Lab-scale performance of selected expandable clays under HLW repository conditions
Smectite clay has been proposed for embedding canisters with highly radioactive waste in deep repositories because of its isolating capacity. Montmorillonite-rich bentonite is a premier buffer candidate for many national organizations that are responsible for disposal of such waste. Experience from the use of drilling mud at large depths indicates that other smectite clay minerals are more stable chemically and saponite is one of them. The physical properties of smectitic mixed-layer minerals like Friedland clay are known to be less sensitive to high salt contents and such clay may also be a buffer candidate. Montmorillonite-rich MX-80 clay, Greek saponite with a minor amount of palygorskite, and Friedland clay were investigated in hydrothermal tests with dense samples confined in oedometers with 95 °C temperature at one end, which was made of copper, and 35 °C at the other, for 8 weeks. A 1 % CaCl2 solution was circulated through a filter at the cold end. At the end of the tests, the samples were sliced into three parts, which were tested with respect to expandability, hydraulic conductivity, and chemical composition. The tests showed that while the saponite was hardly changed at all and did not take up any copper, MX-80 underwent substantial changes in physical performance and adsorbed significant amounts of copper. The Friedland clay sample was intermediate in both respects.
KeywordsSmectite alteration THMC-framework Mineralogy Hydraulic conductivity Swelling pressure
Smectite clay has been proposed for embedding canisters with highly radioactive waste (HLW) in deep repositories because of its tightness, ability to expand for establishing tight contact with the canisters and the surrounding rock, potential to sorb cationic radionuclides, and ductility for minimizing tectonically induced stresses in the canisters. Following the theory of diffuse ion layers, smectite acts as a swelling phase in the direction of its crystallographic c-axis. Na-bearing montmorillonite can develop a strong swelling pressure of up to 3.92 × 108 N/m² (Madsen and Mueller-Vonmoos 1988). This expandability is the origin of the self-sealing properties of smectite. Smectite as buffer material is applied as a compacted block. The high swelling pressure and the compaction of smectite guarantee the required low permeability. The permeability is expressed as hydraulic conductivity (in m/s). Any alteration of smectite will affect the hydraulic conductivity. Under defined circumstances (e.g. with hydraulic conductivity ~10−13 m/s), smectite buffer (e.g. 1 m thickness) could offer safe waste isolation for at least 1 million years (Pusch and Yong 2006). The bulk hydraulic conductivity should be lower than 10−12 m/s in order to avoid advective transport. Bentonite should also be characterized by a swelling pressure >1 MPa to ensure tightness and self-sealing (Andrews 1992).
The clay, termed “buffer”, has to retain these physical properties and must survive the hydrothermal impact associated with hydration of the initially only partly water saturated clay under a temperature gradient of up to 2 °C per centimetre, and the thermally generated chemical processes in the fully hydrated clay in the subsequent period of successively decreasing temperature gradients. The first period lasts for 10 to 100 years in crystalline rock and for 100 to 1,000 of years in argillaceous rock that provides less water for hydration of the buffer clay. The period of dropping temperature lasts for a few 1,000 years after which the clay will not be exposed to heat generated by the radioactive decay.
The impact of heating in conjunction with water uptake of the buffer clay may change its physical properties significantly but no systematic long-term tests have been conducted so far with the exception of ongoing field experiments at AEspoe Underground Laboratory in Sweden, where repository-like conditions have been created for a few years and where the impact of thermal/hydraulic/mechanical/chemical (THMC) processes is currently recorded (Svemar 2005a). Longer experiments under controlled conditions in the laboratory would be invaluable but meet practical difficulties.
