Experimental Basis

Abstract

A research programme has been conducted jointly by GRS together with BRIUG and BGR to characterize GMZ bentonite as buffer material in comparison with the well-known MX80 bentonite.

3.1 Geotechnical Investigation on GMZ

Chun-Liang Zhang, Ju Wang, Stephan Kaufhold, Yuemiao Liu, Oliver Czaikowski

A research programme has been conducted jointly by GRS together with BRIUG and BGR to characterize GMZ bentonite as buffer material in comparison with the well-known MX80 bentonite. With help of newly developed test methods, geotechnical properties and behaviour of the bentonites were determined, including chemical and mineralogical composition, water retention capacity, swelling capacity, water permeability, gas migration, self-sealing capacity, deformability, and thermal effects. A wide range of valuable results were obtained and are presented in detail in (Zhang et al., 2022). The essential results are extracted and represented here.

3.1.1 Sample Materials

Two types of the natural GMZ bentonite, GMZ01 and GZM02, were extracted from the deep and shallow positions of the bedded deposit in the northern China—Inner Mongolia autonomous region. As reference for comparison, the commercial MX80 bentonite from Wyoming in USA was taken, too.

The sample materials GMZ01 and GMZ02 were powder with fine grains d < 74 $\upmu$m, whereas MX80 was granulated by crushing compacted blocks of a dry density of 2.0 $\mathrm{g/cm}^{3}$ to grains d $=$ 0.2–2 mm. Their initial water contents were measured after drying at 105 $^{\circ }$C for 2 days to values of 10.3% at GMZ01, 8.0% at GMZ02 and 12.6% at MX80, respectively. The grain densities were determined on the dried samples using a helium gas pycnometer. Similar values were obtained to 2.67 $\mathrm{g/cm}^{3}$ for MX80, and 2.66 $\mathrm{g/cm}^{3}$ for GMZ01 and GMZ02.

The chemical components, exchangeable cations and cation exchange capacity of bentonite GMZ01 and GMZ02 were determined by BRIUG and summarized in Tables 3.1 and 3.2, respectively. The mineralogical compositions were determined by BGR by means of X-ray diffraction and Rietveld analysis. The mineralogical analysis was performed on the samples without preheating (A) and with preheating at 105 $^{\circ }$C for 14 days (B) to examine the effect of the heat treatment. The measured results show that the bentonites consist predominantly of montmorillonite amounting to 86% in MX80, 71–74% in GMZ01 and 56% in GMZ02, respectively (Table 3.3). Accessory minerals are quartz, cristobalite, plagioclase, K-feldspar, etc. It is to be pointed out that the preheating did not affect the mineralogical compositions of the bentonites.

Table 3.1 Chemical components of bentonites GMZ01 and GMZ02

Table 3.2 Exchangeable cation and cation exchange capacity (CEC) of bentonites GMZ01 and GMZ02

Table 3.3 Mineralogical components of bentonites GMZ01, GMZ02 and MX80

For hydraulic testing, synthetic Beishan site groundwater (BSW) was manufactured. Its density and viscosity were measured at 20–50 $^{\circ }$C as shown in Fig. 3.1. Both parameters decrease with increasing temperature. The viscosity of BSW is consistent with that of the distilled water ($\textrm{H}_{{2}}$O), which can be approached (UPC, 2015) by

$$\begin{aligned} \mu _{w} = A\ exp\left( \frac{B}{273.15 + T} \right) \end{aligned}$$

(3.1)

where ${\mu }_{w}$ is the dynamic viscosity of the water ($\textrm{Pa}\cdot$s), T is the temperature ($^{\circ }$C), and the parameters A $=$ 2.05 $\times$ 10$^{-6}$ $\textrm{Pa}\cdot$s and B $=$ 1805.5 K.

Fig. 3.1

Density and viscosity of the synthetic (BSW) and distilled water ($\textrm{H}_{2}$O)

3.1.2 Experimental Results

3.1.2.1 Water Retention

As a key property, water retention capacity of the bentonites was determined using vapour transfer technique. Loose and compacted specimens were placed in desiccators, in which the relative humidity was adjusted by different salt solutions. The loose specimens with an initial weight of 50 g each were placed in bowls, while the compacted ones were confined in stainless-steel cells of 50 mm diameter and 20 mm height. The compacted GMZ01 and GMZ02 specimens had dry densities of 1.7 and 1.8 $\mathrm{g/cm}^{{3}}$, respectively, which are consistent with the designed buffer blocks (Liu et al., 2014). The compacted MX80 specimens had a dry density of 1.5 $\mathrm{g/cm}^{{3}}$, which is nearly the same as that emplaced in a horizontal drift at Mont-Terri Rock Laboratory (Müller et al., 2018). These compacted specimens in the cells were covered with sintered porous discs allowing exchange of water vapour in- and outside. Under applied vapour-pressure or suction gradients between the specimens and the surrounding air, water molecules migrate from the humid air into the pore space in relatively dry specimens, or inversely. Whereas the relative humidity in each desiccator was continuously recorded by transistor psychrometer sensors, the amount of water uptake was measured by weighing the specimens at time intervals of 1–2 months until equilibrium. The relationship between relative humidity and suction is determined by the psychrometric law (Fredlund and Rahardjo, 1993).

$$\begin{aligned} s\ = - \frac{\rho _{w}\ R\ T}{M_{w}}\ln (RH) \end{aligned}$$

(3.2)

where T is the absolute temperature, R is the universal gas constant, $\rho _{w}$ and ${M_{w}}$ are the density and the molecular mass of water, respectively. A range of RH–values of 22–100% was applied, corresponding to s $=$ 207 to 0 MPa at the testing temperature of 24 $^{\circ }$C.

The amounts of water uptake reached at equilibrium are depicted in Fig. 3.2 as a function of suction. As expected, water content of each bentonite increases with decreasing suction to a maximum at zero suction. The constant volumes of the compacted specimens limited the water uptake at low suctions s < 5–10 MPa. Therefore, each loose bentonite can take more water. The maximum water content reached at s $\approx$ 0 increases with increasing montmorillonite content (${f}_{m}$) due to its high capacity of water adsorption (Bradbury and Baeyens, 2002b; Yong et al., 2012; Birgersson et al., 2017; Villar et al., 2020).

There is a broad consensus that three types of water exist in bentonite: interlayer water, double-layer water and free water. The interlayer water is adsorbed on internal surfaces of montmorillonite units and immobile. The double-layer water exists in electrical double layers associated with external surfaces of the clay stacks, which is bounded and immobile too. The free water appears in interparticle pore space surrounding the clay stacks and other mineral particles. The distributions and proportions of the different types of water in a bentonite depend on the amount of permanent charge, exchangeable cations, soluble components, dry density (porosity) and water content (Kaufhold et al., 2010; Jenni et al., 2019). According to the calculations (Muurinen et al., 1987; Bradbury and Baeyens, 2002b), the free water content in the saturated bentonite MX80 is smaller than 5% at dry densities of 1.2–1.8 $\mathrm{g/cm}^{{3}}$. In comparison, the bentonites GMZ01 and GMZ02 with such densities must comprise more free water due to their smaller montmorillonite contents.

Fig. 3.2

Water retention curves of the bentonites in free and constraint conditions

3.1.2.2 Swelling Pressure

A test setup was developed and used for sequential measurements of swelling pressure, water permeability and gas migration of the bentonites. It consists of 14 stainless-steel cells, a syringe pump, hydraulic lines and instruments. Different kinds of specimens were prepared for testing as shown in Fig. 3.3. More than 40 specimens were compacted in the cells to a same size of 50 diameter and 15 mm height but to different dry densities of 1.3–1.8 $\mathrm{g/cm}^{{3}}$. Moreover, the specimens of GMZ01 pellets were produced by crushing compacted blocks (dry density $=$ 1.7 $\mathrm{g/cm}^{{3}}$) to grain sizes of 2–4 mm and loosely emplaced to a low dry density of 0.87 $\mathrm{g/cm}^{{3}}$. The assembled blocks consisted of a central vertical gap of 3 mm width and had an average dry density of 1.65 $\mathrm{g/cm}^{{3}}$. A mixture of pellets-blocks was prepared to a height of 15 mm for each layer and to an average dry density of 1.31 $\mathrm{g/cm}^{{3}}$. All specimens were installed in the cells and heated at 105 $^{\circ }$C for 2–3 days to match the initially heated and dried conditions in HLW buffer.

Figure 3.4 shows the development of swelling pressure with water uptake observed on some compacted specimens, loose pellets, assembled blocks and pellets-blocks mixture (cf. Fig. 3.3). The specimens took up water quickly for the first 3–5 days due to high suction effect and then slowly over time. Over 2–3 weeks, full saturation was reached at all the specimens. In correspondence to the water uptake, swelling pressure built up in the bentonites. They exhibited a typical swelling pressure evolution with a double-peak shape, independent of the initial inner structures. Generally, the double-peak evolution is attributed to variations of micro- and macrostructures in the compacted bentonite during saturation process (Pusch et al., 1990; Zhu et al., 2013; Imbert and Villar, 2006). The initial increase in swelling pressure is a direct consequence of water uptake and associated expansion of montmorillonite interlayers (interlayer swelling pressure). The interlayer expansion also destabilizes the macrostructure between clay particles. As the local swelling pressure reaches the first peak, the macrostructure in the wetted area becomes weaker, leading to a pressure relaxation. Simultaneously, the hydration in the electrical double layers between clay particles results in overlapping of thin water films at narrow spaces and generates repulsive forces (double-layer swelling pressure). This increases the global swelling pressure again to a constant value as the bentonite is fully saturated.

