Suction in shales: consequences on triaxial testing

. Significant interest has been devoted to claystones and shales in the context of geological radioactive waste disposal at great depth. The determination of their mechanical properties is needed for appropriate design of the underground galleries and tunnels. Given their high sensitivity to changes in water content, special care has to be taken so as to provide characteristics as close as possible to the (saturated) in-situ ones. Based on the hydromechanical path followed by samples from coring to trimming in the lab, that most often lead to some degree of desaturation resulting from evaporation and drying of the samples, some procedures aimed at minimising the resulting perturbations are described. The considerations presented are based on data obtained on two (swelling) claystones considered in Europe for deep geological disposal, i.e., the Callovo-Oxfordian claystone (France) and the Opalinus Clay (Switzerland).


Introduction
Due to their very low permeability and high capacity to retain radionuclides, claystones are considered as suitable host rocks for deep geological disposal of high activity radioactive waste in various countries including France (in the Callovo-Oxfordian claystone -COx) and Switzerland (in the Opalinus Clay). Intensive laboratory investigation has hence been conducted on both claystones so as to characterise their thermo-hydromechanical behaviour, based on isotropic, oedometer or triaxial compression tests [1 -6].
Claystone samples are most often water saturated and extracted and trimmed from cores excavated at great depth (500 m for the Underground Research Laboratory -URL -run by Andra, the French Agency for the management of radioactive wastes, in Bure). Given their sensitivity to changes in water content, obtaining responses that are reasonably representative of the insitu saturated claystone behaviour is a challenge that requires special care, that is described in more detail in this paper.

Materials
The investigation was carried out on both the Callovo-Oxfordian claystone (COx) and Opalinus Clay. Their main characteristics are presented in Table 1. Interestingly, both claystones have comparable characteristics.

Water retention properties
It is known, since Skempton and Sowa [9], that saturated clay samples extracted at a given depth are not submitted to a stress release equal to the effective stress supported at that depth. In the case of "ideal sampling", it is admitted that, provided the sample remains saturated during extraction, the suction induced due to effective stress release should exert on the sample, provided it remains saturated, an equivalent effective stress. This is possible if the effective stress is smaller than the air entry value (AEV) of the sample. This was for instance observed in Boom clay (a stiff clay considered for deep geological radioactive waste disposal in Belgium) by Delage et al. [10]. Fig. 1 [11] shows the water retention curve of a sample of a COx sample cored at a depth of 500 m in the Bure URL. The curve has been obtained by using vapour equilibrium with saturated salt solutions (Table 1). The curve in Fig. 1a shows that the initial point has a suction of 18 MPa, corresponding to a degree of saturation Sr = 92%. From this initial point, the sample has been submitted to decreased suction down to 0.1 MPa along a drying path, close to saturation. A subsequent wetting path was then followed up to a maximum suction of 150 MPa, followed by a drying path back to saturation.
The data show that the initial state of the trimmed sample in the laboratory was not fully saturated, due to the successive and cumulative effects of (air) coring, storage, transportation, core cutting and sample trimming in the lab [12,13]. The shape of the curve also shows that the AEV of the COx sample is around 10 MPa. A significant hysteresis is observed. Observation of the curve of Fig. 1b) shows that significant volume changes occur upon wetting, with a volume increase around 6%. This swelling behaviour of the COx claystone, also observed on the Opalinus Clay, is well known. It is due to the presence in both claystones of a significant proportion of montmorillonite, in the mixed layer illitemontmorillonite that constitute the clay matrix, within which the detritic grains of calcite and quartz are embedded (see Table 1 and Fig. 2).

