Abstract

Bentonites are candidate materials for the encapsulation of radioactive waste. In the ‘Alternative Buffer Material test’ (ABM), compacted ring-shaped blocks of eleven different buffer materials (mainly bentonites) were packed vertically on top of each other with an iron tube as heater in the centre. These buffer materials started with various exchangeable cation populations (ECpopulation). The first ‘ABM package’ was terminated 28 months after installation and the bentonites had been exposed to the maximum temperature (130°C) for about one year. The aim of the present study is first: to describe modification of the cation exchange population, and second to understand the influence of the groundwater on cation exchange at different scales. No significant horizontal variation of any exchangeable cation (EC) was detected between 1 and 7 cm distance from contact with the iron tube. Large total differences of the ECpopulations, however, were observed for the individual blocks after the field experiment (n = 21 blocks) with respect to the composition of the reference materials. The average cation exchange capacity (CEC) values of the analysed bentonites (n = 9 blocks) decreased by 5.5 meq/100 (1.1–8.8 meq/100 g) after the experiment. Exchangeable Na+ and Mg2+ decreased on average, whereas Ca2+ increased. This trend was pronounced in the top region of the parcel (upper seven blocks). Although most changes occurred on the large scale of the whole test parcel, small but important changes were also recorded in the vertical direction on the centimetre scale. The observed differences cannot be explained assuming simply that a bentonite reacts only with neighbouring blocks, which would mean that the system was more or less closed. The differences are much larger and the only conclusion from this observation is that the whole package seems to be influenced by the groundwater which was added from a water tank at the experiment site, enabling at least partial equilibration between the different blocks.

Bentonites are candidate materials for the encapsulation of radioactive waste. Concepts for deep geological storage of such waste are based on bentonite buffers which are parts of engineered barrier systems. Under repository conditions, the long term stability of these engineered barrier systems in crystalline as well as in argillaceous rocks depends on the stability of the smectite minerals, particularly with respect to the swelling capacity and cation exchange capacity (CEC).

Experiments for the evaluation of the integrity of bentonite barrier systems have been performed on different scales: (1) lab scale with the possibility to vary different parameters; however, such experiments were often far from realistic. More realistic were (2) field experiments of intermediate size which were performed in hard rock laboratories or underground rock laboratories. Real scale experiments (3) in underground research laboratories have been rarely conducted, mainly because of the enormous amounts of material, preparation and construction work required. All scales are necessary to understand the processes related to cation exchange. The most important scale with respect to long-term safety analysis is the real scale. Exchangeable cation population (ECpopulation) studies are necessary to understand processes such as bentonite erosion in engineered barrier systems. Bentonite erosion is affected by the amount of exchangeable sodium (e.g. Kaufhold & Dohrmann, 2008; Missana et al., 2011), and chemical or physical processes that can affect the erosion such as redistribution of the ECpopulation have to be understood.

Laboratory-scale studies – the small scale

Madsen (1998) studied relevant properties of two bentonites which were well established in radwaste studies: MX80 (Na+–smectite) and Montigel (Ca2+–Mg2+–smectite). He compared their behaviour with altered K-bentonites and concluded that even if strong illitization would occur, “the fixed K+ in the illite-smectite mixed-layer is not hydrated” and that the remaining divalent exchangeable cations would preserve significant swelling capacity.

Other studies concentrated on performance of sodium bentonites in many different fields of application which were relevant for safety assessment. Of particular importance is the sealing capacity, which requires high smectite contents. Yong (1999) pointed out that, on the one hand, the sealing capacity of smectitic clays (bentonites) in the presence of water is “both a design consideration and a problem”. High sealing capacities are favourable because sealing elements will be homogenous and cracks/fissures can be closed via self-sealing. On the other hand, prediction of the extent and rates of swelling, wetting performance when swelling pressures develop, and redistribution of densities and water contents are problematic. Although model developments were quite successful, these problems are still not solved and, as the authors pointed out, all these problems were linked to clay microstructures which were strongly influenced by the distribution of ECs.

The distribution of the ECs is also influenced by accessory minerals of bentonites, particularly when compacted bentonites were subjected to a heat source. Arcos et al. (2000) tried to quantify and to model reactions of typical accessory minerals. In their model they used calcite, siderite, quartz, SiO2 (amorphous) and anhydrite in combination with realistic pore water, cation exchange competition and pyrite as a redox buffer. The relevant output of their model was that cation exchange reactions were still influencing the ionic composition of pore water after 100,000 years. The authors concluded that “The main process affecting the trace mineral behaviour in bentonite is cation exchange”. Montes–H et al. (2005), for example, confirmed this.

Field experiments – the intermediate scale

Plötze et al. (2007) studied the stability of a bentonite buffer (Almeria, Spain) in a small-scale heater experiment which was carried out at the underground laboratory in Mont Terri, Switzerland. The heater element had a diameter of 10 cm and was held at a constant surface temperature of 100°C. The blocks were water saturated for 35 months before an 18 month long heating phase was started. The authors concluded that the bentonite showed only very weak modifications during the heater experiment, such as cementing or aggregation processes. No mineralogical alteration other than cation exchange and very minor changes in layer charge density of smectites could be observed. Samples from the heater region showed an increase in exchangeable magnesium. The total layer charge density of the smectites decreased slightly; however, IR fitting of the OH stretching region did not show significant changes of the octahedral composition.