Fig. 3.3

Different kinds of specimens prepared using GMZ01 bentonite

Fig. 3.4

Development of swelling pressure with hydration observed on the different bentonite specimens

The final swelling pressures $P_{s}$ of all the specimens are summarized in Fig. 3.5 as a function of dry density $\rho _{d}$ together with some data from literature. The data indicate strong dependencies of the swelling pressure on the dry density ${\rho _{d}}$ and the montmorillonite content ${f_{m}}$, which can be approached by an empirical model

$$\begin{aligned} P_{s} = \alpha \ f_{m}^{n}\ exp\left( \beta \ \rho _{d} \right) \end{aligned}$$

(3.3)

where $\alpha$, $\beta$ and n are the parameters. Fitting the present data yields the parameter values of $\alpha =$ 0.0006 MPa, $\beta =$ 6 $\textrm{cm}^{{3}}$/g and n $=$ 3.7. The model curves represent the mean lines through the scattered data for each kind of bentonite. This model can also match the ${P}_{s}$-${\rho _{d}}$ curve of FEBEX bentonite with a montmorillonite content ${f_{m}} =$ 90% (Villar et al., 2010). It can also be seen that the data resulted from the synthetic groundwater (BSW) are consistent with those from the deionized water ($\textrm{H}_{{2}}$O) for GMZ01 (Zhang et al., 2019; Xu et al., 2017) and for MX80 (Bucher and Müller-Vonmoos, 1989; Karnland et al., 2008; Seiphoori, 2015).

It is interesting to point out that the swelling pressure data obtained from GMZ01 pellets and blocks with the large voids/gaps are consistent with those of the compacted homogeneous specimens. This is the consequence of homogenization of the inner structure in the pellets due to the swelling of clay particles with hydration. Figure 3.6 pictures the homogenized pellets and the sealed gap between blocks, compared with the initial states (Fig. 3.3b, c). Because of the low density of the pellets, the large voids between the particles could not be completely sealed. In contrast, the gap between the blocks could be completely sealed by the high swelling of the dense bentonite matrix.

Fig. 3.5

Summary of swelling pressure data for the various bentonites

Fig. 3.6

Pictures of a homogenized pellets and b self-sealed gap between blocks of GMZ01 bentonite after water flow through a month

3.1.2.3 Water Permeability

After the specimens were fully saturated, the water injection pressure was stepwise increased from 0.1 to 1.5 MPa, while water outflow was measured using the scaled burette at atmospheric pressure. Each pressure step lasted over days to a month for reaching steady flow. The hydraulic gradient across a specimen is calculated by

$$\begin{aligned} i = \frac{P_{w} - P_{o}}{\rho _{w} g\,\, L} \end{aligned}$$

(3.4)

where ${P_{w}}$ and $P_{o}$ are the up- and downstream pressures (kPa), $\rho _{w}$ is the water density (kg/$\textrm{m}^{{3}}$), g is the gravitational acceleration (9.81 $\mathrm{m/s}^{2}$) and L is the specimen length (m). The calculated i-values vary in a range of 60–1000. The application of the different hydraulic gradients aimed at validating Darcy’s law for the compacted bentonites. During steady flow, hydraulic conductivity $K_{h}$ (m/s) and/or intrinsic permeability $K_{w}$ ($\textrm{m}^{2}$) can be determined according to Darcy’s law (Liu, 2017)

$$\begin{aligned} K_{h} = \ \frac{v}{i} = \frac{Q_{w}\ {\ \rho }_{w}\ \ g\ \ L}{A\ \ \left( P_{w} - P_{o} \right) } \end{aligned}$$

(3.5)

$$\begin{aligned} K_{w} = \frac{Q_{w\ }\ \mu _{w}\ \ L}{A\ \ \left( P_{w} - P_{o} \right) } = K_{h}\frac{\mu _{w}}{\ \rho _{w}\ \ g} \end{aligned}$$

(3.6)

where v is the average velocity (m/s) of water flux $Q_{w}$ ($\textrm{m}^{{3}}$/s) through the section A ($\textrm{m}^{2}$), $\mu _{w}$ is the water viscosity ($\textrm{Pa}\cdot$s), $P_{w}$ and $P_{o}$ are the up- and downstream pressures (Pa).

Figure 3.7 depicts water flow velocities v measured across the specimens at different hydraulic gradients. At GMZ01 and GMZ02 specimens with dry densities of 1.5–1.8 $\mathrm{g/cm}^{{3}}$, the flow velocity is linearly related with the hydraulic gradient and the v-i lines pass through the origin of the coordinate system. This confirms Darcy’s law for the compacted bentonites GMZ01 and GMZ02. In contrast, the v-i lines for the compacted MX80 specimens are even though linear but do not pass through the origin of the coordinate system. The intersection of the line with i-axis is the threshold gradient I, which are estimated to be ~20 at $\rho {d} =$ 1.5 $\mathrm{g/cm}^{{3}}$, ~150 at $\rho {d} =$ 1.6 $\mathrm{g/cm}^{{3}}$ and ~300 at $\rho _{d} = 1.8\, \mathrm{{g/cm}}^{3}$, respectively. The existence of threshold gradients should be attributed to the immobile state of the strongly bound water in the dense bentonite. At low hydraulic gradients $i < I$, the bound water cannot be moved.

Fig. 3.7

Water flow velocities measured at different hydraulic gradients

The slope of the v-i line is equal to the hydraulic conductivity $K_{h}$ (Eq. 3.5) and the corresponding water permeability $K_{w}$ can be derived (Eq. 3.6). The measured$K_{w}$-data are summarized in Fig. 3.8 as a function of total porosity $\varphi$. As expected, the permeability decreases with decreasing porosity and with increasing montmorillonite content, which can be approximately approached by

$$\begin{aligned} K_{w} = \kappa \ \left( 1 - f_{m} \right) ^{m}\ exp(\omega \ \phi ) \end{aligned}$$

(3.7)

where $\alpha$, $\beta$ and $\eta$ are the parameters. Fitting the data yields $\alpha = 8 \times 10$ $^{-22}$ $\textrm{m}^{2}$, m $=$ 3.2 and $\omega = 20$. The calculated curves agree well with the data.

Fig. 3.8

Summary of water permeability data for the bentonites as a function of total porosity and montmorillonite content

In fact, the water permeability is determined by effective porosity rather than total porosity. The effective porosity mostly consists of macropores containing free water. The size and connectivity of the macropores are dominated by the swelling pressure of the neighbouring clay particles. Therefore, the water permeability shall be related to the swelling pressure, as shown in Fig. 3.9. It is evidence that the water permeability of the bentonites decreases linearly with increasing swelling pressure. Additionally, the water permeability also decreases with increasing montmorillonite content, which determines the proportions of free and bound pore water. The relationships can be expressed by

$$\begin{aligned} K_{w} = \gamma \ \left( 1 - f_{m} \right) ^{\eta }\ P_{s} \end{aligned}$$

(3.8)

where $\gamma$ and $\eta$ are the parameters. $\gamma = 1.2 \times 10$ $^{-18}$ $\textrm{m}^{2}$ and $\eta = 2$ are determined based on the data. A comparison indicates a reasonable agreement between the model and the data.

As mentioned earlier, the loose pellets and assembled blocks of GMZ01 bentonite became homogenized under swelling effect. Even though some big voids remained in the saturated pellets (Fig. 3.6a), the water permeability of the whole specimen is still very low. This is because the permeability is mainly governed by the narrowed throats of the pathway network.

Fig. 3.9

Water permeability of the bentonites in relation with swelling pressure and montmorillonite content

3.1.2.4 Gas Penetration

Gas penetration testing followed with some water-saturated specimens. Before testing, the water remaining in the inlet and outlet reservoirs was removed by vacuum pumping. Helium gas was then injected to the specimens at controlled flow rates of 0.02–0.2 ml/min, while gas outflow was accumulated in a steel vessel with a volume of 380 ml. Gas pressures in the up- and downstream were monitored by pressure transducers. Stress reaction was recorded by the load cell at the opposite side. The gas injection lasted for sufficient time periods to examine long-term gas flow process and to determine gas breakthrough pressure and permeability. The gas outflow rate can be estimated from the pressure difference across the specimen with elapsed time (Gutiérrez-Rodrigo et al., 2021)

$$\begin{aligned} Q_{g} = V_{g}\ \frac{P_{g}\left( t_{j} \right) \ - P_{g}\left( t_{i} \right) }{P_{m}}\ \frac{1}{\mathrm {\Delta }t} \end{aligned}$$

(3.9)

where $Q_{g}$ is the mean gas flux ($\textrm{m}^{{3}}$/s), $V_{g}$ is the volume of the outlet reservoir ($\textrm{m}^{{3}}$), $P_{g}$($t_{j}$) and $P_{g}$($t_{i}$) are the in- or outlet gas pressures (Pa) at time points j and i, $P_{m}$ is the average gas pressure between the measurement interval $\Delta t = t_{j}$—$t_{i}$ (s). Effective gas permeability $K_{g}$ ($\textrm{m}^{2}$) is then estimated by Darcy’s law:

$$\begin{aligned} K_{g} = \frac{{2\ \ Q}_{g\ }\ \mu _{g}\ \ P_{o}\ \ L}{A\ \ \left( P_{g}^{2} - P_{o}^{2} \right) } \end{aligned}$$

(3.10)

where $\mu _{g}$ is the dynamic viscosity of helium gas (1.96 $\times$ 10$^{-5}$ $\textrm{Pa}\cdot$s).