Sensitivity of shales to changes in water content
As seen in the SEM photo of Fig. 2 [15,16], that was taken on a horizontal sample hydrated under zero suction and no stress, sub-horizontal swelling microcracks (lengths of dozens of µm and thickness of around 1 µm) are clearly apparent within the clay matrix. These micro-cracks would obviously result in some damage of the swollen claystone.
Some pyrite crystals can also be seen in the photo, together with the footprints of some detritic grains that have been pulled out during the freeze-fracturing set up of the observed plane.
. ters deterer content, ing suction changes in n ( Fig. 9 b), Fig. 9 c) and rain versus according to a bilinear function. Along the drying path, at water contents smaller than the initial one (w i ¼ 7.5%), the porosity decrease is of 0.2% for a water content increment of 1%, whereas the corresponding porosity increase along the wetting path (w4 w i ) is of 1.2%. These slopes are similar for the various drying and wetting paths followed along the cycles. This change in slope is not evident in a porosity versus logarithm of suction diagram ( Fig. 9 b). A decrease in porosity of approximately 2% is observed when suction is increased ten times. Fig. 9 b also shows that, once the sample dried (respectively wetted) at the largest suction (respectively smallest), the subsequent wetting (respectively drying) path is located above the previous drying (respectively volumetric behaviour obtained by the suction controlled method.
A. Initial state B. s = 0.1 MPa C. s = 150 MPa imen state parameters deterstate were the water content, uration for tests using suction shown in terms of changes in a) and versus suction ( Fig. 9 b), sus water content ( Fig. 9 c) and es in volumetric strain versus decrease is of 0.2% for a water content increment of 1%, whereas the corresponding porosity increase along the wetting path (w4 w i ) is of 1.2%. These slopes are similar for the various drying and wetting paths followed along the cycles. This change in slope is not evident in a porosity versus logarithm of suction diagram ( Fig. 9 b). A decrease in porosity of approximately 2% is observed when suction is increased ten times. Fig. 9 b also shows that, once the sample dried (respectively wetted) at the largest suction (respectively smallest), the subsequent wetting (respectively drying) path is located above the previous drying (respectively ater retention curve and volumetric behaviour obtained by the suction controlled method.  [15,16]. Together with swelling along wetting paths, claystones are also significantly sensitive to drying, that results in a significant increase in strength. This is shown in Fig. 3 with the data of unconfined compression tests run by Pham et al. [17] on COx samples submitted to various relative humidity between 98% (suction s = 2.5 MPa with a water content w = 5.24%, close to saturation according to  As a consequence, preventing as much as possible tested samples from evaporation starting from coring to trimming in the lab is of outmost importance. In this regard, the measurement of their degree of saturation is mandatory. Suction measurements, for instance by using a chilled mirror tensiometer (e.g., WP4 Decagon) are even better, due the high sensibility of suction to changes in Sr.

