Fracture smectites in bedrock at the Kivetty and Romuvaara sites, Finland, as natural analogues for radioactive waste repository buffers

Bentonite is used as a buffer and backfill material in repository designs for geological disposal of spent nuclear fuel (e.g. SKB 2011; RWM 2017; Nagra 2021; Posiva 2021a) (Fig. 1a, b) and in some designs for low- and intermediate-level waste (e.g. Pusch 2003). Montmorillonite, the predominant mineral in most bentonites (e.g. Meunier 2005), is found in the Finnish bedrock, along with other smectites, as an alteration mineral from a range of geological processes (e.g. Uusinoka and Alkio 1976; Marcos 1989; Gehör et al. 1997a, b; Kärki et al. 1997; Front et al. 1998; Vartiainen 2005; Gehör 2007; Viola et al. 2011, 2013; Posiva 2021b; Nordbäck et al. 2023), providing potential analogues for buffer materials.

In Finland, the bentonite surrounding the waste container (see Fig. 1b) is planned to be of Na-type when installed in the deposition hole (Posiva 2021c) with c. 87 wt% of montmorillonite (e.g. Kiviranta et al. 2017) plus illite, muscovite, quartz, feldspars, carbonates, pyrite and gypsum. The exchangeable cation composition (ECC) of the reference bentonite is dominated by Na: e.g. 74% Na, 18% Ca, 7% Mg and 1% K (Kiviranta et al. 2017). However, groundwater–bentonite interaction will alter the ECC in the repository over time (see e.g. Posiva 2021d).

Chemical erosion, a process that may lead to the loss of swelling clay from repository components under low ionic strength conditions, may occur when sufficiently dilute groundwater comes into contact with bentonite in bedrock with advective pathways (fractures). In general, the processes of erosion and sedimentation within the bedrock fractures are taken into account in the safety cases that assess the performance of the geological repositories (e.g. Hjerpe et al. 2021). The erosion threshold or clay dispersion process depends on the ionic strength and composition of the groundwater, the initial composition of the clay, and the groundwater-flow conditions (e.g. Posiva 2021d). Generally, Na-bentonites are more prone to chemical erosion than Ca-bentonites (e.g. Birgersson et al. 2008; Kaufhold and Dohrmann 2008; Hedström et al. 2023). The total charge equivalent of cations in groundwater should remain above 8 mM to prevent chemical erosion, based on the cautious assumption that these cations are predominantly Na⁺ (Posiva 2021a). Extensive experimental (e.g. Baik et al. 2007; Birgersson et al. 2008; Vilks and Miller 2010; Schatz et al. 2013, 2016; Svoboda 2013; Reid et al. 2015; Hedström et al. 2016; Börgesson et al. 2018; Alonso et al. 2018, 2019) and modelling (Liu et al. 2009; Neretnieks et al. 2009; Neretnieks et al. 2017; Börgesson et al. 2018) work has been carried out to assess the potential risk of the chemical erosion processes on the waste repositories (for a summary see Posiva 2021d); however, uncertainties still remain, some of them specific to the high water/rock ratios (e.g. Muurinen and Lehikoinen 1999) that may be encountered in bedrock fractures surrounding the bentonite in the repositories. Observations from field conditions (Posiva 2021b) do not support the likelihood of erosion of the bentonite buffer (Posiva 2021c, e).

Given the concern (e.g. Posiva 2021d) about potential bentonite loss under dilute groundwater conditions, the stability of the natural smectites at the Kivetty and Romuvaara sites in Finland (Fig. 1) are examined here as potential natural analogues (see Reijonen et al. 2023 for a definition of, and a discussion of the uses of, natural analogues) for the long-term performance of engineered bentonite barriers in a repository. The presence of smectite (especially montmorillonite) in bedrock fractures at these sites (e.g. Gehör et al. 1997a; Kärki et al. 1997) is a powerful argument for the longevity of bentonite stability under similar conditions, while detailed data for repository safety cases are currently lacking (e.g. Reijonen and Alexander 2015; Reijonen and Marcos 2016). The main objective of this study is to assess the mode of occurrence and mineralogy of the natural smectites in such localities.

