- © The Mineralogical Society
The Boom Clay Formation of early Oligocene age, which occurs underground in northern Belgium, has been studied intensively for decades as a potential host rock for the disposal of nuclear waste. The goal of the present study is to determine a reference composition for the Boom Clay using both literature methods and methods developed during this work. The study was carried out on 20 samples, representative of the lithological variability of the formation. The bulk-rock composition was obtained by X-ray diffraction using a combined full-pattern summation and single-peak quantification method. Siliciclastics vary from 27 to 72 wt.%, clay minerals with 25–71 wt.% micas, 0–4 wt.% carbonates, 2–4 wt.% accessory minerals (mainly pyrite and anatase) and 0.5–3.5 wt.% organic matter. This bulk-rock composition was validated independently by major-element chemical analysis. The detailed composition of the clay-sized fraction was determined by modelling of the oriented X-ray diffraction patterns, using a larger sigma star (σ*) value for discrete smectite than for the other clay minerals. The <2 μm clay mineralogy of the Boom Clay is qualitatively homogeneous; it contains 14–25 wt.% illite, 19–39 wt.% smectite, 19–42 wt.% randomly interstratified illite-smectite with about 65% illite layers, 5–12 wt.% kaolinite, 4–17 wt.% randomly interstratified kaolinite-smectite and 2–7 wt.% chloritic minerals (chlorite, “defective” chlorite and interstratified chlorite-smectite). All modelled clay mineral proportions were verified independently using major-element chemistry and cation exchange capacity measurements. Bulk-rock and clay mineral analysis results were combined to obtain the overall detailed quantitative composition of the Boom Clay Formation.
The Boom Clay Formation is a clay-rich Oligocene marine sedimentary deposit in Belgium that has been studied intensively for several reasons: it represents the historical unit stratotype of the Rupelian stage; it has geotechnical properties, often key in civil engineering projects and the clay has been used extensively in coarse ceramics in the past and is still used today. The Boom Clay deposit is now best known for its underground laboratory and research site for long-lived intermediate and high-activity nuclear waste disposal in Belgium (ONDRAF/NIRAS, 2013). Therefore, the mineralogy and in particular the quantitative clay mineralogy of the Boom Clay Formation is of great practical importance and has been studied extensively in the past (Thorez, 1976a; Vandenberghe, 1978; Decleer, 1983; Vandenberghe & Thorez, 1985; Bonne, 1989; Goemaere, 1991; Bouchet & Rassineux, 1993; Merceron et al., 1993; Laenen, 1997; Wouters et al., 1999).
The large differences in bulk mineralogy, reported by these studies (Table 1) can be attributed in the first place to sedimentological differences in the silt and carbonate content, which will determine the proportions of quartz, clay minerals, calcite and feldspars (see Vandenberghe et al., 2014).
However, the reported clay fraction compositions vary greatly, not only quantitatively, but also qualitatively (Table 1) and this latter variability is too great to give credibility to the value of these analyses. Most of these results were generated using X-ray diffractometry (XRD) and suffer from: (1) the lack of standardized sample preparations; (2) the use of qualitative methods that were highly descriptive of 00l reflections upon different treatments, but did not take into account the physical nature of the clay phases present, particularly the interstratified minerals (e.g. Thorez, 1976b); and (3) the use of semi-quantitative analysis methods that relied only on the 00l peak height or approximate area measurements.
Thus, the present study was undertaken to address these issues by applying to a representative set of Boom Clay samples: (1) internationally accepted and well established standardized sample-preparation methods (Jackson, 1975; Moore & Reynolds, 1997; Środoń et al., 2001); (2) a bulk analysis method with a proven record of highly accurate and repeatable results (e.g. Kleeberg, 2005; Omotoso et al., 2006); (3) modelling of the oriented diffraction patterns of the extracted clay fractions (e.g. Sakharov et al., 1999); and (4) independent verifications of the bulk and clay-fraction results from supplemental major element chemical analyses and cation exchange capacity (CEC) measurements.
The primary purpose of applying the above methods was to determine the qualitative and quantitative bulk and clay-mineral compositions of the Boom Clay. In addition, by applying the independent verification approaches, secondary information on mineral chemical compositions and calculated petrophysical properties were obtained. All these methods were applied to characterize the Boom Clay as a material only, without discussing the geological implications of the analyses further. The relationship between mineralogy and sedimentology has been discussed in a recently published review of the geology of the Boom Clay (Vandenberghe et al., 2014).
The Boom Clay Formation and the Selection of Samples
The Belgian Boom Clay is a marine sediment deposited during the Lower Oligocene (Rupelian) in the North Sea Basin at 50–100 m depositional depth (Vandenberghe & Mertens, 2013). Shortly after deposition the environment became reducing as is shown by the common occurrence of pyrite in the Boom Clay. At present the clay contains about 20 wt.% of water and the samples analysed were never buried deeper than about 300 m. The outcrop area constitutes the Rupelian historical unit stratotype succession. At the time of deposition, tectonic subsidence affected a large area of northern Europe as a distal reaction to on-going Alpine compression and at the same time water depths varied due to fluctuating global sea-level in response to the build-up of a significant Antarctic cryosphere.
