installed at the bottom of the sample. Some aspects related to hydromechanical couplings are examined through the investigation of the changes in suction during oedometer compression, a somewhat delicate and poorly documented experimental approach. A comparison is also made with a completely different soil sample under suction, i.e. a statically compacted unsaturated low plasticity silt. Some technical difficulties typical of this new experimental approach are first described in detail so as to optimize the interpretation of the data obtained. The experiment allows the determination of the point at which suction is changed to positive pressure during compression. Below this point, the ratio between the vertical stress and the change in suction are determined. Above this point, the data show that positive pore pressures are dissipated in a common way. The suction/stress behavior during unloading is also described and discussed. Finally, an interpretation in terms of microstructure effects is provided for both samples. The experimental approach initiated here seems to provide interesting further application to better understand hydromechanical couplings in natural soils in relation with suction increase during stress release.

Keywords

Oedometer / tensiometer / swelling / physicochemical and mechanical effects / stress/suction coupling / soil plasticity

Introduction

In the context of research into deep nuclear waste disposal [1], several investigations have been carried out on Boom clay, a stiff clay from the Underground Research Laboratory (223 m deep) in Mol, Belgium. Most works have been done on triaxial specimens that were trimmed from blocks extracted during the excavation phase in the Underground Research Laboratory. Many questions have been arisen as how to reproduce the soil field condition in the laboratory after the complex thermo-hydro-mechanical process to which the soil sample has been subjected: unloading and temperature decrease due to sample extraction at 223 m depth, possible water loss occurring before protecting the sample from evaporation and possible mechanical disturbance during transportation. Because of excavation, air venting and re-saturation, the soil mass close to the gallery wall is submitted to unloading, drying and then wetting. Once radioactive waste is disposed, the soil is submitted to temperature elevation due to the heat emitted by the waste.

Hydro-mechanical coupling in Boom clay has been studied by various authors. Among them, Sultan et al. [2] showed that soil swelling possibly erased the soil mechanical memory resulting in a decrease of the preconsolidation pressure. This phenomenon, described by Gens and Alonso [3], has also been observed on most expansive soils. Various works have also showed a suction increase in samples when stress is released and vice versa. Despite the experimental difficulties related to suction measurements, various authors have studied suction variation under loading. Tang et al. [4] have measured the total suction changes in triaxial tests using a thermocouple psychrometer placed in the soil sample. Caruso and Tarantino [5] built a direct shear box equipped with a tensiometer to monitor suction changes during shear tests at constant water content. The suction – stress relation has also been commented in Rahardjo and Fredlund [6] using a K₀-cylinder. The suction increase due to the release of total stress has been also studied by Ridley et al. [7], Chandler and Gutierrez [8].

Various results on the suction- stress relation have been obtained. Chandler and Gutierrez [8], Skempton and Sowa [9] showed that the suction generated in an initially saturated isotropic sample is theoretically equal to the mean effective in situ stress supported by the sample in the case of “perfect sampling.” Under such conditions, this means that the suction change is opposite to the isotropic stress change leading to ds/dp’ = -1 (s is the suction and p’ the mean effective stress). Doran et al. [10] extended this relation to anisotropic soils by accounting for the effects of the release of in situ deviator stress. Actually, the ds/dp’ slope experimentally deduced from triaxial tests [4] or the ds/dσ_v slopes obtained from K₀-cylinder tests [6] are different and comprised between -0.1 and -1. Table 1 summarizes the slopes values obtained by different authors on soils of different clay fractions and plasticity index (I_p). It appears that the slopes are function of the nature of the soil, the state of sample saturation, the clay fraction and the plasticity index.

In this paper, the hydro-mechanical couplings in soils are further investigated by running oedometer compression tests with suction measurements. The suction changes during oedometer compression are monitored using a high capacity tensiometer installed at the bottom of the sample and the relation between suction and vertical stress changes are investigated. In parallel, similar tests conducted on unsaturated samples made up of compacted Jossigny silt are presented. The comparison of the responses obtained from these two significantly different samples, being Boom clay a saturated plastic clay under suction and the compacted Jossigny sample an unsaturated low plastic soil, helps to better understand some of the mechanisms involved at the microstructure level in the stress/suction coupling during compression.

Materials and methods

Tests on Boom clay samples

A first series of experiments was carried out on intact Boom clay samples trimmed from blocks extracted at a depth of 223 m during the excavation of the connecting gallery in the underground research laboratory in Mol, Belgium. The Boom clay formation belongs to the Rupelian period in the Tertiary era, from 30 to 36 million years before present. Table 2 presents the mineralogical composition of Boom clay, taken from Decleer et al. [11] and Al-Mukhtar et al. [12]. The dominant clay fraction (50% to 60%<2 μm) contains mostly smectite (33%). The geotechnical properties of Boom clay are presented in Table 3 [13,14].