Test data are illustrated in Fig. 3.10 for the homogeneous specimens and in Fig. 3.11 for the homogenized pellets and sealed blocks (cf. Fig. 3.6) in terms of applied gas flow rate ($Q_{g}$), induced up- and downstream pressures ($P_{g}$, $P_{o}$), axial total stress ($\sigma _{a}$) and outflow rate (${Q}_{o}$) versus elapsed time. It is evident that the gas penetration process was qualitatively quite similar with multiple gas breakthrough events and unstable post-breakthrough flow, but quantitatively different in each specimen due to the different material properties and gas injection rates.

Fig. 3.10

Gas penetration in the water-saturated and homogeneous bentonite specimens

Fig. 3.11

Gas penetration in the water-saturated, homogenized pellets and sealed blocks

The gas injection led to a gradual build-up of the upstream pressure, accompanied by an increase of the total stress over the initial swelling pressure. This implies that the water-saturated bentonites were gas tight and subjected to compression by the gas pressure against the rigid boundary. As a peak pressure $P_{gb}$ was reached, gas breakthrough occurred, yielding a rapid gas release and hence a rapid drop of the upstream pressure as well as a rapid rise of the downstream backpressure. When the upstream pressure dropped down to a minimum value ${P}_{gc}$ and the backpressure P_o became constant, the gas flow ceased due to self-sealing of the pathways. The upper peak P_gb is usually defined as the gas breakthrough pressure, whereas the lower peak P_gc is referred to as the gas seal pressure. In correspondence to the drop of the gas pressure from P_gb to P_gc, the total stress decreased from $\sigma$_ab to $\sigma$_ac. Following the first gas breakthrough/ sealing event, further gas injection caused multiple rising/dropping of the upstream pressure, reflecting opening/closing cycles of the pathway network. The temporal disconnections of the network tended to disappear with time and then a continual gas flow followed. In the closed downstream, the backpressure increased with gas accumulation, leading to an increase in average pore pressure (P_gp $=$ (P_g + P_o)/2) and a decrease in pressure gradient ($\Delta P$_g $=$ P_g − P_o). The unsteady pressure conditions caused variations in features of the pathway network (aperture, connectivity, etc.) and thus the gas flowed unsteadily.

Extensive investigations (e.g. Pusch et al. (1987); Sellin (2014); Harrington et al. (2017); Levasseur et al. (2020)) suggested that gas entry and penetration in water-saturated bentonite is governed by dilating local pores in weak regions to form micro-fissures, which needs a high gas pressure exceeding the sum of the minor principal stress and tensile strength of the material. This can be expressed as

$$\begin{aligned} P_{gb} \ge \sigma _{\min } + \ \sigma _{T} \end{aligned}$$

(3.11)

where $\sigma$_min is the minor principal stress and $\sigma$_T is the tensile strength of the material. In case of the oedometer conditions, the radial stress should be the minor one $\sigma _{r}=$ $\sigma _{min}$, which can be assumed being nearly equal to the axial one, $\sigma$_r $=$ $\sigma$_min $\approx$ $\sigma$_a. Thus above Eq. (3.11) can be rewritten by

$$\begin{aligned} P_{gb} \ge \sigma _{rb} + \ \sigma _{T} \approx \sigma _{ab} + \ \sigma _{T} \end{aligned}$$

(3.12)

The tensile strength can be estimated by $\sigma _{T} = P_{gb} -$ $\sigma$_ab, which depends on the confining stress. A linear relationship is here assumed, $\sigma _{T} ={b} \times \sigma _{\text {ab}}$. Thus, equation (3.12) can be expressed by

$$\begin{aligned} P_{gb} \ge (1 + b)\ \sigma _{ab} \end{aligned}$$

(3.13)

and the gas breakthrough boundary condition is defined by

$$\begin{aligned} P_{gb} = \sigma _{ab} + \ \sigma _{T} = (1 + b)\ \sigma _{ab} \end{aligned}$$

(3.14)

The first gas breakthrough pressures observed on each specimen are depicted in Fig. 3.12a as a function of the total stress. Fitting the data from GMZ01, GMZ02 and MX80 specimens yields the parameter b $=$ 0.09. It is obvious that this model matches the data well. Additionally, the model can also predict the gas breakthrough pressures observed in other lab tests (Graham et al., 2016; Harrington et al., 2017) and in a full-scale buffer test Lasgit (Cuss et al., 2014), as shown in Fig. 3.12b. It is to be noted that this gas breakthrough criterion is adequate for water-saturated bentonite under volume-constrained conditions. If the boundary is not rigid, the bentonite specimen is allowed to deform by gas pressuring, which can result in micro-fissuring at pressures below the confining stresses. This has been evidenced in tests on MX80 bentonite in oedometer cell under controlled axial load (Zhang, 2021) and on GMZ01 bentonite in triaxial cell under isostatic load (Cui et al., 2021).

Effective gas permeability K_g was determined from the data of gas outflow Q_o for each specimen. Figure 3.13 shows some typical results of evolution of gas permeability in correlation with effective stress ($\sigma$_eff $=$ $\sigma -$ P_gp). It is obvious that the gas permeability in log(K_g) varies inversely with the periodic variation of the effective stress. In fact, the effective stress controls the geometric features of the network and thus the permeability. Final K_g-values at the end of each test are depicted in Fig. 3.14 as a function of effective stress, which can be approximately approached by

$$\begin{aligned} K_{g} = K_{go}\ exp\left( - \xi \ \sigma _{eff} \right) \end{aligned}$$

(3.15)

where $K_{\text {go}}$ is the gas permeability at $\sigma$_eff $=$ 0 and $\xi$ is a parameter. Fitting the data yields $K_{\text {go}}$ $= 7 \times 10^{-17}$ ${\textrm{m}}^{2}$ and $\xi = 4$, with which the model curve matches the data well.

Fig. 3.12

Relationship of gas breakthrough pressure with total stress: a obtained from the saturated specimens of GMZ01, GMZ02 and MX80 bentonites and b compared with literature data from small-scale specimens (Graham et al., 2016) and full-scale buffer of MX80 bentonite (Cuss et al., 2014)

Fig. 3.13

Evolution of gas permeability in correlation with effective stress

After testing, the residual water content was determined on each specimen. Most of the tested specimens showed the full saturation after gas flowing through, except for the pellets and pellets-blocks mixture. The full saturation implies that no or negligible amount of pore water were displaced out by gas injection and the gas penetration was caused by generating and dilating micro-fissures. In contrast, the desaturation of the pellets indicates that the water in some large voids (Fig. 3.6a) was partially removed by the injected gas.

Fig. 3.14

Gas permeability of the compacted bentonites as a function of effective stress

3.1.2.5 Thermal Effects

Two coupled load rigs were used for thermal testing. Each rig allows two specimens in stainless-steel cells of 120 mm diameter and 160 mm height one upon another. They are installed in separated thermal chambers and can be heated up to 110 $^{\circ }$C. Each specimen is covered with sintered porous plates and connected to thin holes through the upper load piston and the lower basic plate for fluid flow. A syringe pump is used for fluid injection at the bottom, while fluid outflow is measured at the top by means of scaled burette at atmospheric pressure. Gas testing is carried out by injecting helium gas at controlled flow rates, during which gas outflow is accumulated in a steel vessel and the outlet pressure is recorded. Axial load is applied by means of another syringe pump. Axial deformation is monitored using two linear variable differential transducers (LVDTs) mounted between the upper piston and the cell top.

Two specimens of each bentonite GMZ01 and GMZ02 were pre-compacted in the cells to respective dry densities of 1.71 and 1.85 $\mathrm{g/cm}^{{3}}$. Their initial characteristics are summarized in Table 3.4.

Table 3.4 Initial characteristics of the bentonite specimens for thermal testing

Fig. 3.15

Thermo-mechanical compaction and swelling with hydration observed on bentonite specimens at temperatures of 30 and 90 $^{\circ }$C and axial load of 4 MPa

A test procedure was carried out with sequence steps:

• I. Application of axial stress $\sigma$_a $=$ 4 MPa, under which the temperature was increased to 90 $^{\circ }$C at GMZ01a/02a and 30 $^{\circ }$C at GMZ01b/02b, respectively.

• II. Hydration followed by infiltration of the synthetic water BSW into the specimens at atmospheric pressure, whereby their water uptake and axial strain were monitored.

• III. Unloading to $\sigma$_a $=$ 2 MPa and then reloading to $\sigma$_a $=$ 4 MPa again, under which axial strain and water permeability of the saturated specimens were recorded.

• IV. Gas injection at controlled flow rates to determine gas breakthrough pressure and permeability of the specimens at different temperatures.

• V. Examination of self-sealing of gas pathways by measuring water permeability of a specimen at $\sigma _{a} =$ 4 MPa and T $=$ 100–30 $^{\circ }$C.

Figure 3.15 illustrates the water uptake and the porosity variation of the specimens during hydration. Initially, the specimens were thermo-mechanically loaded to T $=$ 90 $^{\circ }$C at GMZ01a/02a and 30 $^{\circ }$C at GMZ01b/02b and to $\sigma$_a $=$ 4 MPa over 7 days. The load resulted in a rapid compaction in the beginning and then followed by a slight reduction in porosity with time at each specimen. The compaction at the high temperature is relatively larger due to more thermal mobilization of the pore water, which reduces the cohesion between clay particles and thus easier collapse of the pore structure.

As water was introduced into the specimens, a hydration process began quickly with rapid water uptake and then slowed down with time until fully saturated. The full saturation was reached relatively earlier at 90 $^{\circ }$C but the amount of water uptake was relatively lower than at 30 $^{\circ }$C. This indicates that the water absorption capacity of the bentonites decreases with increasing temperature. Furthermore, the water uptake caused a gradual swelling of GMZ01 bentonite to a porosity increase of 1.5% at 90 $^{\circ }$C and 3.5% at 30 $^{\circ }$C over 12 days, respectively. Obviously, the swelling capacity decreases with increasing temperature. Compared to GMZ01, GMZ02 showed the lower capacities of water absorption and swelling because of its lower montmorillonite content.