The main lithological properties of the Boom Clay are the rhythmic alternation of silty- with more clay-rich layers of some tens of dm thickness and the occurrence of organic-rich layers of similar thickness and of carbonates, mostly unmixed into horizons with septarian carbonate concretions. Based on these lithological properties the Boom Clay Formation is subdivided into four Members; all layers are numbered and all characteristic septaria horizons have a separate S-number (Fig. 1). Details of the lithostratigraphy of the Boom Clay can be found on the Belgian National Stratigraphic Commission website (http://ncs.naturalsciences.be).
The Boom Clay Formation crops out along the Rupel River and along the Scheldt River south of Antwerp. It dips slightly to the northeast where it is covered by Neogene sediments; the maximal burial depth of the top of the Boom Clay in Belgium – outside the Roer Valley Graben – is slightly more than 300 m and the maximum thickness is almost 150 m (maps in Welkenhuysen et al., 2012). Towards the southwest, the Boom Clay has been eroded and in the outcrop area less than half of the originally deposited clay has been preserved.
The layers in the Boom Clay are highly continuous and laterally and lithologically homogenous (Vandenberghe et al., 2001); therefore the sample selection was focused on the vertical lithological variations within the Boom Clay. Twenty samples representative of the vertical lithological variations known from previous sedimentological studies (Vandenberghe, 1978; Vandenberghe et al., 2001) were collected from the walls of exploitation pits in Sint-Niklaas (Belsele-Waas, Waasland), in Rumst-Terhagen (near Boom, along the Rupel) and in Kruibeke (south of Antwerp along the Scheldt River) and from the core of the ON-Mol-1 well (Fig. 1; Supplement 1). Files containing supplementary information have been deposited with the Principal Editor and are available at www.minersoc.org/pages/e_journals/dep_mat_cm.html. This sample set can be considered representative as the mineralogical variation in the preserved Boom Clay deposits.
The sample preparation method proposed by Środoń et al. (2001) was followed for bulk mineralogical analysis. Air-dried samples were hand-ground to <0.5 mm and milled in methanol for 5 min in a McCrone Micronizing Mill together with 10 wt.% of a zincite internal standard. The resultant powder was side-loaded into a specially designed sample holder for XRD analysis.
The hand-ground <0.5 mm material was also used to extract the clay fractions (<2 μm and <0.2 μm). The extraction techniques involve the removal of carbonates, organic matter and Fe-(oxy-)hydroxides followed by grain-size fractionation by centrifugation (Jackson, 1975). The extracted clay fractions were converted to their Ca-form; as expandable layers in this cation state, clay minerals have low susceptibility to ambient humidity variations (Eberl et al., 1987; Sakharov et al., 1999). In addition, cation saturations with Mg and K were prepared for specific identification purposes.
Smear mounts and sedimentation slides were prepared from the <2 μm fractions for pattern-quality comparison purposes. Only small differences on experimental diffraction patterns were observed for clay mineral ratios between these different types of preparation. However, the sedimentation slides produced stronger intensities and less noise because of the better preferred orientation of the clay minerals. Therefore, they were chosen as the preferred oriented-slide-preparation method, for both the <2 μm and <0.2 μm size fractions. Infinite thickness was achieved by depositing 11–12 mg of clay per cm2 of glass slide (Moore and Reynolds, 1997).
X-ray diffraction analysis
All samples were recorded on a Philips PW1830 X-ray diffractometer with: CuKα radiation, goniometer radius = 173 mm, proportional detector PW3011/00, graphite crystal monochromator, divergence slit 1°, receiving slit 0.1° and soller slit 2.3°, scanning step size of 0.02°2θ, scanning time of 2 s/step and scanning range for powder samples of 5–65°2θ and for oriented slides of 2–47°2θ.
For the bulk samples, in a first step the different mineral phases were identified using standard identification procedures as described by Moore and Reynolds (1997). The quantitative composition was determined using the Quanta software (Chevron ETC proprietary). In contrast to other full-pattern methods in clay mineralogy such as the Rietveld approach (Ufer et al., 2012), Quanta is based on the summation of pre-recorded standards and combines the advantages of both full-pattern summation and single-peak quantification methods. In this method phyllosilicates are quantified based on their 060 reflections and are reported in groups. Both the method and the software have been demonstrated to be highly accurate and repeatable (Kleeberg, 2005; Omotoso et al., 2006).
The qualitative and quantitative composition of the clay fractions was determined by modelling the experimental diffraction patterns obtained from oriented slides of the extracted clay fractions using the Sybilla software (Chevron ETC proprietary). Sybilla combines the formalism for calculating the diffraction patterns of interstratified clay minerals presented by Drits & Sakharov (1976), with a user-friendly graphical interface. During modelling the “multi-specimen” method proposed by Sakharov et al. (1999) was used. This method requires that a model should produce satisfactory fits to the experimental patterns in at least two states, glycolated and air-dry. A fit is considered satisfactory if a close match is achieved for peak positions, peak ratios and peak shapes using realistic input parameters for each layer type and proportions of the different layer-types in interstratified minerals. Modelled layer-type parameters include: mean crystal size thickness, d spacing, Fe content, K content (illitic layer-types) and water and glycol contents for expandable layer types. Besides satisfactory fits in both glycolated and air-dry states, it is also necessary that these fits produce the same or similar quantitative compositions in both states, while using the same layer-type parameters and proportions. The only exception is the proportion of one and two layers of the saturating fluid (water and ethylene glycol), which are expected to change significantly between the two states due to a larger expression of charge heterogeneity and strong dependence on relative humidity in the air-dry state (e.g. Ferrage et al., 2007). In addition, specific additional treatments have to be applied and combined to identify the different chloritic minerals and to investigate whether vermiculite is present (see the Clay mineral identification section below).