Olsson & Karnland (2011) studied the mineralogical and chemical characteristics of bentonite blocks (MX80, Wyoming, USA) used in the ‘Long Term Test of Buffer Material’ (LOT) project at the Äspö hard rock laboratory (HRL), Sweden. This project was initiated by the Swedish Nuclear Fuel and Waste Management Co. (SKB). In that project, samples were kept under repository-like conditions: The A2 test parcel was subjected to elevated temperatures (up to 130°C) and hydration by a Na-Ca-Cl type groundwater for almost six years before it was retrieved and analysed. The experimental design was such that nearly 40 bentonite blocks with a diameter of 28 cm and a height of 10 cm were packed on top of each other to give a 4 m high column. The blocks in the lower 2 m were heated to 130 °C by heater elements inside a central Cu tube, simulating the heat plume of the waste. The bulk density after hydration was reported as 2000 kg m–3. The main results were that (a) sulfate was redistributed horizontally in the heated part of the buffer – anhydrite accumulated in the warmer parts, whereas gypsum was dissolved in the peripheral parts of the buffer; (b) Cu was incorporated in the bentonite matrix at the surface of the Cu tube, indicating corrosion of the tube; (c) exchangeable sodium was replaced by exchangeable calcium and magnesium in the warmest zone; and (d) magnesium was also enriched by other processes which could not yet be clarified, possibly by incorporation of magnesium into the structure of montmorillonite.

Wersin et al. (2007) reviewed the knowledge on the performance of the bentonite barrier at temperatures above 100°C and concluded that for “more reliable information at temperatures beyond 130°C, a series of long-term experiments examining hydraulic, mechanical and mineralogical changes under realistic conditions would be useful”.

Real-scale studies

In the Canadian Underground Research Laboratory, a bentonite-based tunnel sealing experiment (TSX) was conducted between 1998 and 2004 in an intact volume of rock within a granitic pluton. Dixon et al. (2007) reported on the evolution of two bulkheads, one of which consisted of high-performance concrete and the other of blocks of compacted sand-bentonite material. The TSX was installed in a 3.5 m high by 4.25 m wide tunnel located 420 m below the surface. Heating occurred for a period of approximately one year, and after five years of operation the TSX was dismantled and sampled. To date, no mineral or ECpopulation analyses from this experiment have been published.

Dueck et al. (2011) reported on hydro-mechanical and chemical-mineralogical analyses of the bentonite buffer from a full-scale field experiment called ‘Canister Retrieval Test’ (CRT), which was larger than the LOT test. The compacted bentonite surrounding a copper canister equipped with heaters had been subjected to heating at temperatures up to 95°C and hydration by natural Na-Ca-Cl type groundwater for almost five years at the time of retrieval. The bentonite blocks used were 1640 mm in outer diameter and 1050 mm in inner diameter. The same kind of bentonite (MX80, Wyoming, USA) was used but, even at these lower temperatures, sulfate in the bentonite was redistributed and accumulated as anhydrite close to the canister. The authors observed a loss in exchangeable Mg2+ in the outer parts of the blocks.

Such real-scale experiments are helpful to scale up lab study and field study results to repository-like conditions. However, in all former real-scale and intermediate scale experiments, only one type of bentonite was used. SKB started an intermediate scale field experiment called ‘Alternative Buffer Material test’ (ABM) project (SKB, 2007) to overcome this problem.

ABM design

The setup of the ABM experiment is similar to the Swedish KBS-3 concept with a metal canister surrounded by clay situated in crystalline bedrock at approximately 500 m depth (SKB, 2007; Eng et al., 2007). This experiment was similar to the LOT experiment; the differences are mainly the scale, which is smaller for ABM, and that the canister is made of common carbon steel instead of copper. The reason for not using a copper pipe, as in most of the experiments atÄspö HRL, was to be able to study the effects of corroding steel in close contact with the buffer material.

The bore holes had a diameter of 30 cm and a depth of 3 m. The outer diameter of the blocks was 280 mm, the inner diameter was 110 mm, and the height was 100 mm. In each experimental package, three electrical heaters were installed to yield the target temperature in the bentonite blocks. A main heater ran along the entire package length (Fig. 1). Two additional heaters were installed, one at the bottom and another at the top to compensate for the temperature loss at the top and bottom and to give a more homogenous temperature distribution throughout the package length.

The experiment consisted of three packages in three separate boreholes; the duration was planned for one, three and five years followed by excavation. ABM aimed at using greatly different buffer materials (mostly bentonites) packed on each other. Eleven different clays were compacted to rings (with the exception of four steel cages containing granulated material) positioned on top of each other, encapsulating the tube. All reference materials except MX80 were installed two times in the test parcel, and all but MX80 were separated by other blocks (Table 1). MX80 was installed six times as pure MX80 block, two times as MX80 granulate, and two times as ‘MX80 granulate + quartz’. The ECpopulations of the different reference materials were significantly different.