Fig. 3.16

Deformation and water permeability changes of saturated bentonite specimens at temperatures of 30 and 90 $^{\circ }$C and axial loads of 2–4 MPa

Thermal effects on deformation and water permeability of the saturated bentonites are shown in Fig. 3.16. Keeping the applied temperatures and axial stress, the water injection pressure was increased to P_w $=$ 0.4 and 0.8 MPa. This caused a slight increase in porosity at GMZ01a/b but not at GMZ02a/b. When the stress was lowered to 2 MPa, all specimens expanded in the beginning and then tended to constant. Subsequent reloading caused reconsolidation. The deformation of each bentonite appeared quite similar at the different temperatures, indicating a negligible temperature influence on deformation. Moreover, the tests also showed insignificant temperature influence on water permeability of the bentonites. However, the permeabilities obtained from GMZ01a/b are about one order of magnitude lower than that of GMZ02a/b. This is attributed to the different montmorillonite contents. The higher montmorillonite content in GMZ01 means the higher fraction of absorbed immobile pore water and thus the lower effective porosity.

Results of the gas penetration tests are illustrated in Fig. 3.17. Under the constant load and temperature, gas injection led to a gradual build-up of the upstream pressure P_gi until breakthrough occurred at a peak P_gb. The breakthrough yielded a rapid gas release and hence a rapid drop of the upstream pressure to a minimum P_gc as well as a rapid rise of the backpressure P_go. When P_go maintained constant, no gas flowed through the specimen due to sealing of the gas pathways. The gas breakthrough/sealing event repeated itself with continued gas injection, reflecting a periodic process of the path opening/closing under interactions between gas pressure and confining stress.

Fig. 3.17

Gas pressures and associated axial strains observed on the water-saturated bentonite specimens at axial load of 4 MPa and different temperatures of 30–100 $^{\circ }$C

The tests showed influences of different factors on the gas breakthrough and seal pressures P_gb and P_gc. For instance, GMZ01a/02a at 90 $^{\circ }$C showed that both the pressures P_gb and P_gc increase with increasing backpressure P_go. The breakthrough pressure can exceed the confining stress at elevated backpressures (e.g. GMZ01a at 90 $^{\circ }$C). When the injection pressure was limited to P_g $=$ 3.9 MPa slightly below the confining stress at GMZ01b/02b, no gas breakthrough was observed when P_go > 0. This means that the gas backpressure shall be taken in Eq. (3.11) by

$$\begin{aligned} P_{gb} \ge \sigma _{\min } + P_{go} + \sigma _{T}\end{aligned}$$

(3.16)

$$\begin{aligned} P_{gb} - P_{go} \ge \sigma _{\min } + \sigma _{T} \end{aligned}$$

(3.17)

where P_gb − P_go $=$ $\Delta$P_gb can be taken as the effective gas breakthrough pressure.

Thermal effect on gas penetration in the water-saturated bentonites was examined on specimen GMZ01b by increasing temperature from 30 to 50, 70, 90 and 100 $^{\circ }$C. It was observed that the gas breakthrough pressure decreased with increasing temperature. This is attributed to the thermally reduced viscosity and absorption forces in the pore water.

The tests also showed some influences of gas pressure on deformation of the bentonites, particularly at gas breakthrough event. The increase in gas injection pressure can dilate the pore space and then create micro-fissures for gas flowing. After breakthrough, the release of the gas pressure increases the effective stress, leading to the closing and sealing of the pathways.

Figure 3.18 shows the self-sealing effect of the gas pathways by measuring water permeabilities of two specimens GMZ01a/02a after gas penetration. The data obtained at decreased temperatures of 100, 80, 55 and 28 $^{\circ }$C indicate no significant influence of temperature on water permeability. Average values are determined to K_w $= 2 \times 10^{-20}$ for GMZ01a and $3 \times 10^{-20}$ $\textrm{m}^{2}$ for GMZ02a, which are closely consistent with those obtained before the gas generation (cf. Fig. 3.16).

Fig. 3.18

Water permeabilities measured on bentonite specimens after gas penetration

3.1.3 Conclusions

As the buffer material for the potential repository in China, two types of GMZ bentonite were investigated in comparison with the well-known MX80 bentonite. It was found out that the geotechnical properties and behaviour of the bentonites are dominated by the montmorillonite content and the dry density. Major findings are listed as

• (1) The water absorption capacity increases with increasing montmorillonite content. The maximum water content is reached at zero suction but can be limited at low porosities.

• (2) The swelling pressure builds up with hydration, which is characterized by a typical double-peak pattern due to variations of the inner structure. The maximum swelling pressure at full saturation increases exponentially with increasing montmorillonite content and dry density.

• (3) Threshold hydraulic gradients were observed on compacted MX80 bentonite at dry densities above 1.5 $\mathrm{g/cm}^{{3}}$, but not on GMZ01 and GMZ02. The water permeability decreases exponentially with increasing montmorillonite content and linearly with swelling pressure.

• (4) Gas penetration into the water-saturated bentonites under rigid boundary conditions requires high overpressures that exceed the total stress and tensile strength of the material to create local microcracks for gas passage. After gas breakthrough, the pathway network tends to self-seal under swelling effect of the surrounding bentonite.

• (5) Large voids/gaps remaining in the bentonite pellets and assembled blocks tend to disappear with homogenization process during hydration and swelling.

• (6) Various thermal effects were observed: (a) increasing temperature reduces the water absorption capacity and thus the swelling capacity; (b) the hydraulic conductivity is influenced by temperature due to thermally induced variations in viscosity and density of the flowing water, whereas the intrinsic water permeability is almost independent of temperature but dependent on thermally induced variation in effective porosity; (c) increasing temperature increases the mobility of adsorbed pore water and thus lowers the pressure threshold for gas penetration and (d) no negative thermal impact on the bentonite buffer performance was found.

3.2 Mineralogical Investigation on GMZ01/GMZ02

S. Kaufhold

3.2.1 Sample Material and Methods

Samples GMZ 01 and GMZ 02 were received from GRS in 2019 and characterized using a set of methods which are introduced in the following.

The mineralogical composition was determined based on X-ray diffraction (XRD), infrared spectroscopy (IR) and simultaneous thermal analysis (STA). The chemical composition was determined both by X-ray fluorescence (XRF) and C-/S-analysis (LECO).

XRD patterns were recorded using a PANalytical X’Pert PRO MPD $\Theta$–$\Theta$ diffractometer (Co-$\textrm{K}\alpha$ radiation generated at 40 kV and 40 mA), equipped with a variable divergence slit (20 mm irradiated length), primary and secondary soller, diffracted beam monochromator, point detector and a sample changer (sample diameter 28 mm). The samples were investigated from 1$^{\circ }$ to 75$^{\circ }$ 2$\Theta$ with a step size of 0.03$^{\circ }$ 2$\Theta$ and a measuring time of 12 sec per step. For specimen preparation the back loading technique was used.

From XRD patterns the mineralogical composition was calculated by using the Rietveld software $\textrm{AutoQuan}^{\textcircled {R}}$ (AQ; Bergmann and Kleeberg (1998)). After a detailed qualitative analysis, appropriate structural models were selected and applied to all samples. The montmorillonite (smectite) content was calculated by using the structural model published by Ufer et al. (2004). Refined values were obtained based on the $\textrm{AutoQuan}^{\textcircled {R}}$ results by considering the chemical composition and the variability of clay minerals. Additionally, the montmorillonite content was calculated by considering the CEC assuming 15% variable charge as well as a layer charge density of $-0.30$ eq/formula unit according to Kaufhold et al. (2002). The CEC was measured using the Cu-Triethylenetetramine method (Kahr et al., 1996; Lorenz et al., 1999).

For measuring mid (MIR)-infrared spectra the KBr pellet technique was applied. Spectra were collected on a Thermo Nicolet Nicolet iS50 FTIR spectrometer (MIR beam splitter: KBr, detector DTGS TEC). The resolution was adjusted to 2 $\textrm{cm}^{-1}$. Details about the analytical procedure allowing to quantify the mineralogical composition were published by Kaufhold et al. (2012). The spectrometer was built by Nicolet Instruments, Madison, Verona Road, Wisconsin, USA.

Thermoanalytical investigations were performed using a Netzsch 449 F3 Jupiter thermobalance equipped with a DSC/TG sample holder linked to a Netzsch QMS 403 C Aeolus mass spectrometer (MS). 100 mg of powdered material previously equilibrated at 53% relative humidity (RH) was heated from 25 to 1150 $^{\circ }$C with a heating rate of 10 K/min. The devices were manufactured by Netzsch (Selb, Germany).

The organic carbon (Corg) content was measured with a LECO CS-444-Analysator after dissolution of the carbonates. Carbonates had been removed by treating the samples several times at 80 $^{\circ }$C with HCl until no further gas evolution could be observed. Samples of 170–180 mg of the dried material were used to measure the total carbon (Ctotal) content. Total inorganic carbon (Ccarb) was calculated by the difference of Ctotal and Corg. The samples were heated in the device to 1800–2000 $^{\circ }$C in an oxygen atmosphere and the amount of CO2 and SO2 was detected by an infrared detector. The device was built by LECO (St. Joseph, Michigan U.S.A.).

3.2.2 Results

Results of XRD analysis are shown in Fig. 3.19 and the mineralogical composition as determined by Rietveld is given in Table 3.5.