Finally, the sigma star orientation parameter (σ*) was taken into account. Although Reynolds (1986) suggested that σ* is set at 12° for all clay minerals during the modelling of oriented diffraction patterns, Dohrmann et al. (2009) demonstrated that for unconsolidated and poorly consolidated samples (such as this Boom Clay), the value of σ* for a discrete smectite is clearly larger than for the other clay minerals. Not taking this effect into account would lead to a significant underestimation of the smectite content. Therefore, in this study the following σ* values were used in the Sybilla software, based on a previous study by Zeelmaekers (2011) on unconsolidated samples: 11° for discrete smectite and 7° for all other phases for the <2 μm fraction and 8° for discrete smectite and 7° for all other phases for the <0.2 μm fraction. Using a trial-and-error approach employing various σ* ratios, the above values gave a consistent match between the modelled clay content of these samples and independent validations based on major element chemistry and CEC.
The CEC was determined on bulk samples using the copper(II) triethylenetetramine-method (Amman et al., 2005; Meier & Kahr, 1999; Dohrmann et al., 2012). The standard deviation of the CEC value (±1σ) was calculated from three independent measurements (three variable weight loads) of the individual sample.
Chemical analysis of both major and trace elements was carried out on bulk fractions (<2 μm and <0.2 μm) a commercial service by Activation Laboratories (Ontario, Canada); the quality of the results was monitored by adding the NIST70a and NIST76a reference standards to the sample set: details of the instruments used can be found at www.actlabs.com. Major elements and some trace elements were analysed by a lithium metaborate/tetraborate fusion method; a more extensive set of trace elements was analysed by inductively coupled plasma mass spectrometry (ICP-MS). Boron was determined by using prompt gamma neutron activation analysis (PGNAA), N by thermal conductivity detection after combustion, Cl by instrumental neutron activation analysis (INAA), weight loss at 110°C (“H2O−”) and at 1000°C (“H2O+”) by gravimetry, FeO by cold acid digestion and titration, C-total/C-graphitic/C-organic/CO2/S/SO4 by combustion/infrared detection and coulometry. The factor to convert the C-organic% to the organic matter% (OM%) in a sample (calculated from the H/C and O/C ratios in Boom Clay kerogen (Deniau et al., 2008)) is 1.47, well in line with the conversion factor proposed for more evolved non-soil organic matter (Ignasiak et al., 1985). Additional K2O analyses, used for the independent checks of the modelling results of the oriented clay fractions, were done on the extracted clay fractions using flame photometry (Sherwood model 420 UK).
The chemical analysis data were used to validate the Quanta modelling results by first comparing the actual data with the calculated expected chemical composition from the XRD-derived mineralogical analysis, based on the typical formulae and amounts of the minerals identified in the samples. After an initial agreement between the bulk-rock mineral and chemical compositions has been determined, these two factors can be combined in order to refine the detailed mineral formulae. The Bestrock program (Derkowski et al., 2008; Chevron ETC proprietary software), a modification of the Bestmin program (see Środoń & Kawiak, 2012; Chevron ETC proprietary software), was used to optimize the bulk-rock analysis results and to calculate the structural formulae of the minerals with a variable chemical composition, using intrinsic relationships and limitations of element contents in each mineral formula. Bestrock calculation uses a multi-variable non-linear optimization software package with the Solver® engine (Frontline Systems Co., Incline Village, Nevada, USA). In its optimization routine, Bestrock is used to combine, statistically, over a set of comparable samples, the elemental chemistry, the CEC and the H2O adsorbed at 110°C, together with any other mineralogical information available to help constrain the variable compositions of the minerals. As a second step of the Bestrock workflow, the optimized mineral quantities are used to distribute trace elements using linear least-square-based equations with no constraints on element concentrations. Using these results in the final step of the workflow, a set of petrophysical parameters, typically recorded by borehole logging tools, are calculated for the formation studied. The Bestrock-calculated petrophysical properties for the Boom Clay were verified by multiple runs with various boundaries and assumptions that returned very similar mineral formulae. Major elements fit the calculated contents with a relative average error of 2%. Trace-element distributions were validated by correlating the measured and calculated bulk-rock values and obtaining R2 values of 0.97–0.99 with a trend-line slope of 0.988–1.000. The practical value of the Bestrock results with respect to wire-log data interpretation and formation evaluation, has been presented in detail by Środoń & Kawiak (2012) and McCarty et al. (2015).
Bulk Rock Analysis
The bulk-rock analysis results are reported in Table 2. The wt.% of organic matter (OM), calculated from the C-organic analysis and the wt.% of apatite, calculated from P2O5 analysis, were combined with the bulk quantitative results, which were re-normalized to 100 wt.%.