Both research sites (Fig. 1) selected for this study record the present-day dilute conditions at repository-relevant depths (Pitkänen et al. 1998; Anttila et al. 1999b). The Kivetty site lies in the Svecofennian granitoid complex of central Finland (with an age of c. 1880 Ma). The bedrock at Kivetty consists mostly of porphyritic or equigranular granodiorite–granite with minor occurrences of gabbro (Anttila et al. 1992). These have been metamorphosed and deformed in two phases, and brittle deformation features are common but decrease with depth. Based on previous studies (Lindberg and Paananen 1989, 1990, 1992; Gehör et al. 1995, 1996a, 1997a, 1998), the most common fracture minerals at Kivetty are calcite, iron sulfides and iron oxyhydroxides (Fig. 1c). Clay minerals, iron oxides and quartz occur in lesser quantities. The bedrock at Romuvaara (2800 Ma) is in the Archean basement complex of eastern Finland. Typical rock types are migmatitic banded gneisses (tonalite, leucotonalite and mica gneiss) that are cross-cut by granodiorite and metadiabase dykes (Anttila et al. 1999b). The reported fracture mineral studies have focused on existing/geologically recent water-conducting fractures (Gehör et al. 1996), and the most common fracture minerals are calcite, iron sulfides, iron oxyhydroxides and clays (Fig. 1d). At both sites some uncertainty exists in the knowledge of paragenetic relationships, and minerals observed in the same assemblage are not necessarily formed synchronously (Gehör et al. 1995, 1996a, b).

The hydrogeochemical evolution at both sites is well known (Pitkänen et al. 1998; Anttila et al. 1999b). Salinity and the degree of water–rock interaction are low. The low salinity is typical for an old crystalline bedrock environment that has experienced episodes of meteoric water infiltration (including glacial melting) over an extended period during the Quaternary. Based on thermodynamic stability calculations by Pitkänen et al. (1996, 1998), the groundwaters at the study sites in general would favour the alteration of the bentonite buffer mass (assuming an original Na-form) towards Ca-, Mg- and K-smectite but at a negligible level, based on the consideration of buffering capacity in the large mass of buffer filling the repository cavities (Melamed et al. 1992; Muurinen et al. 1996; Pitkänen et al. 1996, 1998). However, the solid/liquid ratio can be much smaller in bedrock fractures than in the bentonite buffer, potentially leading to a different and ranging exchangeable cation composition (ECC) of montmorillonite at lower densities (cf. Muurinen and Lehikoinen 1999) than calculated for compacted bentonite buffers. At both sites, while smectite is observed at all depths (down to c. 1 km), calcite is almost absent from the topmost parts of the bedrock (Fig. 1c, d) due to dissolution by meteoric water close to the bedrock surface (Pitkänen et al. 1996, 1998). This would suggest a higher relative stability of smectite over calcite, even in conditions of relatively recent meteoric water infiltration.

The fracture clays are likely to represent very old alteration phenomena (e.g. Hall et al. 2021; Nordbäck et al. 2023). The most recent geological event potentially increasing the temperature of the bedrock was related to burial under sedimentary cover, which thinned out towards the east, during the Paleozoic: that is, at least 250 Ma (e.g. see Larson et al. 1999). However, the ages reported from apatite fission-track dating for Kivetty and Romuvaara show even older ages for low-temperature conditions: c. 605–290 and c. 900–543 Ma, respectively (Larson et al. 1999). Even older ages are known for fault gouge illites from south to southern Finland, which range from 967.6 ± 19.7 to 697.3 ± 14.1 Ma from Myras (Elminen et al. 2018) and from 1549.7 ± 31.7 to 561 ± 11.2 Ma from Olkiluoto (Mänttäri et al. 2007; Viola et al. 2011, 2013; Nordbäck et al. 2023). Many of the samples from Olkiluoto (Viola et al. 2011, 2013; Nordbäck et al. 2023) contain smectites, smectite–illite or smectite–vermiculite, constrained not just for illitization but also indicating likely significantly older (>250 Ma) ages for the formation of smectites (hypothesized to be retrograde and likely to have been formed subsequent to illite at lower temperatures). Thus, smectite formations can be considered to be at least 250 Ma, especially at depth where weathering phenomena can be ruled out.