The gravimetric water contents of the Boom clay samples used were comprised between 20.10% and 21.53% (note that all water contents in the text are gravimetric). According to Delage et al. [15], the suction value of the soil samples in this range of water contents is included between 2 MPa to 3 MPa. Indeed, suction measurements directly carried out on intact Boom clay samples using high capacity tensiometers (HCT, see below) led to cavitation, showing that the initial suction was higher than 1.5 MPa (the maximum suction measurable by the HCT).

Four oedometer samples were machined on a lathe to reach 20 mm thick and 70 mm in diameter. To decrease the sample suction below 1.5 MPa so as to carry out HCT measurements during oedometer compression, the samples were smoothly wetted by wrapping them in wet filter papers for 10 days. Some swelling was observed during the wetting phase. Hence, the swollen samples were adjusted by hand-trimming to reduce their diameter to 20 mm and then manually thrust into the ring. The water used to wet the Boom clay samples was reconstituted to the in situ chemical concentration following a procedure described in detail in Decraen et al. [16]. In brief, the pore water composition was determined from water obtained by squeezing and leaching in situ clay cores. Table 4 presents the chemical composition of the pore water of Boom clay.

Tests on compacted Jossigny samples

A second series of experiments was carried out on a compacted Jossigny silt, a low plasticity eolian silt from the east of Paris (w_p = 19, w_l = 37). The clay fraction of Jossigny silt (34%) is mainly composed of illite, kaolinite and interstratified illite-mectite [17]. The geotechnical properties of Jossigny silt are given in Table 5. To prepare the compacted sample, Jossigny silt extracted from the field was air-dried, manually ground and passed through a 400 μm sieve. The dry powder was then carefully wetted to a gravimetric water content of 13% and stored in a hermetic box for moisture equilibration during at least 24 h. The powder was then statically compacted in the oedometer ring (diameter 70 mm) to get a sample with a dry unit weight γ_d = 14.5 kN/m³ and a degree of saturation S_r close to 40%. The hydraulic conductivity of saturated compacted sample is about 10^-8 m/s.

High capacity tensiometer

Experiments were carried out in an oedometer cell equipped with a high capacity tensiometer (HCT) in order to allow suction measurements up to 1.5 MPa during compression. The HCT used was developed at CERMES (Ref. [18,19], see Fig. 1) according to the principle described by Ridley and Burland [20] in which a very thin water reservoir is located between a saturated ceramic porous stone of high air entry value (1.5 MPa) and a deformable diaphragm on which a strain gauge is glued. In HCTs, very careful saturation procedures have to be followed so as to avoid cavitation during suction measurements (see also Ref. [21,5,22]). The tensiometer was saturated in a specially designed saturation cell as shown in Fig. 2. To do so, the cell was filled with purified water and vacuum was applied during one hour to eliminate air bubbles. A positive water pressure was afterwards applied from 0 to 2000 kPa in steps of 200 kPa every two hours by means of a pressure-volume controller (GDS). The saturation procedure was carried out after each test. The calibration of the HCT was made in the positive range by applying pressure to the water in the cell and by measuring the response of the HCT. The relationship between the applied water pressure and the electrical signal of the strain gauge of the tensiometer was calibrated. Once saturation completed, the tensiometer was put in a water container.

Oedometer tests

An oedometer cell was modified to host the tensiometer as shown in Fig. 3, as previously reported in Delage et al. [15] and suggested in Dineen and Burland [23]. As seen in the figure, a standard porous disk is placed on top of the sample. To avoid any perturbation on samples under suction, the porous disk is initially kept dry. Given its high porosity and its low water retention capacity, the porous disk does not absorb any water as long as the compressed sample remains in a suction state. To place the HCT, a hole was machined on the cell base to place the tensiometer. To prevent any evaporation and possible sample drying during the test, the cell and piston were covered with a neoprene membrane and grease was added as shown in the figure. Similar dispositions have also been taken by Caruso and Tarantino [5].

It should be pointed out that suction measurements in oedometer (or triaxial apparatus) appear to be somewhat more delicate than the standard measurements of positive pressure (see for instance Ref. [21]). When measuring a positive pressure, water is expelled from the sample, giving rise to a pressure that is transmitted through the porous disks and connected ducts to the gauge diaphragm. Contrarily, in suction measurements, the pressure gauge of the HCT has to measure the negative pressure of water stuck to the sample. In this case, particular caution has to be devoted to the contact between the HCT and the sample. To ensure water continuity and to avoid any cavitation at the sample/HCT interface, a thin layer of paste made of the same soil is put in the HCT porous ceramic disk [7,21,24]. However, further precautions have to be taken to ensure a good contact between the specimen and the HCT, leading to the following procedures:

1) A screw was placed in the HCT location prior to placing the specimen in the oedometer.