Quartz and the 2:1 clay minerals and micas group (illite, smectite, interstratified illite-smectite, muscovite, glauconite, etc., or “2:1 clays” for brevity) are the dominant minerals. Quartz and “2:1 clays” are inversely correlated because they express the alternation between silty and clay-rich layers. K-feldspar, plagioclase and kaolinite are also present in amounts up to 10–15 wt.%. Samples with the smallest quartz contents also contain minimal amounts of feldspars, confirming that the latter occurs in the very fine sand and coarse silt fractions as was observed previously under the microscope (Vandenberghe, 1978). Calcite is generally absent or present only in amounts of less than 4 wt.%. Dispersed fine carbonate in the Boom Clay is known to occur only in specific stratigraphic intervals or in particular layers that also contain septarian carbonate concretions (Vandenberghe, 1978). That traces (<0.5 wt.%) of siderite and dolomite were detected is in line with the occurrence of very early diagenetic siderite in a few of the septaria levels in the Boom Clay (Laenen & De Craen, 2004) and with geochemical studies that suggested that Boom Clay pore fluids are in near-chemical-equilibrium with siderite, dolomite, calcite and rhodochrosite (Beaucaire et al., 2000). Pyrite is a well-known constituent of the Boom Clay and was detected in all samples; up to 3 wt.%. However, it is difficult to estimate the abundance of pyrite in the entire formation as it occurs heterogeneously from μm-scale framboids, to mm–cm-scale bioturbation traces and up to cm–dm-scale nodules. Traces of gypsum (<0.7 wt.%) are probably alteration products of pyrite. Secondary gypsum occurs in outcrops of the Boom Clay. Moreover, in many cores tiny gypsum crystals develop readily after sampling due to pyrite oxidation. Remarkably, anatase was detected in all samples, even up to 1 wt.% and in the <2 μm and <0.2 μm fractions of several samples. Authigenic anatase is known to participate in the heavy minerals of the Boom Clay (Vandenberghe, 1978). Traces of apatite (<0.5 wt.%) occur in all samples. Apatite in the Boom Clay is known from the transgressive horizon at the base (Vandenberghe, 1978; Vandenberghe et al., 2002); its occurrence throughout the clay deposit is attributed to fish teeth and other fossil carapaces that are common in the clay (e.g. Laenen, 1998).
The shape of the 060 reflection of the ‘2:1 clays’ indicates dioctahedral minerals with variable Fe content. During the Quanta analysis the 060 reflections were fitted by a combination of five clay mineral standards, ranging from Al-rich/Fe-poor to Al-poor/Fe-rich end members and intermediates, indicating a mixture of Al-rich and Fe-rich(er) clay phases. Such an assemblage is also consistent with the glauconites and muscovite reported in the coarser fractions of the Boom Clay by earlier studies (Vandenberghe, 1978; Laenen, 1997). The same holds for small amounts of chlorite minerals identified in the bulk rock (see discussion below). Note that the only information retained from the fit is the quantitative sum of the different 2:1 clay mineral species used for the fit, not the clay mineral species themselves. A test with one of the samples showed that using three rather than five clay mineral standards led to almost identical results for the total amounts of ‘2:1 clays’. However, the 060 reflection range of the full set of samples analysed was modelled consistently using five standards. As an example, in Fig. 2a the 060 reflection zone of sample EZE61 is shown together with the individual model XRD traces of the five clay mineral standards used to fit the recorded sample XRD trace. The intensities of the individual traces are proportional to their contributions in the composite fit. This composite model trace almost overlies the recorded sample trace. The sum of the quantities of the individual standard minerals used is expressed in the bulk composition as ‘sum 2:1 clays’ minerals (Table 2). In Fig. 2b a comparison is shown over the full 2θ range between a bulk sample diffraction pattern (EZE43) and the pattern summation fit. The similarity between the pattern summation fit and the sample diffraction pattern is exemplary for all samples.
Validation of the bulk quantitative results using major-element chemistry
The Boom Clay sample set is suitable for Bestrock optimization as the mineralogy is similar for all samples and only the mineral abundances vary. The chemical analyses for the samples studied are shown in Supplement 2.
Based on analyses in the outcrop area in Vandenberghe (1978), Mg and Ca were assumed to be the only exchangeable cations present in the clay minerals; although more recent work on subsurface samples does show Na, Mg, Ca and K in decreasing order of abundance at the exchangeable sites (Griffault et al., 1996; Honty, 2010). The variability of exchangeable cations in the Boom Clay is probably a function of carbonate content and different compositions of leaching water in the outcrop zone and in the subsurface. Therefore, in reality there will not be a single exchangeable cation composition for the Boom Clay. The impact of not using Na as an exchangeable cation in the overall calculations proved to be negligible as test runs showed that the optimization process would assign only ~0.03 Na atoms per half unit cell (phuc) to the average structural formula of the ‘2:1 clays’. The impact of excess Mg assigned to exchangeable cations is equally negligible, as most Mg is assigned to the octahedral sheet of the ‘2:1 clays’ and to chlorite and hence has no impact on the estimation of the (very low) amounts of dolomite. The octahedral composition of the ‘2:1 clays’ was constrained partially by an estimate of the Fe content, obtained from the chemical analysis of the extracted clay fraction (see below).
In the Bestrock optimization routine, trace phases such as siderite, dolomite and gypsum may appear as valid results in the computed composition even though these phases were not identified during regular bulk-rock XRD analysis as they could occur in quantities below the detection limits of the XRD/Quanta method. Thus, the optimization results for these phases should be considered as less reliable than those from regular bulk-rock analysis; final results are shown in Table 3.