In this study we record the mineral composition of factures in granitoid rocks of Finnish bedrock, and attempt to document the geological processes that reflect the existence and long-term stability of clay minerals in conditions that correspond to those in the buffer and backfill material of repository designs.

Two boreholes from Kivetty (KI-KR1 and KI-KR5) and one borehole from Romuvaara (RO-KR2) were sampled for fracture-filling clays (n = 35). An analysis of the <63 µm clay fraction was carried out using X-ray diffraction (XRD) and scanning electron microscopy (SEM). Thin sections and fracture surface samples were prepared for microscopy and SEM-EDS (energy dispersive X-ray spectroscopy) examination. The samples and analytical techniques are listed in Table 1.

X-ray computed tomography (XCT) was carried out with a GE Phoenix V|tome|x S, using a 240 kV microfocus tube and with 0.5 mm of Cu used as a beam filter. At each angle the detector waited for a single exposure time and then took an average over three exposures. Some samples were scanned as vertical multiscans, where several scans were combined into a single 3D image. Samples for the XCT were selected based on several chosen fracture types of interest in hand specimens in order to obtain information on the fracture-filling material inside the drill core.

From the sheared and fractured surfaces of the Kivetty KR1 and KR5 and the Romuvaara KR2 drill core samples, powder of the secondary alteration minerals was extracted manually under a binocular microscope. The aim was to collect as much homogeneous clay material from each fracture as possible whilst trying to avoid sampling the underlying igneous or metamorphic silicates and the possibly earlier crystallized carbonate or chlorite layers underlying the clay filling.

For X-ray powder diffraction (XRPD) measurements, part or all the fracture material was ground manually in an agate mortar in an acetone suspension, poured onto a glass slide, spread evenly and then dried. Samples were measured using a Bruker D8 Discover Bragg-Brentano powder diffractometer equipped with a Cu tube, 0.3° fixed divergence slit, 2.5° sollers, beam knife, spinner, Ni filter and Lynxeye silicon strip detector. Powder diffractograms were measured from the 4°–71.5° 2$θ$ range using 40 kV and 40 mA power settings in continuous mode, 0.02° 2$θ$ s⁻¹ for 1 h/sample. Relative humidity was measured during the analysis and varied between 30 and 40%. Phase identifications were carried out using Bruker EVA software and the International Centre for Diffraction Data (ICDD) PDF-4 (Powder Diffraction File) Minerals 2020 database. The presence of smectite-group clay was tested by wetting the sample with a few drops of ethylene glycol (EG), stored in a EG desiccator overnight, and then analysing using the same settings as above. For samples with very small amounts of powder, the excess EG layer resulted in a high amorphous background noise level. Nonetheless, in the low 2$θ$ range, clear peaks were detected.

Three polished thin sections (30 $μ$m thick) of selected samples with an abundance of fracture-filling minerals (KI-KR1_56.71m, KI-KR5_781.83m and RO-KR2_203.70m) were prepared at the Thin Section Lab, Toul, France. The thin sections were made using oil instead of water during grinding and polishing to preserve the swelling clays in the samples. The thin sections were studied for petrology and mineralogy using polarized light and scanning electron microscopes.