The study reported here is an attempt to determine changes in the physico/chemical properties of three potential buffer candidates in accelerated tests with higher temperature gradients than those in a repository but with relevant absolute temperatures. The three selected samples represented different degrees of occupation by Fe or Mg instead of Al in the octahedral sheet. Larger cation’s radii in the octahedral sheet as for Fe and Mg in comparison to Al could develop higher stresses, which would be expected to lead to a higher degree of alteration for smectites with remarkable amount of Fe or Mg. The trioctahedral saponite mirrored de-facto an end-member in this octahedral occupation. Earlier experiments have shown higher stability for a Fe-rich bentonite composed of illite–smectite mixed layers than Al-rich bentonites like MX-80 and GMZ (Xiaodong et al. 2011). In contrast, the Fe-rich Czech RMN-bentonite has shown remarkable alteration in hydraulic conductivity and swelling pressure in MOCK-Up-field experiments (Pusch et al. 2007). Lantenois et al. (2005) found in their experiments that trioctahedral smectites are essentially unaffected in comparison to dioctahedral smectites. Different authors concluded in their literature studies that the published experiments give only an equivocal indication to alteration processes of bentonite that may occur in an engineered backfill system, especially for experiments under moderate pH ranges (e.g. Kaufhold and Dohrmann 2009; Wilson et al. 2011). They assumed occurrence of few coupled processes (Wilson et al. 2011).
This matter needs further investigation and experiments conducted under repository-like conditions with respect to water saturation under temperature gradients and with canister material like copper present have given a deeper insight into the alteration processes than simple autoclave testing. They have, among other things, visualized the different intensities of bentonite alteration caused by different buffer clays.
Major minerals and chemical composition (in wt%) of the clays
Minerals: +++rich, ++intermediate, +traces
M+++, Q+, F+, C+
S+++, Pg+(+), Q+
MX+++, M+, Q+
Granulometry of the air-dry clay materials (in wt%) used for preparing compacted samples
The temperature gradient of slightly less than 15 °C per centimetre clay was maintained through the saturation period, which is estimated to have lasted for about 3 weeks, and in the subsequent 5-week long THMC period. Samples of the solution at the cold end were extracted twice, firstly after 3 weeks and at the termination of the 8-week long experiment.
At the end of the tests, the samples were directly transferred to oedometer cells with the same diameter as the hydrothermal cells in conjunction with sectioning into three parts for determining the hydraulic conductivity and swelling pressure at different distances from the hot end.
For X-ray diffraction (XRD) of randomly oriented powder samples, a SIEMENS D5000 equipment with a Cu-Kα wavelength and secondary curved graphite-monochromator (40 kV, 30 mA, variable slits: 20 mm, soller: 0.5/25, step and time: 0.02° 2Θ for 3 s) was used. Oriented mounts were analysed in a Präzitronic Freiberg HZG four diffractometer controlled by SEIFERT-C3000 unit (30 kV, 30 mA, Co-beam, Fe-filter, fixed slits: 1.09 mm/6.0 mm, soller: 0.5/25, detector slot: 0.35 mm; step and time: 0.03° 2Θ for 2 s).
Clay suspension was prepared after 10 min of ultrasonic treatment on carbon-coated Cu-grids by air drying. A large number of individual particles were characterized with respect to morphology, electron diffraction, and element distribution by use of a JEOL microscope equipped with a JEM-1210 unit (voltage 120 kV, LaB6-cathode). An OXFORD LINK energy-dispersive X-ray (EDX) system was connected to the transmission electron microscopy (TEM) equipment. The morphology was described according to Henning and Störr (1986). The element distribution was determined by using a JEOL JXA 840 scanning electron microscope (SEM) equipped with an EDX analyser KEVEX 8005.
All equipment was located at the Institute of Geography and Geology (University of Greifswald).
Physical properties and processes
MX-80: montmorillonite-dominated clay
Properties of MX-80, DA0464 and Friedland clays at different distance from the hot boundary (results from 8 weeks-experiments)
Distance from hot boundary, mm
Dry density, kg/m3
Hydraulic conductivity (K), m/s
Swelling pressure (ps), kPa
DA0464: saponite-rich clay1
The tests showed that the dry density increased from initially about 1,240 to 1,380 kg/m3 next to the hot boundary and dropped to 1,175 kg/m3 in the coldest part (Table 3). The swelling pressure remained high in the hottest part but was nearly the same as for untreated saponite clay in the central and cold parts. As for the montmorillonite-rich clay, the increase in density of the hottest part is ascribed to compression in the early hydration phase. The swelling pressure of the most heated part probably caused some compression of the softer, colder parts of the sample, but was not sufficient to totally even out the differences in density. The hydraulic conductivity of the hot part was about 70 times higher than that of untreated saponite while it was insignificantly higher than that of untreated clay with corresponding densities for the central and cold parts.