Fig. 3.19

XRD pattern and qualitative analysis of the two samples

Table 3.5 Cation exchange capacity (CEC) and mineralogical composition of the two samples according to Rietveld refinement, IR spectroscopy, and after correction of the XRD and IR results using XRF

According to XRD-Rietveld refinement and CEC results, sample GMZ 01 contains more smectite than GMZ 02. Assuming a mass of a formula unit accounting for 365 g/mol (which is a reasonable assumption because of low Fe content), 15% variable charge, a layer charge density (LCD) accounting for 0.33 eq/FU, it is possible to explain 70 mass% smectite and a CEC of 75 meq/100g (sample GMZ 01). The combination of 52 mass% smectite and a CEC of 50 meq/100g could be explained by a somewhat lower layer charge density (0.30 eq/FU) of sample GMZ 02. This comparison, however, is not suitable to determine the layer charge density which, therefore, has to be investigated in future.

Sample GMZ 02 contains less smectite than GMZ 01 but more feldspar and quartz. This sample also contains muscovite and clinoptilolite which were not found in sample GMZ 01. The chemical composition is given in Table 3.6.

Table 3.6 Chemical composition of both samples based on XRF and LECO analysis

The main elemental composition of both samples is similar and corresponds well to the mineralogical composition: The higher SiO₂ content of GMZ 02 corresponds to the higher quartz content, the higher K₂O content of GMZ 02 corresponds to the higher muscovite content and the higher MgO content of sample GMZ 01 corresponds to the higher smectite content. The C- and S-contents are low in both samples which according to Kaufhold and Dohrmann (2016) can be considered a favourable property for HLRW bentonites. Moreover, the measured values are almost identical to those measured by Briug (Zhang et al., 2022).

Fig. 3.20

Thermal analysis of the two bentonites (a calorimetric curves, DSC; b mass spectrometry mass 18; c mass spectrometry mass 44; d thermal gravimetry, TG)

The main elemental composition was quantitatively compared with the mineralogical composition by assuming ideal chemical compositions of the minerals. Based on this ideal composition and the mineral contents determined by Rietveld refinement and IR, a theoretical chemical composition was obtained which was compared with the measured one. For the main component smectite, the exchangeable cations measured by Briug (Zhang et al., 2022) were considered. Both samples showed negative CaO and Na₂O contents but positive SiO₂ contents which suggests that the content of clay minerals and feldspar could be lower by 2–4 mass% and that of quartz and cristobalite 2–4 mass% higher. This finding is in-line with IR spectroscopy which provided significantly less smectite (−10 mass%) and more cristobalite ($+$10 mass%). Comparing the XRD and IR results with the chemical composition as explained above suggests that the real values are in between both methods. Results obtained by this chemical correction are given in Table 3.5 (right). Overall, however, a good correlation of the theoretical mineralogical composition and the XRD results indicates the validity of Rietveld refinement results.

The trace elemental composition of both samples is quite similar. The somewhat higher Rb content of sample GMZ 02 can be explained by the higher muscovite content (muscovite is known to incorporate Rb). All other trace elements are quite low in concentration. The low concentration of the transition metals, as an example, can be explained by the low Fe content of the material.

Differential scanning calorimetry expectedly showed an endothermic peak at about 130 $^{\circ }$C (Fig. 3.20a) corresponding to dehydration. The shoulder at higher T points towards a mixed interlayer composition. At 573 $^{\circ }$C a small peak was observed which can be explained by the transition of $\alpha$- to $\beta$-quartz. The dehydroxylation (DHX) temperature is at about 670 $^{\circ }$C which is commonly explained by cis-vacant smectites. This DHX temperature is typical of Fe-poor smectites (Kaufhold et al., 2017) which, therefore, is in accordance with the chemical data presented above.

Fig. 3.21

IR spectra of both GMZ samples (green: dioctahedral smectite, yellow: kaolinite, white: organic material, blue: quartz)

The IR spectra of the samples are shown in Fig. 3.21. The quantitative data is shown in Table 3.5. By IR less smectite was found in both samples but higher cristobalite contents. Comparing this data with the chemical composition, however, shows that too few SiO₂ is present in the samples to validate the high cristobalite contents. Therefore, the XRF-corrected XRD-Rietveld numbers (Table 3.5) are closer to the Rietveld results. IR spectroscopy, on the other hand, proved the presence of a trace of kaolinite (0.5 mass%) in sample GMZ 01. Such small amounts are below the XRD detection limit and hence could only be identified by IR. A couple of bands are indicative of some organic materials present but they are rather low in intensity. Accordingly no CO₂ was found in the mass spectrometer curve (Fig. 3.20c) which points to organic matter contents well below 0.1 mass% (which in turn corresponds to the LECO data, Table 3.6).

3.3 Geochemistry (Bentonite)

Janis Pingel, Thorsten Schäfer, Yasmine Kouhail, Muriel Bouby, Frank Heberling, Nikoletta Morélova, Madeleine Stoll, Stephanie Kraft, Nadine Gill, Claudia Joseph, Horst Geckeis

Bentonites are considered as reference buffer materials for high-level radioactive waste repository concepts notably in crystalline rock. In this regard, they represent essential barriers in the repository near-field. Within this section, various investigations on GMZ bentonite (see Sect. 3.2), which is discussed for Chinese disposal concepts, are described:

• related to the potential erosion behaviour when contacted with simulated granitic or natural groundwaters,

• dealing with radionuclide ($^{{3}}$H, $^{36}$Cl, $^{137}$Cs, $^{60}$Co) diffusion through a magnetite/GMZ layer simulating radionuclide migration under conditions where the waste container is corroded.

Experimental results are partly compared with results obtained for similar investigations using Wyoming bentonite (MX80).

3.3.1 Erosion Experiments

A repository concept in crystalline rock relies on the bentonite barrier in different respects: (1) the swelling clay restricts the transport of water and potentially corroding groundwater constituents such as sulphide to the canister surface and (2) in case of canister failure it acts as a strong sorbent against radionuclide transport. Together with the corrosion-resistant container, the bentonite buffer represents an essential barrier responsible for the isolation and containment of highly radioactive material. In Scandinavian safety case considerations, the potential erosion of the bentonite buffer due to the intrusion of low mineralized glacial melt or meteoric water is considered a relevant scenario SKB (2011).

Within the present study GMZ bentonite erosion is examined

• (1) for raw and homoionized GMZ (GMZ 01 and GMZ 02) samples contacted with natural low mineralized groundwater from the Grimsel Test Site (GTS, Switzerland) (I $=$ 1.6 mmol $\textrm{L}^{-1}$; see composition in Schäfer et al. (2004), Table 1) and

• (2) by contacting raw and impurified bentonite (GMZ 01) with a simulated groundwater from the Beishan site (I $=$ 39 mmol $\textrm{L}^{-1}$; [Ca$^{2+}$] $=$ 2 mmol $\textrm{L}^{-1}$; see composition in Table 3.9).

3.3.1.1 GMZ Erosion in Low Mineralized Water

Experiments conducted Five erosion experiments were conducted within the scope of this work (see Table 3.7). Experiments were set up with pellets of natural GMZ 01, GMZ 02 bentonite (comp. Sect. 3.2), or GMZ 01-derived clay fraction sodium-homoionized smectite (Na-smc) mixed with different quantities of accessory minerals (SF800 quartz, Ø $=$ 2 $\upmu$m (HPF, 2015); AGR40 bassanite, Ø $=$ 10.8 $\upmu$m (CASEA, 2013). Here, the aim was to investigate the different swelling and erosion behaviour of the natural (raw) bentonites compared to the impact of accessory minerals on the artificial Na-smc. With that, 10 wt.% quartz was chosen for each of the three artificial samples as it resembles a typical amount of quartz in many natural bentonites (see Sect. 3.2 or e.g. IAEA (2013)). In the case of the accessory bassanite, the goal was to supply the systems with an additional interior calcium source for the cation exchange process to positively influence the system concerning erosion resistance and therefore barrier integrity. Samples were placed into an artificial Plexiglas-built parallel plate, non-sloped chamber of 85 mm diameter and 1 mm aperture and were continuously flushed for 77 days with a constant flow rate of 50 ± 1 $\upmu$l/min of low mineralized (calcium concentration: 5.73 ± 0.46 mg/L) natural Grimsel Ground Water (GGW), extracted from the LIT Pinkel surface packer at the Grimsel Test Site (GTS). Sample water was extracted from the individual setups periodically and was stored refrigerated for further treatment or analysis, respectively. Hydrochemical (IC, ICP-OES, pH, electrical conductivity) and particle characterization analyses (NTA; Nanoparticle Tracking Analysis—Malvern Panalytical NanoSight NS300) were conducted continuously on sampling aliquots. Swelling pressure time-resolved evolution was monitored in situ ($\textrm{TekScan}^{\textcircled {R}}$ pressure sensors). Sample pellets, in turn, were made from ca. 1 g material with a target bulk density of 1.8 ± 0.05 $\mathrm{g/cm}^{3}$, referring to previously conducted experiments, e.g. (Kiviranta and Kumpulainen, 2011; Svensson et al., 2011; Bouby et al., 2020).

