Scientific consensus exists that using deep geological repositories (DGRs) is a suitable approach to permanently dispose of spent nuclear fuel and high-level radioactive wastes (HLRWs; e.g. IAEA, 2003). Central to the safety concepts of DGRs is the implementation of a multi-barrier system, in some concepts with a buffer – typically composed of highly compacted and/or granular bentonite – playing a crucial role (Sellin & Leupin, Reference Sellin and Leupin2014).
Understanding the long-term behaviour of this clay-based barrier is essential, and greater understanding of such behaviour can be obtained using various approaches: (1) knowledge of the fundamental chemical–physical properties of clay minerals (Dohrmann et al., Reference Dohrmann, Kaufhold, Lundqvist, Bergaya and Lagaly2013a); (2) natural analogue studies (Miller et al., Reference Miller, Alexander, Chapman, McKinley and Smellie2000); (3) natural tracer studies in clay barriers (e.g. Mazurek et al., Reference Mazurek, Alt-Epping, Bath, Gimmi, Waber and Buschaert2011); and (4) laboratory tests. For example, examining factors such as swelling capacity (e.g. Komine, Reference Komine2004; Gens et al., Reference Gens, Sánchez, Guimarães, Alonso, Lloret and Olivella2009), solute transport in engineered clay barriers (e.g. Idiart et al., Reference Idiart, Laviña, Cochepin and Pasteau2020), alteration processes in contact with cement (e.g. Mäder et al., Reference Mäder, Jenni, Lerouge, Gaboreau, Miyoshi and Kimura2017) or steel (e.g. Wersin et al., Reference Wersin, Johnson and McKinley2007) or erosion (e.g. Moreno et al., Reference Moreno, Liu and Neretnieks2011) aims to verify the set of requirements for a DGR. Moreover, large-scale experiments, such as FEBEX at the Grimsel Test Site (GTS; Kober et al., Reference Kober, García-Siñeriz, Villar, Lanyon, Cloet and Mäder2021), the Prototype Repository at Äspö (Svemar et al., Reference Svemar, Johannesson, Grahm, Svensson, Kristensson, Lönnqvist and Nilsson2016), HE-E (Lanyon & Kober, Reference Lanyon and Kober2024) and FE at Mont Terri (Müller et al., Reference Müller, Garitte, Vogt, Köhler, Sakaki, Weber and Cloet2017), offer invaluable insights under more realistic conditions.
Modern sensor monitoring techniques enable precise tracking of changes within a buffer material (e.g. Müller et al., Reference Müller, Garitte, Vogt, Köhler, Sakaki, Weber and Cloet2017; Kober et al., Reference Kober, Schneeberger, Vomvoris, Finsterle and Lanyon2023). However, mineralogical changes occurring over the lifespan of such experiments may only be discernible through post-dismantling analysis and comparison with an initial characterization (e.g. Karnland et al., Reference Karnland, Olsson, Dueck, Birgersson, Nilsson and Hernan-Håkansson2009; Dohrmann et al., Reference Dohrmann, Olsson, Kaufhold and Sellin2013b; Kaufhold et al., Reference Kaufhold, Dohrmann, Sandén, Sellin and Svensson2013a, Reference Kaufhold, Dohrmann, Götze and Svensson2017, Reference Kaufhold, Dohrmann, Ufer and Kober2018, Reference Kaufhold, Dohrmann, Ufer, Svensson and Sellin2021, Reference Kaufhold, Dohrmann, Wallis and Weber2023; Dohrmann & Kaufhold, Reference Dohrmann and Kaufhold2014; Kumpulainen et al., Reference Kumpulainen, Kiviranta and Korkeakoski2016; Fernández et al., Reference Fernández, Kaufhold, Sánchez-Ledesma, Rey, Melón and Robredo2018). Additionally, large-scale experiments necessitate substantial bentonite emplacement, providing opportunities to broaden knowledge on bentonite production, including regarding quality control and bentonite performance (e.g. Kaufhold & Dohrmann, Reference Kaufhold and Dohrmann2016).
The HotBENT experiment, a recent in situ test at the GTS (www.grimsel.com), focuses on understanding potential alterations of the bentonite buffer under elevated temperatures and at relevant scales and gradients (Kober et al., Reference Kober, Schneeberger, Vomvoris, Finsterle and Lanyon2023). This experiment, spanning up to two decades, utilizes electrically powered heaters (running at 200°C and 175°C; Fig. 1) to mimic the heat generated by waste canisters, with granular bentonite material (GBM) backfilling the space (Kober et al., Reference Kober, Schneeberger, Vomvoris, Finsterle and Lanyon2023). GBM was subsequently used to backfill the space between the heater and pedestal and the tunnel wall (Fig. 1). Two different types of bentonites were utilized, with a Na-dominated bentonite stemming from Wyoming, USA, being used for three heater modules of the experiment and a Czech Mg/Ca bentonite (BCV, bentonite Černý vrch; Laufek et al., Reference Laufek, Hanusová, Svoboda, Vašíček, Najser and Koubová2021; Najser et al., Reference Najser, Mašín, Svoboda, Vašíček, Hanusová and Hausmannová2023; Svoboda et al., Reference Svoboda, Mašín, Najser, Vašiček, Hanusova and Hausmannová2023) being used for the fourth heater (Fig. 1). Four boreholes drilled parallel to the tunnel, which can be pressurized with Grimsel groundwater, ensure sufficient water supply to the geosphere. The detailed composition of the Na-dominated, slightly alkaline groundwater is given by Schneeberger et al. (Reference Schneeberger, Kober, Lanyon, Mäder, Spillmann and Blechschmidt2019).
Figure 1. Sketch showing the experimental setup of the HotBENT experiment (Kober et al., Reference Kober, Schneeberger, Vomvoris, Finsterle and Lanyon2023) ©Nagra.
Other artificially hydrated experiments (e.g. ‘LOT’, Karnland et al., Reference Karnland, Olsson, Dueck, Birgersson, Nilsson and Hernan-Håkansson2009; ‘ABM’, Kaufhold et al., Reference Kaufhold, Dohrmann, Sandén, Sellin and Svensson2013a; ‘PTR’, Dohrmann & Kaufhold, Reference Dohrmann and Kaufhold2014) showed that cation exchange readily takes place in these processes, changing the chemical composition without affecting the mineralogical composition (e.g. Dohrmann et al., Reference Dohrmann, Olsson, Kaufhold and Sellin2013b; Dohrmann & Kaufhold, Reference Dohrmann and Kaufhold2014; Kaufhold et al., Reference Kaufhold, Dohrmann, Wallis and Weber2023). Apart from cation exchange, mineral alterations were observed such as gypsum dissolution/precipitation, carbonate recrystallization, and trioctahedralization (formation of trioctahedral domains; Kaufhold et al., Reference Kaufhold, Dohrmann, Sandén, Sellin and Svensson2013a, Reference Kaufhold, Dohrmann, Götze and Svensson2017, Reference Kaufhold, Dohrmann, Ufer, Svensson and Sellin2021; Kumpulainen et al., Reference Kumpulainen, Kiviranta and Korkeakoski2016; Fernández et al., Reference Fernández, Kaufhold, Sánchez-Ledesma, Rey, Melón and Robredo2018). Relevant (thermally induced) reactions observed so far in heater experiments have been summarized by Kaufhold et al. (Reference Kaufhold, Dohrmann, Wallis and Weber2023).
Our study aims to provide mineralogical and chemical reference data for the different bentonites used in the HotBENT experiment, emphasizing the importance of quality control for post-mortem analysis in underground research laboratories. The emplacement of large quantities of bentonite in the HotBENT experiment (>150 tons) allows for extensive characterization of samples at different points in the experiment. Factors such as mineralogical and chemical composition, exchangeable cations (ECs) and cation-exchange capacity (CEC) are evaluated to assess the homogeneity of the bentonite at the time of emplacement. Important barrier parameters such as hydraulic conductivity or swelling pressure were not considered in the present study (which will be addressed in future studies). The aim, therefore, was not to compare the suitability of bentonites or discuss possible specifications of HLRW bentonites but to characterize the bentonite homogeneity and provide reference data for the later comparison of run products.