Friedland clay: mixed-layer mineral-dominated clay
The tests showed that the dry density increased from initially about 1,530 to 1,725 kg/m3 next to the hot boundary, while it remained almost unchanged in the central part and dropped to 1,360 kg/m3 in the coldest part (Table 3). The swelling pressure had dropped strongly in the hottest part and to about 70–80 % of that of untreated Friedland clay in the central and cold parts. As for the other clays, the increase in density of the hottest part is ascribed to compression in the early hydration phase. This indicates that the swelling pressure of the most heated part, which should have been sufficient to compress the softer, colder parts of the sample, had become strongly reduced. The hydraulic conductivity of the hot part was about 50 times higher than that of untreated Friedland clay while it was the same as for untreated Friedland clay of corresponding densities for the central and cold parts.
The solution percolated through the filter at the cold end was sampled after 3 and 8 weeks. The analyses showed very small concentration changes for all cations except for Mg, Na and K, which were released from the clays as a result of exchange of the initially adsorbed Na, Mg and K by Ca. After 3 weeks the Na concentration had increased from 1.6 to 460 ppm, while Mg had increased from 1.2 to 21 ppm and K from 0.5 to 22 ppm. Si had increased from virtually 0 to 0.8 ppm indicating some slight dissolution of the silicate minerals. The copper concentration remained below 0.001 ppm showing that no migration of Cu from the clays to the fluid on the cold side took place in the 8-week test period. Finally, the solution after 3 weeks has shown no significant deviation from the composition of the solution after 8 weeks.
Concentration of elements (in ppm) in the differently heated clays (analyses by LMI AB, Helsingborg, Sweden)
Parts exposed to different temperatures (°C)
Mineral changes of MX-80
Overview about mineral formulae of MX-80 montmorillonite minerals after termination of the experiments
(A) MX80, montmorillonite, original (Herbert et al. 2004)
Ca0.07 Mg0.00 Na0.22 K0.04 Al1.54 Fe0.173+ Mg0.26 Ti0.00 (OH)2 Si3.95 Al0.05 O10
(B) MX80, montmorillonite, after experiments (end-member of illite–smectite mixed-layer series)
Ca0.14 Mg0.00 Na0.05 K0.04 Al1.55 Fe0.123+ Mg0.26 Ti0.03 (OH)2 Si3.97 Al0.03 O10
Ca0.00 Mg0.00 Na0.30 K0.01 Al1.53 Fe0.183+ Mg0.28 Ti0.01 (OH)2 Si3.97 Al0.03 O10
(C) MX80, montmorillonite, after experiments (end member of dioctahedral vermiculite-smectite mixed-layer series)
Ca0.01 Mg0.04 Na0.04 K0.04 Al1.63 Fe0.173+ Mg0.18 Ti0.02 (OH)2 Si3.98 Al0.02 O10
Ca0.04 Mg0.00 Na0.00 K0.06 Al1.69 Fe0.173+ Mg0.12 Ti0.01 (OH)2 Si3.98 Al0.02 O10
The diversity of the diVS-ml series indicates obvious trends for Al, Fe and Mg in the composition of the octahedral sheet especially for the cold regions of the experiment. Al followed a decreasing trend with increasing vermiculitic ratio. Fe and Mg showed reverse behaviour with minimum for Fe, and maximum for Mg, for a dioctahedral vermiculite:smectite ratio of 70:30. The maturity of this process, i.e. illitization and evolution of octahedral composition, was much lower for the hot regions (Table 5).