Calculation of the petrophysical properties of the Boom Clay
The Bestrock analysis tool was designed to calculate the petrophysical properties of the formation studied, by combining the calculation of the structural formulae of variable composition minerals with a statistical distribution of trace-elements over all the minerals present, for the entire sample set (see Środoń & Kawiak, 2012). The trace element distribution over the minerals in the Boom Clay and their calculated petrophysical properties are shown in Supplement 3. Note that the results do not imply that these minerals actually contain these trace elements in the reported proportions, they only imply a statistical association, due to the possibility of trace element-rich heavy minerals being deposited together with certain phases. The high probability of trace-element association with phases such as apatite or anatase does not imply that some of the Boom Clay petrophysical parameters are controlled by these minerals, as they are only present in very small amounts (<1%). The availability of these data can improve the interpretations of well logging in the Boom Clay (see e.g. Wouters et al., 1999).
As expected from the similar qualitative bulk mineralogy of the samples, the clay mineralogy of the Boom Clay is qualitatively homogeneous (Fig. 3).
Illite is identified clearly by its 10 Å and higher-order reflections being almost unaffected by glycolation (Fig. 4). Also, kaolinite is recognized easily from its 7.16 Å and higher-order reflections and their disappearance after heating at 550°C for 1 h. After glycolation the low-angle tail of the 7.16 Å peak is affected slightly, suggesting an admixture of interstratified kaolinite-smectite. The presence of other expandable minerals is also clearly evident after ethylene glycol solvation. The 17 Å peak suggests smectite and/or randomly interstratified illite-smectite. Diffraction effects upon glycolation are manifested by smeared slopes and plateaus rather than by peaks. Their position between illite and smectite end-member reflections indicates interstratified illite-smectite.
To specify the type of smectite involved, its expandability after Li-saturation and heat treatment of the samples was tested (Greene-Kelly test). Although the Greene-Kelly test is not always very reliable (Moore & Reynolds, 1997), Li-saturation followed by brush-glycerol was applied to the main smectite fractions (<0.2 μm) as this method always confirms the presence of montmorillonite (Vandenberghe & Thorez, 1985).
Chlorite was confirmed by the 13.8 Å reflection after heating at 550°C for 1 h (Figs 4 and 5). This chlorite 001 diffraction peak is known to be enhanced by the heat treatment while the intensities of higher-order reflections are reduced (e.g. Moore & Reynolds, 1997).
Previous work has shown disagreement over the presence of vermiculite in the Boom Clay, with some authors reporting very low levels and others abundances up to 10–25% (Thorez, 1976a; Heremans et al., 1980). The relatively high CEC of vermiculites could have a significant impact on the sorption properties of the Boom Clay, thus the presence or absence of vermiculite is of great importance when considering the Boom Clay as a potential host formation for the disposal of nuclear waste. For this reason the minor 14.25 Å peak, visible in the XRD pattern of the Ca-saturated glycolated phase (Fig. 5a), was investigated in detail. A layer charge of >0.6 phuc can be used to distinguish vermiculite from smectite. Operationally, however, the distinction is made by saturation with Mg and expanding with glycerol to distinguish vermiculite and chlorite from smectite, followed by saturation with K and heating to 300°C for 1 h, to distinguish vermiculite from chlorite (Moore & Reynolds, 1997). The diffraction patterns of these tests are shown in Fig. 5. The glycerol was applied directly onto the slide by brush, as higher-charge smectites may not swell fully when exposed to glycerol vapours (B. Sakharov, pers. comm.). Afterwards, a small 14.25 Å shoulder remained and a third order peak of this mineral was visible at 4.73 Å. The non-swelling 14.25 Å peak remaining after Mg-glycerol treatment is not smectite, but it cannot be vermiculite either as the 001/003 intensity ratio for vermiculite should be much greater than was observed for these samples. This would suggest that this phase should be chlorite. However, upon K-saturation and heating at 300°C for 1 h, this phase collapses to 10 Å similar to vermiculite, with total disappearance of the 14.25 Å and 4.73 Å peaks (Fig. 5). This behaviour is typical for a defective chlorite in which the interlayer sheet is incomplete, resulting in poor resistance to heat treatment. Such “defective chlorite” has been described as an Fe-rich tri-octahedral mineral (Renngarten et al., 1978; Varentsov et al., 1983; Sakharov et al., 2004).
In addition, the diffraction patterns of the heat-treated (550°C for 1 h) slides of the Ca-form demonstrate the presence of a randomly interstratified chlorite-smectite phase in some of the samples (e.g. EZE60 in Fig. 5b), based on a diffraction shoulder between the chlorite 13.8 Å reflection and the 10 Å reflection. It is this chlorite-smectite that was described as “degraded” chlorite by Vandenberghe (1978). Figure 5b also shows an example of a sample (EZE45) in which this interstratified phase is absent. The chlorite-smectite phase was identified in several of the outcrop and the core samples, indicating that it is not a recent weathering product, but was already present at the time of deposition (EZE60 shown in Fig. 5b is a core sample). The clay-extraction treatments cannot be responsible for the degradation of the chlorite as the identification of chlorite-smectite (“degraded chlorite”) by Vandenberghe (1978) occurred in samples prepared using de-ionized water only.