Materials and methods
The bentonite samples were provided by two different companies. The Wyoming bentonite material was purchased from Cebo Holland BV. Blocks produced from the products ‘National®WP1802’ and ‘National®WP1000’ (in the following referred to as ‘Wyoming’ bentonites, notably not MX80) were used as GBM for backfilling. The BCV bentonite was purchased from Keramost with two grain-size distributions (Schneeberger et al., Reference Schneeberger, Kober, Kaufhold and Dohrmann2022). The grain size for the BCV GBM was between 0 and 1.0 mm and for the blocks the grain size was between 0.6 and 2.0 mm. Approximately 75% the total raw material was pelletized in large, stainless-steel barrels. The pelletization resulted in highly compacted pellets with grain sizes between 0 and 6.0 mm (Schneeberger et al., Reference Schneeberger, Kober, Kaufhold and Dohrmann2022).
In addition, two different types of blocks were produced for the heater pedestals (see Fig. 1): ordinary blocks, which are essentially rectangular, and top blocks, with a curved side that fits to the curved heater/canister (Schneeberger et al., Reference Schneeberger, Kober, Kaufhold and Dohrmann2022). During installation of the experiment, several samples were taken from the backfilled GBM material as well as from the blocks (not used in the actual experiment but contained in the same shipment).
Common industrial quality criteria (cf. Eisenhour & Reisch, Reference Eisenhour, Reisch and Kogel2006) are not useful for characterization of the HotBENT samples as these were developed for specific product and customer requirements. Suitable criteria for the identification of the homogeneity of bentonites used in the HotBENT experiment are mineralogical and chemical composition, as well as ECs and CEC.
Tables 1 & 2 provide overviews of the samples taken and used for analysis.
Table 1. GBM samples used for backfilling.
Table 2. List of samples taken from the blocks (both top and ordinary blocks).
The GBM arrived in plastic bags and hence these samples were termed ‘bag samples’ in the laboratory, but in this study they are referred to as ‘GBM samples’. We took 10 g aliquots from each plastic bag, which contained ~100 g sample mass. The samples were split using a riffle splitter and carefully ground using a mortar mill without further drying for 10 min.
In addition to the GBM samples, 12 block samples were analysed (Fig. 2). The centre of each half-block was sampled by drilling using a 2 cm-diameter steel drill. Material from the uppermost part was not considered. Only material from the centre of the blocks was used (depth 3–6 cm). Overall, ~10 g of sample was collected from each block (Fig. 3). This material was dried at 60°C (including determination of water content) and finally ground in the same way as the GBM samples. With the exception of Wyoming block 1, which had a water content (60°C) of 16 mass%, all Wyoming block samples had water contents (60°C) between 12 and 14 mass%. Water contents (60°C) of the BCV ordinary blocks were 10–11 mass%, while BCV top blocks had significantly higher water contents (60°C) of 16–17 mass%. These values are somewhat smaller compared to those in Schneeberger et al. (Reference Schneeberger, Kober, Kaufhold and Dohrmann2022), which may be due to some drying having occurred throughout storage and/or compaction.
Figure 2. All split block samples sent to BGR (left), sampling technique (centre) and water contents determined after drying at 60°C for each block (right). Water contents determined at 105°C are given in Table 6.
Figure 3. LOI values of all samples (green = BCV; blue = Wyoming).
The differences between the materials were expected to be small, as inferred from the FE experiment (Kaufhold et al. in Garitte et al., Reference Garitte, Weber and Müller2015). Chemical analysis methods such as X-ray fluorescence (XRF), C/S analysis (LECO) and CEC were considered suitable to detect minor differences, and X-ray diffraction (XRD) was used to explain the causes of any differences related to mineral content.
For XRF analysis of powdered samples, a PANalytical Zetium spectrometer was used (ALMELO, The Netherlands). Samples were prepared by mixing with a flux material (lithium metaborate Spectroflux, Flux No. 100A, Alfa Aesar) and melting into glass beads. The beads were analysed using wavelength-dispersive XRF. To determine loss on ignition (LOI), 1 g of sample material was heated to 1030°C for 10 min, including a ramp at 700°C. Overall, the LOI was quite constant (Fig. 3), but for comparison of the elemental compositions all elements were normalized to LOI = 0.
The organic carbon (Corg) content was measured using a LECO CS-444 analyser after dissolution of the carbonates. Carbonates were removed by treating the samples several times at 80°C with HCl until no further gas evolution was 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 as the difference between Ctotal and Corg. The samples were heated to 1800–2000°C in an oxygen atmosphere, and CO2 and SO2 were detected using an infrared detector. This device was built by LECO (MI, USA).
XRD traces were recorded using a PANalytical X'Pert PRO MPD θ-θ diffractometer (Co-Kα 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 sample changer (sample diameter 28 mm). The samples were investigated from 2° to 80°2θ with a step size of 0.03°2θ and a measuring time of 3 s per step. For specimen preparation the back-loading technique was used. For analysis of selected clay fractions <2 μm, suspensions were transferred to flat ceramic discs and oriented parallel to the basal planes. Such fractions were analysed under air-dried (AD) conditions and after solvation with ethylene glycol (EG) from 2° to 40°2θ with a step size of 0.03°2θ and a measuring time of 6 s per step. Relative humidity (RH) was not controlled during AD measurements. Selected blocks were analysed for detection of expandable clay minerals of Wyoming and BCV bentonite after intercalation of EG vapour in smectitic interlayers using a thin layer of powder in a desiccator and intercalation. These samples were then measured as explained previously.
Thermoanalytical investigations of selected samples were performed using a Netzsch 449 F3 Jupiter thermobalance equipped with a differential scanning calorimetry/thermal gravimetry (DSC/TG) sample holder linked to a Netzsch QMS 403 C Aeolus mass spectrometer (MS). We heated 100 mg of powdered material previously equilibrated at 53% RH from 25°C to 1150°C with a heating rate of 10 K min–1. The devices were manufactured by Netzsch (Germany).
Water contents were determined at both 60°C and 105°C. The water contents reported in this study are, therefore, either given as water content (60°C) or water content (105°C). The reason for using both is that samples dried at 60°C were used for further analysis, whereas 105°C dried samples were discharged because of possible cation fixation and reduction of the CEC/swelling properties. CEC values, as an example, are given in reference to the 105°C state. Therefore, a small aliquot of the 60°C dried material (~0.5–0.6 g of each bentonite) was dried for 4 days in an oven at 105°C. The water was calculated based on the lost mass and referred to the initial weight (sample + water), in contrast to geotechnical water content calculation, which is based on the dry (final mass after drying) sample mass.
The CECs of all samples were measured using the Cu-triethylenetetramine complex (Meier & Kahr, Reference Meier and Kahr1999) with 10.0 mL Cu-trien solution plus 50.0 mL water and two different sample masses of 120 and 160 mg. CEC index cation values were analysed using visible (VIS) spectroscopy (Jenway 6200, Cole-Parmer, UK).
For the BCV block samples, EC analysis was necessary to identify the differences between materials and particularly to identify the presence of soluble components such as Na-carbonate. In interlaboratory round-robin tests, the Cu-trien index cation provided accurate CEC and EC population (ECpopulation) values for calcareous bentonites (Dohrmann et al., Reference Dohrmann, Genske, Karnland, Kaufhold, Kiviranta and Olsson2012a, Reference Dohrmann, Genske, Karnland, Kaufhold, Kiviranta and Olsson2012b), and particularly for the exchangeable Ca2+ values if the Cu-trien5xcalcite variant of the method was used. Using the Cu-trien5xcalcite solution suppresses calcite dissolution during the exchange reaction (Dohrmann & Kaufhold, Reference Dohrmann and Kaufhold2009). The Cu-trien5xcalcite solution was prepared by mixing 2000 mL of 0.01 M Cu-trien solution (cf. Stanjek & Künkel, Reference Stanjek and Künkel2016) with 2 g of fine-grained calcite added to saturate the solution with dissolved calcite as described by Dohrmann (Reference Dohrmann2006a, Reference Dohrmann2006b) and Dohrmann & Kaufhold (Reference Dohrmann and Kaufhold2009). In contrast to the primary publication of Meier & Kahr (Reference Meier and Kahr1999), no additional water was added to the clay, but a greater amount of 0.01 M Cu-trien solution was used, allowing a larger sample mass to be added. This results in a lower factor for the calculation of EC and CEC values based on wet chemical analysis of the supernatants after finishing the exchange experiment and centrifugation, and precision of the analysis is improved. It is possible to use 50.0 mL of exchange solution (hence why it is was called 5x) as in Dohrmann & Kaufhold (Reference Dohrmann and Kaufhold2009) or 30.0 mL, but the sample masses have to be adjusted. The sample principle can be chosen to study the samples without addition of calcite using Cu-trien5x as for the present study. In both cases, two different sample masses were used (200 and 300 mg) and 30.0 mL of Cu-trien5xcalcite. Cu-trien5x exchange solution was added to each sample in an 85 mL centrifuge tube. The slurry was allowed to equilibrate for 2 h in an end-over-end shaker. The solutions were then centrifuged to separate the bentonites and the supernatant solutions, which were diluted and analysed using inductively coupled plasma (ICP) spectrometry (Thermo Scientific ICAP 6300 DUO ICP optical emission spectroscopy (OES); Thermo Fisher Scientific, MA, USA) to measure the ECs and Cu.