2) The ring was firmly fixed to the oedometer base.

3) The sample was carefully placed in the ring and the piston is put in contact with the sample.

4) The HCT was taken out of the water container in which it was stored after saturation, its porous ceramic was covered with a thin layer of soil paste to avoid possible cavitation at the sample/HCT interface during the measurements.

5) The HCT was quickly inserted and screwed into the hole under the cell base. All the process had to be done in less than 30 s to limit any possible soil desaturation.

6) To ensure a good contact between the HCT and the sample bottom, it is necessary to apply an effort with the HCT on the base of the soil specimen. Given the rigidity of Boom clay, it is likely that some small displacement of the specimen occurs when carrying out this operation. Indeed, a small circle corresponding to the HCT diameter has been often observed once the experiment was completed.

The specimen surfaces have to be carefully smoothened to ensure a good contact with the porous ceramic of the HCT.

An initial vertical load of 50 kPa (tests on Boom clay) or 25 kPa (tests on Jossigny silt) was applied during 48 h so as to reach suction equilibrium. After that, vertical stresses were applied by steps in a standard fashion (σ_v(_n) = 2σ_v(_n-1). When the load reached 1600 kPa, unloading in steps with a ratio of σ_v(_n) = 0.5σ_v(_n-1) were carried out. Each loading or unloading step lasted 24 h, the complete oedometer test taking about three weeks. A displacement transducer was used to monitor the vertical displacement of the piston.

The initial and final characteristics of the soil samples tested are summarized in Table 6. Note that the characteristics of Boom clay samples were determined before installing the sample in the cell (for the initial characteristics) and after removing the sample from the cell (for final characteristics). Four tests (B1 to B4) were carried out on Boom clay and three tests on Jossigny silt (J1 to J3). Due to practical problems when taking the soil sample out of the oedometer cell at the end of experiment, the final characteristics of the samples in tests B1 could not be determined. In tests B4 and J3, prior to unloading, the samples were inundated by pouring water on the surface of the piston. They were then loaded and unloaded following different load paths that will not be presented in this paper. Observation of the Table shows that the Boom clay samples have an initial degree of saturation smaller than 100%, the smallest values being 91% for samples B3 and B4, showing that they were initially unsaturated.

Experimental results

Typical results obtained from an oedometer test on Boom clay sample (Test B2) with an initial suction s_i = 180 kPa (w_i = 29.49%) are presented in Fig. 4. Note that this initial suction corresponds to a vertical stress σ_v of 50 kPa.

There are two phenomena with the settlements observed with loads smaller than 400 kPa, in a zone where water remains under suction: i) there is an instantaneous settlement; ii) the increase in positive pressure measured by the HCT is significantly higher than the applied load. Note that the same observation can be made on a similar test carried out with an initial suction of 280 kPa (Test B3) (see Delage et al. [15]).

The stress/suction hydro-mechanical coupling described above is quantified by the ds/dσ_v parameters investigated in Fig. 6. This parameter has the same value during the loading and unloading sequences of Test B1 at the higher initial suction (s = 600 kPa), giving a final suction close to the initial one around 600 kPa. At lower suctions, the ds/dσ_v value appears to be larger during unloading (-0.74 and -0.80 for tests B2 and B3 respectively), giving rise to final suctions higher than the initial ones. In other words, the same stress decrease induces a larger suction increase, showing a change in the hydro-mechanical coupling with larger suction changes during unloading.

1) The unsaturated state of the sample (Table 6), resulting in the compression of the air contained in the sample. The changes in void ratio that would result from the compression of the corresponding air volume are ∆e = 0.077 and 0.015 respectively. Although not negligible, these values are not large enough to account for all the change in void ratio observed before reaching zero suction in the compression curves.

2) Possible gaps existing between the sample and the ring created when manually thrusting the sample into the ring.

3) Inhomogeneous sample hydration by the filter papers, with possible higher water contents on the top and the bottom of the sample, resulting in zones that could be more deformable and more saturated on both sides of the sample.

4) A strain heterogeneity at the sample base created by the HCT when putting it in contact with the sample with no load applied. Indeed, a circle due to slight HCT penetration was observed at the end of the tests.

Whereas the strain due to the non-saturation can be estimated as done above to approximately one third of the total observed strain before zero suction, the estimation of other artifacts is more difficult, with no possibility of previous calibration.