This detailed investigation of the 14 Å phases in the Boom Clay demonstrates that vermiculite is absent from the Boom Clay and only minor amounts of chlorite, defective chlorite and interstratified chlorite-smectite are present.
The oriented slides of the <2 μm fraction of the Boom Clay samples were modelled using the same qualitative approach incorporating six different clay-mineral phases. Four of these were discrete clay minerals: illite, smectite, kaolinite and chlorite; the remainder being randomly interstratified illite- and kaolinite-smectite, both of which were modelled using 3 layer types: illite or kaolinite plus “low charge” or “high charge” smectite layers, in the randomly interstratified structure. In the context of the modelling results described in this section the terms, “low charge” and “high charge”, respectively, refer to the presence of two vs. one water/glycol layers in the interlayer space. Sometimes the “high charge” component is referred to as “vermiculite layers” (Lindgreen et al., 2002; Claret et al., 2004). The high vs. low charge smectite layers were quantified from the glycolated state model as in the water-saturated state their apparent ratio can vary with relative humidity. For simplicity these interstratified phases are denoted as “illite-smectite” and “kaolinite-smectite”.
The results of the clay fraction modelling are gathered in Table 4. Examples of the glycolated and air-dry fits for the same samples are shown in Fig. 6. The discrete illite phase, with a broad 001 reflection and a relatively intense 002 reflection, was best modelled assuming a small amount of smectite layers (1–4%) and no octahedral Fe. Such composition, typical of deep diagenetic and anchizonal illite (Lindgreen et al., 2000), is very common in sedimentary rocks sourced, at least partially, from sedimentary terrains. This mineral is further referred to as illite and its abundance ranges from 14 to 25% with an average of 21%. The averages reported in this study refer to the 20 samples examined and do not represent a weighted average for the total Boom Clay volume.
The discrete smectite phase has a modelled Fe content of ~0.5 atoms phuc, which is consistent with earlier studies that used Mössbauer spectroscopy to show that the smectite in the Boom Clay contained octahedral Fe (Decleer, 1983). The ratio of low/high-charge smectite layers in glycolated traces was 89–96/4–11 (Table 4, as percentage layer charge (%LC)). For air-dried smectite these proportions were modelled close to 60:40 (Fig. 6b), indicating charge heterogeneity. The discrete smectite content varies between 19 and 39% with an average of 27%.
The interstratified illite-smectite phase was modelled with a large illite content (64–69%), a small amount of high-charge expandable layers (4–8%) and Fe content of 0.2 atoms phuc. Following the multi-specimen method, the approximate amount of illite in this phase should be preserved when adjusting the model to air-dry conditions and this is indeed observed (68% vs. 66% illite layers, Fig. 6a,b). Due to charge heterogeneity, under air-dry conditions the ratio of low/high-charge smectite layers is shifted towards the latter (from 27/5 to 17/17, Fig. 6a,b). The illite-smectite content varies between 19 and 42% with an average of 28%. Figure 6 displays the modelled individual phases illite, smectite and illite-smectite vs. the experimental and modelled patterns of a representative Boom Clay sample.
In the kaolinite model only its mean crystal size was adjusted. However, as discussed above, the low-angle tail of its 001 reflection and the high-angle side of its 002 reflection (features G and H on Fig. 6c) indicate the additional presence of a kaolinite-smectite phase. This phase was also modelled using low- and high-charge smectite layer types. However, due to the very small overall contribution of these expandable layers to the samples it is not possible to determine with certainty their exact swelling character. The total expandable content in the kaolinite-smectite was 17–19%. The kaolinite content varies between 5 and 12% with an average of 8% and the kaolinite-smectite content varies between 4 and 17% with an average of 11%.
The distinction between the very small quantities of chlorite, defective chlorite and randomly interstratified chlorite-smectite is beyond the resolution of the modelling method. Therefore only one combined ‘chloritic’ model was used to model the diffraction effects (Table 4). This average model corresponds to a defective chlorite, as a reasonable fit could only be achieved with a partially complete (70%) interlayer octahedral sheet. This structure was modelled with 1 Fe atom phuc in the octahedral sheet and 1.5 Fe atoms phuc in the interlayer octahedral sheet and its abundance in the sample set varies between 2 and 7% with an average of 4%.
Using the same model in similar proportions, the fit of the glycolated and the air-dry models to the experimental patterns is highly satisfactory (Fig. 6a,b). Some areas of mis-fit still exist: the ultra-low angle mis-fit (3–3.5°2θ; Feature A in Fig. 6a) is background only and at the time of the development of the software no mathematical description of the diffraction effects in this region was available (B. Sakharov, pers. comm.). The mis-fit features B and D in Fig. 6a are attributed to natural heterogeneity in the minerals that could not be taken into account sufficiently in the calculated phases. The mis-fit located on the 002 reflection of kaolinite indicated by feature C in Fig. 6c is systematic, but at present unexplained. The mis-fit for the air-dried samples is somewhat larger than for the glycolated samples, as the expandable layer thickness after glycolation is more uniform than after air drying.