The first out of three ‘ABM packages’ was heated from the start and the buffer materials were exposed for the maximum temperature (around 130°C) for about one year. The packages were moistened by groundwater from fractures (’ Äspö water’) and optionally also artificially from an installed wetting system using the same groundwater. Water saturation using titanium pipes is described by Eng et al. (2007) as follows: “The saturation system aims at simulating water bearing fractures in the rock wall. This is done by installing four pipes along the outer edge of each package. The pipes are connected to a water tank at the experiment site. At the location of the simulated fractures... ‘every 10 cm’...small holes are drilled in the pipes, allowing water to leak onto the buffer blocks.” The near-neutral Äspö water used for saturation is a sodium-calcium-chloride dominated groundwater (each ∼2500 mg l–1 Na+ and Ca2+, ∼8500 mg l–1 Cl and ∼500 mg l–1SO42) with minor contents of Mg2+, Br and K+ (all <100 mg l–1).

ABM 1 was terminated 28 months after installation. Almost all geochemical and mineralogical alterations of the different bentonites (apart from ECs) were restricted to the contact between iron and bentonite (Kaufhold et al., 2013); however, cation exchange processes were not included in the first part of the study. The aim of this second part is first to describe modification of the cation exchange population, and second to understand the influence of the groundwater on cation exchange at different scales.

Materials and Methods

Sampling

After excavation, the blocks were sampled at different distances from the contact to the iron tube (Fig. 1b) and from bentonite/bentonite and bentonite/cement interfaces (Fig. 1c). To study the horizontal direction, samples were taken from the central part of the blocks (see also Kaufhold et al., 2013). For buffer materials sampled by the German laboratory, the sample labelled ‘0.1 cm’ was collected by scraping off the surface layer of bentonite blocks at the contact to the iron tube with a sharp knife. The other samples were taken at 1, 3, 5 and 7 cm from the contact with the iron tube (German laboratory) and at 1, 5 and 9 cm (Swedish laboratory). In two selected small-scale studies (blocks 29 and 30 and blocks 10 and 11), the vertical direction was also sampled by taking samples in the direction parallel to the heater at approximately the centre of the blocks. For the two upper blocks 29 and 30, a narrow profile was sampled starting at 1.25 cm from the upper level of block 30 (bentonite/cement interface at the top of the ABM parcel). The other samples were taken in increments of 2 cm from 3.5 down to 19.5 cm from this bentonite/cement interface. For blocks 10 and 11, samples were taken on both sides of the bentonite/bentonite interface at distances of 1, 4 and 6 cm. Excavation of the ABM experiment could not be performed in an O2-free atmosphere. Accordingly no glove box was used in the laboratory, and possible exchangeable Fe(II) could not be determined.

CEC and EC methods

Two different CEC-index cations were used, providing accurate CEC and ECpopulation values even in calcareous bentonites (Dohrmann et al., 2012a, b): ammonium (method 1) and Cu(II)–triethylentetramin (methods 2 and 3). In this joint project, the ammonium chloride method was used in the Swedish laboratory, whereas the other methods were used in the German laboratory. The standard Cu(II)-triethylentetramin (Cu-trien) method has the advantage that it requires only 200 mg of sample, but the EC values suffer from partial carbonate dissolution, leading to overestimated exchangeable calcium values. Therefore, the Cu-trien5 × calcite method, which uses calcite saturation of the exchange solution, is preferable (Dohrmann & Kaufhold, 2009), although typically 1000 mg of sample are consumed. This is a large amount of material for some samples, such as the contact zone between the iron tube and the buffer material. In some experiments, only small sample quantities were available and the advanced Cu-trien5 × calcite method could not be applied. In such cases, only a few selected samples of the horizontal profile were analysed by Cu-trien5 × calcite, (a whole set includes samples from the contact to the iron tube and four samples from the central parts; see Fig. 1b). All measurements were performed in duplicate.

Method 1. EC values were determined using NH4Cl (ammonium chloride, 0.15 m) in ∼83% ethanol following Belyayeva (1967). Ethanol was used instead of water to minimize the dissolution of carbonates and gypsum. NH+4 exchange was repeated three times because the selectivity of NH+4 is low, and in contrast to index cations such as the di-valent Cu-trien (Meier & Kahr 1999) and the mono-valent silver-thiourea (Dohrmann, 2006), a single exchange step does not guarantee complete cation exchange. After NH+4-saturation and washing, ethanol was evaporated from the filtrate/supernatant and the volume adjusted by deionized water addition followed by ICP analysis of the ECs. No ammonium analysis was performed; accordingly no index cation CEC values were calculated. This method uses 800 mg sample for a single measurement.