Hydration phase
Given the swelling behaviour observed in Fig. 1b, the first rule to follow when testing a swelling claystone (either in oedometer, isotropic or triaxial compression) is to avoid any contact with water as long as the sample is not submitted to in-situ stress conditions. To do so, the sample has to be placed in contact with dry porous disks until the sample is under suitable stress conditions. Then, air has to be expelled by applying vacuum to the drainage system, prior to saturate the drainage lines with de-aired water [12]. Note, as initially shown by Mohajerani et al. in the oedometer [18] or Monfared et al. [12] in the triaxial apparatus, some swelling may be observed, even under in-situ stress conditions. This trend has been further investigated by Delage and Belmokhtar [19], who characterised the changes in swelling of the Opalinus Clay (shallow sample from Lausen extracted at a depth of 30 m) under hydration with respect to the isotropic applied stress (Fig. 4).  [19].
The curve clearly shows that significant swelling occurs under 1 MPa (the in-situ stress) and that an effective isotropic stress of 5 MPa is necessary to reduce swelling below 0.5%. No significant swelling is observed under 15 MPa. This shows that the properties of the sample at low stresses (below 5 MPa) may be significantly affected by damage.  [19] shows the stress strain curves obtained under a confining stress of 10 MPa on a sample (38 mm wide, 76 mm high) with bedding perpendicular to the axis (contraction is positive). Strains were locally monitored by using strain gauges directly stuck on the sample at mid-level. The curves evidence some degree of anisotropy, with local radial strains significantly smaller than axial ones, a feature typical of shales. A significant difference is observed between local (green squares) and external (pink triangles) axial strains, confirming the  displacement rate 0.3 lm/min, slow enough to ensure good drainage of the specimen with a 19 mm drainage length [4]. Results are shown in Fig. 7 in terms of changes in shear stress (q = r 1 -r 3 ) with respect to axial strain (r 1 ), radial strain (r 3 ) and volumetric strain (e 1 ? 2e 3 ). The larger axial strains measured from the external LVDT are also plotted. The numbering from a to h is done with respect to the applied confining stress, in order to compare more easily similar tests. Looking at Fig. 7a and b, one can observe a good repeatability of tests HP3-1 and HP3-2 carried out under an in situ Terzaghi effective stress r ' of 1 MPa (r ' = r -u w , with r = 1.3 MPa, u w = 0.3 MPa), with peak strengths at 8.6 MPa and 9.1 MPa, respectively. The peak strength at 5.4 MPa in test HP3-7 (Fig. 7c) under in situ effective stress is lower than that of HP3-1 and HP3-2. Young's moduli, measured in all tests from first loading at an axial strain of 0.1%, are more or less comparable, with values between 820 and 960 MPa (see Table 5).
All Young's moduli are plotted together in Fig. 9 with respect to the effective confining stress, showing a linear increase with effective stress, and a good comparability between tests carried out at same confining stresses. need of local strain measurements, particularly for determining the Young modulus (that may be significantly underestimated when measured from external strain measurements). The volume change curve first exhibits a contracting phase, followed by a dilating one, with a transition at 13 MPa.    [19].
The interest of undrained tests in shales carried out following Ewy et al. [22] approach is that samples keep their initial volume and are not affected by any swelling due to previous hydration (see [22] for more details). Fig. 6 shows quite a good agreement between Giger et al. [20]'s undrained tests and our drained data for p'> 10 MPa. For this stress, Fig. 4 shows that the swelling under hydration is equal to 0.2%. Our drained data at 5 MPa, with a swelling of 0.4%, exhibit a lower failure shear stress than that obtained through Giger's undrained tests, with an even larger under-estimation at lower stress. In spite of their different origin, the same observation apparently holds for the data obtained by Favero et al. [5] and Wild and Amman [21] on samples from the shaly facies of Mt Terri (these undrained tests were apparently not conducted following Ewy's approach [22]). As a consequence, it seems that the shape of the failure criterion at low stresses is not easy to determine, and that undrained tests at lower stress would be necessary to check whether the criterion keeps being linear.

Concluding remarks
The sensitivity to changes in water content (and suction) of some claystones considered as possible host rock for the deep geological disposal of high activity radioactive waste in France (Callovo-Oxfordian claystone) and Switzerland (Opalinus Clay) has been taken into account to optimise the experimental characterisation of their mechanical properties (either through oedometric, isotropic or triaxial compression). Claystones are saturated in their in-situ state, but the successive operations of coring, storage, transport and lab trimming most often result in some degree of desaturation. Because of their sensitivity to changes in water content, special care has to be taken for mechanical testing, because the suction release occurring when putting the sample in contact with water in the cell should be compensated by applying a sufficiently high stress (that can be higher than the in-situ one). Based on this consideration, some data on the mechanical properties of these claystones have been discussed, focussing on possible under-estimation of their properties under smaller confining stresses.
The work presented here has been supported by Andra (France) and Nagra (Switzerland) through various PhD thesis. The author is indebted to Dr. N. Conil and J. Talandier (Andra) and Dr. S. Giger (Nagra) for these supports. Thanks also to former PhD students, Dr. H. Menaceur and M. Belmokhtar, for the excellent work carried out during their PhD thesis. pecimen volume changes to the hydration carried out in he laboratory under (much lower) in situ stress suggests hat the long-term geological uplift and unloading of the palinus Clay layer did not significantly affect the echanical characteristics of the Opalinus Clay [16,34]. In ther words, Opalinus Clay did not swell significantly during the significant, but very long, unloading sequence from an estimated 1000 m depth to around 30 m, as also confirmed by the value of the bulk wet density of 2.47 Mg/ m 3 , comparable to that of specimens from greater depths (e.g. [11]). This means that the diagenetic bonds resulting from the long-term supported high stresses and ig. 12 Comparison of the peak stresses from the present work (drained paths in thick red lines, strain rate of 6.6 9 10 -8 s -1 ) with the data of iger et al.