In the XRD diagrams for oriented mounts of material from the hot region, a small 10 Å peak is seen but no mica-like phases were found in the TEM-investigations. It is believed that a significant part of the montmorillonitic phases collapsed in the hot region, which contributed to the higher density of the material in the hot region.
Montmorillonite in the colder parts was mainly unchanged by the hydrothermal treatment as illustrated by comparison with the original composition of MX-80 (cf. Tables 4, 5A, B). The sum of the Na, Ca and Mg contents is nearly the same in the cold and hot parts. The concentrations of Fe and Al are slightly higher in the colder part of the sample and there is also a weak tendency for the Si concentration to have risen in the central and colder parts.
Copper had dissolved and entered the clay to a significant extent in the hot part. The concentration gradient indicates that the migration is diffusive and curve fitting gives an approximate diffusion coefficient of E−12 m2/s. The general lower degree of alteration for the central and cold parts confirms the conclusions from similar experiments with Al-rich bentonites and Na as natural main cation in the interlayer (e.g. Xiaodong et al. 2011).
Mineral changes of DA0464
Ca replaced the initially adsorbed Na and Mg in the clay but the exchange was far from complete in the hot and central parts. The sum of the Na and Ca concentrations is almost 50 % higher in the coldest part than in the hot and central ones.
The concentrations of Fe, Al, Mg and Si dropped slightly in the direction of the temperature gradient. The conclusion is that dissolution of the minerals was less significant than for the MX-80 clay. Copper did not enter the clay at all. XRD has not shown any significant change in expandability of clays.
Mineral changes of Friedland clay
Ca replaced the initially adsorbed Na to a large extent in the hottest part and also to a high degree in the central and coldest parts. The content of Mg adsorbed by the clay is concluded to have been partly replaced by Ca or Fe set free in conjunction with slight dissolution of iron complexes. The largely uniform Fe, Al and Si contents show that the clay mineral structure remained nearly unchanged throughout the clay.
Copper had dissolved and entered the clay to a significant extent in the hot part and migration of this element had occurred even into the colder parts. The concentration gradient indicates that the migration is diffusive and curve fitting gives an approximate diffusion coefficient that is 10–50 times higher than for MX-80. This suggests effective migration paths, presumably represented by porewater in larger voids or channels. XRD has not shown any significant change in expandability of clays.
Although the agreement between the results of different investigations of MX-80 and Friedland clays is very good, the role of time and scale is not known. For these clays the experiments showed that about 25 % was very significantly degraded in a few weeks, which would be a serious matter if it applies to the buffer in a repository. The high temperature gradient could have exaggerated the degree of degradation, whereas the short testing time to underestimation of it.
Close-ups of the corroded copper plate that had been in contact with MX-80 clay revealed pitting corrosion in the form of caverns with rather distinct edges (Fig. 6b). Most of them had an average diameter of 5–200 μm and a maximum depth of up to 11 μm. The total mass of copper that was released from the clay-contacting area can be estimated at about 5 mg.
The different bentonites have shown clear differences in behaviour under comparable experimental conditions. The study showed that MX-80 underwent considerable changes, particularly in the form of a very significant loss of swelling pressure and a rise in hydraulic conductivity by 100 times in the hottest part. The central and cold parts were significantly less affected with respect to conductivity and expandability.
The physical properties of the saponite changed less than those of MX-80. Thus, the swelling pressure remained high even in the hottest part and was roughly the same as for untreated saponite clay in the central and cold parts. The hydraulic conductivity of the hot part was about 70 times higher than that of untreated saponite while it was insignificantly higher than that of untreated saponite in the central and cold parts.
The swelling pressure of the Friedland clay had dropped strongly in the hottest part but only slightly in the central and cold parts. The hydraulic conductivity of the hot part was about 50 times higher than that of untreated Friedland clay but remained unchanged in the central and cold parts. These results are in complete agreement with the outcome of other investigations of similar type (Xiaodong et al. 2011).
The overall conclusion is that the saponite clay suffered less from the hydrothermal treatment than the montmorillonite clay and that the mixed-layer mineral-dominated clay was intermediate in this respect.