Detailed studies of both the thin sections and the fracture-filling surface samples were carried out by SEM-EDS analyses. The drill-core samples consisted of half-cores that were c. 4 cm in diameter and 20–25 cm in length (Fig. 2). These samples were selected based on the XCT and XRD results in order to obtain data from fractures of interest with different mineralogical properties. Backscattered electron (BSE) images and the chemical compositions of fracture-filling minerals were collected using an Hitachi SU3900 SEM equipped with an Oxford Instruments EDS-spectrometer X-Max 20 mm² (SDD). BSE images were used to document smectite and other clay minerals at ×100 and ×5000 magnifications. EDS was used in the semi-quantitative chemical analysis for the identification of each mineral phase using a 20 kV accelerating voltage and 1 nA probe current.

The XCT 3D imaging confirmed the different fracture types defined during the sampling and guided the thin section location selection to examine open, closed and multiphase fracture-filling materials (see Supplementary material 1). Sample KI-KR1_56.71m showed an open fracture, seen as black in the greyscale image (Fig. 3a), with fine-grained fracture-filling material. Sample KI-KR5_781.83m was confirmed to contain a closed fracture type (Fig. 3b). Sample RO-KR2_203.70m has been assigned as closed based on drill core logging; XCT imaging additionally shows a layered structure in the filling material, suggesting multiple phases of fracture mineral growth (Fig. 3c).

The petrographical thin sections of the granitic rocks (KI-KR1_56.71 and KI-KR5_781.83m) from Kivetty contain the mineral assemblage plagioclase, quartz, biotite, Fe–Ti oxides, hornblende and K-feldspar (Fig. 4a–d). The sample (RO-KR2_203.70m) from the Romuvaara area has a similar mineral composition but lacks hornblende (Fig. 4e, f). Fracture-filling minerals at Kivetty are mainly clay minerals, carbonates (calcite), chlorites, iron sulfides and epidote (Fig. 4a–d). The alteration of mafic minerals such as biotite and hornblende to chlorite (Fig. 4b) and the replacement of plagioclase by calcite and sericite (Fig. 4c, d) can be clearly seen. The sample from Romuvaara (Fig. 4e, f) shows the textural relationship between the secondary mineral formations (clay appearing secondary to carbonate formation). Fragments of the main rock-forming minerals are also seen in the fracture-filling assemblages. There is no significant difference between the origins of the fracture-filling clay for the two sites, both show typical products of hydrolysis reactions and hydrothermal alteration (cf. Weisenberger and Bucher 2011).

The XRD phase identification results from the fracture minerals are provided in Table 2. The sampling was focused on the fracture-filling material but the adjacent wall components were also usually present in the preparation due to the inclusion of very small amounts of host-rock phases in the fracture-filling material (fracture widths were mostly <1 mm; see Fig. 2). The primary minerals observed in XRD included quartz, albite, microcline, biotite, muscovite and amphibole. Some of the analysed quartz, albite and micas can also occur as secondary occurrences in the fracture-filling material. The most common fracture minerals identified by XRD were calcite and clay minerals. Accessory minerals detected included REE fluorcarbonate (synchysite, a common hydrothermal accessory mineral), muscovite, phlogopite, annite, pyrophyllite and hematite.

The clay minerals detected included smectite group minerals, illite, chlorite and mixed-layer phases. Kaolinite/serpentine and talc were identified in some samples. Smectite group minerals were present in almost all samples. Smectite was identified from air-dried samples by its (001) peak at 14–15 Å d-spacing range but, in three samples, the main peak of montmorillonite was close to 12 Å, suggesting potentially Na as the main exchangeable cation (Brindley and Brown 1980). Smectite samples showed a shift of (001) reflection towards 17 Å after ethylene glycol (EG) treatment. Only five samples showed a peak shift to 16.5 Å or more; with most samples the expansion was smaller, to about 15 Å. This could indicate the presence of mixed-layer clays, where illite, chlorite or vermiculite layers are mixed with the smectite layers in the mineral structure. Diffractograms are provided in Supplementary material 2.

SEM-BSE imaging and SEM-EDS analyses were performed on cut drill-core samples and polished thin sections. By observing the clay mineral morphologies and their chemical compositions, it was possible to establish that smectite and chlorite were the dominant components in most of the analysed samples, with a minor amount of illite and kaolinite.