For ICP-OES analysis, the following settings were used: argon radial plasma, nebulizers (cross-flow and modified Lichte), no auxiliary gas flow, gain value for plasma (1.400 W) and calibration every seventh measurement. The Cu-trien complex concentration was also analysed using VIS spectroscopy to cross-check the ICP-Cu concentration. Each CEC value was calculated by averaging four single CEC values (two from ICP analysis and two from VIS spectroscopy). Each EC value was calculated by averaging only two single EC values measured using ICP. As usual, all EC and CEC values were corrected for water loss (4 days in an oven) based on the 105°C water contents. The error (±3σ) of the values determined using the Cu-trien5xcalcite method for bentonites (Dohrmann & Kaufhold, Reference Dohrmann and Kaufhold2009) was different for each of the ECs and the CEC. The scattering of EC values was lowest for K+ (±0.3 meq 100 g–1), followed by Mg2+ (±0.8 meq 100 g–1), Ca2+ (±0.8 meq 100 g–1), Na+ (±1.9 meq 100 g–1) and the CEC (±3.1 meq 100 g–1). No sample had significant exchangeable Fe (all <0.1 meq 100 g–1 Fe3+), and ECs were measured in meq 100 g–1. The ECs of the various materials were calculated as a percentage of the measured CEC value to make the EC values comparable.
The total nitrogen content was measured with an Elementar vario MAX cube. For analysis, ~400 mg of the dried samples and 800 mg tungsten(VI) oxide (WO3) were heated to 1140°C under an oxygen atmosphere. Nitrogen was released from the sample as N2 or nitrogen oxides and transferred to a reduction column to reduce NOX to N2, to reduce SO3 to SO2 and to trap halogens. The gas was dried, and CO2 and SO2 were removed from the gas flow in two separate adsorption columns. N2 was then detected using a thermal conductivity detector (TCD). The system was built by Elementar Analysensysteme GmbH (Germany).
Water extracts of two BCV block samples were analysed with respect to pH, conductivity and chemical composition. We dispersed 3 g of solid in 220 mL of water, and the mixtures were shaken end over end for 20 h at 10 rpm. We then placed 10 mL into glass cylinders to test dispersibility (stability of the suspensions), and the rest was centrifuged at 4600 rpm for 30 min. The supernatant of the ordinary block sample was clear and could be easily filtered through a syringe filter (0.2 μm). The supernatant of the top block, however, was yellowish and had to be filtered by pressure filtration to obtain a solution that could be analysed using ICP. We used 100 mL of each filtered supernatant for ion chromatography (IC), ICP (explained above) and alkalinity measurements, whereas the rest was used to measure pH and conductivity.
Anions of aqueous extracts (F–, Cl–, Br–, NO3– and SO42–) were analysed using IC (Thermo Scientific Dionex ICS-5000+) based on DIN EN ISO 10304-1 (2009). After separating the ions using a guard column (Dionex IonPac AG19) and analytical column (Dionex IonPac AS19), the single-anion peaks were detected according to electrical conductivity following neutralization of the alkaline KOH eluent using a membrane suppressor technique (ADRS600). The limit of quantification for all anions is 0.003 mg L–1.
Alkalinity (acid-neutralizing capacity) was determined using a 50 mL aliquot of the non-filtrated sample, which was titrated (Mettler-Toledo T90) with 0.05 N HCl down to pH 4.3. The endpoint was determined potentiometrically using a two-cell pH-glass electrode.
A Zeiss Sigma 300 VP field emission gun scanning electron microscope (SEM) operating at 15 kV was used to evaluate samples on the micro-scale using the high-vacuum mode. The microscope was equipped with the following detectors: a Bruker Xflash® 6/60 energy-dispersive X-ray spectroscopy (EDX) detector, a high-definition backscattered electron detector (HDBSD), a secondary electron detector (SE2) and a variable-pressure secondary electron detector (VPSE) and an InLens detector for detection of secondary and backscattered electrons, respectively.
Results
Chemical analysis
The LOI values determined based on 60°C dried samples are shown in Fig. 3. LOI values of the Wyoming GBM were ~11 mass%, whereas LOI values of BCV GBM were ~17 mass%. BCV GBM samples revealed two outliers with LOI values slightly above 18 mass% (Fig. 3). Such differences may result from different water contents and do not point to relevant differences of material properties. The block samples, as an example, were obviously not treated in the same way as the GBM material because they showed different values. The LOI of the Wyoming bentonite blocks was slightly above 8 mass% and no difference between top and ordinary blocks could be found. The top and ordinary blocks of the BCV, however, showed different LOI values. The ordinary blocks were similar to the GBM but the top blocks showed a much lower LOI (slightly above 12 mass%). To be able to compare the other major elements it was necessary to eliminate the effect of the different LOI values. All other elements that are discussed in the following were normalized to LOI = 0/sum elements = 100. The data compilation in the Supplementary Material, however, contains element concentrations including the LOI. Notably, the scatter of elements with low contents such as TiO2 was larger compared to the main elements because the low values are closer to the detection limit. To investigate the bentonite homogeneity, the focus was, therefore, on the main elements and on those elements that may be involved in dissolution precipitation processes (mainly C and S). Data on all other elements including main and trace elements are provided in the Supplementary Material.
SiO2 is the main elemental component of bentonite, and variations of it are commonly explained by SiO2-rich phases such as quartz, cristobalite and/or opal (-CT, -A), zeolite, feldspar and clay minerals. All SiO2 contents are shown in Fig. 4. The BCV samples varied from 56 to 58 mass% SiO2, and the Wyoming bentonites ranged from 67 to 70 mass% SiO2. The BCV top blocks showed slightly lower SiO2 contents (~56.5 mass%) compared to the ordinary blocks (~57.5 mass%). The variability of the SiO2 content of the Wyoming GBM samples was slightly larger compared to the BCV samples, and all Wyoming block materials showed slightly lower SiO2 contents compared to the GBM samples.
Figure 4. Distribution of SiO2 contents of all samples (LOI-free; green = BCV; blue = Wyoming).
The Na2O contents of all samples are shown in Fig. 5. The most significant differences were observed for the BCV top blocks, which were significantly larger compared to all other BCV samples. In the Wyoming GBM bentonite samples the Na2O content ranged from 2.7 to 3.0 mass%, but all Wyoming ordinary and top blocks showed values >3.0 mass%.
Figure 5. Na2O contents of all samples (LOI-free; green = BCV; blue = Wyoming).
The organic carbon contents (Corg) of all samples are shown in Fig. 6. Significant differences were found only for a few samples (e.g. NAG 20, NAG 21, KAN 1, BN 16). In addition, all BCV top blocks showed significantly larger values compared to the ordinary blocks. Because of the analytical procedure used to distinguish organic and inorganic carbon (C analyser before and after acid digestion), it is possible that the Corg values are affected by siderite, which is less soluble in acids compared to other common carbonates. Hence, different siderite contents could also explain the differences evident in Fig. 6.
Figure 6. Organic carbon contents of all samples, possibly including inorganic carbon from siderite (LOI-free; green = BCV; blue = Wyoming).
All major element concentrations were assessed with respect to their variability. Block and GBM samples as well as the different sources of the Wyoming bentonites were considered separately (Table 3). In addition, BCV ordinary and top blocks were listed separately because of their significant differences.
Table 3. Statistical assessment of the XRF and C/S analyser (LECO) data of all samples (without LOI, normalized to sum of element oxides = 100 mass%). Averages calculated as $\bar{x} = {1 \over n}\mathop \sum \limits_{i = 1}^n x_i$
and standard deviations as $s = \sqrt {{1 \over {1-n\;}}\mathop \sum \limits_{i = 1}^n {( {x_i-\bar{x}} ) }^2}$
, where n is the number of samples.