The occurrence of instantaneous pressure peaks with a measured pressure value larger than the load applied is another concern. This discrepancy is related to the slight penetration of the HCT in the sample that appeared necessary to ensure a good contact. This penetration is progressively mobilized during the first loading steps, when loads are lower (loading steps of 50, 100 and 200 kPa) and when the sample rigidity is higher (due to higher levels of suction). The measurements carried out in this area could be affected by some stress concentration effects and by an application of total stress that would affect the HCT measurement due to parasite deformations. Things seem to be improved when getting closer from the zero suction state, above 400 kPa in the case of test B2 (see Fig. 4), the sample becoming less rigid due to suction release, with a better adaptation of the HCT/sample interface at higher stress.

A better response is observed on a less stiff sample once water is expelled, probably due to less stress concentration effect thanks to the stable positioning of the HCT at the contact of the sample. Nevertheless, the correspondence between the load applied and the HCT response is not perfect, with an overestimation of 100 kPa when applying the 1600 kPa load with a load increment of 800 kPa. on the contrarily, the coming back to zero of the HCT measurements once zero suction is reached is encouraging. It shows a standard pore pressure dissipation process and a decrease in water content of the sample with some water expelled in the top porous stone. The same remarks can be made on the unloading phases, with instantaneous coupled water pressure decreases observed. An instantaneous suction value of 700 kPa is observed when unloading from 1600 kPa to 800 kPa, to be compared to the load release (800 kPa).

Figures 5(a) and (b) respectively show the oedometer compression curves (e-{\sigma }_v) of Boom clay samples B1 to B3 on one hand and of Jossigny silt sample J2 on the other hand during loading and unloading. Note that the void ratio of Boom clay sample at σ_v of 50 kPa was calculated by taking into account the initial void ratio (Table 6) and the vertical displacement of the loading step of 50 kPa (see Fig. 4). Samples B1 and B2 were not unloaded to the initial loads because the HCT cavitated at a suction comprised between 100 to 200 kPa (much lower than the HCT capacity). This confirmed that the porous ceramic of the HCT could not support negative pressures during too long periods of time. A similar observation was also made by Chiu et al. [25] and Cui et al. [19].

Figure 6 presents the results from oedometer tests on Boom clay (see Figs. 6(a)-(c)) and Jossigny silt (see Fig. 6(d)) in terms of water pressure or suction variations versus vertical stress, once equilibrium was reached. These diagrams show how suction is reduced during loading and increased during unloading. In the case of Boom clay, the extrapolation from the points obtained during compression give an estimate of the value of the stress under which zero suction occurred. Considering Test B1 (see Fig. 6(a)), a slope ds/dσ_v= -0.56 can be estimated. Following this estimation, zero suction was reached at a vertical stress of 1200 kPa. During unloading, a slope ds/dσ_v= -0.59 was deduced, almost equal to that during loading. A final suction close to the initial one was observed.

Tests B2 and B3 have different behavior with respect to Test B1 due to the lower initial suctions. During loading, a slope ds/dσ_v= -0.45 was obtained between 50 to 400 kPa vertical stress for Test B2. Zero suction was reached at 460 kPa vertical stress, remaining zero at higher loads. During unloading, the suction remained zero from 1600 to 410 kPa vertical stress; it increased below 410 kPa vertical stress with a slope ds/dσ_v= -0.74. This slope value is larger than that obtained during loading. Similar behavior can be observed in Test B3 (see Fig. 6(c)), with a loading slope ds/dσ_v= -0.53 and zero suction reached at 590 kPa. A small suction increase of 40 kPa was recorded during the unloading path between 1600 to 430 kPa followed by a more significant suction increase with a slope ds/dσ_v = -0.80 below 430 kPa.

In Fig. 5, the points at which zero suction is reached are determined: σ_v = 1200 kPa (B1, s_i = 600 kPa), σ_v = 600 kPa (B3, s_i = 280 kPa) and σ_v = 460 kPa (B2, s_i = 180 kPa). One observes that zero suction is reached at smaller load when starting from a lower initial suction. This is consistent with the fact that the ds/dσ_v ratios are close in the three tests (-0.56, -0.53 and -0.45 for B1, B3 and B2 respectively), with a tendency to decrease when the soil is further hydrated at lower initial suction.

As commented before, the significance of the compression strain observed before this point has to be cautiously considered. Once zero suction is reached, water is expelled into the upper porous stone. The anti-evaporation system on the top of the oedometer cell minimizes the evaporation of the water contained in the stone.