Table 3.7 Overview of major properties for the conducted erosion experiments. A more detailed overview of the mineralogical composition of the raw bentonites can be found in Sect. 3.2

Results

Swelling and pressure evolution

In the case of the GMZ 01 bentonite, the expanded material showed a mostly homogeneous distribution with an insinuated tear-shaped form towards the outlet (see Fig. 3.22). Thereby, highly visible erosion streaks emerged from the beginning of the experiment. A maximum swelling radius was achieved with ca. 30 ± 4 mm after 56 days (see Fig. 3.23). Initial contact pressure (total force measured relative to the contact area upon the pressure sensor) peaked at up to 532 kPa but continuously declined afterwards, reaching zero pressure after 56 days. The expanded GMZ 02 showed a rather diffuse transition between its internal sections. However, as for the GMZ 01 bentonite, visible erosion streaks were observed from the start of the experiment towards the outlet. A maximum radial swelling was reached at 21 ± 1 mm after 35 days. Due to the comparably lower smectite content, an initial contact pressure peak was measured with 449 kPa, reaching zero pressure after 28 days. With an initially similar swelling pace as observed for the raw GMZ bentonites, maximum radial swelling was reached at 16 mm after 10 days for the 90/10 (GMZ 01) sample. Furthermore, visible erosion started after 7 days, perceptibly visible as an emerging whitish ring at the gel/water interface. Subsequently, significant amounts of clay were washed out after 28 days, leaving a whitish framework of quartz grain skeleton within the chamber. While the erosion process was retarded compared to the 90/10 (GMZ 01) sample, a similarly pronounced erosion event was observed for the 2 wt.% bassanite (GMZ 01) sample, starting after ca. 14 days, although, the gel/water interface seemed to be fringier in the latter case. In contrast, no significant erosion was observed for the 5 wt.% bassanite (GMZ 01) setup. Due to a higher calcium supply by the total mass of dissolving bassanite and an inferior swelling capability of Ca-montmorillonite, the swelling was severely reduced, reaching a maximum swelling distance of ca. 7 ± 0.5 mm. Regarding the contact pressure, the artificial samples were measured with much less pronounced values than the raw GMZ bentonites, peaking between 200 and 300 kPa. Furthermore, samples showed a continuous pressure decline afterwards, albeit slower than the raw GMZ bentonites. Here, zero pressures were reached after 14 or 28 days in the case of the 90/10 (GMZ 01) and 2 wt.% bassanite (GMZ 01) samples, respectively. Of all samples, though, the 5 wt.% bassanite sample was the only stabilizing at around 100 kPa, reaching no zero pressure values during the experiment.

Fig. 3.22

Top view on applied samples after 7 or 28 days, respectively

Fig. 3.23

Radial swelling distance (left) and contact pressure (right) over time

Hydrochemical propagation

No significant pH changes in eluates were observed for the raw GMZ bentonites over time, ranging at ca. ± 0.2 with respect to the applied GGW (see Fig. 3.24). Concerning the artificial samples, though, especially the erosive 90/10 (GMZ 01) and 2 wt.% bassanite (GMZ 01) samples were measured with more pronounced pH value decrease during the erosion events (pH ~ 6.9). While pronounced conductivity increase was measured for the initial phases of Ca-carrier setups at up to 195 $\upmu$S/cm for the 5 wt.% bassanite (GMZ 01) sample, no significant variances were measured for the other setups. Thus, no direct correlation between the electrical conductivity and erosion events could be traced. Eventually, all samples adjusted within proximity of the applied GGW concerning pH: 7.4 ± 0.1 and electrical conductivity: 112 ± 3 $\upmu$S/cm. Note that all experiments were conducted under atmospheric conditions, thus affecting the pH of the applied GGW.

Fig. 3.24

All setups pH and electrical conductivity evolution over time

Due to the cation exchange process, calcium was stripped from the applied GGW for all samples. Thereby, the most pronounced concentration decrease was measured for the 90/10 (GMZ 01) sample (see Fig. 3.25), where almost all calcium was extracted from the aqueous phase (Ca $\approx$ 0.3 mg/L), correlating to the observed erosion event between days 7 and 28 (see Fig. 3.25). Decreasing concentrations were also measured for the 2 wt.% bassanite (GMZ 01) setup (Ca $\approx$ 1.3 mg/L) and the GMZ 01 setup (Ca $\approx$ 3.0 mg/L) after 21 and 35 days, respectively. However, these were less pronounced due to the release of calcium by the dissolving artificial Ca-carrier (dissolution rate ca. 4.5 ± 0.5E-4 $\mathrm{mol/g}\cdot$s (Brandt and Bosbach, 2001) or the initial $\textrm{Na}^{+}$/$\textrm{Ca}^{2+}$/$\textrm{Mg}^{2+}$ surface charge in the case of the raw bentonites. In contrast, the 5 wt.% bassanite (GMZ 01) sample even experienced a slight concentration increase, however within the measurement uncertainty area, likely due to the higher amount of added bassanite and, thus, calcium release.

Fig. 3.25

Sodium, calcium and sulphate concentration over time for all setups

Unlike calcium, the artificial samples were measured with pronounced concentration peaks in the case of sodium during the initial phases of the respective experiments. Due to the cation exchange of the Na-smc, sodium concentration peaked at up to 34.9 mg/L after 3 days in the case of the 5 wt.% bassanite (GMZ 01) setup. Over time, concentrations decreased as the cation exchange proceeded in favour of $\textrm{Ca}^{2+}$, and subsequently, less sodium was released. While slightly raised values were measured at the beginning of the raw GMZ bentonite setups, lower initial surface cation Na-occupancy (see Sect. 3.2) and exchange capacity (see Table 3.8) resulted in the inferior release or exchange, respectively.

While slightly increased values were measured for magnesium and potassium at the beginning of each experiment, values did not exceed 0.5 mg/L with respect to the applied GGW (Mg $=$ 0.014 mg/L, K $=$ 0.141 mg/L). While the GMZ 02 and artificial Na-smc experiments adjusted within proximity of the applied GGW shortly after the beginning, the GMZ 01 setup was measured with an additional magnesium and potassium peak, coinciding with the concentration decrease at around 35 days.

In the case of silicon and aluminium, increased concentration values were correlated with the respective particle releases of all setups, especially pronounced for the 90/10 (GMZ 01) and the 2 wt.% bassanite (GMZ 01) experiment.

While different amounts of bassanite were added for the 2 wt.% and 5 wt.% bassanite experiments, different high sulphate peaks of ca. 30 and 40 mg/L were measured during the initial phase of the experiments. As the added bassanite slowly dissolved, both setups adjusted within proximity of the applied GGW (6.47  ± 0.24 mg/L) after ca. 21 and 70 days, respectively. As the GMZ 01 and GMZ 02 raw bentonites do not contain natural amounts of any soluble sulphate Ca-carrier like gypsum or anhydrite, no SO₄$^{2-}$ release was measured during the experiments.

Fig. 3.26

Median hydrodynamic diameter (left) and colloidal concentration (right), measured by NTA. An intentional flow stop for GMZ 01 and GMZ 02 between days 35 and 56 is marked as dotted lines

Table 3.8 CEC (meq/100g) and element concentration (%) of major exchangeable cations for different sections harvested after the respective experiments. The relative abundance was thereby calculated concerning the ion’s mol-weight

Particle characterization

Concerning particle characterization in the eluates in contact with the bentonite source, similar median hydrodynamic diameters were measured for all setups conducted, ranging primarily between 100 and 200 nm (see Fig. 3.26), which is near the measured background of GMZ 01 colloids (177 ± 9 nm). While no dependencies regarding the swelling and erosion behaviour of the setups were observed, a slightly increasing trend was observed for all setups. This, however, is likely linked to an easier release of smaller particle size during the initial swelling phase of the experiments.

Distinct colloidal particle concentrations were measured for the different setups conducted. Thereby, the raw GMZ 01 and GMZ 02 setups were measured with a similar but stable colloidal release of ca. 2.4 ± 1.7E+09 particles/mL during the first 60 days of the experiments (see Fig. 3.26). Afterwards, a decline towards lower ca. 2.0E+08 particles/mL concentrations was likely induced by an intentional flow stop of 3 weeks (days 35–56) and a prolonged exchange time for the surface cations. In contrast, the 90/10 (GMZ 01) setup was measured with significantly increased concentration values during the erosion event, peaking at 9.7 ± 0.2E+10 particles/mL after 14 days. Afterwards, particle release decreased due to the significant washout of clay material. Thus, reaching concentration values within proximity of the applied GGW after ca. 35 days. In the case of the Ca-carrier experiments, 2 wt.% bassanite was non-sufficient to stabilize the system from erosion. Thus, increased particle concentrations of up to 8.6 ± 0.5E+10 per mL were measured. However, due to the released calcium, the process was retarded compared to the 90/10 (GMZ 01) setup. Eventually, though, a significant amount of clay was also washed out, decreasing the number of emitted colloids over time. Contrastingly, 5 wt.% bassanite successfully enhanced the system integrity by supplying sufficient amounts of calcium, thus increasing the cation exchange in favour of $\textrm{Ca}^{2+}$. Subsequently, low amounts of colloids were released with concentration values close to the natural concentration of the GGW contact water.

By applying the NTA quantification approach of Mehrabi et al. (2017), mass balances were approximated for all experiments conducted. The approach thereby evaluates the NTA-based data by considering the particle size distribution histogram in a given size range (bins). Subsequently, the average number concentration of each bin was multiplied by its associated spherical particle diameter $d_{i}$ and thereof its respective particle volume ${\overline{v}}_{c,i}$. With a typical montmorillonite density of ($\rho$ $=$ 2.35 $\mathrm{g/cm}^{3}$, e.g. (Duda et al., 1990)), mass approximation was performed by summing each bin:

$$\begin{aligned} m_{c} = \rho \sum _{i = 1}^{n}\left( {\overline{v}}_{c,i}\frac{1}{6}\pi d_{i}^{3} \right) \end{aligned}$$

(3.18)

In order to validate the method, a five-point calibration of GMZ 01 clay colloids was conducted, showing an overall uncertainty of ca. ± 10% concerning the target values.