The SEM images of the surface fractures in cut drill cores (Fig. 5) and polished thin sections (Fig. 6) showed morphologies ranging from typical thin platy/flaky shapes to swirly features and honeycomb textures, which are characteristic of clay minerals with a smectitic (often authigenic) composition (cf. Fesharaki et al. 2007; Iacoviello et al. 2012). Kaolinite and illite occur as book-like forms and as platy or scalloped forms, respectively. Mixed-layer clays, intermediate products of clay mineral transition and alteration, are also present in some samples, such as smectite–chlorite (S/C), smectite–illite (S/I) and illite–chlorite (I/C). An overview on the minerals detected with SEM-EDS in the thin sections and drill-core samples is provided in Table 3, while the full dataset of SEM-EDS clay mineral compositions is given in Supplementary material 3.

In sample KI-KR1_56.71m the smectite particles occur as a cellular texture with dimensions ranging from 5 to 10 µm (Fig. 5a). This sample, taken close to the bedrock surface from an open fracture, may have been affected by some weathering reactions. Sample KI-KR1_398.58m shows that smectite aggregates occur as a cornflake or cellular texture associated with radiating platelets of clinochlore (Fig. 5b). A thin platy/flaky morphology of smectite is clearly observable in samples KI-KR1_398.58m and RO-KR2_441.57m (Fig. 5c, d). Samples from both sites have a minor occurrence of disseminated synchysite, confirming the XRD observations (not in the same samples, which is likely to be due to the very small sample amounts available for XRD). Synchysite occurs in the same fracture-filling mass as smectite, suggesting the simultaneous formation of synchysite with late-stage clays, sometimes replacing them (Fig. 5c, d). In addition, the BSE images of samples RO-KR2-306.77m and RO-KR2_203.70m show smectite composed of fine platy particles (1–5 µm) associated with S/I and S/C mixed-layer phases (Fig. 5e, f).

The analysis of clay mineral compositions using SEM-EDS is not an accurate technique. Potential problems in the SEM-EDS interpretation arise from the poor resolution, electronic beam penetration through very thin materials such as clay minerals, and incorrect measurement geometry of the rough mineral surface compared to a flat polished surface. The electron beam is powerful enough to penetrate through a thin mineral (clays) into any underlying mineral grain. This gives rise to the weak detection of elements from the studied mineral grain and the addition of elements from other surrounding mineral grains (cf. Welton 2003). Thus, the SEM-EDS data are treated semi-quantitatively in this study.

Different clay mineral phases, mainly composed of Si, Al, Fe, Mg, K, Ca and Na, were recognized based on the SEM-EDS analysis. The full dataset of SEM-EDS clay mineral compositions is presented in Supplementary material 3. EDS analyses of clay minerals with major elements of Si, Al, Fe, Mg, Ca and Na were interpreted as Ca- and Na-montmorillonites, whereas Si, Al, K and Fe were interpreted to be illitic in composition. SEM-EDS cannot determine whether the analysed Ca and Na are present in the smectite crystal structure or in the surrounding phases, obviously creating uncertainty in the results. However, by targeting the analyses on smectite textures (based on BSE images), it is possible to obtain at least indicative data on the potential smectite composition. In most of the studied samples, the EDS spectrum also indicates the presence of Mg in smectite that is associated with clinochlore (Mg-rich chlorite) to form mixed-layer S/C. Mixed layers of S/I were also identified in some samples, their elemental composition consisting of Si, Al, and minor amounts of K, Ca, Na, Mg and Fe.