Wyo = Wyoming.
The CEC results are shown in Fig. 7, and a statistical assessment of the data is provided in Table 4. The reproducibility of the traditional Cu-trien method after Meier & Kahr (Reference Meier and Kahr1999) as carried out in the BGR laboratory (S = ±0.4 meq 100 g–1) was calculated based on the differences of repeated measurements (n = 186; 1σ) following Köster (Reference Köster1979). The statistical parameter S is defined as per Equation 1:
with X′ being measurement 1, X″ being measurement 2 of the double determination and K being the number of similar objects. S is useful for CEC analysis because it is determined based on paired data for a large group of similar samples and is not dependent on randomly distributed differences of only a few pairs of data (cf. Dohrmann, Reference Dohrmann2006b). The CEC differences observed were above the reproducibility value, hence indicating the existence of measurable compositional variability, which is discussed later. The CEC data derived from the block samples were not included in the statistical assessment because of there being too few data. The CEC of the BCV ordinary blocks was larger than the that of the GBM samples, and the CEC of the BCV top blocks was significantly lower. The CEC values of each of the three samples of both types of blocks, however, were consistent. The CEC values of the Wyoming blocks were 3 meq 100 g–1 lower compared with the average GBM value, but the CEC values of the four investigated Wyoming blocks were similar. No differences between the ordinary and top blocks could be found in the case of the Wyoming bentonite.
Figure 7. CEC results for all GBM samples (Cu-trien after Meier & Kahr, Reference Meier and Kahr1999; green = BCV; blue = Wyoming).
Table 4. Statistical assessment of all CEC values of the GBM samples using the traditional Cu-trien method after Meier & Kahr (Reference Meier and Kahr1999). Precision of the method = ±0.4 meq 100 g–1 (n = 186; 1σ).
Wyo = Wyoming.
Mineralogical composition (XRD and SEM)
Powder XRD analyses followed by Rietveld refinement using appropriate disordering models (Ufer et al., Reference Ufer, Roth, Stanjek, Dohrmann, Kleeberg and Bergmann2004, Reference Ufer, Stanjek, Roth, Dohrmann, Kleeberg and Kaufhold2008) are quantitatively still less precise than XRF. As an example, a difference of 0.2 mass% SiO2 could result from a positive or a negative difference of 0.2 mass% of quartz. Such small differences are difficult to resolve using XRD. XRF was therefore preferred for the investigation of material homogeneity, and XRD was only used for selected samples that showed the most significant chemical differences (e.g. BCV top and ordinary block samples, the latter being similar to the GBM samples). The results were also compared with previously published data on MX80 and BCV bentonite (Table 5).
Table 5. Mineralogical composition of selected bentonite samples determined by Rietveld refinement (‘0’ = <1 mass% but present) and data derived from the literature. Cristobalite, opal-CT and opal-A (amorphous silica) were grouped together because they cannot be distinguished using XRD. Clinoptilolite, heulandite and analcime were grouped together because their differentiation was difficult in the analysed samples using XRD. In the following, this group is referred to as ‘zeolite’. Values were partly rounded.
All samples were dominated by the 2:1-layered silicate smectite. However, minor amounts of mica (also a 2:1-layered silicate) were detected by XRD and SEM (Fig. 8). It turned out to be more suitable to sum the values for both 2:1-layered silicates (Table 5), the reasons for which are discussed later.
Figure 8. SEM images of rounded mica particles of samples (a) NAG 43 and (b) NAG 21.
Mica particles were mostly between 100 and 200 μm, curved and well preserved (no indication for chemical weathering). The structural formula was derived from EDX assuming an ideal trioctahedral composition (Mg + Al + Fe2+ + Ti + Si = 7.0). The presence of a trioctahedral mica was indicated by high Fe and Mg contents. The Fe, therefore, was assumed to be ferrous iron. The resulting structural formula of the particle shown in Fig. 8a of (Na0.1K0.8)(Al0.5Mg0.7Fe1.7)(Ti0.2Al1.5Si2.3)O10(OH)2 would correspond to Fe-rich biotite. All of the six mica particles of sample NAG 43 that were investigated using SEM showed similar compositions, with, however, slightly variable Fe contents. In sample NAG 21 (Fig. 8b) mica particles with much lower Fe but higher Mg contents were also found, which could be referred to as phlogopite: (Na0.1K0.7)(Al0.3Mg1.9Fe0.6)(Ti0.3Al1.2Si2.5)O10(OH)2. Dioctahedral mica particles were not observed.
Detailed analysis of the BCV blocks
Most of the chemical differences summarized in Table 3 were close to reproducibility, reflecting compositional variation (discussed later). The BCV top block samples, however, significantly differed from the other BCV materials (GBM and ordinary blocks). The top blocks showed higher inorganic carbon values than the ordinary blocks and the GBM samples but only slightly greater CaO values, which is interesting because normally both parameters are related by the calcite/dolomite content. At the same time, greater K2O, P2O5 and Corg values were found in the top blocks compared to the ordinary blocks and GBM samples (Fig. 9). The BCV top blocks also contained more Na2O (Fig. 5) and more water and showed reduced CEC values.
Figure 9. Differences of BCV top and ordinary blocks (green line = average values of GBM samples) determined by XRF and LECO; y-axis = mass%.
Such significant differences were not expected and hence we investigated them in more detail. Accordingly, additional analyses were performed on both types of BCV block material.
The analysis of ECs of the BCV block samples was performed using two different exchange solutions. Exchange solutions with and without pre-treatment with calcite were prepared and used (Table 6) to detect the effects of soluble phases other than calcite.
Table 6. EC population, CEC and water loss up to 105°C (mass%) of the BCV ordinary and top block samples using Cu-trien5xcalcite and Cu-trien5x after Dohrmann & Kaufhold (Reference Dohrmann and Kaufhold2009). For water loss analysis up to 105 °C, samples previously dried at 60°C were used.
av. = average.
The water loss between ‘as delivered’ and 105°C is similar for both BCV variations (ordinary and top), but the water contents at 60°C differed. BCV ordinary blocks contained ~11 mass% water after drying at 60°C. Additional drying at 105°C yielded 11 mass% water. The total water content at the reference temperature 105°C hence amounted to 22 mass% water (Table 6). BCV top blocks contained ~16 mass% water (60°C) plus 6 mass% water (105°C), with a total amount of 22 mass% water.
ECs are different for the two different materials used for the production of BCV ordinary and top blocks. The differences between the three samples taken from either BCV ordinary or top blocks, respectively, are low. This indicates that the blocks are internally homogeneous (Fig. 9 & Table 6).
BCV ordinary block samples are dominated by exchangeable Mg2+ (59%/CEC), followed by Ca2+ (39%/CEC), K+ (3%/CEC) and Na+ (2%/CEC). CEC values are 72 meq 100 g–1, and the sum of ECs exceeded the CEC by 4%/CEC. Differences between different sample masses for exchangeable Ca2+ are low, indicating that errors typically caused by calcite dissolution are low or absent (Dohrmann, Reference Dohrmann2006a, Reference Dohrmann2006b). The sum of ECs is larger (4%) than the CEC, a range that is well within the analytical error of the method (Dohrmann et al., Reference Dohrmann, Genske, Karnland, Kaufhold, Kiviranta and Olsson2012b). Possible effects of excess electrolytes were not analysed (Bradbury & Baeyens, Reference Bradbury and Baeyens2003; Dohrmann et al., Reference Dohrmann, Genske, Karnland, Kaufhold, Kiviranta and Olsson2012b; Fernández et al., Reference Fernández, Kaufhold, Sánchez-Ledesma, Rey, Melón and Robredo2018; Frederickx et al., Reference Frederickx, Honty, De Craen, Dohrmann and Elsen2018).
BCV top block samples showed a different EC population than BCV ordinary block bentonites. They are dominated by exchangeable Na+ (75%/CEC), followed by Mg2+ (20%/CEC), Ca2+ (8%/CEC) and K+ (4%/CEC). Average CEC values are as low as 58 meq 100 g–1 (compared to the BCV ordinary block samples), and the sum of ECs exceeds the CEC by 7%/CEC. Differences between different sample masses for exchangeable Ca2+ are somewhat larger than for BCV ordinary block samples (Table 6); however, the values are still close.