Figure 3.27 shows the cumulative masses of all experiments conducted to have a high consensus with the observed clay erosion (comp. Fig. 3.22). While the highly erosive setups of 90/10 (GMZ 01) and 2 wt.% bassanite (GMZ 01) were calculated to have lost 94% and 81% ± 10% of their initial clay material, with most material being eroded within the first 20 to 35 days, the 5 wt.% bassanite (GMZ 01) setup was calculated to have lost a mere 0.5% of its initial clay material due to the enhanced calcium supply. In the case of the raw GMZ bentonites, the GMZ 01 was calculated with a lesser clay release of 12.6% compared to the GMZ 02, which eventually lost 32.6% of its original clay material during the experiment.

Fig. 3.27

Approximated clay erosion (%) as cumulative curves based on the calculation using Eq. 3.18

With respect to the increased Al, Si and Mg concentrations measured, which were especially pronounced for the 90/10 (GMZ 01) and 2 wt.% bassanite setups, Fig. 3.28 shows the hydrochemical data correlated with the respective colloidal mass. Thus, the NTA determined colloidal size fraction released during erosion (comp. Fig. 3.26) is clearly not removed during the pre-filtering with 0.45 $\upmu$m cellulose acetate filters and partly measured in the acidified samples. Unfortunately, for these experiments no unfiltered aliquots for quantitative analysis of the erosion masses were measured. However, the clear correlation of structural elements (Al, Si, Mg) with the colloidal mass of the samples indicates the release of clay colloids. However, as higher Si concentrations were measured with respect to the structural Al/Si ratio of montmorillonite (ca. 1:4), an additional but subordinated source of quartz colloids can be expected. In case of the raw bentonites, no such surplus of silicon was measured, though (Al/Si ratio of ca. 1:4).

Fig. 3.28

Correlation diagrams of the Al, Si and Mg concentrations with the calculated erosion masses, exemplified for at 90/10 (GMZ 01) setup

Fig. 3.29

Exemplified for the GMZ 01 bentonite after 28 days (comp. Fig. 3.22); the image highlights the different sections (core, inner ring, outer ring) harvested after the experiments

Cation exchange capacity and ion occupancy

After each experiment, samples were harvested from the remnant material within the erosion chambers.

Thereby, different sample sections were collected (core, inner ring, outer ring) to investigate potential differences within the expanded material (see Fig. 3.29) concerning changes in cation exchange capacity (CEC) and occupancy of major cations at the montmorillonite surface. As only small amounts of sample material were originally emplaced (ca. 1 g), the Cu(II)-trien method after Lorenz et al. (1999) and adjusted by Dohrmann et al. (2012) was applied, as hence even small amounts of sample material (<100 mg) can be used to determine the CEC. Table 3.8 compares measured CEC values and major cation concentrations for all harvested sample sections and the raw material.

CEC measurements of the raw bentonites determined values of 75.6 ± 0.6 meq/100g in the GMZ 01 bentonite and 56.9 ± 0.8 meq/100g in the GMZ 02 bentonite, which is in accordance with the data presented in Sect. 3.2. Furthermore, the initial occupancy of the surface cations was also determined by applying ICP-OES. Thus, showing the $\textrm{Na}^{+}$/$\textrm{Ca}^{2+}$/$\textrm{Mg}^{2+}$ exchangeable cation composition (CEC) of the raw GMZ bentonite material and the artificial montmorillonite being primarily Na-exchanged.

Due to the cation exchange process in contact with the applied GGW and/or bassanite dissolution, surface cations were exchanged differently for the harvested sections. Thereby, the setups solely flushed with GGW were measured to have had an increasingly higher surface cation exchange in favour of calcium towards the outer ring sections due to the selective sorption of calcium from the applied GGW and slow diffusion process through the compacted clay material. However, cations were not completely exchanged during the experiments, leaving $\textrm{Na}^{+}$/$\textrm{Ca}^{2+}$/$\textrm{Mg}^{2+}$ ratios of ca. 14/78/5% in the case of the GMZ 01 and 33/54/10% in the case of the GMZ 02 setup for their respective outer ring sections. Due to the significant clay erosion of the 90/10 (GMZ 01) setup, a single sample of the remaining quartz grains was harvested. Here, a low CEC value of ca. 3 meq/100g was measured, with major cations being below the detection limit of the ICP-OES. Differently, cation exchange was observed to be vice versa for the 2 wt.% bassanite setup, with slightly higher Ca values in the inner ring section. However, as a significant clay release was observed, no core section was harvested (comp. Fig. 3.22), and considerably reduced CEC values were measured. Due to the observed low swelling process of the 5 wt.% bassanite sample, only a single sample was harvested. While the sample CEC was measured to be similar in comparison to the raw bentonites, the exchange in favour of calcium was enhanced (91.8 ± 0.3%) due to applied GGW (external supply) and the bassanite-related calcium (inner supply), demonstrating the efficiency of stabilization against erosion in this sample.

3.3.1.2 GMZ Erosion in Simulated Beishan Groundwater

Experiments conducted Two GMZ bentonite pellets were prepared by hydraulic pressing (19 mm diameter, 10 mm height, dry density: 1,6 $\mathrm{g/cm}^{{3}}$) and emplaced in separate compartments of a transparent double-sided erosion cell (see Bouby et al. (2020) and Fig. 3.30). The pellets were confined between a PEEK spacer and a porous stainless-steel filter (pore width: 20 $\upmu$m) simulating a scenario of a fracture filled with porous fracture-filling minerals.

Fig. 3.30

Double-sided erosion test cells with bentonite pellets

Table 3.9 Composition of simulated Beishan groundwater acc. to Zhu et al. (2022) and as prepared (note that there is a difference of nominal and actual concentrations for Ca); n.a.: not analysed

Synthetic porewater from Gansu Beishan (Table 3.9) was circulated for 155 days with a rate of 3 $\upmu$L/min at pH $=$ 8.5. Visual inspection through the transparent reaction cells shows that hydration of the bentonite is complete after 3 days.

Results Following the evolution of the porewater solution over time shows constant concentrations for most constituents. A slight decrease in $\textrm{Ca}^{2+}$ and $\textrm{Mg}^{2+}$ concentration by ~ 10 and 15%, respectively, suggests sorption to the GMZ bentonite. On the other hand, the $\textrm{Na}^{+}$ concentration in the circulating porewater increases by about 8% indicating an ion exchange reaction where $\textrm{Na}^{+}$ is released and $\textrm{Ca}^{2+}$ is enriched in the bentonite. An initial decrease in $\textrm{K}^{+}$-concentration is followed by a continuous increase. The $\textrm{Sr}^{2+}$ concentration is increasing as well steadily (not shown in Fig. 3.31) over the entire observation period even though at a very low concentration level. All those findings point to an ongoing ion exchange reaction, which is not yet in equilibrium after 200 days, even though bentonite saturation was achieved after 3 days by visual inspection of the transparent cells. No variation could be seen for the anion concentrations and only a minor pH change from 8.54 to 8.37 is observed. Al(III) could be taken as a clay colloid erosion indicator (as, e.g. seen in Bouby et al. (2020)). Respective concentration data analysed by ICP-MS are, however, very much scattered and an elution peak during the initial bentonite swelling process is not detectable. A slightly transient increase of the Si concentration might be interpreted as a temporary release of clay colloids. Bentonites are, however, known to contain reactive and soluble silica components (see, e.g. Kaufhold et al. (2020)) so that the Si-elution peak could also be due to silicate leaching. However, no reactive silica species are reported for GMZ bentonite (see Kaufhold et al. (2020), Sect. 3.2). No other indicators for clay colloids were visible. The initial erosion rate determined with the same setup for raw MX80 bentonite and a low mineralized groundwater simulate (I $=$ 1.16 mmol $\textrm{L}^{-1}$, [$\textrm{Ca}^{2+}$] $=$ 0.05 mmol $\textrm{L}^{-1}$) after 342 days was at 0.043 ± 0.002 kg $\textrm{y}^{-1}$ $\textrm{m}^{-2}$ (Bouby et al., 2020). The findings made in experiments with simulated Beishan groundwater (I $=$ 39 mmol $\textrm{L}^{-1}$, [$\textrm{Ca}^{2+}$] $=$ 2 mmol $\textrm{L}^{-1}$) are in-line with previous studies suggesting that $\textrm{Na}^{+}$ and $\textrm{Ca}^{2+}$ concentrations larger than critical coagulation concentrations ($\textrm{CCC}_\mathrm{{Na}}$ $=$ 10–100 mmol $\textrm{L}^{-1}$; $\textrm{CCC}_\mathrm{{Ca}} =$ 0.1–1 mmol $\textrm{L}^{-1}$, (García-García et al., 2007; Seher et al., 2020) prevent significant bentonite erosion. As a next step, the simulated Beishan groundwater will be exchanged against a low mineralized groundwater from the Grimsel Test Site in order to study GMZ bentonite erosion under the same conditions.

Fig. 3.31

Concentration evolution of some groundwater simulate constituents during the erosion experiment; dashed lines indicate the initial concentration level

3.3.2 Diffusion Experiments

Different materials are considered for waste canisters in deep geological repositories (DGR) in crystalline rock. While Finland and Sweden foresee “corrosion-resistant” copper-coated canisters, other countries discuss as well “corrosion-acceptable” concepts consisting of thick, double-walled steel containers (e.g. outer shell made of carbon-steel, inner shell made of stainless steel) (see, e.g. Pospiskova et al. (2017)). In the latter case, the iron-based canisters are considered to corrode in contact with groundwater over a period of thousands of years partly at the elevated temperatures of the thermal phase. Under those conditions, secondary iron phases such as magnetite are expected to form and dissolved Fe(II) species may react with the bentonite (see, e.g. Chaparro et al. (2021); Féron et al. (2009); Kaufhold et al. (2020)). Thus, radionuclides released from the waste over longer time-scales would not be in contact with pristine bentonite only, but with a corrosion-induced alteration layer adjacent to the bentonite. In the present study, radionuclide diffusion through a magnetite/bentonite interface is studied in order to assess the long-term barrier function of an altered bentonite.