The unloading sequences are conditioned by the limited amount of available water stocked in the upper porous stone. Indeed, due to water limitation and under the hypothesis of the samples remaining saturated during rebound, swelling is limited by a maximum void ratio corresponding to that at which zero suction appeared. Indeed, the three rebound curves exhibit limited swelling, with a good correspondence for Test B1. The correspondence remains satisfactory for Test B3, and is less satisfactory with Test B2, with a too limited swelling with respect to the amount of water available. Some water loss by evaporation could be the reason of this problem. The higher suction value obtained at the end of the test is consistent with this water deficiency.

The results of Test J2 carried out on Jossigny silt (see Fig. 5(b)) shows a standard volumetric elasto-plastic behavior. During loading, a pseudo-elastic behavior is observed till reaching the compaction stress measured during static compaction σ_comp = 360 kPa (see Fig. 5(b)). Significant plastic strain occurs at stress higher than the pre-compaction stress. A compression index C_c = 0.467 can be computed in the vertical load range from 400 to 1600 kPa. A linear relation is observed during unloading from 1600 to 25 kPa. It should be recalled here that, being Jossigny silt unsaturated, compression only occurs due to air expulsion.

In Fig. 6, the results of Tests B2 and B3 show an irreversible s-σ_v relationship during loading and unloading at low initial suctions, with final suctions higher than the initial ones. Note that in both tests, during the unloading path, the suction increases from zero at a vertical stress close to 400 kPa. Interestingly, this is the same vertical stress at which a slope change occurs on the void ratio-vertical stress curve, as shown in Fig. 5(a). Actually, this shows that the load below which no more water is available to hydrate the sample can be detected on the swelling curve. Note that, once no more water infiltrates the sample, there is still a slight rebound, showing that the sample under suction is not fully rigid, unlike undrained samples. As a matter of fact, the amount of this zero suction rebound is similar in the three curves and close to a void ratio increment ∆e = 0.02. This shows that, as compared to the undrained response in the positive range of pressure in which volume change are null, a slight deformability can be observed in a clay sample under suction, although no water exchange occurs. This comment, taken from the observation of the rebound curves, can account for some of the strain observed in the compression curves and commented above.

The results of Fig. 6 also show that the zero suction state is reached at lower load when starting from lower initial suctions: (B2) σ_v = 460 kPa for s_i = 180 kPa; (B3) σ_v = 600 kPa for s_i = 280 kPa ; (B1) σ_v = 1200 kPa for s_i = 600 kPa.

The change in suction of the unsaturated Jossigny silt with respect to changes in load (see Fig. 6(d)) is less pronounced with no zero suction state attained. Note that the first loading step shows an initial increase in suction related to an artifact due to the first adaptation of the HCT/sample interface (also observed in Tang et al. [26]). An almost reversible response can be observed in Test J1 along the loading-unloading cycle between 200 to 400 kPa vertical stress (see Fig. 6(d)). A slope ds/dσ_v = -0.13 was determined during the loading path from 200 to 1600 kPa. A complete unloading path (1600-25 kPa) could not be performed due to the occurrence of tensiometer cavitation. A reversible process and linear relationship with a slope ds/dσ_v = -0.11 was obtained in Test J2 during loading and unloading.

To further investigate the effect of water on the volumetric behavior, two complementary experiments were carried out on Boom clay (B4, s_i = 310 kPa) and Jossigny silt (J3, s_i = 328 kPa) in which soil samples were soaked under the highest load (1600 kPa). The unloading path began once the water pressure and the corresponding vertical displacement were stable.

Figure 7 shows the void ratio change as a function of vertical stress in an e-{\sigma }_{\mathrm{v}} plot for Boom clay and Jossigny silt (see Figs. 7(a) and (b), respectively). The volume change behavior of the soils can be divided into three stages:

(1) During loading on Boom clay, the compression curves of samples B3 and B4 are similar because of close values of initial suctions. Note that both tests B3 and B4 exhibited significant strain before reaching zero suction. The e-{\sigma }_v curves obtained with the Jossigny silt samples in Tests J2 and J3 are almost identical, with slopes C_c equal to 0.467 and 0.529, respectively.

(2) After a soaking period of 48 h, the void ratio of Boom clay sample remained unchanged, confirming that the soil sample had been completely saturated under 1600 kPa. On the contrary, the Jossigny silt sample exhibited a slight collapse, with a volumetric strain of 1.45%. This collapse is typical of unsaturated soils when soaked under a sufficiently high vertical load (see for instance Alonso et al. [27]).

(3) During unloading, the swelling curves of Boom clay samples B3 and B4 (see Fig. 7(a)) between 1600 and 400 kPa are parallel, they correspond to the sample re-hydration by water. As seen before, there is no more water available for sample B3 below 400 kPa, explaining the significant difference observed with sample B4 that goes on swelling due to the infiltration of available water. The high swelling index C_s = 0.08 measured on Test B4 (on the unloading path from 800 to 50 kPa) is typical of swelling soils, like the increase in slope observed below 50 kPa that finally reaches a void ratio value close to the initial one.