The percentage of exchangeable Ca2+ of the top blocks increases by a factor of ~2 when comparing respective values of Cu-trien5xcalcite and Cu-trien5x. Ca2+ values again show significant differences between different sample masses when the Cu-trien exchange solutions are not saturated with calcite, indicating the presence of calcite (Dohrmann, Reference Dohrmann2006b). As for BCV ordinary block samples, the BCV top block samples again show no/minor variation of the other ECs (Na+ 74%/CEC, Mg2+ 21%/CEC, K+ 4%/CEC). The sum of ECs of the BCV ordinary block samples is also significantly (15%) larger than the CEC and is inflated by calcite dissolution. In none of the analysed materials were indications for soluble phases other than calcite found.
All BCV block samples were additionally investigated using simultaneous thermal analysis. Most significant differences were detected in the MS CO2 curve (Fig. 10).
Figure 10. MS CO2 curve of simultaneous thermal analysis of both BCV block type samples (green = top blocks; blue = ordinary blocks).
The top blocks showed some peaks in the MS CO2 curve in the temperature range of 250–600°C, which may be attributed to organic materials of different maturities and the presence of small amounts of siderite. At temperatures of ~100°C, a further difference was observed. Despite the comparably high signal-to-noise ratio, no peaks were observed in the curves of the ordinary blocks. All top blocks, on the other hand, showed small humps at ~100°C, which could be explained by the presence of small amounts of a Na-carbonate phase. Despite their comparably low intensity, Kaufhold et al. (Reference Kaufhold, Emmerich, Dohrmann, Steudel and Ufer2013b) could attribute such peaks to traces of Na-carbonate phases (soda or nahcolite) that can be found in Na-activated bentonites.
Figure 11 shows two suspensions prepared from the ordinary block and the top block BCV samples. The suspension produced with the ordinary block sample showed a sediment volume that is typical of Ca/Mg-rich bentonites, and the suspension produced from the top block of the BCV material was stable and hence can be referred to as a dispersion. Such behaviour is typical of Na-rich bentonites, confirming the interpretation of differential thermal analysis (DTA) MS CO2 signals as possibly showing traces of Na-carbonate phases (soda or nahcolite).
Figure 11. Qualitative test for suspension stability of ordinary (left) and top (right) BCV blocks.
The results of the sediment volume tests (Fig. 11) and thermal analysis (Fig. 10) indicated the presence of soluble components in the BCV samples that could also be confirmed by aqueous extracts. Analyses of two BCV qualities (BCVO and BCVTP) indicated the presence of nitrogen (N) at low concentrations measured after digestion (Table 7). The N values ranged from 0.03 to 0.05 mass%. On average, slightly less N was found in the top block sample, but this difference is barely significant. Additional data obtained from analysing the aqueous extracts are listed in Table 7. The solution derived from the BCV top blocks contained more carbonate, had a higher pH and had greater electrical conductivity. At the same time, Na+ increased by a factor of nearly 20, dominating the solution.
Table 7. Results of N analysis and alkalinity, pH, electrical conductivity and chemical analyses obtained using IC and ICP of aqueous extracts.
Typically, Na-rich smectites can be distinguished from Ca- or Mg-rich smectites by XRD analysis of the basal distances. The <2 μm fractions of one top and one ordinary BCV block sample were investigated using XRD (oriented mounts; Fig. 12).
Figure 12. XRD analysis of oriented mounts of the <2 μm fractions of (a) BCV top block BTB3 and (b) BCV ordinary block OB1, both air-dried (AD; black) and after EG solvation (red).
The d 001 reflection (AD) of the top block sample is centred at 13.0 Å, in contrast to that of the ordinary block sample (15.0 Å). A d 001 value of ~13 Å is commonly observed for Na-smectites rich in monovalent cations, and 15 Å corresponds to Ca/Mg-smectites rich in bivalent cations. Both samples are dominated by smectite, but in the case of the top block sample peaks at 10.0 and 5.0 Å (EG trace) were found, which correspond to some quantity of illite/mica.
Discussion
Compositional variability (chemical composition and CEC)
First, representative chemical compositions of both types of bentonite are provided, including standard deviations (Table 3), which will allow us to determine whether the composition of samples taken after termination of the test shows significant changes or not. Moreover, standard deviations indicate that both types of bentonite differ with respect to their compositional variability. For the BCV samples, the largest standard deviations were detected for SiO2, Al2O3, Fe2O3 and CaO. A comparison of the CEC, which largely reflects the smectite content, with those most deviating elements helps us to understand the type of compositional variation occurring. The CEC does not show any correlation with the SiO2 content but positive correlations with Al, Fe and Mg contents. This means that some variation of the quartz content exists, but not in a way such that higher quartz contents would lead to lower smectite contents (lower CEC). The variability of the SiO2 content hence can be considered to be non-systematic. The positive correlation of other smectite structural elements such as Al, Fe and Mg with the CEC (Fig. 13a), on the other hand, results from the fact that these elements predominantly occur in smectites. This result was not surprising for Al and Mg, but the Fe content could also have varied according to different contents of Fe-oxyhydroxides, which is, however, not the case. Negative correlations were observed for the structural smectite elements (except for Si) and the CEC (representing the smectite content) with both the inorganic carbon content (Fig. 13b) and the CaO content (the latter both being highly correlated in this sample set at R 2 = 1; Fig. 13b). These observations indicate that the smectite content and carbonate content vary systematically, whereas the content of SiO2 phases varies non-systematically.
Figure 13. Comparison of the CEC values of the BCV GBM samples with (a) the Fe2O3 content and (b) the inorganic carbon content.
The largest standard deviations of the Wyoming bentonites were found for SiO2 and Al2O3 (Table 3). In addition, good correlations were found for CEC and SiO2 (negative), SiO2 and Na2O (negative), as well as CEC and Na2O (positive; Fig. 14). XRD (Table 5) indicated the presence of various Na- and Si-rich phases such as montmorillonite, zeolite, Na-feldspar (albite-rich), quartz, cristobalite/opal and mica. Assuming that all differences in CEC (~6 meq 100 g–1) can be explained by quartz + cristobalite/opal contents, then ~2.4 mass% SiO2 would be required. The total difference of SiO2 observed is 2.7 mass%. The correlations, therefore, can be explained by the fact that Na predominantly occurs in smectite and that the major compositional variability of the Wyoming bentonites results from a systematically varying smectite/quartz + cristobalite/opal ratio. Figure 14 also shows that the ratio of Na/Si can be used to distinguish the different sample series (blocks, NAG, KAN, BN). The BN series revealed the lowest Na/Si ratio, whereas the block samples showed the highest Na/Si ratio, which indicates that the Wyoming block samples have the highest smectite contents of the HotBENT Wyoming bentonite samples. This result is consistent with the CEC data.
Figure 14. Comparison of CEC values (meq 100 g–1) using the traditional Cu-trien method after Meier & Kahr (Reference Meier and Kahr1999) with different chemical features (mass%), proving that the CEC does reflect different smectite contents being caused by compositional variability (Kaufhold et al., Reference Kaufhold, Dohrmann, Ufer and Meyer2002; Kaufhold & Dohrmann, Reference Kaufhold and Dohrmann2003).
The content of organic carbon (Corg) in both types of bentonites (Wyoming and BCV) is relatively low (mostly <0.1 mass%; BCV top blocks = <0.2 mass%) and hence probably is less relevant. Organic carbon, however, can lead to gas production and act as microbial feed and hence may be important even at low concentrations. In addition, organic material is a typical possible contaminant of technical bentonite products and hence is worth investigating in more detail. Two BCV GBM samples showed significantly larger Corg values compared to the others (NAG 20 and NAG 21; Fig. 6). These different values can be explained by the different batches of raw material containing different Corg contents. On the other hand, some contamination may have occurred anywhere between mining and production. Note that the Corg values could also reflect the siderite content, because siderite is comparably insoluble in acid and can lead to incorrect organic carbon content measurements based on the analytical procedure used in the present study. The Fe contents of these samples, however, did not vary and hence do not point to higher siderite contents. On the other hand, slightly higher CaO values were observed, which, however, would not explain the larger Corg values. The reason for the slightly higher Corg values of NAG 20 and NAG 21 (0.2–0.3 mass%) cannot be explained as of now.