3.3.2.1 Experiments

Four through-diffusion experiments (Fig. 3.32) through GMZ bentonite and through a mixed porous medium of magnetite and GMZ bentonite were performed using a pre-equilibrated synthetic pore water of Gansu Beishan (composition described in Table 3.10) at room temperature at a reservoir circulation flow rate of 0.4 mL/min. 0.4 mM $\textrm{NaHCO}_{{3}}$ were added to the porewater after the necessary percolation with argon to minimize dissolved oxygen. After a pre-equilibration step of 3 weeks in contact with bentonite in a glove box under Ar atmosphere, the composition of the pore water was measured by ICP-OES and ion chromatography, and pH and Eh were measured and reported in Table 3.10. The pore water in the high reservoirs of the mixed porous media was additionally equilibrated with magnetite at 12.5 mg/L.

Fig. 3.32

Scheme of the setup of through-diffusion experiments

Table 3.10 Composition of the synthetic GMZ pore water in contact with groundwater from Gansu Beishan acc. to Wu et al. (2014) and after equilibration; n.a.: not analysed

Cylindrical pellets of GMZ 01 bentonite (25.64 mm diameter, 10 mm height or 8.04 mm height) and magnetite (25.6 mm diameter, 2.2 mm height) were compacted using a hydraulic press with a dry density of 1.6 and 1.7 $\mathrm{g/cm}^{{3}}$, respectively. Magnetite was prepared acc. to Schwertmann and Cornell (2000). Briefly, equimolar aqueous solutions of $\textrm{FeCl}_{{2}}$ and $\textrm{FeCl}_{{3}}$ were mixed, and pH adjusted to 8.25–9 using a NaOH solution. The precipitate was washed and dialysed. The dried solid was poorly crystalline with crystallite sizes of ~ 10 to 17 nm. The magnetite pellets were compacted in a glove box under argon atmosphere. Stainless steel diffusion cells were packed with a pellet of GMZ 01 bentonite or magnetite and bentonite in each cell with a stainless-steel filter (type 316 L, 10 $\upmu$m pore size) at each end of the cell. The magnetite/bentonite pellets were then pre-equilibrated for 4 weeks with the synthetic pore water.

The diffusion experiments through the bentonite and the mixed porous media of magnetite and GMZ bentonite were started in a glove box after the pre-equilibration by first using non-sorbing radiotracers ($^{36}$Cl and tritiated water HTO with an activity of 1000 Bq/mL in the high concentration reservoirs) to investigate porosity and diffusion coefficients. The non-sorbing radiotracers were measured by liquid scintillation counting (LSC) using an LSC Ultima Gold XR cocktail (Perkin Elmer).

After out-diffusion of the non-sorbing tracers, the diffusion of the sorbing tracers $^{137}$Cs(I) and $^{60}$Co(II) was investigated. The experiments were carried out by using an activity of 10 kBq/mL of $^{137}$Cs(I) (carrier free) or $^{60}$Co(II) (carrier: 17 ng/mL Co) in the high reservoir of each cell. The low reservoirs were sampled periodically and the activities of $^{137}$Cs(I) and $^{60}$Co(II) were measured by gamma spectroscopy and diffusion coefficients were determined.

3.3.2.2 Results and Discussion

Figures 3.33, 3.34 and 3.35 contain experimental data of diffusion experiments performed so far. Only small variations in ion concentrations can be seen in the reservoir solutions after pre-equilibration of the simulated porewater (Wu et al., 2014) with the magnetite/bentonite samples (Table 3.10, 2${\textrm{nd}}$ column). Slightly increasing concentrations for $\textrm{Na}^{+}$, $\textrm{Cl}^{-}$ and SO₄$^{2-}$ suggest the dissolution of soluble compounds from the non-purified GMZ. On the other hand, a moderate initial decrease in $\textrm{Ca}^{2+}$ concentration may point to an ion exchange process. Fe concentration establishes at a concentration level of ~0.01–0.02 mmol $\textrm{L}^{-1}$. Invariant pH points to an overall equilibrium during the diffusion experiment. The Eh value only slightly decreased during the experiment, showing that an overall redox equilibrium with magnetite within the entire diffusion setup was obviously not attained.

Fig. 3.33

Accumulated activity (in Bq) and flux (in Bq.$\textrm{d}^{-1.}\textrm{cm}^{-2}$) as a function of diffusion time (in days) for HTO and $^{36}$Cl through GMZ bentonite, effective diffusion coefficients and rock capacity factors in cell 1 (a & b)

The effective diffusion coefficients and rock capacity factors were calculated from the analytical solution of Fick’s second law using the experimental results of through-diffusion tests (see, e.g. Van Loon and Soler (2004)) and are summarized in Table 3.11. Note that in this simplified view, the heterogeneity of the magnetite/bentonite sample is not considered. Respective parameters have to be considered as a first indication for variations in diffusion properties in the bentonite due to the presence of magnetite. HTO diffusion parameters are comparable with the ones reported in Wu et al. (2014), with GMZ ($D_e$ $=$ (1.12 ± 0.06) x 10$^{-10}$ $\textrm{m}^{2}$/s and $\alpha$ $=$ 0.44 ± 0.02). From the HTO diffusion experiments, a slight increase of porosity can be stated for the magnetite/GMZ system as compared to the diffusion experiment with GMZ alone (0.536/0.448 for GMZ alone and 0.625/0.577 for magnetite/GMZ). This is to be expected as the magnetite will not undergo swelling as the bentonite does. The clear decrease of $\alpha$ derived from the $\textrm{Cl}^{-}$-diffusion data as compared to the HTO experiment with GMZ is clearly due to the anion exclusion effect, restricting the accessible porosity of anions in the bentonite. It is remarkable that the anion exclusion effect strongly decreases in the magnetite/GMZ sample, i.e. the rock capacity factor increases. Such an effect was not expected and potentially points to some kind of a charge screening effect by dissolved Fe species, being released from the magnetite and interacting with bentonite.

Fig. 3.34

Accumulated activity (in Bq) and flux (in Bq.$\textrm{d}^{-1.}\textrm{cm}^{-2}$) in function of diffusion time (in days) for HTO and $^{36}$Cl through the magnetite/GMZ bentonite interface, effective diffusion coefficients and rock capacity factors in cell 3 (a and b)

Fig. 3.35

Accumulated activity (in Bq) and flux (in Bq.$\textrm{d}^{-1.}\textrm{cm}^{-2}$) in function of diffusion time (in days) for $^{137}$Cs and $^{60}$Co through the magnetite/GMZ bentonite interface, effective diffusion coefficients and rock capacity factors in cell 3 (a) and cell 4 (b)

Diffusion experiments with sorbing tracers $^{60}\textrm{Co}^{2+}$ and $^{137}\textrm{Cs}^{+}$ were only performed with GMZ/magnetite samples so far. Sorption coefficients (K_d-values) derived from rock capacity factors $\alpha$ obtained from through-diffusion data should be considered preliminary for the same reason as discussed already above. Post-mortem analyses are under way to analyse diffusion profiles using the abrasive peeling technique (see Van Loon and Müller (2014)) and an improved diffusion modelling approach, taking the different material layers into account is currently tested.

The K_d-value for Cs derived from the rock capacity factor of this study matches perfectly with the sorption model from Bradbury and Baeyens (2002b), and the respective K_d-values derived therein for the diffusion in compacted bentonite (10 L/kg compared to 2.4 L/kg in the present study). Those results demonstrate that Cs-sorption does not differ much in a pure bentonite and a layered magnetite/bentonite system. The slightly lower K_d-value obtained from GMZ/magnetite experiments may as well point to a competing effect of dissolved Fe species. Cs-K_d-values reported in Molera and Eriksen (2002) are way higher (200–580 L/kg) but have been derived considering surface diffusion phenomena, so that those values cannot be compared with those in the present study.

Table 3.11 Effective diffusion coefficients and rock capacity factors of HTO, $^{36}$Cl, $^{137}$Cs and $^{60}$Co through GMZ bentonite and magnetite/GMZ bentonite interface; note that parameters for the GMZ/magnetite samples are the result of a simplified modelling approach (see text for more details)

A K_d-value derived for $\textrm{Co}^{2+}$ from data obtained in diffusion experiments in magnetite/GMZ is way lower (K_d $=$ 4.2) than expected according to sorption modelling and published experimental results obtained for $\textrm{Co}^{2+}$ sorption/diffusion onto/in bentonite (Bradbury and Baeyens (2002a): 400 L/kg for $\textrm{Ni}^{2+}$ and Molera and Eriksen (2002): 2400–2600 L/Kg; Missana et al. (2007): 14.000 L/kg). However, as pointed out earlier, data are not directly comparable ($\textrm{Ni}^{2+}$ has been investigated in one case, a surface diffusion approach has been applied in the other). Nevertheless, the quite low sorption value obtained in the present study may also be a consequence of competitive sorption of Fe species.

Data for $\textrm{Co}^{2+}$ and $\textrm{Cs}^{+}$ sorption onto magnetite have been reported by Ebner et al. (2001) (Cs ~20 L/kg, Co ~90 L/kg) and are also higher than found in our diffusion study. However, they were determined in 0.01 M $\textrm{NaNO}_{{3}}$. The higher ionic strength conditions of the present experiments and the potential competition with dissolved Fe species might explain the differences.