The response of Opalinus Clay when exposed to cyclic relative humidity variations

Clay shale specimens were exposed to cyclic relative humidity variations to investigate the response of the material to natural environmental changes. Opalinus Clay, a clay shale chosen as host rock for nuclear waste disposal in Switzerland, was utilized. The specimens were exposed to stepwise relative humidity cycles where they were alternately allowed to equilibrate at 66 and 93% relative humidity. Principal strains were monitored throughout the experiments using strain gauges. After each relative humidity cycle, Brazilian tensile strength tests were performed to identify possible changes in tensile 10 strength due to environmental degradation. Results showed that Opalinus Clay follows a cyclic swelling-shrinkage behaviour with irreversible expansion limited to the direction normal to bedding, suggesting that internal damage is restricted along the bedding planes. The Brazilian tensile strength in direction parallel and normal to bedding as well as the water retention characteristic remained unaffected by the RH variations. 15

and shrinkage of Opalinus Clay. Möri et al. (2010) also observed a net closure of the cracks over several seasonal cycles, which indicates an irreversible deformation component that is likely associated with time-dependent processes such as consolidation, creep or slaking. These irreversible deformation components can contribute to both long-term tunnel convergence and selfsealing of the EDZ. The self-sealing effect is the ability of clay shales to close previously developed cracks and therefore reducing their permeability by hydro-mechanical, hydro-chemical, and/or hydro-biochemical processes (Bernier et al., 2007). 5 Among others, the adsorption of water on clay minerals and related volumetric expansion can be associated with this effect.
Numerous studies have been conducted to show the influence of drying/wetting cycles on clay or clay shale specimens (e.g. Chu and Mou, 1973;Popescu, 1980;Chen and Ma, 1987;Osipov et al., 1987;Dif and Bluemel, 1991;Day, 1994;Al-Homoud et al. 1995;Basma et al. 1996;Pejon and Zuquette, 2002). In those studies, however, swelling has been performed by allowing the specimens to fully soak in water. Few studies exist where the influence of cycles in RH on the drying/swelling 10 characteristics of clay shales has been investigated (e.g. Grice, 1968;Van Eeckhout, 1976;Olivier, 1979;Farulla et al., 2010;Cardoso et al., 2011;Pineda et al., 2014). Grice (1968) noted that specimens of Utica shale that were immersed in water disintegrated completely after oven drying. Specimens that were exposed to RH fluctuations between 60 and 90% for a period of 9 months, however, showed only minor cracking. Van Eeckhout (1976) equilibrated specimens of Beatrice coalmine shale to various levels of RH to study the mechanisms of reduction in rock strength resulting from variations in RH. During moisture 15 absorption he measured a volumetric expansion in the order of 0.2-1%. The strains were larger in direction normal to bedding and occurred mostly between 48 and 100% RH. Subsequent drying of the specimens to the initial level of RH showed that about 0.25% of the strains were not recoverable. Van Eeckhout (1976) identified these expansion-contraction characteristics and the associated lengthening in internal cracking as possible cause for the lowering in strength he observed due to humidity fluctuations. Similar observations have been made by Olivier (1979) for specimens of a Lower Triassic mudrock. With the 20 help of water retention curves for several wetting/drying cycles between 10 and 99% RH, Cardoso et al (2011) showed that the air entry value of an Upper Jurassic marl decreases with increasing number of cycles. The decrease is accompanied by an increase in void ratio indicating a degradation of the material. Pineda et al. (2014) experimentally investigated the influence of RH cycles on the degradation of Lilla claystone in a long-term RH cycling experiment using ultrasonic wave velocity measurements and Brazilian tensile strength tests. The applied RH cycles caused an irreversible increase in the specimens ' 25 volumes as swelling always exceeded the amount of shrinkage. Pineda et al. (2014) found that higher peak-to-peak amplitudes in RH cycles (cycles between 20 and 99% have been compared to cycles between 50 and 99%) lead to larger volumetric swelling. This effect was less pronounced for specimens that were tested under higher confinements. With the help of microstructural analyses, fissuring has been identified as the main cause for irreversible swelling for Lilla claystone.
Furthermore, the degradation of the material was manifested in a decrease of tensile strength from 2.9 MPa to 0.2 MPa after 30 four cycles and a decrease in dynamic Young's modulus by more than 50%. For both quantities, the reduction was largest for the first cycle, afterwards a decreasing degradation rate has been observed. Earth Discuss., doi:10.5194/se-2016-171, 2016 Manuscript under review for journal Solid Earth Published: 14 December 2016 c Author(s) 2016. CC-BY 3.0 License.