SEM-EDS analysis results from the three thin sections confirm that the dominating fracture-filling clays are smectites occurring with chlorite, pyrite, Fe oxides and calcite. Clear indications of various types of argillic alteration can be observed. In the sample KI-KR1_56.71m, Na–Ca-smectite fracture filling in plagioclase (Pl) was observed (Fig. 6a). It had been formed in the open fracture planes, probably due to the release of lithostatic stress, accompanied by fluid penetration and subsequent alteration via plagioclase dissolution. The feldspar of the host rock in the vicinity of a fracture flow path can be easily dissolved by the attack of H⁺ in flowing fluids (e.g. in hydrothermal waters). Na–Ca-smectite is often presumed to have formed by chemical weathering and dissolution–precipitation from feldspar in the presence of fluids allowing hydrolysis (Fig. 6a, b) (e.g. Kadir 2007). In Kivetty, Na–Ca-smectite, interpreted as montmorillonite, occurs in sealed microfractures (Fig. 6a) as part of a larger open fracture in the sample (Table 1). In addition, Na–Ca-smectite is observed as precipitates within fractures along K-feldspar–plagioclase grain boundaries (Fig. 7a). In the same sample, smectites were observed as alteration minerals of plagioclase and hornblende (Fig. 6b). The sample also show partial chloritization of biotite and precipitation of Fe-smectite in microfractures and in the dissolution pores (Fig. 6c). Smectite was detected as a partial or total transformation product of hornblende, as shown in sample KI-KR5_781.83m, producing Mg-smectites on the centre of the grain and S/I in the rim of the hornblende and plagioclase grains (Fig. 6d).

SEM-EDS results of the RO-KR2_203.70m sample allow an interpretation of the textural relationship between the secondary carbonate and clay in the fracture network. Smectite, mixed-layer S/C and calcite are the main secondary mineral deposits sealing these fractures (Fig. 6e, f). Similar dissolution features as those for Kivetty are seen as Na–Ca-smectite clusters in pore-/fracture-filling material and replacing feldspars (Fig. 7b). The smectites occur due to hydrothermal alteration paragenesis, and are a typical product of hydrolysis reactions at the expense of microcline and plagioclase. Alkaline pH and low redox potential in the groundwater are factors that may favour the formation of smectite and calcite (Lima et al. 2011). Furthermore, XCT imaging of the sample RO-KR2_203.70m (Fig. 3c) shows that calcite has formed in association with smectite, producing a layered mass of denser (lighter grey in XCT image) calcite as coarse-grained fracture-filling and less dense (darker grey) smectite and S/C.

Based on the SEM-EDS analysis of thin sections, the clay phases result from intensive leaching of the primary minerals of the granitic host rocks by the infiltration of the hydrothermal fluids, as illustrated in Figures 6 and 7.

In understanding the uncertainties mentioned above regarding the semi-quantitative SEM-EDS data, clay analytical data from SEM-EDS were used to examine the overall chemical compositions (Fig. 8) to check the overall assessment of smectites present and to exhibit the overall similarity of duplicate samples analysed both from cut rock samples and thin sections. The chemical composition of most samples reflects advanced alteration and the presence of smectites (Fig. 8a). Large variations are seen in the overall percentiles of K, Na and Ca, highlighting the difficulty in assessing the smectite composition based solely on these chemical data and reinforcing the need for textural and petrographical support analysis (Fig. 8b).

Hydrothermal alteration at both sites was mainly focused on the fractures with previously known alteration overprints of muscovite/sericite, carbonate and epidote, with the Romuvaara site exhibiting more extensive alteration than that found at Kivetty (e.g. Pitkänen et al. 1996). At the Kivetty site, there is clear evidence of altered, porous (e.g. Anttila et al. 1999a) hydrothermal zones from several drill holes (Gehör et al. 1995), including the drill hole KI-KR5 that is included in this study. However, our sampling was targeted on fractures located outside of this porous zone. At Kivetty, hydrothermal alteration has been strongly oxidizing, leading to the thorough hematization of biotite and the alteration of feldspars to clays in the affected zones (Gehör et al. 1995).