In the Jossigny silt, a higher rebound is observed on the soaked sample (J3, C_s = 0.01), as compared to sample J2 unloaded at constant water content (C_s = 0.005). This illustrates a slightly enhanced swelling tendency when inundating a slightly denser sample.

Figure 8 presents the suction/vertical stress relation along the load-soaking-unload path (Tests B4 and J3) and the load-unload path (Tests B3 and J2). For both soils, the results obtained from the two samples with comparable initial suction values are presented.

Figure 8(a) shows the s-{\sigma }_v curve for Boom clay during loading path before soaking. Similar slopes ds/dσ_v = -0.53 and ds/dσ_v = -0.45 are obtained from Tests B3 and B4, respectively. Suction values equal to zero were reached at 600 kPa vertical load in Test B3 and at 800 kPa in Test B4. This observation demonstrates a good repeatability of the s - σ_v results. During the unloading path, a quite different behavior can be observed on the s - σ_v curve, since Sample B4 was soaked.

Figure 8(b) presents the s - σ_v curve of Jossigny silt during the loading and unloading path. Identical slopes ds/dσ_v = -0.11 were determined for Tests J2 and J3 before soaking. An almost reversible behavior was observed during the unloading path in test J2 (no soaking). Sample J3 shows that after soaking the suction decreased from 170 kPa to zero.

Discussion

To further investigate the hydro-mechanical couplings occurring in Boom clay samples during oedometer compression, a comparison has been made with constant water content oedometer compression tests in compacted Jossigny silt. The significantly different results obtained in terms of stress/suction couplings are due to the great difference between the two soils in terms of microstructure. As observed by Dehandschutter et al. [14], Boom clay microstructure is of a matrix type, i.e., the proportion of clay particle is large enough to form a matrix in which other minerals are embedded. The clay matrix is also characterized by an apparent bedding of the clay particles that is typical of stiff clays. The sub-horizontal bedding and the clay particle reorientation is due to the magnitude of the stress applied, significant in the case of deep Boom clay samples excavated at a depth of 223 m. In such a configuration and in the absence of any significant capillary effect due to the fact that Boom clay is saturated, suction is caused by the physico-chemical adsorption of water molecules along clay particles. Among the clay particle present in Boom clay, smectites have a predominant role due to their larger physico-chemical activity. At the macroscopic scale, this activity is quantified by the soil plasticity (I_p = 40-50 with 50% to 60% of the particles smaller than 2μm).

In such an oriented clay matrix microstructure, hydration occurs by the infiltration of water molecules along some of the bedding planes, leading to the physico-chemical adsorption of some water molecules along the surface of the more active clay particles. During unloading, the swelling observed under a given stress is an indication that the level of attraction exerted by the clay particles on the water molecules is high enough to allow, under the applied stress, water infiltration in the inter-particle porosity so as to separate the particles and induce some swelling. In other words, the clay-water attraction is strong enough to overcome the mechanical stress that tend to bring the particles together and to reduce the pore space. Observation of Fig. 7 shows that a significant swelling can occur under a low vertical stress. Under 25 kPa, the B4 sample came back and almost recovered its initial void ratio.

The stress/suction hydro-mechanical coupling described above is quantified by the ds/dσ_v parameters investigated in Fig. 6. This parameter has the same value during the loading and unloading sequences of Test B1 at the higher initial suction (s = 600 kPa), giving a final suction close to the initial one around 600 kPa. At lower suctions, the ds/dσ_v value appears to be larger during unloading (-0.74 and -0.80 for tests B2 and B3 respectively), giving rise to final suctions higher than the initial ones. In other words, the same stress decrease induces a larger suction increase, showing a change in the hydro-mechanical coupling with larger suction changes during unloading.

On the contrary, the compacted Jossigny silt is known to have an aggregate microstructure typical of samples compacted dry of optimum [28,29]. The aggregates have a diameter included between 10 and 40 μm and the inter-aggregates pores have a mean entrance diameter of 1-10 μm. In aggregate microstructure, compression occurs by affecting the inter-aggregates links and by reorganizing the aggregates in a denser microstructure with smaller inter-aggregate porosity. The changes in suction of compacted samples have recently been investigated by Tarantino and De Col [30] who also showed that significant changes in stress (from 300 to 1200 kPa) induced comparably small suction changes (from 770 to 660 kPa respectively). Thus, the tiny suction changes observed during the compression of dry compacted soils is linked to the aggregate microstructure. Being the inter-aggregate pores dry in specimens compacted dry of Optimum, suction is mainly controlled by adsorption and capillary effects inside the aggregates. As seen above, since compression does not significantly change the morphology of the aggregates in terms of porosity, there is no significant change in suction during compression. Similarly, unloading mainly induces a release of the inter-aggregate contact with little influence inside the aggregates, leading to tiny suction changes, reasonably reversible according to Fig. 6.