Solid
All studies mentioned above showed that cyclic variations in RH can have a significant influence on rock mechanical parameters such as tensile strength and can lead to irreversible volume changes which might contribute to the destabilisation of underground excavations but also favour processes that are considered to control self-sealing.
This study aims at contributing to the understanding of the influence of RH variations on the mechanical and hydro-mechanical behaviour of Opalinus Clay. A series of specimens were exposed to RH variations under unstressed conditions and tested for 5 their Brazilian tensile strength. The study focuses on answering the question if Opalinus Clay shows a damage evolution when exposed to RH cycles that affects the tensile strength and causes irreversible volumetric expansion that might be relevant for long-term deformations and/or self-sealing.
2 Tested material and experimental procedure

Sampling and specimen preparation
31 specimens were taken from two 67.5 mm diameter bore cores obtained from a 25 m long borehole (BHM-1) that was drilled in Gallery 08 in the shaly facies of Opalinus Clay at the Mont Terri URL (Fig. 1). A triple-tube core barrel with compressed air cooling was used. The bore axis of BHM-1 was oriented parallel to the bedding. The samples were immediately sealed in 25 vacuum-evacuated aluminium foil after core extraction. Core samples were cut under dry conditions to a diameter to length ratio of approximately 2:1. Additionally, a cubic specimen (S4) was cut in such a way as to allow for the measurement of the principal strains (i.e. the strains perpendicular (ε1) and parallel (ε2) to the bedding plane orientation, and the strain in the plane of isotropy (ε3)). The layout of the electronic strain gauges (HBM type: K-LY41-2/120) that were directly glued on the specimen's faces is shown in Fig. 2a. Furthermore, one cylindrical specimen (E19) was used to measure the strain 30 Solid Earth Discuss., doi: 10.5194/se-2016-171, 2016 Manuscript under review for journal Solid Earth Published: 14 December 2016 c Author(s) 2016. CC-BY 3.0 License. perpendicular and parallel to bedding (Fig. 2b). The environmental exposure time of the specimens during the installation of the strain gauges was minimized to about 30 minutes.

Experimental layout
The specimens were exposed in desiccators to an alternating sequence of low and high RH levels under unstressed conditions. The maximum and minimum level of RH chosen for the laboratory experiment (i.e. 66 and 93% respectively) are based on the 5 seasonal variations at the Mont Terri URL between 1997 and 2011 reported by Swisstopo (2014). It was found that the RH follows a cyclic annual variation following a sine curve with maximum humidity values of 93.8% during summer and minimum humidity values of 62.6% during winter. The RH was controlled by using supersaturated salt solutions. Sodium nitrite (NaNO2, RH = 66% at 20°C) and ammonium di-hydrogen phosphate (NH4H2PO4, RH = 93% at 20°C) were used. At each level of RH, the specimens were allowed to equilibrate with the environment. Equilibration was achieved when the weight of the specimens 10 (periodically measured) remained constant. Brazilian tensile strength tests were conducted after each cycle when the specimens were equilibrated at 66% RH. In total 4.5 cycles were applied, starting with an equilibration phase at 93% RH to establish the same initial conditions for all specimens. The experimental setup is schematically shown in Fig. 3. For the monitoring of the RH, a Honeywell HIH-4000-001 sensor (accuracy ±3.5 %) was used. Temperature was monitored by a resistance thermometer (Pt100). 15

Suction and strain calculations
The water content was determined according to the ISRM suggested methods (ISRM, 1979). From the RH and the temperature that were monitored during the experiment, the total suction can be calculated according to Kelvin's relationship: where ψ is the suction in Pa, R is the ideal gas constant (i.e. 8.314 J/mol/K), T the absolute temperature in Kelvin, Vwo the 20 specific volume of water (i.e. about 0.001 m 3 /kg), ωw the molecular mass of water vapour (i.e. 0.018 kg/mol), p the vapour pressure of water in the system in MPa, and p0 the vapour pressure of pure water in MPa. The ratio p/p0 equals the RH.
On the cubic specimen (S4), strains were measured in all three principal directions. Thus, the volumetric strain (εv) can be calculated by adding all three principal strains (i.e. εv = ε1 + ε2 + ε3). On the cylindrical specimen (E19), the strain parallel (ε2) and perpendicular (ε1) to the bedding was recorded. Assuming a transversely isotropic material, the strain parallel to bedding 25 (ε2) equals the strain in the plane of isotropy (ε3). Hence, the volumetric strain can be calculated from the sum of the strain perpendicular to bedding and twice the strain parallel to bedding (i.e. εv = ε1 + 2ε2). For all strains, expansion is taken as positive.

Mechanical testing procedure
Brazilian tensile strength tests were conducted at ETH Zurich utilizing a modified 2000 kN servo-hydraulic rock testing machine (Walter and Bai, Switzerland). The tests were conducted according to the ISRM suggested methods (ISRM, 1978) immediately after removal from the desiccator. Load was applied parallel (A-specimens) or normal to bedding (E-specimens) (Fig. 4) using a constant loading rate between 0.05 and 0.08 kN/s. 5

Results
The specimens' dimensions and initial properties are given in Table 1. The initial water contents of the specimens range between 6.95 and 7.34 %, which is comparable with the water content measured on cores right after core extraction (Pearson et al. 2003). The initial saturation was estimated from the initial water content, the bulk dry density, and the porosity of the specimens according to the ISRM suggested methods (ISRM 1979). A grain density 10 of 2.73 g/cm 3 was used to calculate the porosity (Pearson et al., 2003;Bossart, 2005; own data). Values of saturation that exceed 100% can be related to the uncertainty in the grain density (± 0.03 g/cm 3 ) and the specimen's volume. Fig. 5 shows the results of the RH, temperature and strain measurements for the E-specimens. The specimens were first equilibrated to a RH of 93% and then subjected to 4.5 cycles with peak-peak amplitudes between 30 and 36% (i.e. RH variation between 63% and 94%; Fig. 5a). The temperature was kept between 19 and 23°C throughout the experiment. The resulting 15 suction applied to the specimens was calculated according to Eq. (1) and is plotted together with the corresponding response of the water content in Fig. 5b. Similar trends with respect to the water content changes were observed for all specimens. A constant water content and a small change in volumetric strain were observed during the first equilibration phase indicating a high initial saturation degree of the specimens. During cycling, the water content changed by ±2.2-2.4%. Except for the first drying period where 0.6-0.8% water content was lost, the water content was reversible. A comparable response was observed 20 for the series of A-specimens (Fig. 6) which were subjected to the same testing procedure, although RH or strains were not measured explicitly. Strain measurements on the specimens E19 and S4 showed an immediate response to changes in RH (Fig.   5c, d). Swelling occurred during wetting, shrinkage during drying phases. The magnitude of swelling exceeded the shrinkage, which accumulated to an irreversible volumetric strain of 0.55-0.75% at the end of the experiment (Fig. 5d). The individual strain measurements showed that mainly deformations normal to bedding contributed to the overall expansion of the rock 25 specimen. The strain measured parallel to bedding was significantly smaller and approximately reversible.
Although irreversible strain was measured, no significant change in Brazilian tensile strength was observed (Fig. 7) Figure 8 shows the relationship between suction and water content for specimen E7 during the stepwise cyclic RH experiment.
The system is in equilibrium at the highest and lowest suction values (turning points between wetting and drying paths) but not in between. Also shown are the main drying and wetting paths reported by Wild et al. (2015). 5 The first drying path for the specimen follows the main drying path as it represents the drying of the intact rock starting from initial conditions which were comparable to the study of Wild et al. (2015). Since the specimens were not dried to their residual water content, the following scanning curves lie between the main drying and main wetting paths. Hysteresis can be observed between drying and wetting path caused by non-homogeneous pore size distribution, different contact angles between wetting and drying, or entrapped air bubbles during wetting (Birle et al., 2008). Therefore, the initial water content cannot be re-10 established anymore and a water loss of 0.6-0.8% occurred. However, the scanning curves of the specimen subjected to the stepwise RH cycles approximately lie within the main drying and wetting paths, indicating that the water retention characteristics are not significantly affected by the variations in RH. This is consistent with findings by Pineda et al. (2014).

Strain and damage
Strain results of the dynamic and stepwise cyclic RH experiments (Fig. 5c) indicate that the Opalinus Clay follows a cyclic 15 expansion and contraction associated with water absorption and desorption processes. Thereby, the Opalinus Clay shows a strongly transversely isotropic deformation behaviour where the strain in direction normal to bedding (ε1) dominate the bulk deformation. These results are consistent with findings by Minardi et al. (2016) who also found an anisotropic response of strain for an Opalinus Clay specimen subjected to one cycle of wetting and drying. This observation can be related to the absorption of water into parallel orientated clay interlayers (i.e. parallel to the bedding planes) leading to swelling in normal 20 direction. Moreover, according to Houben et al. (2013), the pores of Opalinus Clay are elongated along the bedding.
Irreversible volumetric expansion took place during the stepwise cyclic RH exposure. Many studies on single mineral types (e.g. Na-montmorillonite) have demonstrated that clay minerals show distinct hydration states when exposed to different levels of RH (e.g. Mering, 1946;Mooney et al., 1952;Gillery, 1959;Emerson, 1962;Van Olphen, 1965;Glaeser and Mering, 1968;Chipera et al., 1997;Ferrage et al., 2005;Likos and Lu, 2006). These hydration stages reflect the intercalation of one to four 25 discrete layers of water molecules between the mineral interfaces and is driven by the hydration of the cations of the clay minerals (Norrish, 1972). During transition between these stages, the interlayer spacing can increase up to a factor of two (Norrish, 1954). For the hydration of Na-montmorillonite, for example, the interlayer spacing increases from 10 to 12.5 Å between 0 and 20% RH, from 12.5 to 15.5 Å between 50 and 70% RH, and further to about 19 Å for RH>98.5% (Mooney et al., 1952;Gillery, 1959;Emerson, 1962;Glaeser and Mering, 1968). Between 70 and 95% a two-layer hydration state is present 30 for both Ca-and Na-montmorillonite (Seedsman, 1985). Similar stages for other clay minerals are given by Gillery (1959). Earth Discuss., doi:10.5194/se-2016-171, 2016 Manuscript under review for journal Solid Earth Published: 14 December 2016 c Author(s) 2016. CC-BY 3.0 License.

Solid
They all indicate a relative stable state between 80 and 90% RH. Furthermore, sorption and adsorption paths for clay minerals show hysteresis indicating that crystalline swelling is an irreversible thermodynamic process (Laird et al., 1995).
This might explain the accumulated irreversible volumetric expansion as most of the clay minerals transition between the oneand two-layer hydration state in the RH range covered within the experiments in this study. This is also supported by macroscopically detectable fissuring that was observed during the experiment. 5 Although slight fissuring of the specimens was detected and irreversible volumetric strain was observed, no significant influence on the Brazilian tensile strength was observed after three to five cycles for both experiments. It is therefore concluded, that the observed degradation caused by the cyclic variations of RH in this study is not sufficient to cause severe damage that influences the strength of the material. The lower degradation potential for Opalinus Clay compared to other clay shales when subjected to RH cycling is in agreement with findings reported by Pineda et al. (2011). Compared to Lilla Claystone (Pineda 10 et al., 2008;Pineda et al., 2014), the (tensile) strength and (dynamic) stiffness of Opalinus Clay is significantly less affected by cyclic RH variations.

Conclusions
This study demonstrates that cyclic RH variations have the potential to internally damage the Opalinus Clay leading to irreversible volumetric expansion. Internal damage is mainly taking place along the bedding, supported by the fact that 15 irreversible strain was almost exclusively observed in direction normal to the bedding.
The Brazilian tensile strength of Opalinus Clay seems to be unaffected by cyclic RH variations (i.e. a change was not measurable or insignificant). The Brazilian tensile strength parallel to bedding remained constant over three to five cycles while corresponding values for the direction normal to bedding only indicate insignificant decreasing trends. Water retention characteristics of Opalinus Clay were not significantly altered by the observed environmental degradation. 20 The experimental study demonstrates that environmental variations, in particular long-term variations in RH can lead to irreversible volumetric strains that contribute to long term deformations of underground excavations and favour processes that are considered to control self-sealing in Opalinus Clay.    In (a) the load is applied normal to bedding allowing the measurement of the Brazilian ten-sile strength parallel to bedding (σt,p). In (b) the load is applied parallel to bedding allowing the measurement of the Brazilian tensile strength normal to bedding (σt,n).

Figure 5:
Results of the stepwise cyclic RH experiment for the E-specimens, including (a) the changes in relative humidity (RH) and temperature, (b) the water content and apparent suction calculated from the measured RH, (c) the principal strains normal to bedding (ε1) and parallel to bedding (ε2 and ε3) of specimens S4 and E19, and (d) the volumetric strains of specimens S4 and E19.