Fractures with an alteration halo, filled with clays and secondary minerals (calcite, iron sulfide, iron oxide and hydroxides) were found in the studied samples. Hydrothermal phases observed consisted mainly of carbonates and clay minerals (smectite, S/I and S/C). Of these, smectites were detected along both open and closed fractures, although most of the fractures examined in the drill cores were classified as open. XCT was found useful for thin section location selection. However, thin-section analyses indicate that there are several types of small-scale fractures found in the samples when examined at the microscopic level that are not detectable using XCT.

The most common fracture minerals observed by XRD were calcite and smectite. Smectite occurs in the samples and also commonly as the mixed-layer clays (S/C and S/I) that are found in all samples. Based on the SEM-EDS results, smectites in various compositions can be present in one sample and in the same fractures. This is typical of smectites formed via in situ alteration of different minerals, where smectite formation reflects the composition of the parent mineral. SEM-BSE imaging of the fracture filling of cut drill-core samples showed that smectite is the main pure clay mineral, and most of the smectite particles displayed both honeycomb/cellular and cornflake textures, as well as flaky shapes.

Petrographical analyses support the hydrothermal origin of the smectites in the fractures and their in situ occurrence as alteration products of the parent minerals. The alteration patterns observed followed the typical patterns (e.g. Fulignati 2020) of hydrolysis reactions at the expense of silicate phases, (plagioclase, feldspar and amphiboles being the most important ones at the sites studied). In the thin sections examined in this study, replacement of parent minerals by clays were observed in situ via:

• feldspars or amphibole → Na–Ca-smectite;

• amphiboles and chlorite → Mg-smectite; and

• biotite → chlorite → Fe-smectite.

Defining the exact smectite mineralogy has uncertainties due to the limitations of the methods used. However, the SEM images in Figures 5–7 and the SEM-EDS analyses (see Supplementary material 3) show considerable compositional differences between the various types of clay minerals in the samples studied, allowing the observation of Na–Ca-smectites, Mg-smectites and Fe-smectites supported by the typical smectite honeycomb and cellular textures in the BSE images. In addition, compositions of S/C and S/I were observed, sometimes displaying a clear alteration pattern from the altered parent mineral, through an interstratified phase to a pure smectite composition. Pure smectites were also observed to be directly replacing minerals without interstratified minerals present. Based on the XRD and SEM examinations, the Na–Ca-smectites were identified as montmorillonite, confirming the presence of bentonite buffer-relevant smectites in bedrock fractures.

Most of the samples obtained were located in fractures that have had recent groundwater circulation (open fractures). As both sites are likely to have had fresh groundwater conditions prior to the latest glaciation and subsequent glacial retreat, meteoric water infiltration has been ongoing for extended periods of time (e.g. Anttila et al. 1999a, b). In Table 2, the known hydrogeological conditions around the sampled core are provided for context with regard to groundwater flow, showing that smectite is (including Na-montmorillonite) observed in situ, even in fractures with measured groundwater flow, supporting the stability of smectites. As various types of smectites are possible in bedrock fractures, it is difficult to assess to what extent subsequent alteration of ECC in montmorillonite has taken place. In the samples analysed, Ca-form is more pronounced, complying with the overall understanding from previous studies (Pitkänen et al. 1996, 1998).

To assess the natural analogue of fracture-filling clays to bentonite buffer potentially intruding into/in contact with fractures in the host rock in the geological repository conditions, the overall mineralogy needs to be considered. At both sites, mineralogical parageneses are complex and smectites do not occur as monomineralic masses in any of the fracture fillings studied. However, bentonites that may be used as the buffer material in a radioactive waste repository are also known to have accessory minerals, and the differences in the overall mineralogical composition between bentonites in the repository and the fracture-filling materials analysed in the present study are not expected to be unreasonably large. Although no quantitative data are available for the fracture minerals present, a rough comparison can be made between the XRD major and minor/trace phases and bentonite composition. The same minerals can be found in the smectite-bearing fractures as in the bentonites (smectite, muscovite, quartz, plagioclase, K-feldspar, pyrite, calcite and illite) (Fig. 9). The occurrence of abundant chlorite is the most significant difference compared with the reference bentonite buffer overall mineralogy discussed here (Kiviranta et al. 2017), although chlorite can be a component in bentonites in general. The natural analogue presented is conceptualized schematically in Figure 9. Similar conceptualization has been used previously as a basis for natural analogues for bentonite stability by, for example, Reijonen et al. (2024) for fresh to saline groundwater environments and Reijonen and Alexander (2015) for brine conditions.