Similarly, two samples (KAN 1 and BN 16) of the Wyoming bentonites were found that contained more Corg than the others (>0.15 mass% compared to <0.1 mass% for most of the others). As in the case of the BCV samples, the Fe contents of both of these Wyoming samples were not higher compared to the others and hence higher siderite contents can be excluded. For both types of bentonites it was not possible to conclude whether the outliers (samples with especially high Corg values) resulted from sample contamination or natural compositional variability.
Mineralogical composition (XRD)
Selected Wyoming bentonite samples were investigated using XRD/Rietveld analysis (Table 5). The samples were selected based on the largest chemical differences (e.g. Na and SiO2 contents, as explained above). Wyoming bentonites are known for their variable mica contents (Slovinsky, Reference Slovinsky1958). A couple of mineralogical mica-free compositions of Wyoming bentonites have been published previously (Liu, Reference Liu2010; Kaufhold et al., Reference Kaufhold, Hein, Dohrmann and Ufer2012). Others, however, have proved the presence of minor amounts of mica (Karnland et al., Reference Karnland, Olsson and Nilsson2006). According to Knechtel & Patterson (Reference Knechtel and Patterson1962), biotite, which was identified by SEM and EDX in the samples used in the present study (Fig. 8), may be either of volcanic origin or be detrital material. Angular particles were considered to be of volcanic origin, whereas curved particles might be detrital. The mica particles detected in the present study (Fig. 8) are curved (bent and folded) and hence are probably of detrital origin. Rietveld refinement is based on a set of structure models, one for each phase. Accordingly, separate sets of refinement values for smectite and mica were calculated. SEM images (Fig. 8) show that trioctahedral mica was mostly present in the >100 μm fraction. Mica in a phase mixture or as a pure sample often withstands grinding, which affects XRD quantification, for which all particles should be <20 μm. For validation of mica quantities, several tests were performed that varied the sample treatment (grinding and preparation of XRD mounts). The results of these tests showed that mica contents varied within 3–9 mass%, although no significant variation of the K2O content (indicative of mica) could be detected (Table 3 & Supplementary Material). This indicates that the preferred orientation or even single-crystal reflexes of the larger mica particles may have affected the quantification. Notably, erroneous mica values also affect the other mineral contents because the sum of minerals is always 100%. For a more detailed investigation, one sample (Wyoming top block 1) was selected after intense grinding, and it was then dispersed and passed through a <20 μm sieve. Approximately 1 mass% of particles >20 μm was separated from this powder consisting of quartz, feldspar, glass, organic material, Fe-oxyhydroxide and mica (Fig. 15). The few mica particles were much larger compared to the other phases, which supports the notion that they are difficult to grind, in turn causing difficulties regarding phase quantification. Most of the mica particles were identified as biotite based on EDX data, but even a few large muscovite particles were detected.
Figure 15. SEM image of the >20 μm fraction of intensely powdered Wyoming top block sample 1.
To quantitatively distinguish smectite from mica based on Rietveld refinement one could repeatedly grind the >20 μm fraction to pass through a 20 μm sieve and then add the >20 μm fraction to the sample again. This procedure, however, has two problems: first, such small amounts of sample (20 mg in this case) can hardly be ground to the desired particle size without some mass loss in the mill, at the pestle and in the sieve. Second, it is difficult to finally mix the two powders in such a way that an entirely homogeneous mixture is obtained. To date, mica has not proven to be of particular importance in barrier systems. Therefore, it was decided not to separate the <20 μm fraction of all samples but to provide the sum of 2:1-layered silicates instead. Presenting the sum of 2:1-layered silicates, which in the case of the samples used in the present study represents the sum of the 2:1 minerals smectite and minor amounts of mica, is therefore more plausible. An estimation of the mica content can be based on chemical analysis, more specifically the K2O content, which does not vary significantly, indicating relatively constant mica and K-feldspar contents. The samples contained ~0.7 mass% K2O according to XRF analysis and ~4 mass% K-feldspar according to Rietveld refinement (in Table 5 only the sum of feldspar is given). Assuming a K content of 13 mass% K2O of the 4 mass% K-feldspar leads to 0.2 mass% remaining of K2O, which, depending on the K content of the mica, results in 1–2 mass% mica. These values, however, only provide an estimation because of the analytical error of the XRD K-feldspar quantification resulting from the complex (low-symmetry) and possibly disordered structure, but also from the unknown K contents of the actually present mica and K-feldspar. A more accurate quantification would be based on determining the chemical formulas of all K phases in the bentonite, which, however, has to be based on a solid statistical basis and hence the analysis of many particles, which is time-consuming. Because of the relatively constant K contents, it was decided not to go into more detail and to present the sum of 2:1 layers instead, which largely represents the smectite content with minor amounts of mica.
For the comparison of the chemical composition with the smectite content, the CEC can be used as an alternative that correlates well with the smectite content because differences in the layer charge density, variable charge and minor constituents are supposed to be low because of the similarity of the samples (in contrast to different samples from different deposits). Note that the Wyoming bentonites used in this study contain zeolites that may be described as size-selective cation exchangers (e.g. Barrer, Reference Barrer, Sand and Mumpton1976). Zeolites may have large CEC values exceeding smectite CECs by far; however, in this study, CEC was determined with the large index cation Cu-triethylenetetramine that is not able to reach the inner parts of zeolites for cation exchange. Meier & Kahr (Reference Meier and Kahr1999) reported CEC values of 5 meq 100 g–1 for pure clinoptilolite with Cu-trien only. For ammonium acetate, CECs of 160 meq 100 g–1 were measured for this zeolite. The zeolite contribution to CEC values as used in this study can be regarded as low (<1% of the CEC values).
The mineralogical compositions of the four Wyoming bentonites investigated were similar. Block samples and GBM samples showed minor differences that are barely significant. The largest difference was found for the sum of 2:1-layered silicates (blocks: 73 mass%; GBM: 68–70 mass%). The K content of the GBM samples, which reflects the sum of K-feldspar and mica content, was slightly higher (0.04 mass% K2O) compared with the block samples. This difference corresponds to ~0.5 mass% mica and hence does not explain the difference of 3–5 mass%.
The differences in the SiO2 contents of the Wyoming samples correspond to differences in cristobalite/opal contents, with the largest amounts found for KAN 1, which was chemically characterized as relatively rich in SiO2 and poor in Na2O (Fig. 14).
Differences of the Na2O content and CEC could also only partly be explained based on the Rietveld analysis results. The block samples contain slightly more Na2O, which would indicate greater contents of either Na-rich smectite, Na-feldspar or zeolite or a varying smectite interlayer cation composition. Assuming differences of 10 meq 100 g–1, exchangable Na+ would occur between the various Wyoming bentonite samples studied. This would sum up to 0.31 mass% Na2O, which is ~10% of the average total Na2O in the Wyoming bentonite samples studied. Using Rietveld refinement, a greater content of 2:1-layered silicates (dominated by Na-rich smectite) was found, which is in agreement with the chemical data for Na2O. The slightly lower Na-feldspar and zeolite contents of the block samples, however, contradict this trend. In addition, the slightly lower CEC (65 ± 1 meq 100 g–1) of the block samples compared to the GBM samples (68 ± 1 meq 100 g–1) contradicts the slightly larger sum of 2:1-layered silicates of the block samples.
Na2O contents, representing smectite and also Na-feldspar, were larger for Wyoming block samples, explaining the slightly larger sum of feldspars in these materials. In addition, the SiO2 content is lower and the Al2O3 and Fe2O3 contents are larger for Wyoming block samples. The CEC values of the block samples, representing the smectite content, were slightly lower and hence contradict the difference between the sums of 2:1-layered silicates of the block and GBM samples. The CEC differences were much larger than the precision of the CEC method in the laboratory (±0.4 meq 100 g−1, 1σ; n = 186) and could also be explained by the different layer charge density of the smectites, which, however, was not investigated in the present study. Small differences in phase contents based on XRD Rietveld refinement, the CEC values and the chemical compositions of the major oxides may indicate compositional variability of the various Wyoming bentonites studied originating from two different industrial sources.