Method 2. The CEC was determined according to Meier & Kahr (1999) using the standard Cu-trien method as discussed and modified by Kaufhold & Dohrmann (2003). After Cu-trien saturation, centrifuged solutions were diluted followed by ICP analysis of the ECs as well as of Cu, which allowed calculation of CEC values. Cu-trien was also analysed by VIS spectroscopy to cross-check the ICP-Cu concentration. This method uses a total of 200 mg sample for two repetitions.

Method 3. The Cu-trien5 × calcite method was used to minimize carbonate dissolution. The Cu-trien5 × calcite solution was prepared by mixing 300 mL of 0.01 m Cu-trien solution with 1500 mL of deionized water in a 2 L beaker and 2 g of fine-grained calcite to presaturate the solution and suppress calcite dissolution as described by Dohrmann & Kaufhold (2009). After Cu-trien5 × calcite saturation, centrifuged solutions were diluted followed by ICP analysis of the ECs and Cu, allowing calculation of CEC values. The Cu-trien complex concentration was also analysed by VIS spectroscopy to cross-check the ICP-Cu concentration. This Cu-trien5 × calcite approach does not prevent dissolution of gypsum (Dohrmann & Kaufhold, 2010) and uses 1000 mg sample for two repetitions.

For ICP-OEC analysis the following techniques were used: argon radial plasma, nebulisers (cross-flow and modified Lichte), no auxiliary gas flow, gain value for plasma (1.400 W), calibration every 7th measurement.

Results and Discussion

Horizontal variation of the CEC and the ECs within individual blocks

Horizontal variation of the ECpopulation was expected to be clearly visible (as for the MX80 bentonite in the LOT project) because the temperature was higher at the contact with the iron tube than at the outer parts of the bentonite blocks (Table 1). Exchangeable cation (EC) values of block No. 17 (#17), Kunigel V1 (JNB), are given in more detail in Table 2, because #17 is a typical example and this block was sampled by both laboratories.

Reference material. The JNB reference material is a Na+-bentonite (86–100% exchangeable Na+) with minor amounts of exchangeable Ca2+ (6–9%, only methods 1 and 3), and trace amounts of exchangeable Mg2+ (1–3%) and K+ (0–2%) (Table 2). The CEC values varied between 59–64 meq/100 g.

The JNB reference material contains approximately 3 wt.% calcite. Calcite dissolution inflates exchangeable calcium values of method 2. This can be checked by the parameter ‘sum/CEC’, which was calculated by adding up all ECs (in meq/100 g) and dividing this sum by the CEC (in meq/100 g). Ideally this control value should give 1.00, which is more or less achieved using method 3, Cu-trien5 × calcite. For calculation of the exchangeable Ca2+ population (%/CEC), therefore only methods 1 and 3 were used. No CEC results were available for method 1, ethanolic NH4Cl. Accordingly ECpopulation was calculated by dividing individual EC values by the parameter ‘sum of exchangeable cations’. For method 2 (standard Cu-trien) individual EC values were divided by the CEC because the ‘sum’ was inflated by calcite dissolution. For method 3, Cu-trien5 × calcite, the quotient ‘sum/CEC’ was close to the ideal value, which was expected to be similar for method 1 (ethanolic NH4Cl) as well (Dohrmann et al., 2012b).

Samples retrieved after the field experiment. In contrast to what was assumed, no significant horizontal variation of ECs was detected for block No. 17 between distances of 1 and 7 cm from the contact to the iron tube (Table 2, Figure 2). Initially dominating exchangeable Na+ was reduced by half (44%, down from 94). The lack of horizontal variation of the ECs was not expected but it enabled a calculation of average values of the whole block. A second advantage was that the statistical basis of the averaged values of the reacted samples was improved (n = 15 for Na+, K+ and Mg2+; n = 7 for Ca2+ and n = 12 for the CEC). The sample at the direct contact to the iron tube was not included in these average values, because the total mass of this layer was very small and it does not represent the bulk composition of the block. This contact region is important for understanding initial processes; however, it represents only part of the 0.1 cm thin surface layer of ∼8% of the total surface of the blocks, and much less of the total volume.

The average exchangeable Ca2+ proportion increased from 7 to 45% and was at the same level as the average exchangeable Na+ proportion. The average exchangeable Mg2+ proportion increased by the same factor of ∼5–6 as the average exchangeable Ca2+ proportion, from 2 to 10% exchangeable Mg2+. A small increase in exchangeable K+ was observed; however the exchangeable K+ values were too close to the detection limit to draw conclusions from this increase. CEC values were close to the reference CECs. The average CECs were 2 meq/100 g lower than the average CEC of the reference bentonite. Such a small deviation cannot be interpreted as a real difference.

Only the sample which was collected close to the bentonite/heater contact differed significantly. This sample had 10% (relative) less exchangeable Na+ than the other samples of the same block. The CEC of the contact sample decreased from 61 to 49 meq/100 g, which was probably caused by clinoptilolite dissolution (Kaufhold et al., 2013).