Hydrothermal microcrystalline clays are found in the fractures of granitic bedrock in situ at two sites, Kivetty and Romuvaara, in Finland (ranging to well below repository-relevant depths: i.e. below 500 m). They consist of smectites, illite and chlorite, and associated mixed-layer clays (S/C and S/I), and reflect their parent mineral compositions. They were formed in hydrothermal low-temperature environments (estimates ranging from the Precambrian to at least 250 Myr ago), prior to development of the current groundwater regime. The results presented here further define the alteration mineralogy reported from both sites (Gehör et al. 1996a, b), expanding the clay mineral observations from kaolinite, montmorillonite and smectite clay mineral mixtures towards specific smectite species. The occurrence of Na- and Ca-type smectites/montmorillonites indicates that both compositions may prevail in the current fresh groundwater regimes. Moreover, the in situ occurrence of smectite, shown in the petrographical assessment, suggests that no erosion/resedimentation of clays in fractures has occurred since the hydrothermal alteration event.

While no simple montmorillonite locality was observed, smectites are present throughout the system, including Na-bearing smectites (here assigned to be montmorillonites). In open fractures, where groundwater movement is slow – as is to be expected around the deposition hole of a geological radioactive waste repository – results presented here indicate smectite stability in low-salinity groundwater conditions.

Further method development is required to investigate small quantities of smectite occurring in fracture systems to obtain detailed quantitative descriptions of smectite mineralogy and exchanger compositions. One option for obtaining such samples would be direct sampling from, for example, tunnel excavations, which would remove the possibility of potential loss of samples by drilling.

Heikki Nurmi (Geological Survey of Finland) is thanked for help in the initial sampling campaign at the Loppi drill-core storage facility in Finland. Russell Alexander (Bedrock Geosciences) and Pasi Heikkilä (Geological Survey of Finland) are acknowledged for their useful comments during the writing of this paper. The anonymous reviewers of Geoenergy are thanked for their fruitful comments on the manuscript.

HMR: conceptualization (lead), data curation (equal), formal analysis (equal), funding acquisition (lead), investigation (lead), methodology (equal), project administration (lead), software (lead), visualization (lead), writing – original draft (lead), writing – review & editing (lead); TA-A: data curation (equal), formal analysis (equal), investigation (supporting), methodology (supporting), visualization (supporting), writing – original draft (supporting), writing – review & editing (supporting); TE: data curation (equal), formal analysis (equal), investigation (supporting), methodology (supporting), writing – original draft (supporting), writing – review & editing (supporting); JK: data curation (equal), formal analysis (equal), investigation (supporting), methodology (equal), visualization (supporting), writing – original draft (supporting), writing – review & editing (supporting); KI: data curation (supporting), formal analysis (supporting), methodology (supporting), writing – review & editing (supporting); RL: conceptualization (supporting), writing – original draft (supporting), writing – review & editing (supporting).

Posiva Oy, Eurajoki, Finland partially funded this research.

The authors declare the following financial interests that may be considered as potential competing interests: the work was conducted partially within a company-funded (Posiva Oy) project. No other competing interests are identified.

One dataset generated during and/or analysed during the current study are available in the Fairdata.fi repository, https://doi.org/10.23729/1ce75faa-d52a-43cd-af40-3961b3081f70. Two other sets of supplementary material (Supplementary material 2 and Supplementary Material 3) generated or analysed during this study are included in this published article (and its supplementary information files). Raw data are available from the corresponding author on reasonable request.