The modelling results are supported by the fact that, essentially, a single clay-mineral assemblage model could be used to model the entire sample set satisfactorily. The same set of minerals gave a satisfactory fit to the glycolated diffraction patterns of the <0.2 μm fraction (Fig. 7) after making suitable slight adjustments to certain values, such as: decreasing the mean crystal thicknesses for all phases in this fraction and increasing the expandability of the illite-smectite phase (34% vs. 32%) and its high-charge smectite content (14% vs. 7%).
Determining the ‘2:1 clay’ mineral formula from elemental analysis of the <0.2 μm fraction
The <0.2 μm fractions of the Boom Clay are not mono-mineralic as demonstrated above, but dominated by the ‘2:1 clays’. Therefore, it is possible to propose an approximate average ‘2:1 clay’ mineral structural formula.
The chemical analyses of the <0.2 μm are shown in Supplement 4. The P and Mn oxides are not commonly associated with clay mineral lattices and Ti oxide is usually associated with anatase rather than with clay minerals. Chlorite-type minerals are absent from the <0.2 μm fraction. The kaolinite and kaolinite-smectite content obtained from the modelling of the oriented diffraction patterns was used to calculate the amount of SiO2 and Al2O3 to be subtracted from the chemical analysis results to correct for their presence in the samples. The remaining major-element oxides were used to calculate the structural formulae of the combined ‘2:1 clays’ (illite + smectite + illite-smectite) (Table 5), following the structural formula calculation outlined by Moore & Reynolds (1997). Varying the kaolinite content by a few % did not change the calculated 2:1 clay mineral structural formula substantially. The average structural formula calculated is: K0.33Na0.02Ca0.15(Si3.69Al0.31)(Al1.14Mg0.38Fe0.49)O10(OH)2.
Independent validation of the clay mineral quantitative results for the <2 μm fraction
The reliability of the quantity of illitic components modelled in the clay fraction, i.e. the discrete illite phase and illite layers from the interstratified illite-smectite, can be evaluated from the K2O content of this fraction. This is indeed possible because these illitic layers are the only potassium-bearing mineral phases present, as was verified by the random powder recordings of this size fraction. The independent validation was done by comparing the measured K2O content of the <2 μm fraction to the K2O content calculated using the Sybilla software for the corresponding model. The software makes this calculation by taking into account the modelled illite content in discrete and interstratified minerals, the modelled number of K atoms in each illitic layer type and the modelled crystal size distribution for each K-bearing mineral. The measured K2O content of the <2 μm fraction (Supplement 5) needed correction as chemical analyses were reported on an air-dry basis, which varies due to the room temperature and relative humidity during the analysis, as opposed to the K2O content calculated for models in Sybilla, which refer to an absolutely dry state. The correction was based on the mass loss of the sample equilibrated at 47% relative humidity under ambient conditions after heating in a thermobalance to 200°C, which is considered a close approximation of the absolutely dry state (Środoń & McCarty, 2008). The calculated and the analytically measured K2O contents (after correction) for the <2 μm of all samples are compared Supplement 6. The relative differences are generally <10%. The modelled parameters for the illitic layers and the illite content estimation in the illite-smectite can all be varied somewhat while still producing good diffraction-pattern fits; tests showed that this range of variability leads to relative variations in calculated K2O content of ~10%. Therefore differences of 10% or less between calculated and measured values are considered a close agreement and an independent validation of the modelling results. This is further supported when considering the very narrow range of the actual K2O content in the sample set (3.2 to 4.0 wt.%) compared to the absolute possible range of 0 wt.% (a sample containing no illite or illite-smectite) to 11 wt.% (a sample containing only illite).
The total smectitic content was verified independently from CEC measurements. Analogous to the validation of the illitic content a validation was sought for the total smectitic content, being the sum of the discrete smectite phase and the smectite layers from the interstratified illite-smectite. This is based on the assumption that the CEC will be controlled effectively by the smectitic surfaces of these phases and to a much lesser extent by other clay mineral surfaces. The contribution of the expandable layers in kaolinite-smectite was ignored as it only represents a very small fraction of the total clay content.
Due to the relatively large amounts of material required (0.5 to 1 g), CEC measurements could only be performed for the bulk-rock samples. Therefore, the smectitic contents obtained from modelling of the <2 μm fraction were recalculated to give the smectite wt.% contribution to the bulk rock sample. This was done by first splitting the interstratified illite-smectite content into illite and smectite equivalent wt.% based on their modelled proportions. These proportions were added to the wt.% of discrete illite and smectite, respectively, allowing establishment of a smectitic vs. illitic content ratio. Finally, this ratio was used to split up the ‘2:1 clays’ content from the bulk analysis into smectitic and illitic content in the bulk rock. This procedure assumes that the ratios between ‘2:1 clays’ in the bulk rock and those in the <2 μm fraction are comparable, which is considered reasonable in the absence of large amounts of coarse-grained micas.
The CEC for an average charged smectite is close to 100 meq/100 g. For example the experimental results in Środoń & McCarty (2008) show an average smectite CEC of 103–110 meq/100 g. Therefore, the CEC value can be considered as an approximation of the smectitic content of the sample and be compared to the smectitic content calculated from the clay-fraction modelling results, as explained above (Table 6, Fig. 8). Table 6 also shows an alternative method of calculating the smectitic content in the bulk rock directly from the CEC using the relationship: smectitic content = 0.91×CEC (Środoń, 2009), which is based on the assumption of a constant charge of 0.41 phuc for the smectite.
The agreement is close for all three methods, except for two samples (EZE 59 & 60) in which the CEC appears too high compared to the smectitic content. No evident explanation was found for these two exceptions. The linear fit equation shows that between 3 and 4 meq/100 g of the CEC are unexplained by smectitic minerals. This ‘supplementary’ CEC is assumed to come from illitic surfaces/edges, the smectite layers in the kaolinite-smectite and perhaps the small amounts of organic matter present in the samples.
The kaolinite and chlorite contents were also verified independently by applying the bulk-analysis quantification method described above on the <2 μm fraction. It has been well established that the bulk analysis method used is highly accurate (Kleeberg, 2005; Omotoso et al., 2006); however, 2–3 g of the <2 μm sample would be necessary for this analysis and this was unavailable for most of the samples in this set. Therefore, the validation test was performed on five additional samples of Boom Clay taken over a short interval in the exploitation pit at Rumst-Terhagen in the type area along the Rupel River. To preserve the concept of an independent validation, the exact same detailed model as discussed above was used to model the <2 μm fraction of these samples. The kaolinite and chlorite content obtained was then compared to the values determined by applying the bulk-rock method on the same fraction. Supplements 7 and 8 show that the values for both methods are in good agreement. This agreement is increased even further when the kaolinite-smectite phase is corrected for its smectite content using the expandability determined by modelling (last column, Supplement 7). This correction is necessary as the bulk-analysis method is based on the 060 reflection area where kaolinite and smectite are quantified separately. For both kaolinite and chlorite the independent validations show absolute differences <3 wt.%.
Integration of the bulk-rock and clay-fraction analyses results
As a final step, results from both bulk-rock analysis and the modelling of the extracted clay fraction were integrated by splitting the bulk-rock number for the ‘2:1 clays’ into proportions of illite, smectite and interstratified illite-smectite according to their relative proportions in the <2 μm fraction extracted (Table 7). Further mineralogical specification can be added by detailing the kaolinitic content as consisting of discrete kaolinite and interstratified kaolinite-smectite with ~17–19% expandable layers and detailing the chloritic content as a mixture of chlorite, defective chlorite and a randomly interstratified chlorite-smectite.
This integration of the data again assumes that the quantitative proportions in the clay minerals in the >2 μm fraction are the same as or very close to those in the <2 μm fraction. This assumption is expected to overestimate the smectitic components and underestimate the illitic components slightly due to the smaller grain size of smectitic minerals compared to illitic. In these coarser size fractions, minerals like glauconite and muscovite, identified in microscopy work (Vandenberghe, 1978), will add to the underestimation of the illitic components in our analysis, thus in the current study, muscovite and glauconite were included in the ‘illite’ fraction. In the same microscopy study some chlorite flakes were also identified in the coarser fraction. The observed quantities of each of these coarser minerals were small in comparison to the overall clay content of the rocks and therefore their impact on the proportions determined will be small.
It is clear from the results presented that, qualitatively, the different Boom Clay samples contain the same minerals in variable proportions. These variations can be understood easily in terms of the lithology types observed in the formation, which are the result of its sedimentological history; while at the same time the provenance of the minerals remains approximately constant during the entire depositional history. The main variations controlled sedimentologically are the proportions of quartz and clay minerals and the contents of carbonates and organic matter.
The applied methodology, both for the bulk-rock and the extracted clay-fraction analysis, relies on consistent sample pre-treatments and proven quantitative analysis methods. It has permitted a reliable and robust total qualitative and quantitative mineral composition determination for the Boom Clay. All quantitative results were validated independently by chemical and CEC analyses. In addition, cation saturations followed by swelling and heating treatments were used to solve a long-standing discussion regarding the exact nature of the 14 Å components present in the Boom Clay.
The results presented contrast with several previous clay-mineral studies of the Boom Clay which all reported different clay minerals, in widely varying quantities, using non-standardized methods. Such non-standardized approaches led to poor quality diffraction patterns and uncontrolled cation saturation states, thereby impacting severely on reproducibility. Accuracy was also poor because clay-mineral quantification was based solely on a comparison of 001 intensities with standard minerals and the use of correction parameters that were too general. The use in this study of a higher σ* for smectite than for the other phases in these kinds of poorly consolidated samples is considered a critical factor in being able to determine accurate quantitative proportions successfully from modelling.
Considering the marked improvement in applied methodology, the results of the study presented can become the reference for the mineralogical composition of the Boom Clay in different application fields.
This study is, in part, the result of a doctoral thesis by EZ at the University of Leuven, Belgium, financed by IWT Flanders and supervised by NV and JŚ. Chevron ETC, Houston is thanked for permission to use the proprietary software Quanta, Sybilla and Bestrock. Victor Drits and Boris Sakharov are thanked for help with the analysis of the exact nature of the 14 Å components. Dorota Bakowska is acknowledged for performing the K2O analyses. This work was performed in close cooperation with and with the financial support of NIRAS-ONDRAF, the Belgian Agency for Radioactive Waste and Fissile Materials, as part of their program on geological disposal of high-activity and long-lived radioactive waste. The reviewers and editor are thanked sincerely for their constructive comments, which led to considerable improvements of the manuscript.
- Received September 17, 2014.
- Revision received June 14, 2015.