Average EC values of each retrieved and analysed block were calculated in the same way as described for block No. 17, representing the cation exchange population of the block. The average values are listed for the reference clays as well as for the samples from the field experiment (Table 3). As only two blocks were analysed by all three methods, the statistical basis is smaller for the other blocks: n=12 for Na+, K+, and Mg2+, and for the CEC; n = 4 for Ca2+ for the Cu-trien and Cu-trien5 × calcite methods; n = 3 for Na+, K+, Mg2+ and Ca2+ for the NH4Cl method).

Average CEC and EC values of the whole parcel and the influence of the groundwater

This section describes the vertical distribution of the ECs of the whole parcel in order to evaluate the influence of the groundwater at this larger scale. Prior to the field experiment, the ECpopulations differed from one block interface to the next. Compared to the reference materials, significant ECpopulation changes were observed after the field experiment (Table 3).

Large-scale variations of the whole test parcel (all 30 blocks) can be identified by a mass balance. ECpopulations of the blocks were widely different. To identify trends with respect to large scale variations, average values of the ECs and CECs were calculated (Table 3, lower three rows). Such average values only have a meaning for mass-balance estimations. The data set is not complete because not all blocks were analysed after retrieval (ECs: n = 21; CECs: n = 12). Average values were calculated only for samples with complete sets of data in both groups; i.e. reference materials and samples of the field test.

Exchangeable Na+ made up nearly half (48%) of the ECpopulation of the reference blocks, but only 39% of the samples after the field experiment Exchangeable Mg2+ decreased as well, from 18% to 15%. Exchangeable Ca2+, on the other hand, increased from 32% to 45%, whereas exchangeable K+ remained stable. Initial average values of all 30 blocks were relatively close to those of the selected 21 blocks: Na+ (53%), Mg2+ (16%), Ca2+ (29%) and K+ (3%),

It may be argued whether the quantity “percentage EC/CEC” would be suitable for mass balance estimations, because absolute differences of ECs are not considered. Low-CEC materials with high Ca2+ percentages of the ECpopulation would overprint the high-CEC samples. To identify such influences, average EC values of the 21 blocks were calculated, confirming the %EC/CEC values listed above: Exchangeable Na+ averaged 36 meq/100 g in the reference blocks, but only 26 meq/100 g in the samples after the field experiment. Exchangeable Mg2+ decreased as well, from 13 meq/100 g to 10 meq/100 g). Exchangeable Ca2+ on the other hand increased from 24 meq/100 g to 34 meq/100 g, and exchangeable K+ remained stable (from initially 1.7 to 1.5 meq/100 g). Initial average exchangeable cation values of all 30 blocks were relatively close to the selected 21 blocks: Na+ (39 meq/100 g), K+ (1.6 meq/100 g), Mg2+ (11 meq/100 g) and Ca2+ (21 meq/100 g).

Average CEC values of all analysed blocks decreased slightly from initially 66 meq/100 g down to 62 meq/100 g after the experiment. Among the group of “true” bentonites, initial CECs vary from approximately 60–100 meq/100 g with an average of 82 meq/100 g. The drop in average bentonite CEC after retrieval was 5.5 meq/100 g (1.1–8.8 meq/100 g). The precision of the methods (1 sigma) was 1–2 meq/100 g (Dohrmann & Kaufhold, 2009; Dohrmann et al., 2012), which means that the measured CEC differences indicate genuine decreases for at least some of the bentonites. Such CEC decreases can be interpreted in several ways: (1) a 10% decrease of CEC values was typical for bentonites in laboratory experiments of 38 different bentonites with NaCl, KCl, and Ca(OH)2 solutions (Kaufhold & Dohrmann, 2009, 2010a,b, 2011); (2) with decreasing exchangeable sodium content the pH decreased as well (Kaufhold et al., 2008), which reduced the amounts of variable charge and lowered the measurable CEC. The same mechanism could have also lowered the CECs in the ABM samples. On the other hand, only 12 blocks were analysed with respect to the CEC, which was therefore possibly less representative than the EC values. Within these 12 blocks, the correlation between the difference in sodium concentration (in %) of the ECpopulation and the CEC decrease (in %) was very poor (R2 = 0.36). This does not support the hypothesis of pH influence.

Along with the overall exchange reactions, minimum and maximum values of individual ECs in individual blocks converged after the field experiment (Table 3, lower two rows). The lowest concentrations of the reference materials were close to zero (1–2%) for any EC, whereas maximum values varied significantly: the highest concentrations measured for exchangeable K+, Mg2+, Ca2+ and Na+ were 11%, 39%, 76% and 90% were. After retrieval of the field samples, all minimum EC values apart from K+ increased as follows: Mg2+ from 1 to 7%, Ca2+ from 2 to 13%, and Na+ from 1 to 23%. All maximum EC concentrations decreased thus: K+ from 11 to 9%, Mg2+ from 39 to 26%, Ca2+ from 76 to 65%, and Na+ from 90 to 60%.

The upper part (blocks 24–30) of the ABM 1 parcel has lost (desorbed) more sodium and incorporated (adsorbed) more calcium than the middle and the lower parts. Exchangeable sodium was initially (63%) more concentrated in the upper part than in the whole parcel (53%). Exchangeable calcium concentration of blocks 24–30 was initially (20%) lower than in the whole parcel (29%). The lower twenty-three blocks averaged 50% exchangeable Na+ and 32% exchangeable Ca2+. After the field experiment these blocks (blocks 1–23) had an average of 41% exchangeable Na+ and 42% exchangeable Ca2+. The upper seven blocks (blocks 24–30) started with even more exchangeable Na+ (63%) but ended with less exchangeable Na+ (33%) than the lower twenty-three blocks. Exchangeable Ca2+, on the other hand, was initially lower in this upper part (20%) and subsequently higher (53%). The average exchangeable Mg2+ loss in the upper part of the parcel was small (-3%) and the exchangeable K+ loss was close to detection limit (–1%). In the lower part (blocks 1–23), no loss of exchangeable Mg2+ and K+ was detected. These pronounced differences can only be explained by taking into account interaction with the groundwater, which was added from a water tank at the experiment site using perforated titanium pipes allowing water to leak onto the buffer blocks.

Redistribution of exchangeable cations of the same materials in different parts of the parcel

During the ABM experiment, a significant redistribution of exchangeable cations of the individual blocks occurred (Table 3, Fig. 3).The general trend for most blocks (all bentonites) was:

  • The higher the initial exchangeable sodium (calcium, magnesium), the more sodium (calcium, magnesium) was lost during the experiment.

  • The target cation exchange population has a typical range for each EC: ∼8–15% exchangeable Mg2+, ∼25–30% exchangeable Na+ and ∼45% exchangeable Ca2+. Below this target concentration, a block typically gained the respective cation by ion exchange (net adsorption), whereas above this target concentration, a block lost significant amounts by ion exchange (net desorption).

  • In total fourteen blocks lost exchangeable Na+ and Mg2+ and seven blocks gained exchangeable Na+ and Mg2+. Exchangeable Ca2+ was desorbed only from five blocks; all other sixteen sampled blocks were enriched in exchangeable Ca2+. All of the seven upper blocks gained exchangeable Ca2+, but most of these blocks were initially very rich in Na+.

Influence of initial composition, neighbouring blocks, and relative position within the test parcel on cation redistribution. Seven blocks (apart from MX80, which was used more than two times) were sampled twice after retrieval of the ABM parcel. These blocks were always ‘sandwiched’ by other blocks. The question arises whether the initial composition of the neighbouring blocks of the ‘doubly analysed blocks’ had an influence on the cation redistribution of the differently sandwiched blocks. If the differently sandwiched blocks would not have been influenced by their neighbouring blocks, gains or losses of the three major ECs Na+, Ca2+, and Mg2+ should have the same trend and occur to approximately the same degree in both sampled blocks. This is the case for Calcigel (CAL), FRI, IBE and Rokle, but Asha 505 (Asha), Dep.CAN, and Febex behaved differently, having net gains or losses of these major cations in both sampled blocks with opposite signs (Fig. 4).

Example 1 (Fig. 4): The Ca2+(Mg2+)-rich buffer material Calcigel (CAL) gained large amounts of exchangeable Na+ (+40%) but lost exchangeable Mg2+ (–11%) and large amounts of exchangeable Ca2+ (–29%) at block position 5. The same material gained less exchangeable Na+ (+19%) at block position 23, exchangeable Mg2+ loss was similar (-9%) and loss of exchangeable Ca2+ (–7%) was less pronounced than at position 5. However, gains and losses showed the same trends at both block positions. This trend was confirmed for FRI, IBE, and Rokle. The most important reason for this trend could be that all of these bentonites were initially dominated by a single exchangeable cation, Ca2+ (Febex and Rokle) or Na+ (FRI and IBE).

Example 2 (Fig. 4): The buffer material Febex with an initial composition of Mg2+⩾Ca2+>Na+ gained large amounts of exchangeable Ca2+ (+26%) but lost exchangeable Mg2+ (–23%) with a minor loss of exchangeable Na+ (–3%) at block position 21. The same material gained large amounts of exchangeable Na+ (+15%) at block position 8 whereas exchangeable Ca2+ was at the same level (–1%) as the initial material. Apart from the low concentration of exchangeable K+, the only cation which showed a similar loss at both positions 8 and 21 was exchangeable Mg2+ (–14%).

Asha and Dep.CAN also showed a non-uniform behaviour similar to Febex (Fig. 4). Dep.CAN had no single dominating EC, which is similar to Febex. Asha, on the other hand, was dominated by Na+. Both Asha blocks were embedded between two Mg2+–rich blocks; however at position 14, Asha gained exchangeable Mg2+ whereas at position 24 exchangeable Mg2+ was lost. Block 24 belongs to the upper part of the parcel which gained an excess of exchangeable Ca2+. It is obvious that the block position was more important than the compositions of the directly neighbouring blocks.

Dep.CAN (blocks 15 and 27) with an intermediate composition of the ECpopulation (27% Na+, 46% Ca2+ and 27% Mg2+) was sandwiched between Asha (block 14) and Rokle (block 16) and IBE (block 26) and IKO (block 28). At position 15, the contrast of the ECpopulations of the neighbouring blocks to that of Dep.CAN was very large. The lower block 14 was very Na+-rich (Ca2+-poor), and the upper block 16 was Ca2+-rich (Na+-poor). Here Dep.CAN (block 15, in the middle of the ABM parcel) lost exchangeable Mg2+ (–10%) and gained minor amounts of exchangeable Na+ (+9%) and exchangeable Ca2+ (+4%). In contrast, in the upper parcel position (block 27), Dep.CAN gained 17% exchangeable Ca2+, although both neighbouring blocks were very poor in exchangeable Ca2+ (7% and 19%) and although the lower IBE-block 26 also gained significant amounts of exchangeable Ca2+ (+57%). The second difference was that exchangeable Na+ (–2%) was desorbed, although both adjacent blocks were initially much richer in exchangeable Na+ (factor 2 and 3). The observed differences cannot be explained assuming simply that a bentonite reacts only with neighbouring blocks, which would mean that the system was more or less closed. The differences are much larger, and the only conclusion from this observation is that the whole package seems to be influenced by the groundwater, which enabled at least partial equilibration between the different blocks.

Absolute gains and losses of exchangeable cations

In this section, absolute concentrations of ECs in meq/100 g are discussed (Fig. 5). Two clays with low absolute CEC values (12–23 mq/100 g, COX and FRI) were packed between bentonite blocks with approximately 5 times higher CEC values. The ECs of the reference and the reacted materials thus differed significantly (Fig. 5a,b).

The LOT experiment had shown that sulfate becomes redistributed horizontally in the heated part of the MX80 buffer (Olsson & Karnland, 2011). Other studies described Mg2+ enrichment in the hot parts (e.g. Plötze et al., 2007). The most significant change of the ECpopulation occurred in Ibeco Seal (IBE) (blocks 8 and 26), which was the sodium-richest clay within the reference materials (Fig. 5c). The two IBE blocks were in relatively cold parts of the test parcel, one in the upper and the other in the lower third of the package. The peak temperature at the contact to the heater was obviously not the driving force for the strong EC redistribution in the two IBE blocks. Asha showed pronounced changes at position 24 (in total 79 meq/100 g which equals 87% of the ECpopulation) and smaller changes at position 14 (in total 59 meq/100 g which equals 65% of the ECpopulation). At both positions the temperature at the contact to the heater was approximately the same. No correlation of the extent of the redistribution of the ECpopulation with the peak temperatures at the contact with the iron tube could be observed for the other blocks of the studied ABM 1 parcel.

The total amounts of redistributed cations (all absolute cation differences were summed up) do not correlate at all with the CEC (R2 = 0.01). The smallest absolute redistribution of all bentonites was found in MX80 at position 2 (14 meq/100 g), however for the same material up to 46 meq/100 g were recorded at the other end of the test parcel.

These examples underline the hypothesis that material properties were not responsible for the observed EC redistribution, but changes can be explained by interaction with groundwater.

Vertical distribution of ECs between different blocks and at the cement/bentonite interface on the centimetre scale

Although most changes occurred on the large scale of the whole test parcel, small but important changes were also recorded in the vertical direction on the centimetre scale. Two hypotheses were studied. (i) Did all ECs equilibrate to the same degree at the direct contact between two blocks? (ii) Did the cement plug on top of the ABM test parcel influence the pronounced gain of exchangeable Ca2+ in the upper part?

Equilibrium of ECs at the direct contact between two blocks. Two adjacent blocks (MX80, No. 11, and IKO, No. 10) were used for a selected small scale study aiming at identifying vertical equilibrium conditions between neighbouring blocks. The CEC of IKO is approximately 15–20% larger than that of MX80. In the reference material, exchangeable Na+ (57 meq/100 g) was at the same absolute level in both blocks but differed with respect to the percentage of the ECpopulation. Exchangeable Ca2+ was larger in MX80 (24 meq/100 g; 29%) than in IKO (18 meq/100 g; 20%), but exchangeable Mg2+/K+ was lower in MX80 (7/1.6 meq/100 g; 8/2%) than in IKO (24/2.2 meq/100 g; 29/2%). After the experiment, exchangeable Na+ had decreased significantly and exchangeable Ca2+ increased significantly. Both were at the same absolute level in both blocks (Fig. 6, right), but differed with respect to the percentage of the cation exchange population (Fig. 6, left). Exchangeable K+ and Mg2+ were at different levels in both blocks after the experiment. Exchangeable K+ remained unchanged compared to the starting material, whereas exchangeable Mg2+ increased in MX80 (from 7 to 14 meq/100 g or from 8 to 18%) and decreased in IKO (from 24 to 19 meq/100 g or from 26 to 20%). Both blocks were sandwiched between FRI and COX with very low amounts of exchangeable Mg2+ (max. 5 meq/100 g). These blocks could not act as an exchangeable Mg2+ source, and the FRI block adjacent to the Mg2+-rich block IKO had approximately the same amount of exchangeable Mg2+ after retrieval as before. Exchangeable Mg2+ did not migrate into the FRI block but it was transported from the IKO block into the MX80 block.

All ECs show bimodal concentration distributions with gaps of 2%, 5% and 7% for Mg2+, Ca2+ and Na+, respectively. Assuming that equilibrium would have been established and assuming that selectivities would have been equal for both materials, the resulting distributions should have been equal for all ECs if expressed as %/CEC (Fig. 6 left). This is obviously not the case and it remains unclear if either the reaction time was too short or the selectivities are different.

Influence of the cement plug on top of the ABM test parcel

The bentonites of the top region of the ABM experiment were subjected to cement for a relatively short period of time. During concrete degradation, the early pore water leached may release sodium and potassium hydroxides (Sanchez et al., 2006). The question was if the cement plug on top of the ABM test parcel had influenced the pronounced gain of exchangeable Ca2+ in the upper part. The distribution of ECs along a profile from the contact between the cement and the ‘top blocks’ (MX80, No. 30 and 29) was as expected. The hypothesis was that the cement acted as a Na+ and K+ source, which should have caused a pronounced effect on the uppermost parts of the bentonites. In the direct contact layer exchangeable Na+ and K+ concentrations should have been larger than below that layer, with decreasing concentrations downwards. After a few centimetres to decimetres, the general trend of exchangeable Ca2+ increase in the upper part of the whole test parcel should have caused an increase of exchangeable Ca2+ and Mg2+ with distance from the cement/bentonite interface. This was confirmed in the ABM experiment (Fig. 7).

Exchangeable Ca2+ increased from 40% at the top to 44% at the bottom of the lower block. The most significant increase was observed for exchangeable Mg2+, starting from 6% at the top to 14% in the lowest sample. Exchangeable Na+, on the other hand, decreased from 51% to 40%, and a minor decrease from 3% to 2% was observed for exchangeable K+. Visual inspection indicated that iron corrosion has influenced the upper 3–4 cm. This was confirmed by XRF analysis. The iron content (Fe2O3) was ∼5 wt.% in the cement/bentonite contact zone, and decreased to ∼4 wt.% in the lower parts of the two top blocks (Kaufhold et al., 2013). The sodium (Na2O) content decreased continuously from 1.9 wt.% down to 1.4 wt.%. This continuous sodium decrease confirms the observed decrease in exchangeable Na+. XRF analyses also confirmed the continuously increasing portions of exchangeable Mg2+ and Ca2+ (MgO: 2.3 to 2.6 wt.%; CaO: 1.6 to 2.3 wt.%). The structural elements of the smectites, SiO2 (67.1 wt.%) and Al2O3 (21.1 wt.%), remained unchanged within 0.1 wt.%. Obviously the cement plug above the ABM test parcel directly influenced the top blocks, but the influence was limited to a range of only a few centimetres to decimetres.

Summary and Conclusions

The results of previous field studies such as the LOT project led us to expect that horizontal variation of the ECpopulation would also be clearly visible in this field experiment. As in other heater tests, the temperature in the buffer materials was higher at the contact with the iron tube than with the outer parts. The horizontal variation of the ECs, however, was so small that no horizontal variation of any EC was detected for the retrieved blocks between 1 and 7 cm from contact. Instead it was decided to calculate average values for the whole blocks of the field experiment and to use these average values for mass balance estimations.

Large total differences of the ECpopulation, however, were measured for the individual blocks after the field experiment (n = 21 blocks) with respect to the composition of the reference materials. During the field experiment the bentonites had lost sodium and magnesium, and losses and gains were observed in all parts of the parcel, but the upper part had lost (desorbed) more sodium and incorporated (adsorbed) more calcium than the middle and lower parts. Two possible Ca2+ sources can be excluded, Ca-carbonates and Ca-sulfates, because Kaufhold et al. (2013) showed that both inorganic carbon and sulfur concentrations varied only insignificantly in the bulk blocks. The source of the observed gain in exchangeable Ca2+ was obviously groundwater. In general, the observed differences are too large to be explained assuming a closed system. The differences are much larger, and the only conclusion from this observation is that the whole package seems to have been influenced by groundwater which was added from a water tank at the experiment site, enabling at least partial equilibration between the different blocks.

Although most changes occurred on the large scale of the whole test parcel, small but sigificant changes were also recorded in the vertical direction on the centimetre scale. The bentonites of the top region of the ABM experiment were subjected to cement for a relatively short period of time, which influenced the ECs over a distance of two blocks against the overall trend of exchangeable Ca2+ enrichment. In the vertical EC profiles of two adjacent blocks (block 10 and 11), all ECs show bimodal concentration distributions with gaps of a few percent for exchangeable Mg2+, Ca2+ and Na+. It remains unclear if the reaction time was too short or if the selectivities are different.

The authors are grateful to Natascha Schleuning and Wolfgang Glatte for their great analytical work. Comments and suggestions by four anonymous reviewers (partly of an earlier version of the paper) were valuable and helped to improve the manuscript.