The MX-80 clay adsorbed substantial amounts of copper to a significant distance from the copper plate. Saponite did not take up any copper at all and FIM clay was similar in this respect. The distribution of Cu in MX80 mirrored advective transport (also for FIM clay) and additionally local channel-type migration occurred. The reason for this is probably that the wider voids in MX-80 clay formed more continuous channels.
A very important conclusion of the study is that montmorillonite caused more pronounced corrosion of contacting copper metal than saponite, which was almost inert under the testing conditions. The clay dominated by mixed-layer minerals was intermediate in this respect.
The mentioned changes in physical parameters and the intensity of Cu corrosion like saponite < FIM clay ≪ MX-80 are not in agreement with the expected behaviour taking the octahedral Fe-content into consideration since this would imply MX-80 < FIM clay ≫ saponite, hence illustrating the uncertainty pointed out by Wilson et al. (2011) about the unequivocal indications from all published experiments.
In this context the results from Lantenois et al. (2005) experiments concerning the stability of saponite in similar experiments with Fe-powder should be considered. They concluded that the full occupation of all three octahedral sites by cations and missing octahedral Fe3+-cations in saponite inhibits the assumed mechanism of smectite alteration caused by electrochemical processes of corrosion. The lowest alteration of hydraulic conductivity and especially in swelling pressure for saponite sample may confirm Lantenois conclusions.
The different behaviour between MX-80 bentonite and FIM clay could be caused by different initial pH-values in the respective experiment and by different buffering capacities for dissolved Si. Wyoming bentonites in contact with water give a pH range >9, whereas, FIM clay gives pH <8 (e.g. Nguyen Thanh Lan 2012). The small pyrite content is responsible for this lower pH range of FIM clay. Beverskog and Puigdomenech (1998) modelled an immobile behaviour for Cu at 80 and 100 °C for an Eh range <−200 mV in case of pH <9, while Cu is known to corrode in this temperature zone in case of higher alkalinity than 9. That means that the stronger Cu-corroding effect by MX-80 bentonite can be explained by its high initial pH value.
Experiments by Nguyen Thanh Lan (2012) have shown a high alteration and degree of FIM clay accompanied by a high buffer capacity for dissolved Si. Dissolved Si seals decomposed smectite layers of illite–smectite mixed-layer series associated with partial smectitisation. The montmorillonitic character of smectite in MX-80 bentonite inhibits this buffer option and dissolved Si precipitates and cements adjacent smectite aggregates. It is assumed by the author that Cu can migrate in MX-80 bentonite only in channel-like pathways surrounding and separating cemented “micro-regions”.
Mineralogical composition of smectite expressed as the ratio of montmorillonitic layers in illite–smectite mixed-layer series, the amount of octahedral Fe3+, di- or trioctahedral type of smectite and the occurrence of pH-affecting admixtures (e.g. pyrite) seem to control the degree of corrosion of Cu.
Following Lantenois et al. (2005), the vacant sites in dioctahedral sheet and occurrence of octahedral Fe3+ act as catalysts for Cu corrosion explaining the inert behaviour of trioctahedral smectites in mineralogy and physical performance.
Illite–smectite mixed-layer phases offer a high buffer potential for dissolved Si in comparison to montmorillonitic end members of this mixed-layer series (Nguyen Thanh Lan 2012). This high buffer potential could compensate the higher degree of alteration for Fe-rich smectite like FIM clay. As a consequence of all that the thermal-coupled processes increasing hydraulic conductivity and reducing swelling pressure have a lower impact in barrier material with illite–smectite mixed-layer phases than in those with pure montmorillonite. This is hence believed to be the reason why MX-80 bentonite has shown stronger differences in barrier performance between less and more intense hydrothermal treatment than FIM clay.
Additionally, copper can better resist electrochemical corrosion processes better at low alkalinity or neutral pH conditions. High alkaline pH conditions, e.g. developed by MX-80 bentonite, plus the described catalytic function of dioctahedral smectite promote Cu corrosion. Cu oxidation is also increasing the pH range in a reducing environment that lead to higher dissolution potential of smectite. Precipitated Si cementing neighboured smectite particles is believed to support the observed development of channel-like advective transport of Cu in MX-80 bentonite.