Knechtel & Patterson (Reference Knechtel and Patterson1962) discussed the mineral abundance of various Wyoming bentonite beds, but they did not identify any zeolite that was detected in the present study. In addition, Sauer et al. (Reference Sauer, Caporuscio, Rock, Cheshire and Jove Colón2020) reported on the presence of up to 13 mass% clinoptilolite in a Wyoming bentonite sample, and Cheshire et al. (Reference Cheshire, Caporuscio, Jové-Colón and McCarney2013) found that the naturally present clinoptilolite of a Wyoming bentonite sample recrystallized to analcime under repository-like conditions. Similar reactivity was reported for cristobalite as a minor component of the bentonites studied. A two-stage dissolution pattern of Wyoming bentonite was observed by Karnland et al. (Reference Karnland, Olsson and Nilsson2007), as this bentonite also contains cristobalite, which is known to be unstable under high-pH conditions. Kaufhold et al. (Reference Kaufhold, Dohrmann, Sandén, Sellin and Svensson2013a) showed that amongst all bentonite constituents only clinoptilolite and cristobalite reacted in a medium-scale test, which was also found by Bao et al. (Reference Bao, Jiaxing and Huixin2016) for Gaomiaozi (GMZ) bentonite. When analysing the samples of the HotBENT experiment, it may therefore be particularly interesting to consider the zeolite content.
BCV block samples
The chemical composition of the ordinary BCV block sample was similar to that of the BCV GBM sample. Therefore, no additional BCV GBM samples were analysed using XRD. Because of the significant differences of the BCV top and ordinary blocks, however, separate XRD analyses were performed on these two types of material (Table 5). The top block material, which contained Na-rich smectites, showed a lower smectite content (73 mass% sum 2:1) than the ordinary block sample (81 mass% sum 2:1), which is in accordance with a lower CEC (15 meq 100 g–1 less) and is similar to the smectite content published previously by Svoboda et al. (Reference Svoboda, Šťástka, Vašíček, Špinka, Bureš and Pospíškova2022; Table 5). The ordinary block BCV sample contained less feldspar and quartz, but any other differences were not significant. In addition, a greater siderite content was found for the top block sample, which is in agreement with the data obtained using simultaneous thermal analysis (STA).
The ordinary blocks are not identical in composition but are similar to the BCV GBM samples. Significant chemical and mineralogical differences, however, were found between the BCV top and ordinary blocks, as partly discussed earlier. In the pedestals of the HotBENT experiment, both of these materials are in (close) contact. This interface will therefore be particularly interesting when dismantling the experiment because cation exchange was found to be rapid in previous experiments (Dohrmann & Kaufhold, Reference Dohrmann and Kaufhold2014), and a re-equilibration of ECs is expected to occur in the HotBENT experiment as well. This situation is comparable to the Swedish/SKB ABM project using similar material contrasts in a heater experiment in crystalline rock. First, the top blocks contain less smectite, which explains the much lower CEC. In addition, the dominating EC of the top block BCV material is Na+ (75%) in contrast to Mg2+ + Ca2+ (98%) and 2% Na+ for the other BCV materials. Using STA, traces of a Na-carbonate phase were identified, which, based on Kaufhold et al. (Reference Kaufhold, Hein, Dohrmann and Ufer2012), indicates that the material was possibly industrially soda-activated. This would also explain the composition of the aqueous extracts containing more carbonate species and Na+, as well as the higher pH and the greater electrical conductivity. Soda activation would also explain the good dispersibility of the top block BCV sample (Fig. 11). All analyses provided consistent data showing that the BCV top blocks contain Na-rich bentonite in contrast to the other BCV materials.
Comparison of the HotBENT bentonites with previously used similar materials
Wyoming bentonites are widely used in large-scale in situ tests in rock laboratories, particularly in Äspö, Sweden. Much more information is therefore available for Wyoming bentonite compositions compared to the BCV. For the MX80 product, Karnland (Reference Karnland2010) defined quality criteria of the bulk material used for block and backfill production in Äspö based on chemical composition, mineralogical composition, original ECs, CEC, grain density, specific surface area, particle size, water content and geotechnical parameters. For five different MX80 samples (Wyoming bentonite), the CEC values ranged from 71 to 78 meq 100 g–1, with an average of 75 meq 100 g–1 (n = 16). In the present study, the Wyoming GBM CEC values were significantly lower, ranging from 64 to 71 meq 100 g–1, with an average of 68 meq 100 g–1 (Fig. 14 & Table 4). The precision of the analysis, on the other hand, was as low as ±0.4 meq 100 g–1 (n = 186; 1σ). This is an interesting aspect for studies that use CEC differences for the calculation of smectite alteration processes. If CEC values could be determined with high precision for reference materials and reacted materials, it would be possible to evaluate smectite alteration processes (Kaufhold et al., Reference Kaufhold, Dohrmann, Wallis and Weber2023) even based on small changes to the CEC values.
In the PTR experiment in Äspö, Dohrmann & Kaufhold (Reference Dohrmann and Kaufhold2014) reported on differences of reference Wyoming bentonite materials that were available for each bentonite ring – a similar approach to that used in the present study. CEC values were on average 86.1 ± 0.9 meq 100 g–1 (n = 8; 2σ). Average CEC values were 8 meq 100 g–1 larger than those studied by Karnland (Reference Karnland2010) and 18 meq 100 g–1 larger than for the granular Wyoming bentonite used in the present study. Na+ and Ca2+ populations were determined as 68%/CEC for Na+ and 26%/CEC for Ca2+. Exchangeable Mg2+ concentrations were larger both for PTR (10%/CEC) as well as in the round-robin study (8%/CEC), whereas Karnland (Reference Karnland2010) reported a value of 6%/CEC only. Wyoming bentonites showed relatively similar EC populations in all three of the studies. CEC values, however, differed by 10–20%. These differences can be explained by the varying smectite contents. In summary, the smectite content in commercially available bentonites from Wyoming shows quite some variation.
XRF chemical analyses reported by Karnland (Reference Karnland2010) showed concentrations ranging from 66 to 69 mass% SiO2 (n = 6) on a LOI-free basis, representing variations to the bentonite batches (industrial material qualities) delivered over 20 years. These absolute SiO2 concentrations (LOI-free) are on average 1 mass% lower compared to Wyoming bentonites used in the HotBENT experiment (67–70 mass% SiO2; Table 3). Greater concentrations were reported for Al2O3 (1.0 mass%), MgO (0.8 mass%) and CaO (0.1 mass%), whereas average concentrations were lower for Na2O (0.6 mass%), Fe2O3 (0.1 mass%) and K2O (0.1 mass%). PTR reference samples (Dohrmann & Kaufhold, Reference Dohrmann and Kaufhold2014), on the other hand, were much closer to those studied by Karnland (Reference Karnland2010), but with lower scattering of the data, most probably because only four reference samples from the same batch were used and analysed in the same run compared to samples from over ~20 years as in Karnland (Reference Karnland2010). Such differences of ~1 mass% are relatively small and confirm the homogeneity of the Wyoming-type bentonites used in the three studies with respect to their chemical composition.
The total C of the samples studied by Karnland (Reference Karnland2010) was on average higher (0.4 mass%) compared to the Wyoming GBM samples used in the present study (0.2 mass%). Inorganic C followed the same trend, at 0.16 mass% compared to 0.11 mass%. Total S was on average larger at 0.34 mass% compared to 0.21 mass% for the Wyoming bentonites used in the present study. Karnland (Reference Karnland2010) also differentiated S in sulfide-S (pyrite) and sulfate-S (soluble, gypsum). PTR reference samples were in-between both datasets for inorganic C (0.7 mass%) and S (0.25 mass%). Svensson et al. (Reference Svensson, Eriksson, Johannesson, Lundgren and Bladström2019) reported on the quality control for various bentonite batches in Äspö, differing in size from 20 kg up to 20 tons, and acceptable parameter ranges were based on technical design requirements for industrialization of a repository (e.g. Posiva SKB, Reference Posiva2017).
In Table 5, the mineralogical composition of the HotBENT samples is compared with previously published MX80 data. The data provided by Karnland et al. (Reference Karnland, Olsson and Nilsson2006) and Kaufhold et al. (Reference Kaufhold, Hein, Dohrmann and Ufer2012) are similar, but Liu (Reference Liu2010) found more quartz, less smectite, less clinoptilolite and no cristobalite (or opal) in their samples. For a 200 kg charge of Wyoming bentonite they found little variability in the chemical composition (XRF) and a CEC of 84 meq 100 g–1 (n = 2). In their study, Liu (Reference Liu2010) highlighted that some bentonite qualities varied significantly, reflecting heterogeneities in bentonite mines that were not compensated for with suitable homogenization procedures being conducted by mining operators. With approximately identical major oxide concentrations of Si and Al, the CEC values ranged from 73 79 meq 100 g–1, and precision was calculated from 0.4 to 1.1 meq 100 g–1. Overall, the smectite contents of the Wyoming bentonites used in the HotBENT experiment were in the range for previously published MX80 data (~5 mass% lower; Table 5). The ranges of concentrations are also similar for zeolite and feldspar (~5 mass% higher; Table 5). Such differences cannot be accounted for by measuring typical bentonite properties such as water uptake or rheological parameters. In addition, important barrier properties such as hydraulic conductivity or swelling pressure depend more on the compaction than on small differences in the smectite content (Kaufhold & Dohrmann, Reference Kaufhold and Dohrmann2016).