Conclusions

Oedometer compression tests carried out in a cell equipped with a high capacity tensiometer for suction monitoring were performed on hydrated Boom clay samples so as to investigate the hydromechanical couplings in Boom clay, based on a comparison with similar tests run on samples that were chosen quite different in nature, i.e., samples of compacted unsaturated low plasticity silt (Jossigny silt). Soaking sequences were also carried out in some tests before unloading.

A detailed examination of the experimental data revealed that the measurements of suction changes during oedometer compression are not easy, particularly in the area where hydrated Boom clay samples are under suction state. It seems that the dynamic effort corresponding to step loading induces some instantaneous parasite effects on the response of the high capacity tensiometer, followed by reasonable suction equilibration. In the positive range, the tensiometer provides satisfactory response in terms of pore pressure dissipation measurements. In Boom clay samples, the experiments provided the description of the transition from suction to positive pressure and the estimation of the amount of expelled water that afterwards controlled the maximum possible swelling.

It appeared that the stress/suction responses of the two soils are completely different in nature, due to significant difference in terms of microstructure. As compared to the matrix microstructure of Boom clay in which a predominant clay fraction is embedding the other soil components, the compacted Jossigny silt is characterized by an aggregate microstructure that is known to exhibit slighter changes in suction during compression, being suction changes governed by phenomena that occur inside aggregates that are not significantly affected by the compression. Whereas capillarity is likely to play an important role in the unsaturated compacted silt, suction in Boom clay samples close to saturation or saturated is mainly governed by physico-chemical clay water interactions, with a predominant role of the smectite phase, estimated equal to 33% in the literature. During oedometer compression with pore water under suction – a stress path apparently not often investigated experimentally – the increase in density results in a decrease of the inter-layer spaces in the clay matrix that results in a suction decrease. In other words, the hydro-mechanical coupling is operating by transferring the mechanical energy applied during the compression step to the change in water potential that is quantified by the suction decrease. When suction is afterwards released, the mobilization of the clay-water physico-chemical interactions induce some swelling, depending of the stress applied. It is likely that the hydro-mechanical couplings acting along the two hydric paths investigated (compression with suction decrease and swelling with suction increase) are of a similar nature, as indicated by the same slopes observed in Fig. 6(a), at least with the higher initial suction (sample B1, s_i = 600 kPa). The difference in slopes observed at lower initial suctions (samples B2 and B3, s_i = 180 and 280 kPa respectively) could be an effect of higher microstructure perturbation due to stronger initial hydration and swelling.

The exploration allowed by the first experimental results obtained on Boom clay along provided a series of preliminary conclusions. Obviously, it also raised some important questions that certainly deserve further investigations with improved suction monitoring.

References

[1] ONDRAF/NIRAS. Aperçu technique du rapport SAFIR 2. Safety assessment and feasibility interim report 2. Publication NIROND 2001–05 F, 2001, 280 (in French)

[2] Sultan N, Cui Y J, Delage P. Swelling properties of Boom clay. In: Proceedings of IS-TOHOKU’98, International Symposium on Problematic Soils. Sendai, 1998, 2: 253–256

[3] Gens A, Alonso E E. A framework for the behavior of unsaturated expansive clays. Canadian Geotechnical Journal, 1992, 29(6): 1013–1032

[4] Tang G X, Graham J, Blatz J, Malcom G, Rajapakse R K N D. Suctions, stresses and strengths in unsaturated sand-bentonite. Engineering Geology, 2002, 64(2–3): 147–156

[5] Caruso M, Tarantino A. A shear box for testing unsaturated soils at medium to high degrees of saturation. Géotechnique, 2004, 54(4): 281–284

[6] Rahardjo H, Fredlund D G. K₀-volume characteristics of an unsaturated soil with respect to various loading paths. Geotechnical Testing Journal, 2003, 26(1): 79–91

[7] Ridley A M, Dineen K, Burland J B, Vaughan P R. Soil matrix suction: some examples of its measurement and application in geotechnical engineering. Géotechnique, 2003, 53(2): 241–253

[8] Chandler R J, Gutierrez C I. The filter-paper method of suction measurement. Géotechnique, 1986, 36(2): 265–268