These experiments demonstrate the high complexity of coupled processes. The approach of catalytic mechanism of dioctahedral smectites described by Lantenois et al. (2005), the Si buffer potential introduced by Nguyen Thanh Lan (2012), and the pH-related different behaviour of Cu, give suitable tools now to explain the observed experimental data without excluding any of the processes that have been identified or proposed.
One effect, not in focus of the study but of great importance for the performance of the buffer in a repository, is the role of cementation by precipitated compounds. It may hence be that the assumed congruent dissolution leading to precipitation of Fe and Si/Mg complexes as spot weldings that prevented the expandable components of the clays to swell after complete hydration (Pusch et al. 2010). If sufficiently extensive this process can increase the stiffness of the buffer causing less ductility and overstressing of the waste containers at seismically or tectonically induced shearing.
A final comment is that the finding that saponite-rich clay neither suffered significantly from the hydrothermal treatment nor interacted to a measurable degree with the copper would make it a strong buffer candidate in HLW repositories. This conclusion is also in agreement with earlier findings from deep oil drilling and some national buffer studies (Pusch and Yong 2006).
The clay can be taken as saponite despite the presence of palygorskite since it is judged from the comprehensive microstructural analyses that saponite forms the clay matrix and controls the physical properties.
Support to complete the manuscript from the National Foundation for Science and Technology Development, Vietnam (project code 105.02.54.09) is gratefully acknowledged.
- Andrews PRA (1992) Summary report No. 17: Bentonite, Fuller’s Earth and Kaolinite, CANMET, OttawaGoogle Scholar
- Beverskog B, Puigdomenech I (1998) Pourbaix diagrams for the system copper-chlorine at 5–100°C. SKI Rapport 98:19. Swedish Nuclear Power InspectorateGoogle Scholar
- Henning K-H, Störr M (1986) Electron micrographs (TEM, SEM) of clays and clay minerals. Verlag, BerlinGoogle Scholar
- Madsen FT, Mueller-Vonmoos M (1988) Das Quellverhalten der Tone. Mitteilungen des Instituts für Grundbau und Bodenmechanik der ETH—Zürich, 133:39–56Google Scholar
- Nguyen Thanh Lan (2012) Mineralogical Characterization of Fe-driven Alteration in Smectites. Inauguraldissertation, University of GreifswaldGoogle Scholar
- Pusch R (2002) Handbook of buffer and backfill. SKB Technical Report TR-02-12. Swedish Nuclear Fuel and Waste Management Co, StockholmGoogle Scholar
- Pusch R, Kasbohm J (2002) Alteration of MX-80 by hydrothermal treatment under high salt content conditions. SKB Technical Report TR-02-06. Swedish Nuclear Fuel and Waste Management Co, StockholmGoogle Scholar
- Pusch R, Yong RN (2006) Microstructure of smectite clays and engineering performance. Taylor & Francis, LondonGoogle Scholar
- Pusch R, Kasbohm J, Pacovsky J, Cechova Z (2007) Are all smectite clays suitable as “buffers”? Phys Chem Earth Parts A/B/C 32:116–122 (Clay in natural and engineered barriers for radioactive waste confinement—Part 1). doi:10.1016/j.pce.2006.03.008
- Svemar C (2005a) Prototype repository project. Final Report of European Commission Contract FIKW-2000-00055, Brussels, BelgiumGoogle Scholar
- Svemar C (2005b) Cluster repository project (CROP). Final Report of European Commission Contract FIR1-CT-2000-20023, Brussels, BelgiumGoogle Scholar
- Wilson J, Savage D, Bond A, Watson S, Pusch R, Bennett D (2011) Bentonite—A Review of key properties, processes and issues for consideration in the UK context. Quintessa Ltd., Henley-on-ThamesGoogle Scholar