The BCV bentonite was mined in the Czech bentonite mine Černý vrch and, amongst other materials, is considered to be a possible barrier material in the Czech Republic (Hausmannová et al., Reference Hausmannová, Hanusová and Dohnálková2018). It is described as Ca/Mg-rich bentonite that was industrially processed at Keramost, Ltd. In 2018, this material was proposed to be used in various large-scale tests (interaction experiments, corrosion experiments, the BEACON project, engineered barrier 200°C; Hausmannová et al., Reference Hausmannová, Hanusová and Dohnálková2018). In other large-scale tests (e.g. the ABM test series at Äspö), a Czech bentonite from the Rokle deposit was used, which is a different material from a different deposit. Much less experience with and fewer results are available for BCV bentonite in comparison with Wyoming bentonite because the BCV bentonite only recently became a focus of interest, whereas Wyoming bentonites have been investigated for decades, and not only within the framework of HLW repository research. However, some studies on BCV bentonite are available focusing on its performance under repository-like conditions. As an example, Svoboda et al. (Reference Svoboda, Mašín, Najser, Vašiček, Hanusova and Hausmannová2023) studied the hydromechanical behaviour of BCV bentonite, and Najser et al. (Reference Najser, Mašín, Svoboda, Vašíček, Hanusová and Hausmannová2023) studied its homogenization behaviour. Mineralogical compositions were published by Hausmannová et al. (Reference Hausmannová, Hanusová and Dohnálková2018), Červinka et al. (Reference Červinka, Večerník, Kašpar and Vašíček2018) and Laufek et al. (Reference Laufek, Hanusová, Svoboda, Vašíček, Najser and Koubová2021). These results are compared in Table 5. Two different batches of the BCV, namely BCV 2017 and BCV 2018, have been reported to date. Svoboda et al. (Reference Svoboda, Vašíček, Pacovská, Šťástka, Franěk and Rukavičková2019) compared these two different BCV batches, and, amongst other differences, they found a varying carbonate content. For BCV 2017, Svoboda et al. (Reference Svoboda, Vašíček, Pacovská, Šťástka, Franěk and Rukavičková2019) reported an inorganic carbon content of 0.5 mass%, which approximately corresponds to 4 mass% carbonate, and for the BCV 2018 sample less than half of this value was reported. Statistical assessment of the samples investigated in the present study also revealed some variation of the carbonate content. These results indicate the variable carbonate contents of the materials from this mine. The mineral compositions of BCV published to date, however, indicate that mineral contents other than carbonates also vary (e.g. the smectite, goethite and kaolinite contents; Table 5).
Often, the chemical and mineralogical compositions of bentonite samples from a given deposit published in different studies are not identical. This can be explained by different analytical methods used and the specific errors of the different methods (including different methods for interpretation of XRD data, and even different Rietveld codes), but also by the compositional variation of the different batches of a product resulting from the presence of different materials in the mine. The names of the bentonite (e.g. MX80 or Calcigel) are mostly brand names that are used to trade a bentonite product, often for one specific application. Accordingly, the bentonite product has to fulfil some specifications for industrial use (e.g. smectite content >70 mass%, S content <0.5 mass%, specific rheology of a mud), but normally this does not include the entire mineralogical composition. Bentonite products are produced based on naturally occurring raw materials, which are rarely homogeneous in composition over an entire deposit. Comprehensive characterizations of the homogeneity of a bentonite deposit, as provided by Knechtel & Patterson (Reference Knechtel and Patterson1962) for the Wyoming bentonites, are rare. The companies involved follow different strategies to provide as homogeneous a material as possible, even over several years, but variations in composition may still exist. Purchasing a bentonite with a brand name guarantees that the material will fulfil its given specifications, but it may not be identical to another charge of the product with the same name. Therefore, it is essential to analyse materials prior to employing them in specific experiments (as intended in the present study) instead of relying on previously published data. From the above explanations one can understand the necessity of selecting and analysing a large sample set for assessing bentonite homogeneity if large amounts of material are to be used in an experiment (as is the case for the HotBENT experiment).
Summary and conclusions
The present study provides reference data and a statistical assessment of the chemical and mineralogical composition of the bentonite material used to set up the HotBENT project, which will be particularly useful for analysis of samples taken after termination of the experiment. Two different bentonites were used: natural Na-rich Wyoming bentonite from the USA and BCV bentonite from the Czech Republic.
The Wyoming bentonite samples were quite similar, although they were delivered in different batches and prepared in different places. The ordinary and top block materials, both from the same charge, were identical. Minor compositional variations in the GBM and block materials could be resolved using both the Na/Si ratio and the CEC/Si ratio. Despite some overlap of these parameters, even the different sources (industrial batches) of the Wyoming bentonites could be distinguished based on these two ratios.
For single Wyoming bentonite samples, exceptionally high Corg contents were determined. Other single samples showed high CaO contents, which should be considered when analysing the run products. Similarly to the Wyoming bentonites, BCV bentonites also produced a few samples with relatively high Corg contents. All BCV top block samples showed greater Corg values compared to the other BCV samples, but this can be explained by the slightly larger siderite contents of the different raw materials. These samples are therefore not considered outliers, in contrast to two GBM samples. As with the case of the two Corg-rich Wyoming bentonites, it was not possible to conclude whether the Corg outliers resulted from contamination or from compositional variability. When analysing the run products, however, a significant range of Corg contents has to be considered.
The CEC values of granular Wyoming bentonite materials were lower than those used for the block production of in situ tests in Äspö (using industrial MX80 product qualities). The CEC values measured for the HotBENT samples may be helpful for identifying mineralogical changes (e.g. smectite alterations) or cation exchange.
The Wyoming bentonite material used for setting up the HotBENT experiment is comparable to other Wyoming bentonite products, such as MX80. The presence of zeolites in the samples used to set up the HotBENT project, which was rarely reported in MX80 (e.g. Acikel et al., Reference Acikel, Rowe, Brachman, Baral and Su2019), is worth mentioning because it previously turned out to be more reactive compared to other silicates contained in the bentonite, possibly similarly to cristobalite. This may be relevant when comparing run products with the precursor material.
The compositions of the BCV GBM samples and ordinary blocks were similar. Some compositional variation could be resolved using the ratio of CEC (or major oxides predominantly occurring in the smectites, including Fe) to inorganic C or CaO, with the quantities of the latter two being highly correlated. This result indicates that the compositional variations of the BCV bentonites resulted from different quantities of smectite and carbonate minerals. Most of the significant differences were observed between the BCV ordinary or BCV GBM samples and the BCV top blocks. The BCV top blocks differed with respect to chemical composition (higher Na content, amongst others) and smectite content, which was indicated both by XRD and CEC. Most importantly, exchangeable Na+ was found to dominate the smectites of the top blocks, whereas both ordinary blocks and GBM samples were dominated by Mg2+ and Ca2+. All data (STA, aqueous extracts and EC population) are consistent, indicating that Na-activated bentonite was possibly used for the production of BCV top blocks. This finding is of particular interest because both materials are in contact with each other, and the top blocks contain more water-soluble phases than the ordinary blocks. The interface of BCV top and ordinary blocks is therefore an interesting location for sampling after termination of the experiment. However, this interface represents only a small fraction of the volume of the entire experiment.
The results presented in the study reveal a high degree of chemical and mineralogical homogeneity for the GBM sampled across 20 m of backfilled tunnel with ~147 tons of GBM, as most measured differences were close to the accuracy limits of the methods.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1180/clm.2024.25.
Acknowledgements
Natascha Schleuning, Annette König, Niko Götze and the BGR XRF laboratory are acknowledged for providing analyses.
Conflicts of interest
The authors declare none.
Open access