[9] Skempton A W, Sowa V A. The behavior of saturated clays during sampling and testing. Géotechnique, 1963, 13(4): 269–290

[10] Doran I G, Sivakumar V, Graham J, Johnson A. Estimation of in-situ stresses using anisotropic elasticity and suction measurements. Géotechnique, 2000, 50(2): 189–196

[11] Decleer J, Viane A, Vandenberghe N. Mineralogical characteristics of the Rupelian Boom clay. Clay Minerals, 1983, 18(1): 1–10

[12] Al Mukhtar M, Belanteur N, Tessier D, Vanapalli S K. The fabric of a clay soil under controlled mechanical and hydraulical stress states. Applied Clay Science, 1996, 11(2–4): 99–115

[13] Belanteur N, Tacherifet S, Pakzad M. Étude des comportements mécanique, thermo-mécanique et hydro-mécanique des argiles gonflantes et non gonflantes fortement compactées. Revue Française de Géotechnique, 1997, 78: 31–50 (in French)

[14] Dehandschutter B, Vandycke S, Sintubin M, Vandenberghe N, Wouters L. Britlle fractures and ductile shear bands in argillaceous sediments: inferences from Oligocen Boom Clay (Belgium). Journal of Structural Geology, 2005, 27(6): 1095–1112

[15] Delage P, Le T T, Tang A M, Cui Y J, Li X L. Suction effects in deep Boom clays samples. Géotechnique, 2007, 57(2): 239–244

[16] Decraen L, Wemaere I, Labat S, van Geet M. Geochemical analyses of Boom Clay pore water and underlying aquifiers in the Essen-1 borehole. External report of the Belgian Nuclear Research Centre, Belgium. SCK-CEN-ER, 2004, 19: 24

[17] Cui Y J, Delage P. Yielding and plastic behavior of an unsaturated compacted silt. Géotechnique, 1996, 46(2): 291–311

[18] Mantho A T. Echanges sol-atmosphère application à la sécheresse. Dissertation for the Doctoral Degree. Paris: Ecole Nationale des Ponts et Chaussées, 2005 (in French)

[19] Cui Y J, Tang A M, Mantho A T, de Laure E. Monitoring field soil suction using a miniature tensiometer. Geotechnical Testing Journal, 2008, 31(1): 95–100

[20] Ridley A M, Burland J B. A new instrument for the measurement of soil moisture suction. Géotechnique, 1993, 43(2): 321–324

[21] Tarantino A. Direct measurement of soil water tension. In: Juca J F T, de Campos T M P, Marinho F A M, eds. Proceedings of the 3rd International Conference on Unsaturated Soils. Recife: Balkema, 2004, 3: 1005–1017

[22] Delage P, Romero E, Tarantino S. Recent developments in the techniques of controlling and measuring suction in unsaturated soils. In: Toll D G, Augarde C E, Gallipolli D et al, eds. Proceedings of the 1st European Conference on Unsaturated Soils. Durham: CRC Press, 2008, 33–52

[23] Dineen K, Burland J B. A new approach to osmotically controlled oedometer testing. In: Alonso E E, Delage P, eds. Proceedings of the 1st International Conference on Unsaturated Soil. Balkema, 1995, 2: 459–465

[24] Marinho F A M, Take A, Tarantino A. Tensiometric and axis translation techniques for suction measurement. Geotechnical and Geological Engineering, 2008, (In press)

[25] Chiu C F, Cui Y J, Delage P, Haza E, de Laure E. Lessons learnt from suction monitoring during centrifuge modeling. In: Proceedings of the International Symposium on Advanced Experimental Unsaturated Soil Mechanics. Trento, 2005, 3–8

[26] Tang A M, Cui Y J, Barnel N. Compression-induced suction change in a compacted expansive clay. In: Toll D G, Augarde C E, Gallipolli D, et al, eds. Proceeding of the 1st European Conference on Unsaturated Soils E-UNSAT. Durham: CRC Press, 2008, 369–374

[27] Alonso E E, Gens A, Hight D W. Special problem soils. In: Proceedings of Comptes Rendus de la 9ème Conference Européenne de Mécanique des Sols et des Travaux de Fondations, Dublin, 1987, 3: 1087–1146

[28] Ahmed S, Lovell C W, Diamond S. Pore size and strength of compacted clay. Journal of the Geotechnical Engineering Division, 1974, 100(GT4): 407–425

[29] Delage P, Audiguier M, Cui Y J, Howat M D. Microstructure of compacted silt. Canadian Geotechnical Journal, 1996, 33(1): 150–158

[30] Tarantino A, de Col E. Compaction behaviour of clay. Géotechnique, 2008, 58(3): 199–213

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg