- Email: [email protected]
Na-montmorillonite
Physics and Chemistry of the Earth 36 (2011) 1564–1571
Contents lists available at SciVerse ScienceDirect
Physics and Chemistry of the Earth journal homepage: www.elsevier.com/locate/pce
Role of cation mixing in the sol formation of Ca/Na-montmorillonite Magnus Hedström ⇑, Martin Birgersson, Ulf Nilsson, Ola Karnland Clay Technology AB, IDEON Science Park, SE-223 70 Lund, Sweden
a r t i c l e
i n f o
Article history: Available online 31 July 2011 Keywords: Montmorillonite Sedimentation Sol formation Swelling pressure Turbidity
a b s t r a c t Compound Ca/Na-montmorillonites were prepared by mixing given amounts of homoionic Ca-montmorillonite with given amounts of homoionic Na-montmorillonite. Colloidal sol formation in deionized water of these mixed Ca/Na-montmorillonites were investigated by means of turbidity measurements or through swelling pressure evolution in specially constructed swelling pressure test cells. It is concluded that the sol phase is absent in homoionic Ca-montmorillonite, which is interpreted as a manifestation of ion-ion correlations that are strong for divalent ions. Using the swelling pressure test cells it was found that Na-montmorillonite readily penetrates filters with pore size 2 lm, whereas in the case of Camontmorillonite no material was lost through the filters even at a pore size of 100 lm. However, mixed Ca/Na-montmorillonite with 20% or more sodium swells extensively and give rise to rapid increase in turbidity and behave qualitatively the same as homoionic Na-montmorillonite with respect to sol formation. This is also confirmed in erosion experiment using the swelling pressure test cells as well as in sedimentation tests monitored with a turbidimeter. Montmorillonite with 90% or more calcium in the interlayer behaves similar to homoionic Ca-montmorillonite. Thus it is concluded that 90% of the clay layer charge needs to be compensated by Ca2+ for correlation interactions to prevent sol formation. Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction In the Swedish KBS-3 concept for a geological deep storage of spent nuclear fuel, bentonite of high montmorillonite content is proposed to serve as a buffer surrounding copper canisters containing the spent fuel. Montmorillonite, which is the main constituent in bentonite, has an exceptional affinity for water which results in the build-up of a swelling pressure when bentonite (with access to water) is placed in a confined volume. There may be fractures intersecting the deposition hole and at those fractures the bentonite is not restricted but can continue to swell until equilibrium or steady state is reached. Under present day Swedish ground water conditions the swelling into fractures would be limited because the montmorillonite at the swelling front will eventually coagulate owing to the composition of dissolved salts. However, at the end of a glaciation it cannot be excluded that glacial melt water of low ionic strength will permeate the bedrock. This could cause significant erosion of the bentonite due to colloidal sol formation at the swelling front especially if the bentonite has been depleted of soluble accessory minerals. It is important to note that the gel-sol transition is not only a function of the ionic strength but also depends on the composition of counterions in the montmorillonite as well as the makeup of ions in the external solution (Bir-
⇑ Corresponding author. E-mail address: [email protected] (M. Hedström). 1474-7065/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.pce.2011.07.019
gersson et al., 2009, 2011). A homoionic Ca-montmorillonite would not pose any problem with respect to colloidal sol formation because it has limited swelling most likely due to attraction forces caused by ion correlations (Kjellander et al., 1988). In homoionic Na-montmorillonite, on the other hand, the correlation interactions are weak and cannot prevent the sol formation in case the montmorillonite is contacted with water of low ionic strength. Colloid sol formation may be stopped or reduced by the presence of excess salt. High concentrations, in the range 1–2 M, of a 1:1 salt, e.g. NaCl would screen the double-layer repulsion effectively and the van der Waals attraction between the clay layers would dominate the system and limit the swelling of the interlayers to about 1 nm (Norrish, 1954). Coagulation in Na-montmorillonite systems is also observed at lower NaCl concentrations, in the range 5–25 mM, which is attributed to edge-face interactions (Lagaly and Ziesmer, 2003; Birgersson et al., 2011). However, the topic of this paper is the investigation of sol formation in the absence of excess salt in the system, i.e. the only cations present are those needed to compensate the clay layer charge. Numerous experiments have shown that montmorillonite with calcium or magnesium as counterion does not form colloidal sols, even if placed in deionized water. The problem with extensive swelling and colloidal particle release only occurs when monovalent counterions are present. What fraction of divalent counterions in the clay is needed to prevent colloidal sol formation? To answer this question compound Ca/Na-montmorillonites were prepared by mixing given amounts of pure Ca-montmorillonite with given
1565
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
amounts of pure Na-montmorillonite. It is noted in passing that the absence of excess salt is too conservative for the worst-case scenario of glacial meltwater, but is a highly relevant test system as it allows for isolating the effect of correlation mediated by divalent ions in the interlayer. 2. Experimental 2.1. Materials Two types of montmorillonite have been used in the investigations, Wyoming-type (Wy) obtained from MX-80 bentonite (American Colloid Co.) and Kutch-type (Ku) from the Indian bentonite Asha 505 (Ashapura Minechem Co.). The montmorillonites are extracted and ion-exchanged to either sodium or calcium form according to the procedure described in Karnland et al. (2006). The chemical formula for an ideal (sodium) montmorillonite, interlayer water excluded, can be written as
ðSi8a Ala ÞðAl4b Mgb ÞO20 ðOHÞ4
ðaþbÞ
Naþaþb ;
ð1Þ
where by definition the tetrahedral charge is lower than the octahedral charge (a < b) and the sum of tetrahedral and octahedral charges fulfil 0.4 < a + b < 1.2. It has previously been established that the Wy-montmorillonite is dominated by octahedral charges b = 0.54–0.6 and the total charge is 0.65e, whereas Ku-montmorillonite has a much higher portion of the charge in the tetrahedral layer, a = 0.38, as well as a higher total charge of 0.79e (Karnland et al., 2006). Both for Wy and Ku, mixed Ca/Na clays were made at mass fractions 100/0, 80/20, 60/40, 40/60, 20/80 and 0/100, where 100/0 and 0/100 means pure Ca-montmorillonite and pure Na-montmorillonite respectively. In addition to the above mass fractions a batch of Wy-montmorillonite at mass fractions 98/2, 95/5 and 90/10 was made. The mass fractions also reflect the fraction of the cation exchange capacity (CEC) compensated by Ca2+ and Na+, e.g. in the 80/ 20 montmorillonite 80% of the surface charge is compensated by Ca2+ and 20% by Na+. The experiments to be monitored with a turbidimeter need only very small amounts of montmorillonite. In total 2 g of clay was dispersed in 250 ml deionized water for each Ca/ Na combination. To facilitate the mixing, the suspension was agitated with ultrasound for 15 min and then mixed with a magnetic stirrer for at least 12 h, which is a long equilibration time considering that Shainberg and Kaiserman (1969) showed that the properties of mixed Ca/Na-montmorillonite suspensions were stabilized within 10 min of mixing. After the mixing the water was removed from the suspension in an oven at 60 °C. 2.2. Turbidity measurements The swelling and sedimentation behaviour of Wy and Ku montmorillonite are investigated with respect to counterion composition in the interlayer; either Na+ or Ca2+ or mixtures of the two ions. The experiments are performed using cylindrical glass vials, with outer dimensions: height 35 mm and diameter 23 mm. In swelling experiments a given amount of montmorillonite is dried-in at the bottom of the vials and 10 ml of deionized water is added. The swelling and colloidal particle release vertically is monitored by measuring the change in turbidity with time at approximately 12–13 mm above the bottom. At low particle concentrations the turbidity is directly proportional to the concentration. Similarly for sedimentation the same equipment is used. The sedimentation tests are prepared by dispersing up to 50 mg clay in 10 ml water overnight. Prior to the first turbidity measurement the vials are shaken to ensure a homogeneous distribution of clay.
Turbidity is measured with a portable turbidimeter TN-100 from Eutech Instruments. The light source in TN-100 is an infrared-emitting (850 nm) diode. Light scattered from the clay particles in the suspension is detected at an angle of 90 ° with respect to the light source, thus formally the TN-100 is a Nephelometer. The intensity of the scattered light is directly converted to nephelometric turbidity units (NTU) and the TN-100 operates in the turbidity range 0–1000 NTU. As mentioned above the standard turbidimeter measures the turbidity at 12–13 mm height in the vials. A second TN-100 instrument was rebuilt to allow the registration of turbidity at different heights in the vials, thereby offering the possibility of obtaining time-resolved swelling or sedimentation profiles. Turbidity at low colloidal particle concentrations is proportional to the concentration or equivalently the particle number density. It is found that different clays give different turbidity responses. Up to a certain concentration the turbidity response is linear and beyond the linear regime the turbidity increases less than the increase in clay concentration. For Wy-Na the turbidity response is linear up to 5 g clay/litre deionized water (114 NTU per g/l). KuNa shows linear response to approximately 2.5 g/l (315 NTU per g/l). 2.3. Erosion tests Erosion of montmorillonite was monitored by measuring the swelling pressure response using a swelling pressure test cell (e.g. Karnland et al., 2007) as shown in Fig. 1. Highly compacted discs, radius 35 mm and height 5 mm, of montmorillonite were placed in pressure cells with permeable 2 mm thick filters of different pore sizes ranging from 0.2 to 100 lm. After water saturation, 100 ml deionized water was continuously circulated behind the filters and the swelling pressure was monitored during the test period that lasted about a month. Since swelling pressure depends strongly on the montmorillonite density any loss of clay through the filters due to sol formation, would be detected as a drop in swelling pressure. 3. Results and discussion 3.1. Turbidity evolution Before discussing the results from the experiments it is instructive to reflect upon the quantities of montmorillonite used and the dimensions of the vials. To cover the bottom area of the vials by one clay layer of thickness d = 1 nm, requires the mass, m = qcpr2d of montmorillonite. With a grain density of montmorillonite of qc = 2750 kg/m3 and the vial inner radius r 11 mm, m becomes approximately 1 lg. Typically in the experiments 20–100 mg of clay is used. If these amounts of clay were evenly dried-in at the bottom the height would be 0.02–0.1 mm or equivalently 20,000 to 100,000 layers of clay. After water uptake to a basal distance of 200 nm, roughly the distance where free rotation of a typical
100 mL circulating solution Force transducer Clay sample
Filters (titanium or stainless steel) Fig. 1. Schematic drawing of the swelling pressure test cell.
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
Wy-Na particle could be expected (Michot et al., 2004), the height of the clay phase in the vials would be 4 mm in the case of 20 mg and 20 mm for 100 mg montmorillonite, respectively. The height of 10 ml liquid in the vials is 25 mm so swelling to the point of free rotation means that 100 mg of clay essentially fills the whole sample, 50 mg barely reaches the turbidity detection height at 12– 13 mm in the vial and 20 mg reaches one third of the detection height.
500 450 400 350 300 250 200 150 100 50 0
25
D1 100/0 D3 80/20
20
D5 60/40
15
D7 40/60
10
D9 20/80
5
D11 0/100
0
0
100
200
300
400
Time [Days] Fig. 3. Swelling series D: Turbidity at vial mid-height vs. swelling time for Wy-Ca/ Na montmorillonite. Clay content is 20 mg per 10 ml.
after 57 days. If the de-mixing hypothesis were the dominating mechanism here, a much slower increase in turbidity would be expected for sample B4 because clay particles would not fully exfoliate, which would have two effects that both reduce the transport of clay against gravity. First, clay particles would be heavier (‘‘stacks’’ of Ca-montmorillonite with Na-ions on the outer surfaces) and therefore more affected by gravity. Second, osmotic pressure would be lower. This follows from the fact that the osmotic pressure (P) at large particle separations is inversely proportional to the square of the separation (h) (Evans and Wennerström, 1999, Eq. 5.1.18)
P¼
p2 o r ðkB TÞ2 1 2ðzeÞ2
h
2
ð2Þ
;
and distances between ‘‘stacks’’ are necessarily larger than they would be between fully exfoliated clay layers at a given water to solid mass ratio. Undoubtedly Wy-80/20 clay enters the osmotic swelling regime where the basal distance surpasses by far 20 Å and water is adsorbed in a continuous manner (Brindley and Brown, 1980). With that in mind it is also clear that an increase of the percentage of sodium should give higher swelling pressures at long distance due to a larger number of ions in the interlayer. This is reflected at the initial stage of the turbidity response, where higher maximum turbidities are noticed the higher the sodium fraction. Below the correlation between swelling pressure and maximum turbidities is discussed in more detail. With higher total clay content the initial swelling-dominated increase in turbidity is more rapid and with 100 mg clay/10 ml the turbidity for Wy-0/100 reaches a maximum after 8 days (Fig. 2) whereas with 20 mg/10 ml the maximum occurs after 57 days (Fig. 3). It comes as no surprise that an intermediate clay content of 50 mg/10 ml, presented in Fig. 4, gives maximum turbidities and
120
B2 100/0 B4 80/20 B6 60/40 B8 40/60 B10 20/80 B12 0/100
0
100
200
300
400
Time [Days] Fig. 2. Swelling series B: Turbidity at vial mid-height vs. swelling time of Wy-Ca/Na montmorillonite for various Ca/Na combinations. Clay content is 100 mg per 10 ml.
Turbidity [NTU]
Turbidity [NTU]
3.1.1. Swelling The swelling of mixed Ca/Na montmorillonite is monitored by measuring the turbidity at vial mid-height as the clay expands and disperses against gravity. Initially the clay is dried-in at the bottom and 10 ml deionized water is added. The results for clay contents of 100 mg and 20 mg are displayed in Figs. 2 and 3, respectively. Pure Wy-Ca (100/0) behaves totally different from the other clays with 20% Na+ or more in exchange position. In practice there is no release of colloidal particles from Wy-Ca and the measured low turbidity reflects the fact that Wy-Ca swells to a maximum distance. Furthermore, the original grains of Ca-montmorillonite on the bottom of the vials are intact, albeit considerably swollen compared to the dry state. Inclusion of as little as 20% Na+ makes the Wy-Ca/Na clay display predominantly sodium montmorillonite behaviour and no sign of the original grains is visible in the vials. It is instructive to compare the curve for Wy-80/20 (B4) in Fig. 2 with the curve for Wy-0/100 (D11) in Fig. 3. Sample B4 consists of 80 mg Wy-Ca and 20 mg Wy-Na, whereas sample D11 contains 20 mg Wy-Na. Thus both samples have the same amount of Na+ but their swelling behaviours are entirely different. An assumption that mixed Ca/Na smectite systems segregate into Ca-layers and Na-layers, a process termed de-mixing, is sometimes encountered in the literature (Bar-On et al., 1970; Iwasaki and Watanabe, 1988; Verburg and Baveye, 1994; Laird 2006). For de-mixing to occur there must exist an interaction energy that dominates over the entropy of mixing as described in e.g. (Evans and Wennerström, 1999). It is very hard to envisage any such force that would make e.g. Ca2+ ions to prefer to be surrounded by other Ca2+ ions over Na+ ions, but this point will be revisited in the summary and conclusions in Section 4. However, the results presented here strongly suggest that the mixing of Ca-montmorillonite with Na-montmorillonite results in a clay where the ions are mixed in the interlayer and not to a situation where there are Ca-layers and Na-layers. The latter situation would correspond to 80% WyCa that do not release colloidal particles and 20% Wy-Na that would give a maximum turbidity of approximately 25 NTU in accordance with Fig. 3. Instead the Wy-80/20 clay gives a turbidity maximum above 160 NTU as seen in Fig. 2. Furthermore, the turbidity maximum of series B4 in Fig. 2 occurs after 27 days, whereas the turbidity maximum in swelling series D11 in Fig. 3 takes place
30
Turbidity [NTU]
1566
100
D2 100/0 D4 80/20
80
D6 60/40
60
D8 40/60
40
D10 20/80
20
D12 0/100
0
0
100
200
300
400
Time [Days] Fig. 4. Swelling series D: Turbidity at vial mid-height vs. swelling time of Wy-Ca/Na montmorillonite for various Ca/Na combinations. Clay content is 50 mg per 10 ml.
1567
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
turbidity-increase rates in between the values obtained at 20 and 100 mg/10 ml respectively. It is also clear that the maximum turbidities at vial mid-height do not vary linearly with clay content. For a diffusion-driven process a linear variation of turbidity with the amount of dried-in clay is expected at all times. However, at the termination of the experiments (after 1 year) the mid-height turbidities appear proportional to the amount of clay in the vials. The initial non-linearity can be explained on the basis of the osmotic pressures in lamellar systems at low densities. According to the Poisson–Boltzmann equation the swelling pressure is inversely proportional to the square of the mean interlayer distance (see Eq. (2)) or, equivalently, proportional to the square of the solid to water mass fraction which follows from the relation
qclay d 1 mclay ¼ / ; mwater qwater h h where d = 1 nm, i.e. the thickness of a single clay layer and h is the interlayer separation. Therefore the swelling pressures in the systems discussed here with 100, 50 and 20 mg total amounts of montmorillonite obey the relation P(100 mg) = 4P(50 mg) = 25P(20 mg). Let NCa=Na denote the maximum turbidity for a given Ca/Na clay in m the swelling experiments. From Figs. 2–4 it is concluded that the turbidities for the 0/100 clays at maximum obey the relation N 0=100 ð100 mgÞ ¼ 4:4N 0=100 ð50 mgÞ ¼ 18:3N 0=100 ð20 mgÞ and for the m m m 80/20 clays the relation is similar; N 80=20 ð100 mgÞ ¼ 3:9N 80=20 ð50 mgÞ ¼ 25:1N 80=20 ð20 mgÞ: The fact m m m that the maximum turbidities behave as the swelling pressures can be rationalized as follows: Turbidity is proportional to clay concentration, therefore Nm is proportional to the mass of clay in a cross-section of thickness d at detection height, where d is the width of the scattering region. The weight of this mass is balanced by the swelling pressure exerted by the clay below, i.e. P qdg, where g is the acceleration of gravity. Thus the ratio of maximum mid-height turbidities or densities for tests with initially dried-in amounts m1 and m2 of clay is envisaged to approximately obey the relation
qðm1 Þ Pðm1 Þ m1 qðm2 Þ Pðm2 Þ m2
2 :
This relation is of course only approximate and one source of imprecision is the neglect of the weight of the clay above the detection height. On the other hand, the maximum turbidities show that the concentrations at mid-height are much lower than the average concentration, which is a sign that most of the clay remains below detection height. This is also confirmed by measuring the turbidity profiles along the height of the vials using the modified turbidimeter. The turbidity profiles also show that the decrease in mid-height turbidity after the maximum value is primarily due to sedimentation and not to further swelling or diffusion of clay upwards.
Ku-Ca/Na behaves similarly to Wy-Ca/Na as can be seen in swelling series E in Fig. 5. Kutch is more turbid than Wyoming montmorillonite, which is reflected in the high turbidities despite the fact the clay content in series E is 50 mg/10 ml, i.e. half of the clay content in Series B (Fig. 2). In addition a larger correlation between Na content and turbidity at maximum for Ku-Ca/Na than for Wy-Na/Ca is also noticed. As time progress the initial differentiation among the clays containing sodium is faded out both for Ku and Wy. Swelling experiments on Ku-Ca/Na were also performed for 20 mg/10 ml and the non-linear dependence on Nm with clay content seen for Wy-Na/Ca was confirmed and for the pure KuNa clay the relation was found to be N 0=100 ð50 mgÞ ¼ 9:1N 0=100 ð20 mgÞ and similar for the other Na+ m m containing clays. Increasing the calcium content beyond 90% of CEC makes the limited swelling behaviour typical for calcium montmorillonite to dominate as can be seen in Fig. 6. Note that when water is added some grains of clay are stirred up which gives relatively high initial turbidities. These grains are quite large and sediment rapidly leading to a minimum in the turbidity after a few days. As can be seen in Fig. 6, Wy-90/10 still shows some active colloidal particle release but the turbidity increase at vial mid position is very minor and reaches a maximum of 5.6 NTU after 130 days. In comparison Wy-80/20 at 5 g/l gives a turbidity maximum above 40 NTU in 40 days (Fig. 4). Furthermore Wy-80/20 appears similar to pure Wy-Na with no sign of limited swelling, whereas in free swelling of Wy-90/10 original grains of clay are present at the bottom of the vial. 3.1.2. Sedimentation Sedimentation of Wy-Ca/Na confirms the above picture that 90% Ca or more is needed for behaviour similar to pure Wy-Ca to occur, as shown in Fig. 7. With 60% Ca or less the sedimentation follows the same pattern as for pure Wy-Na. Wy-80/20 sediments slightly faster and Wy-90/10 falls in between the fast sedimentation of Wy-Ca and the slower sedimentation of Wy-80/20. The gradual increase in turbidity for Wy-100/0, 98/2 and 95/5 after 90 days has only been observed in Wyoming montmorillonite and not in Kutch. We have at present no explanation for this pattern but the turbidity at maximum after 250 days (Wy-100/0) is only 10 NTU and obviously these particles aggregate in suspension and re-sediment. The same effect is seen in the swelling experiments of Wy-Ca/Na, but appears more pronounced in Fig. 7 due to the logarithmic scale. In order to gain more insight into this somewhat unexpected turbidity increase in Wy-Ca (Wy-100/0) the concentration profiles were measured after 214 days, using the modified turbidimeter, and the results for sedimentation series C are displayed in Fig. 8. Zero means the topmost measurement position and 18 mm is the
14
400
Turbidity [NTU]
E2 100/0
300
E4 80/20
250
E6 60/40
200
E8 40/60
150
E10 20/80
100
E12 0/100
50 0
Turbidity [NTU]
12
350
10 Ca/Na 98/2
8
Ca/Na 95/5
6
Ca/Na 90/10
4 2 0
0
100
200
300
400
Time [Days] Fig. 5. Swelling series E: Turbidity at vial mid-height for Ku-Ca/Na montmorillonite. Clay content is 50 mg per 10 ml.
0
100
200
300
Time [Days] Fig. 6. Turbidity at vial mid-height as a function of swelling time for Wy-Ca/Na at a clay content of 50 mg per 10 ml.
1568
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
C1 100/0
1.E+05
Ca/Na 98/2
1.E+04
Ca/Na 95/5
100
P [kPa]
Turbidity [NTU]
1000
Ca/Na 90/10 C3 80/20
C9 20/80 100
200
300
y = 0.5153e
R = 0.9993 0
0.5
1
1.5
dry [g/cm
400
6.3739x
2
3
C11 0/100 0
1.E+02
1.E+00
C7 40/60
1
Exp. Regr.
1.E+03
1.E+01
C5 60/40
10
P
2
]
Fig. 10. Swelling pressure (kPa) vs. Wy-Ca dry density (g/cm3).
Time [Days] Fig. 7. Turbidity at vial mid-height as a function of sedimentation time for Wy-Ca/ Na at initial concentration of 5 g/l. Note that turbidities are displayed on a logarithmic scale.
Turbidity [NTU]
1000
C1 100/0 C3 80/20 C5 60/40 C7 40/60 C9 20/80 C 11 0/100
100
10
1
0
5
10
15
20
Depth [mm] Fig. 8. Concentration profiles after 214 days of sedimentation for Wy-Ca/Na at initial concentration of 5 g/l.
deepest measurement position. However, from the fact that initially (time 0) after shaking the vials the concentration is even and gives a turbidity of 570 NTU for Wy-0/100 (sample C11) it is realized that most of the montmorillonite is located below the 18 mm depth. The turbidity N is displayed on a log scale so straight lines mean exponential decay in accordance with
mgz NðzÞ / N0 exp ; kB T
ð3Þ
where m is the buoyant mass and z, the height above a reference level, kB Boltzmann’s constant and T is the absolute temperature. The samples containing some 20% or more sodium all show the ex-
pected exponential concentration profiles and from the data in Fig. 8 m can be evaluated. Furthermore, assuming that the clay particles are 1 nm thick circular discs the extracted buoyant mass would translate to clay particles with radii of approximately 100 nm. The profile for the 100/0, i.e. Wy-Ca, shows the presence of colloidal particles that are evenly distributed throughout the vial and therefore not affected by gravity on the length scale of the vials (h = 25 mm). These particles must therefore have a very small buoyant mass that can be roughly estimated from m 6 0.1kBT/gh, which becomes in the order of 1021 kg, or converted to montmorillonite, clay particles with radii of less than 10 nm for Ca-montmorillonite. It is possible that these particles instead are silica residues (or other minerals) that have not been completely removed in the purification. The slow increase in mid-height turbidity seen in the experiments on Wy-100/0, shows that this process is diffusion governed and not related to swelling. Furthermore, no loss of Wy-Ca has been observed in swelling pressure test cells, even at filter pore size of 100 lm as shown in Section 3.2. 3.2. Erosion deduced from swelling pressure response The pressure cell setup in Fig. 1 has been used in numerous experiments in our laboratory (e.g. Karnland et al., 2005, 2006, 2007) and for homoionic Na-montmorillonites it has been observed that the circulating solution, especially in the case of deionized water, becomes turbid due to loss of clay through the filters. Loss of clay through the filters also results in a reduction of swelling pressure. In the previous applications this erosion of clay has been an unwanted effect and handled in various ways, e.g. by avoiding continuous circulation of deionized water or by placing osmotic membranes between the clay and the filters. Ca-montmo-
Fig. 9. Swelling pressure as a function of time for Wy-Ca samples contacted with circulating deionized water over 40 and 100 lm filters respectively.
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
1569
Fig. 11. Swelling pressure in Wy-Na as a function of time.
Fig. 12. Swelling pressure as a function of time for Wy-80/20 erosion test.
rillonites on the other hand have never been found to erode. First this observation was reconfirmed using filters with pore sizes up 10 lm, i.e. pore sizes through which Na-montmorillonite readily erode when deionized water is circulated. Further tests on WyCa montmorillonite were made using coarser filters with pore sizes of 40 and 100 lm. As can be seen in Fig. 9, the swelling pressure remained constant in both cases. A small initial drop can be observed in the test with the 100 lm filter and could be explained with some clay that propagates into the filter (the swelling pressure is after all above 5 MPa) but no colloidal particles could be observed in the circulating solution. This test shows that the sol phase is absent in Ca-montmorillonite. Even the tiniest loss of montmorillonite would be manifested as a drop in swelling pressure as exemplified in the calculation below. In Fig. 10 measured swelling pressure, P is plotted versus dry density, qdry for Wy-Ca. The data points, taken from Table 4–7 in Karnland et al. (2006), can be perfectly fitted (solid line in Fig. 10) to an empirical swelling pressure function of the form
P ¼ A expðBqdry Þ;
ð4Þ 3
1
where A = 0.5153 kPa and B = 6.3739 cm g . By differentiating Eq. (4) an equation that relates a change in swelling pressure, DP to a clay mass loss, Dm is obtained,
DP ¼ BPDqdry ¼ BP
Dm ; V
ð5Þ
where V is the volume of the clay sample, in this case equal to p 3.52 0.5/4 = 4.8 cm3. A drop in swelling pressure of 100 kPa would be easily detected and at a swelling pressure of 5.5 MPa as in Fig. 9 such a decrease would according to Eq. (5) correspond to a mass loss of 14 mg. Furthermore, 14 mg clay in 100 ml (circulating) solution is well above the turbidity detection limit. For WyNa a clay concentration of 0.14 g/l gives a turbidity of 16 NTU and Wy-Ca is found to be more turbid. Thus the absence of measurable turbidity at the initial drop in swelling pressure corroborates the explanation that clay penetrates into the filter but is not further eroded by the circulating deionized water. As shown in the swelling and sedimentation experiments monitored by turbidity measurements Wy-Na montmorillonite actively disperses, which is in agreement with e.g. the atomic force microscopy study by Ploehn and Liu (2006) showing that fully exfoliated Na-montmorillonite dominates in dilute suspensions. This property becomes even more evident in the erosion tests using swelling pressure cells. Fig. 11 shows the swelling pressure development for two tests on Wy-Na using filters with pore sizes 2 and 10 lm respectively. The drop in swelling pressure or equivalently the loss of clay follows the same rate in both experiments showing that the
1570
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
5
-131 Ų/e
Posm /RT (M)
4
-96 Ų/e
3
-81 Ų/e
2 1 0 -1 -2
0
5
10
15
h (Å) Fig. 13. Calculated osmotic pressure (Turesson et al., 2004) as a function of interlayer distance for different charge densities given as area per charge. The chemical potential difference between mono- and divalent ions is held constant Dl = 5kBT.
clay particles as readily penetrate a 2 lm filter as a 10 lm filter. Further tests have shown that pore sizes of 0.5 lm effectively stops erosion of Wy-Na. To test how an 80/20 Wy-Ca/Wy-Na mixed clay behaves compared to the homoionic materials an experiment was run with this compound clay and 2 lm filters. Swelling pressure as a function of time is shown in Fig. 12. The rate of decrease in swelling pressure is comparable to that of Wy-Na in Fig. 11 despite the high content of Ca2+ in the clay. This confirms the earlier experimental and theoretical findings that an 80/20 montmorillonite behaves similar to pure Na-montmorillonite in deionized water. The Wy-80/20 clay appears to completely delaminate as it is flushed out through a 2 lm filter. Had there been grains (stacks) of Wy-Ca present some reduction in the erosion rate would have been expected, bearing in mind that pure Wy-Ca does not erode even through a 100 lm filter. In summary, it is found that at very low or negligible concentrations of excess salt close to 90% of the CEC needs to be compensated by Ca2+ to prevent colloidal sol formation in mixed Ca/Namontmorillonite.
4. Summary and conclusions The swelling and sedimentation behaviours were studied for mixed Ca/Na-montmorillonites in order to determine at what percentage of sodium the limited swelling behaviour seen in pure Camontmorillonite is lost. This property of Ca-montmorillonite has been described in the literature and stems from ion-ion correlations in the interlayer that strongly influence the swelling pressure so it actually becomes negative beyond interlayer distance of approximately 10 Å (Guldbrand et al., 1984; Kjellander et al., 1988). In order to separate the correlation effect of divalent ions from ionic strength effects or edge-face coagulation, these experiments were performed in deionized water. An important finding is the fact that approximately 90% of CEC needs to be compensated by Ca2+ in order to have a calcium-dominated system, i.e. a system where correlation effects give a stable minimum. At first it appeared contradictory that as little as 10–20% sodium in the interlayer would make the clay qualitatively similar to a pure Na-montmorillonite. However, these findings are in accordance with results obtained by Jönsson et al. (2009) who demonstrate in numerically exact solutions of the primitive model that the delicate balance between osmotic repulsion and correlation effects indeed gives clays that show ‘‘unlimited’’ swelling when the sodium content is 20% and clays with limited swelling when the sodium content is 10% of CEC. It is argued that the experiments presented here can only be explained under the assumption that the mixing of Ca2+ and Na+ occurs in the interlayer and not as de-mixed Ca- and Na-layers.
Mixing at the interlayer level is also inherent in the calculations by Jönsson et al. (2009). De-mixing is of course present in the formation of illite (Karnland and Birgersson, 2006) and most probably related to the weaker hydration of K+ compared to Na+. Justifying de-mixing of Na+ and Ca2+ is less obvious as both ions hydrates well. However, the stronger correlation forces among divalent ions could possibly lead to segregation into Ca and Na dominated layers but this depends also strongly on the layer charge as shown in Monte Carlo simulations of lamellar systems with different charge densities in contact with a NaCl/CaCl2 solution (Turesson et al., 2004). Consider the curve labelled 81 Å2/e in Fig. 13, which shows the calculated swelling pressure for a lamellar system with negative layer charge corresponding to one elementary charge per 81 Å2. At a certain interlayer distance (3 Å) the swelling pressure becomes negative, i.e. the attractive correlation forces dominate. At 8 Å the swelling pressure becomes positive again and eventually reaches a local maximum. From such oscillation in the pressure vs. interlayer distance the so called Maxwell equal area construction in analogy to the treatment of the van der Waals equation (Atkins and De Paula, 2006) may be used to find two layer separations in coexistence. The system at the shorter separation around 3 Å will have a higher Ca2+ content than the system at the larger separation as shown by Turesson et al. (2004). Thus at high layer charge partial de-mixing may theoretically be accounted for due to correlation effects. However, the upper curve in Fig. 13 denoted 131 Å2/e does not show any oscillatory behaviour and therefore no two-phase region, the two phases being Ca and Na dominated systems. The charge density in Wy-Na corresponds to one negative charge per 145 Å2 which is too low a charge density to give demixing. Both Kutch and Wyoming montmorillonite showed unlimited swelling at 20% sodium content. For Kutch montmorillonite no experiments with 10% sodium were carried out, but nothing suggests that Kutch montmorillonite would behave differently from the Wyoming material. If any difference, theory predicts higher correlation attraction the higher the layer charge, which means that Ku-Ca/Na would possibly need to contain more sodium than Wy-Ca/Na before forming a sol. Montmorillonite with 90% calcium or more in exchange position behaves as pure calcium montmorillonite and does not release colloidal particles. Furthermore it has been shown in the erosion tests that no Ca-montmorillonite is lost even when the filter pore size is as large as 100 lm. This is in contrast to tests with Wy80/20, which erodes through a 2 lm filter as readily as homoionic Wy-Na montmorillonite. Acknowledgement This work has been funded by the Swedish Nuclear Fuel and Waste Management Company (SKB). References Atkins, P.W., De Paula, J., 2006. Physical Chemistry, eighth ed. Oxford Univ. Press, Oxford. Bar-On, P., Shainberg, I., Michaeli, I., 1970. Electrophoretic mobility of montmorillonite particles saturated with Na/Ca ions. J. Colloid Interface Sci. 33, 471–472. Birgersson, M., Börgesson, L., Hedström, M., Karnland, O., Nilsson, U., 2009. Bentonite Erosion – Final Report. SKB Technical Report, TR-09-34, SKB Stockholm.. Birgersson, M., Hedström, M., Karnland, O., 2011. Sol formation ability of Ca/Namontmorillonite at low ionic strength. Phys. Chem Earth. doi:10.1016/ j.pce.2011.07.017. Brindley, G.W., Brown, G. (Eds.), 1980. Crystal Structures of Clay Minerals and their X-ray Identification. Mineralogical Society Monograph No.5. Evans, D.F., Wennerström, H., 1999. The Colloidal Domain, second ed. Wiley-WCH, New York. Guldbrand, L., Jönsson, B., Wennerström, H., Linse, P., 1984. Electrical doube-layer forces. A Monte Carlo study. J. Chem. Phys. 80, 2221–2228.
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571 Iwasaki, T., Watanabe, T., 1988. Distribution of Ca and Na ions in dioctahedral smectites and interstratified dioctahedral mica/smectites. Clays Clay Miner. 36, 73–82. Jönsson, B., Åkesson, T., Jönsson, B., Segad Meehdi, M., Janiak, J., Wallenberg, R., 2009. Structure and Forces in Bentonite MX-80. SKB Technical Report, TR-09-06, SKB Stockholm.. Karnland, O., Muurinen, A., Karlsson, F., 2005. Bentonite swelling pressure in NaCl solutions – experimentally determined data and model calculations. In: Alonso, Ledesma (Eds.), Advances in Understanding Engineered Clay Barriers. Taylor & Francis Group, London. Karnland, O., Birgersson, M., 2006. Montmorillonite Stability. With Special Respect to KBS-3 Conditions. SKB Technical Report, TR-06-11. SKB, Stockholm, and References Therein. . Karnland, O., Olsson, S., Nilsson, U., 2006. Mineralogy and Sealing Properties of Various Bentonites and Smectite-rich Clay Material. SKB Technical Report, TR06-30, SKB Stockholm. . Karnland, O., Olsson, S., Nilsson, U., Sellin, P., 2007. Experimentally determined swelling pressures and geochemical interactions of compacted Wyoming bentonite with highly alkaline solutions. Phys. Chem. Earth 32, 275–286.
1571
Kjellander, R., Marcelja, S., Quirk, J.P., 1988. Attractive double-layer interactions between calcium clay particles. J. Colloid Interface Sci. 126, 194–211. Lagaly, G., Ziesmer, S., 2003. Colloid chemistry of clay minerals: the coagulation of montmorillonite dispersions. Adv. Colloid Interface Sci. 100–102, 105–128. Laird, D.A., 2006. Influence of layer charge on swelling of smectites. App. Clay Sci. 34, 74–87. Michot, L., Bihannic, I., Porsch, K., Maddi, S., Baravian, C., Mougel, J., Levitz, P., 2004. Phase diagrams of Wyoming Na-montmorillonite clay. Influence of particle anisotropy. Langmuir 20, 10829–10837. Norrish, K., 1954. The swelling of montmorillonite. Discuss. Faraday Soc. 18, 120– 134. Ploehn, H.J., Liu, C., 2006. Quantitative analysis of montmorillonite platelet size by atomic force microscopy. Ind. Eng. Chem. Res. 45, 7025–7034. Shainberg, I., Kaiserman, A., 1969. Kinetics of the formation and breakdown of Camontmorillonite tactoids. Soil Sci. Soc. Am. Proc. 33, 547–551. Turesson, M., Forsman, J., Åkesson, T., Jönsson, B., 2004. Simulation of phase equilibria in lamellar surfactant systems. Langmuir 20, 5123–5126. Verburg, K., Baveye, P., 1994. Hysteresis in the binary exchange of cations on 2:1 clay minerals: a critical review. Clays Clay Miner. 42, 207–220.
Contents lists available at SciVerse ScienceDirect
Physics and Chemistry of the Earth journal homepage: www.elsevier.com/locate/pce
Role of cation mixing in the sol formation of Ca/Na-montmorillonite Magnus Hedström ⇑, Martin Birgersson, Ulf Nilsson, Ola Karnland Clay Technology AB, IDEON Science Park, SE-223 70 Lund, Sweden
a r t i c l e
i n f o
Article history: Available online 31 July 2011 Keywords: Montmorillonite Sedimentation Sol formation Swelling pressure Turbidity
a b s t r a c t Compound Ca/Na-montmorillonites were prepared by mixing given amounts of homoionic Ca-montmorillonite with given amounts of homoionic Na-montmorillonite. Colloidal sol formation in deionized water of these mixed Ca/Na-montmorillonites were investigated by means of turbidity measurements or through swelling pressure evolution in specially constructed swelling pressure test cells. It is concluded that the sol phase is absent in homoionic Ca-montmorillonite, which is interpreted as a manifestation of ion-ion correlations that are strong for divalent ions. Using the swelling pressure test cells it was found that Na-montmorillonite readily penetrates filters with pore size 2 lm, whereas in the case of Camontmorillonite no material was lost through the filters even at a pore size of 100 lm. However, mixed Ca/Na-montmorillonite with 20% or more sodium swells extensively and give rise to rapid increase in turbidity and behave qualitatively the same as homoionic Na-montmorillonite with respect to sol formation. This is also confirmed in erosion experiment using the swelling pressure test cells as well as in sedimentation tests monitored with a turbidimeter. Montmorillonite with 90% or more calcium in the interlayer behaves similar to homoionic Ca-montmorillonite. Thus it is concluded that 90% of the clay layer charge needs to be compensated by Ca2+ for correlation interactions to prevent sol formation. Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction In the Swedish KBS-3 concept for a geological deep storage of spent nuclear fuel, bentonite of high montmorillonite content is proposed to serve as a buffer surrounding copper canisters containing the spent fuel. Montmorillonite, which is the main constituent in bentonite, has an exceptional affinity for water which results in the build-up of a swelling pressure when bentonite (with access to water) is placed in a confined volume. There may be fractures intersecting the deposition hole and at those fractures the bentonite is not restricted but can continue to swell until equilibrium or steady state is reached. Under present day Swedish ground water conditions the swelling into fractures would be limited because the montmorillonite at the swelling front will eventually coagulate owing to the composition of dissolved salts. However, at the end of a glaciation it cannot be excluded that glacial melt water of low ionic strength will permeate the bedrock. This could cause significant erosion of the bentonite due to colloidal sol formation at the swelling front especially if the bentonite has been depleted of soluble accessory minerals. It is important to note that the gel-sol transition is not only a function of the ionic strength but also depends on the composition of counterions in the montmorillonite as well as the makeup of ions in the external solution (Bir-
⇑ Corresponding author. E-mail address: [email protected] (M. Hedström). 1474-7065/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.pce.2011.07.019
gersson et al., 2009, 2011). A homoionic Ca-montmorillonite would not pose any problem with respect to colloidal sol formation because it has limited swelling most likely due to attraction forces caused by ion correlations (Kjellander et al., 1988). In homoionic Na-montmorillonite, on the other hand, the correlation interactions are weak and cannot prevent the sol formation in case the montmorillonite is contacted with water of low ionic strength. Colloid sol formation may be stopped or reduced by the presence of excess salt. High concentrations, in the range 1–2 M, of a 1:1 salt, e.g. NaCl would screen the double-layer repulsion effectively and the van der Waals attraction between the clay layers would dominate the system and limit the swelling of the interlayers to about 1 nm (Norrish, 1954). Coagulation in Na-montmorillonite systems is also observed at lower NaCl concentrations, in the range 5–25 mM, which is attributed to edge-face interactions (Lagaly and Ziesmer, 2003; Birgersson et al., 2011). However, the topic of this paper is the investigation of sol formation in the absence of excess salt in the system, i.e. the only cations present are those needed to compensate the clay layer charge. Numerous experiments have shown that montmorillonite with calcium or magnesium as counterion does not form colloidal sols, even if placed in deionized water. The problem with extensive swelling and colloidal particle release only occurs when monovalent counterions are present. What fraction of divalent counterions in the clay is needed to prevent colloidal sol formation? To answer this question compound Ca/Na-montmorillonites were prepared by mixing given amounts of pure Ca-montmorillonite with given
1565
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
amounts of pure Na-montmorillonite. It is noted in passing that the absence of excess salt is too conservative for the worst-case scenario of glacial meltwater, but is a highly relevant test system as it allows for isolating the effect of correlation mediated by divalent ions in the interlayer. 2. Experimental 2.1. Materials Two types of montmorillonite have been used in the investigations, Wyoming-type (Wy) obtained from MX-80 bentonite (American Colloid Co.) and Kutch-type (Ku) from the Indian bentonite Asha 505 (Ashapura Minechem Co.). The montmorillonites are extracted and ion-exchanged to either sodium or calcium form according to the procedure described in Karnland et al. (2006). The chemical formula for an ideal (sodium) montmorillonite, interlayer water excluded, can be written as
ðSi8a Ala ÞðAl4b Mgb ÞO20 ðOHÞ4
ðaþbÞ
Naþaþb ;
ð1Þ
where by definition the tetrahedral charge is lower than the octahedral charge (a < b) and the sum of tetrahedral and octahedral charges fulfil 0.4 < a + b < 1.2. It has previously been established that the Wy-montmorillonite is dominated by octahedral charges b = 0.54–0.6 and the total charge is 0.65e, whereas Ku-montmorillonite has a much higher portion of the charge in the tetrahedral layer, a = 0.38, as well as a higher total charge of 0.79e (Karnland et al., 2006). Both for Wy and Ku, mixed Ca/Na clays were made at mass fractions 100/0, 80/20, 60/40, 40/60, 20/80 and 0/100, where 100/0 and 0/100 means pure Ca-montmorillonite and pure Na-montmorillonite respectively. In addition to the above mass fractions a batch of Wy-montmorillonite at mass fractions 98/2, 95/5 and 90/10 was made. The mass fractions also reflect the fraction of the cation exchange capacity (CEC) compensated by Ca2+ and Na+, e.g. in the 80/ 20 montmorillonite 80% of the surface charge is compensated by Ca2+ and 20% by Na+. The experiments to be monitored with a turbidimeter need only very small amounts of montmorillonite. In total 2 g of clay was dispersed in 250 ml deionized water for each Ca/ Na combination. To facilitate the mixing, the suspension was agitated with ultrasound for 15 min and then mixed with a magnetic stirrer for at least 12 h, which is a long equilibration time considering that Shainberg and Kaiserman (1969) showed that the properties of mixed Ca/Na-montmorillonite suspensions were stabilized within 10 min of mixing. After the mixing the water was removed from the suspension in an oven at 60 °C. 2.2. Turbidity measurements The swelling and sedimentation behaviour of Wy and Ku montmorillonite are investigated with respect to counterion composition in the interlayer; either Na+ or Ca2+ or mixtures of the two ions. The experiments are performed using cylindrical glass vials, with outer dimensions: height 35 mm and diameter 23 mm. In swelling experiments a given amount of montmorillonite is dried-in at the bottom of the vials and 10 ml of deionized water is added. The swelling and colloidal particle release vertically is monitored by measuring the change in turbidity with time at approximately 12–13 mm above the bottom. At low particle concentrations the turbidity is directly proportional to the concentration. Similarly for sedimentation the same equipment is used. The sedimentation tests are prepared by dispersing up to 50 mg clay in 10 ml water overnight. Prior to the first turbidity measurement the vials are shaken to ensure a homogeneous distribution of clay.
Turbidity is measured with a portable turbidimeter TN-100 from Eutech Instruments. The light source in TN-100 is an infrared-emitting (850 nm) diode. Light scattered from the clay particles in the suspension is detected at an angle of 90 ° with respect to the light source, thus formally the TN-100 is a Nephelometer. The intensity of the scattered light is directly converted to nephelometric turbidity units (NTU) and the TN-100 operates in the turbidity range 0–1000 NTU. As mentioned above the standard turbidimeter measures the turbidity at 12–13 mm height in the vials. A second TN-100 instrument was rebuilt to allow the registration of turbidity at different heights in the vials, thereby offering the possibility of obtaining time-resolved swelling or sedimentation profiles. Turbidity at low colloidal particle concentrations is proportional to the concentration or equivalently the particle number density. It is found that different clays give different turbidity responses. Up to a certain concentration the turbidity response is linear and beyond the linear regime the turbidity increases less than the increase in clay concentration. For Wy-Na the turbidity response is linear up to 5 g clay/litre deionized water (114 NTU per g/l). KuNa shows linear response to approximately 2.5 g/l (315 NTU per g/l). 2.3. Erosion tests Erosion of montmorillonite was monitored by measuring the swelling pressure response using a swelling pressure test cell (e.g. Karnland et al., 2007) as shown in Fig. 1. Highly compacted discs, radius 35 mm and height 5 mm, of montmorillonite were placed in pressure cells with permeable 2 mm thick filters of different pore sizes ranging from 0.2 to 100 lm. After water saturation, 100 ml deionized water was continuously circulated behind the filters and the swelling pressure was monitored during the test period that lasted about a month. Since swelling pressure depends strongly on the montmorillonite density any loss of clay through the filters due to sol formation, would be detected as a drop in swelling pressure. 3. Results and discussion 3.1. Turbidity evolution Before discussing the results from the experiments it is instructive to reflect upon the quantities of montmorillonite used and the dimensions of the vials. To cover the bottom area of the vials by one clay layer of thickness d = 1 nm, requires the mass, m = qcpr2d of montmorillonite. With a grain density of montmorillonite of qc = 2750 kg/m3 and the vial inner radius r 11 mm, m becomes approximately 1 lg. Typically in the experiments 20–100 mg of clay is used. If these amounts of clay were evenly dried-in at the bottom the height would be 0.02–0.1 mm or equivalently 20,000 to 100,000 layers of clay. After water uptake to a basal distance of 200 nm, roughly the distance where free rotation of a typical
100 mL circulating solution Force transducer Clay sample
Filters (titanium or stainless steel) Fig. 1. Schematic drawing of the swelling pressure test cell.
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
Wy-Na particle could be expected (Michot et al., 2004), the height of the clay phase in the vials would be 4 mm in the case of 20 mg and 20 mm for 100 mg montmorillonite, respectively. The height of 10 ml liquid in the vials is 25 mm so swelling to the point of free rotation means that 100 mg of clay essentially fills the whole sample, 50 mg barely reaches the turbidity detection height at 12– 13 mm in the vial and 20 mg reaches one third of the detection height.
500 450 400 350 300 250 200 150 100 50 0
25
D1 100/0 D3 80/20
20
D5 60/40
15
D7 40/60
10
D9 20/80
5
D11 0/100
0
0
100
200
300
400
Time [Days] Fig. 3. Swelling series D: Turbidity at vial mid-height vs. swelling time for Wy-Ca/ Na montmorillonite. Clay content is 20 mg per 10 ml.
after 57 days. If the de-mixing hypothesis were the dominating mechanism here, a much slower increase in turbidity would be expected for sample B4 because clay particles would not fully exfoliate, which would have two effects that both reduce the transport of clay against gravity. First, clay particles would be heavier (‘‘stacks’’ of Ca-montmorillonite with Na-ions on the outer surfaces) and therefore more affected by gravity. Second, osmotic pressure would be lower. This follows from the fact that the osmotic pressure (P) at large particle separations is inversely proportional to the square of the separation (h) (Evans and Wennerström, 1999, Eq. 5.1.18)
P¼
p2 o r ðkB TÞ2 1 2ðzeÞ2
h
2
ð2Þ
;
and distances between ‘‘stacks’’ are necessarily larger than they would be between fully exfoliated clay layers at a given water to solid mass ratio. Undoubtedly Wy-80/20 clay enters the osmotic swelling regime where the basal distance surpasses by far 20 Å and water is adsorbed in a continuous manner (Brindley and Brown, 1980). With that in mind it is also clear that an increase of the percentage of sodium should give higher swelling pressures at long distance due to a larger number of ions in the interlayer. This is reflected at the initial stage of the turbidity response, where higher maximum turbidities are noticed the higher the sodium fraction. Below the correlation between swelling pressure and maximum turbidities is discussed in more detail. With higher total clay content the initial swelling-dominated increase in turbidity is more rapid and with 100 mg clay/10 ml the turbidity for Wy-0/100 reaches a maximum after 8 days (Fig. 2) whereas with 20 mg/10 ml the maximum occurs after 57 days (Fig. 3). It comes as no surprise that an intermediate clay content of 50 mg/10 ml, presented in Fig. 4, gives maximum turbidities and
120
B2 100/0 B4 80/20 B6 60/40 B8 40/60 B10 20/80 B12 0/100
0
100
200
300
400
Time [Days] Fig. 2. Swelling series B: Turbidity at vial mid-height vs. swelling time of Wy-Ca/Na montmorillonite for various Ca/Na combinations. Clay content is 100 mg per 10 ml.
Turbidity [NTU]
Turbidity [NTU]
3.1.1. Swelling The swelling of mixed Ca/Na montmorillonite is monitored by measuring the turbidity at vial mid-height as the clay expands and disperses against gravity. Initially the clay is dried-in at the bottom and 10 ml deionized water is added. The results for clay contents of 100 mg and 20 mg are displayed in Figs. 2 and 3, respectively. Pure Wy-Ca (100/0) behaves totally different from the other clays with 20% Na+ or more in exchange position. In practice there is no release of colloidal particles from Wy-Ca and the measured low turbidity reflects the fact that Wy-Ca swells to a maximum distance. Furthermore, the original grains of Ca-montmorillonite on the bottom of the vials are intact, albeit considerably swollen compared to the dry state. Inclusion of as little as 20% Na+ makes the Wy-Ca/Na clay display predominantly sodium montmorillonite behaviour and no sign of the original grains is visible in the vials. It is instructive to compare the curve for Wy-80/20 (B4) in Fig. 2 with the curve for Wy-0/100 (D11) in Fig. 3. Sample B4 consists of 80 mg Wy-Ca and 20 mg Wy-Na, whereas sample D11 contains 20 mg Wy-Na. Thus both samples have the same amount of Na+ but their swelling behaviours are entirely different. An assumption that mixed Ca/Na smectite systems segregate into Ca-layers and Na-layers, a process termed de-mixing, is sometimes encountered in the literature (Bar-On et al., 1970; Iwasaki and Watanabe, 1988; Verburg and Baveye, 1994; Laird 2006). For de-mixing to occur there must exist an interaction energy that dominates over the entropy of mixing as described in e.g. (Evans and Wennerström, 1999). It is very hard to envisage any such force that would make e.g. Ca2+ ions to prefer to be surrounded by other Ca2+ ions over Na+ ions, but this point will be revisited in the summary and conclusions in Section 4. However, the results presented here strongly suggest that the mixing of Ca-montmorillonite with Na-montmorillonite results in a clay where the ions are mixed in the interlayer and not to a situation where there are Ca-layers and Na-layers. The latter situation would correspond to 80% WyCa that do not release colloidal particles and 20% Wy-Na that would give a maximum turbidity of approximately 25 NTU in accordance with Fig. 3. Instead the Wy-80/20 clay gives a turbidity maximum above 160 NTU as seen in Fig. 2. Furthermore, the turbidity maximum of series B4 in Fig. 2 occurs after 27 days, whereas the turbidity maximum in swelling series D11 in Fig. 3 takes place
30
Turbidity [NTU]
1566
100
D2 100/0 D4 80/20
80
D6 60/40
60
D8 40/60
40
D10 20/80
20
D12 0/100
0
0
100
200
300
400
Time [Days] Fig. 4. Swelling series D: Turbidity at vial mid-height vs. swelling time of Wy-Ca/Na montmorillonite for various Ca/Na combinations. Clay content is 50 mg per 10 ml.
1567
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
turbidity-increase rates in between the values obtained at 20 and 100 mg/10 ml respectively. It is also clear that the maximum turbidities at vial mid-height do not vary linearly with clay content. For a diffusion-driven process a linear variation of turbidity with the amount of dried-in clay is expected at all times. However, at the termination of the experiments (after 1 year) the mid-height turbidities appear proportional to the amount of clay in the vials. The initial non-linearity can be explained on the basis of the osmotic pressures in lamellar systems at low densities. According to the Poisson–Boltzmann equation the swelling pressure is inversely proportional to the square of the mean interlayer distance (see Eq. (2)) or, equivalently, proportional to the square of the solid to water mass fraction which follows from the relation
qclay d 1 mclay ¼ / ; mwater qwater h h where d = 1 nm, i.e. the thickness of a single clay layer and h is the interlayer separation. Therefore the swelling pressures in the systems discussed here with 100, 50 and 20 mg total amounts of montmorillonite obey the relation P(100 mg) = 4P(50 mg) = 25P(20 mg). Let NCa=Na denote the maximum turbidity for a given Ca/Na clay in m the swelling experiments. From Figs. 2–4 it is concluded that the turbidities for the 0/100 clays at maximum obey the relation N 0=100 ð100 mgÞ ¼ 4:4N 0=100 ð50 mgÞ ¼ 18:3N 0=100 ð20 mgÞ and for the m m m 80/20 clays the relation is similar; N 80=20 ð100 mgÞ ¼ 3:9N 80=20 ð50 mgÞ ¼ 25:1N 80=20 ð20 mgÞ: The fact m m m that the maximum turbidities behave as the swelling pressures can be rationalized as follows: Turbidity is proportional to clay concentration, therefore Nm is proportional to the mass of clay in a cross-section of thickness d at detection height, where d is the width of the scattering region. The weight of this mass is balanced by the swelling pressure exerted by the clay below, i.e. P qdg, where g is the acceleration of gravity. Thus the ratio of maximum mid-height turbidities or densities for tests with initially dried-in amounts m1 and m2 of clay is envisaged to approximately obey the relation
qðm1 Þ Pðm1 Þ m1 qðm2 Þ Pðm2 Þ m2
2 :
This relation is of course only approximate and one source of imprecision is the neglect of the weight of the clay above the detection height. On the other hand, the maximum turbidities show that the concentrations at mid-height are much lower than the average concentration, which is a sign that most of the clay remains below detection height. This is also confirmed by measuring the turbidity profiles along the height of the vials using the modified turbidimeter. The turbidity profiles also show that the decrease in mid-height turbidity after the maximum value is primarily due to sedimentation and not to further swelling or diffusion of clay upwards.
Ku-Ca/Na behaves similarly to Wy-Ca/Na as can be seen in swelling series E in Fig. 5. Kutch is more turbid than Wyoming montmorillonite, which is reflected in the high turbidities despite the fact the clay content in series E is 50 mg/10 ml, i.e. half of the clay content in Series B (Fig. 2). In addition a larger correlation between Na content and turbidity at maximum for Ku-Ca/Na than for Wy-Na/Ca is also noticed. As time progress the initial differentiation among the clays containing sodium is faded out both for Ku and Wy. Swelling experiments on Ku-Ca/Na were also performed for 20 mg/10 ml and the non-linear dependence on Nm with clay content seen for Wy-Na/Ca was confirmed and for the pure KuNa clay the relation was found to be N 0=100 ð50 mgÞ ¼ 9:1N 0=100 ð20 mgÞ and similar for the other Na+ m m containing clays. Increasing the calcium content beyond 90% of CEC makes the limited swelling behaviour typical for calcium montmorillonite to dominate as can be seen in Fig. 6. Note that when water is added some grains of clay are stirred up which gives relatively high initial turbidities. These grains are quite large and sediment rapidly leading to a minimum in the turbidity after a few days. As can be seen in Fig. 6, Wy-90/10 still shows some active colloidal particle release but the turbidity increase at vial mid position is very minor and reaches a maximum of 5.6 NTU after 130 days. In comparison Wy-80/20 at 5 g/l gives a turbidity maximum above 40 NTU in 40 days (Fig. 4). Furthermore Wy-80/20 appears similar to pure Wy-Na with no sign of limited swelling, whereas in free swelling of Wy-90/10 original grains of clay are present at the bottom of the vial. 3.1.2. Sedimentation Sedimentation of Wy-Ca/Na confirms the above picture that 90% Ca or more is needed for behaviour similar to pure Wy-Ca to occur, as shown in Fig. 7. With 60% Ca or less the sedimentation follows the same pattern as for pure Wy-Na. Wy-80/20 sediments slightly faster and Wy-90/10 falls in between the fast sedimentation of Wy-Ca and the slower sedimentation of Wy-80/20. The gradual increase in turbidity for Wy-100/0, 98/2 and 95/5 after 90 days has only been observed in Wyoming montmorillonite and not in Kutch. We have at present no explanation for this pattern but the turbidity at maximum after 250 days (Wy-100/0) is only 10 NTU and obviously these particles aggregate in suspension and re-sediment. The same effect is seen in the swelling experiments of Wy-Ca/Na, but appears more pronounced in Fig. 7 due to the logarithmic scale. In order to gain more insight into this somewhat unexpected turbidity increase in Wy-Ca (Wy-100/0) the concentration profiles were measured after 214 days, using the modified turbidimeter, and the results for sedimentation series C are displayed in Fig. 8. Zero means the topmost measurement position and 18 mm is the
14
400
Turbidity [NTU]
E2 100/0
300
E4 80/20
250
E6 60/40
200
E8 40/60
150
E10 20/80
100
E12 0/100
50 0
Turbidity [NTU]
12
350
10 Ca/Na 98/2
8
Ca/Na 95/5
6
Ca/Na 90/10
4 2 0
0
100
200
300
400
Time [Days] Fig. 5. Swelling series E: Turbidity at vial mid-height for Ku-Ca/Na montmorillonite. Clay content is 50 mg per 10 ml.
0
100
200
300
Time [Days] Fig. 6. Turbidity at vial mid-height as a function of swelling time for Wy-Ca/Na at a clay content of 50 mg per 10 ml.
1568
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
C1 100/0
1.E+05
Ca/Na 98/2
1.E+04
Ca/Na 95/5
100
P [kPa]
Turbidity [NTU]
1000
Ca/Na 90/10 C3 80/20
C9 20/80 100
200
300
y = 0.5153e
R = 0.9993 0
0.5
1
1.5
dry [g/cm
400
6.3739x
2
3
C11 0/100 0
1.E+02
1.E+00
C7 40/60
1
Exp. Regr.
1.E+03
1.E+01
C5 60/40
10
P
2
]
Fig. 10. Swelling pressure (kPa) vs. Wy-Ca dry density (g/cm3).
Time [Days] Fig. 7. Turbidity at vial mid-height as a function of sedimentation time for Wy-Ca/ Na at initial concentration of 5 g/l. Note that turbidities are displayed on a logarithmic scale.
Turbidity [NTU]
1000
C1 100/0 C3 80/20 C5 60/40 C7 40/60 C9 20/80 C 11 0/100
100
10
1
0
5
10
15
20
Depth [mm] Fig. 8. Concentration profiles after 214 days of sedimentation for Wy-Ca/Na at initial concentration of 5 g/l.
deepest measurement position. However, from the fact that initially (time 0) after shaking the vials the concentration is even and gives a turbidity of 570 NTU for Wy-0/100 (sample C11) it is realized that most of the montmorillonite is located below the 18 mm depth. The turbidity N is displayed on a log scale so straight lines mean exponential decay in accordance with
mgz NðzÞ / N0 exp ; kB T
ð3Þ
where m is the buoyant mass and z, the height above a reference level, kB Boltzmann’s constant and T is the absolute temperature. The samples containing some 20% or more sodium all show the ex-
pected exponential concentration profiles and from the data in Fig. 8 m can be evaluated. Furthermore, assuming that the clay particles are 1 nm thick circular discs the extracted buoyant mass would translate to clay particles with radii of approximately 100 nm. The profile for the 100/0, i.e. Wy-Ca, shows the presence of colloidal particles that are evenly distributed throughout the vial and therefore not affected by gravity on the length scale of the vials (h = 25 mm). These particles must therefore have a very small buoyant mass that can be roughly estimated from m 6 0.1kBT/gh, which becomes in the order of 1021 kg, or converted to montmorillonite, clay particles with radii of less than 10 nm for Ca-montmorillonite. It is possible that these particles instead are silica residues (or other minerals) that have not been completely removed in the purification. The slow increase in mid-height turbidity seen in the experiments on Wy-100/0, shows that this process is diffusion governed and not related to swelling. Furthermore, no loss of Wy-Ca has been observed in swelling pressure test cells, even at filter pore size of 100 lm as shown in Section 3.2. 3.2. Erosion deduced from swelling pressure response The pressure cell setup in Fig. 1 has been used in numerous experiments in our laboratory (e.g. Karnland et al., 2005, 2006, 2007) and for homoionic Na-montmorillonites it has been observed that the circulating solution, especially in the case of deionized water, becomes turbid due to loss of clay through the filters. Loss of clay through the filters also results in a reduction of swelling pressure. In the previous applications this erosion of clay has been an unwanted effect and handled in various ways, e.g. by avoiding continuous circulation of deionized water or by placing osmotic membranes between the clay and the filters. Ca-montmo-
Fig. 9. Swelling pressure as a function of time for Wy-Ca samples contacted with circulating deionized water over 40 and 100 lm filters respectively.
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
1569
Fig. 11. Swelling pressure in Wy-Na as a function of time.
Fig. 12. Swelling pressure as a function of time for Wy-80/20 erosion test.
rillonites on the other hand have never been found to erode. First this observation was reconfirmed using filters with pore sizes up 10 lm, i.e. pore sizes through which Na-montmorillonite readily erode when deionized water is circulated. Further tests on WyCa montmorillonite were made using coarser filters with pore sizes of 40 and 100 lm. As can be seen in Fig. 9, the swelling pressure remained constant in both cases. A small initial drop can be observed in the test with the 100 lm filter and could be explained with some clay that propagates into the filter (the swelling pressure is after all above 5 MPa) but no colloidal particles could be observed in the circulating solution. This test shows that the sol phase is absent in Ca-montmorillonite. Even the tiniest loss of montmorillonite would be manifested as a drop in swelling pressure as exemplified in the calculation below. In Fig. 10 measured swelling pressure, P is plotted versus dry density, qdry for Wy-Ca. The data points, taken from Table 4–7 in Karnland et al. (2006), can be perfectly fitted (solid line in Fig. 10) to an empirical swelling pressure function of the form
P ¼ A expðBqdry Þ;
ð4Þ 3
1
where A = 0.5153 kPa and B = 6.3739 cm g . By differentiating Eq. (4) an equation that relates a change in swelling pressure, DP to a clay mass loss, Dm is obtained,
DP ¼ BPDqdry ¼ BP
Dm ; V
ð5Þ
where V is the volume of the clay sample, in this case equal to p 3.52 0.5/4 = 4.8 cm3. A drop in swelling pressure of 100 kPa would be easily detected and at a swelling pressure of 5.5 MPa as in Fig. 9 such a decrease would according to Eq. (5) correspond to a mass loss of 14 mg. Furthermore, 14 mg clay in 100 ml (circulating) solution is well above the turbidity detection limit. For WyNa a clay concentration of 0.14 g/l gives a turbidity of 16 NTU and Wy-Ca is found to be more turbid. Thus the absence of measurable turbidity at the initial drop in swelling pressure corroborates the explanation that clay penetrates into the filter but is not further eroded by the circulating deionized water. As shown in the swelling and sedimentation experiments monitored by turbidity measurements Wy-Na montmorillonite actively disperses, which is in agreement with e.g. the atomic force microscopy study by Ploehn and Liu (2006) showing that fully exfoliated Na-montmorillonite dominates in dilute suspensions. This property becomes even more evident in the erosion tests using swelling pressure cells. Fig. 11 shows the swelling pressure development for two tests on Wy-Na using filters with pore sizes 2 and 10 lm respectively. The drop in swelling pressure or equivalently the loss of clay follows the same rate in both experiments showing that the
1570
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571
5
-131 Ų/e
Posm /RT (M)
4
-96 Ų/e
3
-81 Ų/e
2 1 0 -1 -2
0
5
10
15
h (Å) Fig. 13. Calculated osmotic pressure (Turesson et al., 2004) as a function of interlayer distance for different charge densities given as area per charge. The chemical potential difference between mono- and divalent ions is held constant Dl = 5kBT.
clay particles as readily penetrate a 2 lm filter as a 10 lm filter. Further tests have shown that pore sizes of 0.5 lm effectively stops erosion of Wy-Na. To test how an 80/20 Wy-Ca/Wy-Na mixed clay behaves compared to the homoionic materials an experiment was run with this compound clay and 2 lm filters. Swelling pressure as a function of time is shown in Fig. 12. The rate of decrease in swelling pressure is comparable to that of Wy-Na in Fig. 11 despite the high content of Ca2+ in the clay. This confirms the earlier experimental and theoretical findings that an 80/20 montmorillonite behaves similar to pure Na-montmorillonite in deionized water. The Wy-80/20 clay appears to completely delaminate as it is flushed out through a 2 lm filter. Had there been grains (stacks) of Wy-Ca present some reduction in the erosion rate would have been expected, bearing in mind that pure Wy-Ca does not erode even through a 100 lm filter. In summary, it is found that at very low or negligible concentrations of excess salt close to 90% of the CEC needs to be compensated by Ca2+ to prevent colloidal sol formation in mixed Ca/Namontmorillonite.
4. Summary and conclusions The swelling and sedimentation behaviours were studied for mixed Ca/Na-montmorillonites in order to determine at what percentage of sodium the limited swelling behaviour seen in pure Camontmorillonite is lost. This property of Ca-montmorillonite has been described in the literature and stems from ion-ion correlations in the interlayer that strongly influence the swelling pressure so it actually becomes negative beyond interlayer distance of approximately 10 Å (Guldbrand et al., 1984; Kjellander et al., 1988). In order to separate the correlation effect of divalent ions from ionic strength effects or edge-face coagulation, these experiments were performed in deionized water. An important finding is the fact that approximately 90% of CEC needs to be compensated by Ca2+ in order to have a calcium-dominated system, i.e. a system where correlation effects give a stable minimum. At first it appeared contradictory that as little as 10–20% sodium in the interlayer would make the clay qualitatively similar to a pure Na-montmorillonite. However, these findings are in accordance with results obtained by Jönsson et al. (2009) who demonstrate in numerically exact solutions of the primitive model that the delicate balance between osmotic repulsion and correlation effects indeed gives clays that show ‘‘unlimited’’ swelling when the sodium content is 20% and clays with limited swelling when the sodium content is 10% of CEC. It is argued that the experiments presented here can only be explained under the assumption that the mixing of Ca2+ and Na+ occurs in the interlayer and not as de-mixed Ca- and Na-layers.
Mixing at the interlayer level is also inherent in the calculations by Jönsson et al. (2009). De-mixing is of course present in the formation of illite (Karnland and Birgersson, 2006) and most probably related to the weaker hydration of K+ compared to Na+. Justifying de-mixing of Na+ and Ca2+ is less obvious as both ions hydrates well. However, the stronger correlation forces among divalent ions could possibly lead to segregation into Ca and Na dominated layers but this depends also strongly on the layer charge as shown in Monte Carlo simulations of lamellar systems with different charge densities in contact with a NaCl/CaCl2 solution (Turesson et al., 2004). Consider the curve labelled 81 Å2/e in Fig. 13, which shows the calculated swelling pressure for a lamellar system with negative layer charge corresponding to one elementary charge per 81 Å2. At a certain interlayer distance (3 Å) the swelling pressure becomes negative, i.e. the attractive correlation forces dominate. At 8 Å the swelling pressure becomes positive again and eventually reaches a local maximum. From such oscillation in the pressure vs. interlayer distance the so called Maxwell equal area construction in analogy to the treatment of the van der Waals equation (Atkins and De Paula, 2006) may be used to find two layer separations in coexistence. The system at the shorter separation around 3 Å will have a higher Ca2+ content than the system at the larger separation as shown by Turesson et al. (2004). Thus at high layer charge partial de-mixing may theoretically be accounted for due to correlation effects. However, the upper curve in Fig. 13 denoted 131 Å2/e does not show any oscillatory behaviour and therefore no two-phase region, the two phases being Ca and Na dominated systems. The charge density in Wy-Na corresponds to one negative charge per 145 Å2 which is too low a charge density to give demixing. Both Kutch and Wyoming montmorillonite showed unlimited swelling at 20% sodium content. For Kutch montmorillonite no experiments with 10% sodium were carried out, but nothing suggests that Kutch montmorillonite would behave differently from the Wyoming material. If any difference, theory predicts higher correlation attraction the higher the layer charge, which means that Ku-Ca/Na would possibly need to contain more sodium than Wy-Ca/Na before forming a sol. Montmorillonite with 90% calcium or more in exchange position behaves as pure calcium montmorillonite and does not release colloidal particles. Furthermore it has been shown in the erosion tests that no Ca-montmorillonite is lost even when the filter pore size is as large as 100 lm. This is in contrast to tests with Wy80/20, which erodes through a 2 lm filter as readily as homoionic Wy-Na montmorillonite. Acknowledgement This work has been funded by the Swedish Nuclear Fuel and Waste Management Company (SKB). References Atkins, P.W., De Paula, J., 2006. Physical Chemistry, eighth ed. Oxford Univ. Press, Oxford. Bar-On, P., Shainberg, I., Michaeli, I., 1970. Electrophoretic mobility of montmorillonite particles saturated with Na/Ca ions. J. Colloid Interface Sci. 33, 471–472. Birgersson, M., Börgesson, L., Hedström, M., Karnland, O., Nilsson, U., 2009. Bentonite Erosion – Final Report. SKB Technical Report, TR-09-34, SKB Stockholm.
M. Hedström et al. / Physics and Chemistry of the Earth 36 (2011) 1564–1571 Iwasaki, T., Watanabe, T., 1988. Distribution of Ca and Na ions in dioctahedral smectites and interstratified dioctahedral mica/smectites. Clays Clay Miner. 36, 73–82. Jönsson, B., Åkesson, T., Jönsson, B., Segad Meehdi, M., Janiak, J., Wallenberg, R., 2009. Structure and Forces in Bentonite MX-80. SKB Technical Report, TR-09-06, SKB Stockholm.
1571
Kjellander, R., Marcelja, S., Quirk, J.P., 1988. Attractive double-layer interactions between calcium clay particles. J. Colloid Interface Sci. 126, 194–211. Lagaly, G., Ziesmer, S., 2003. Colloid chemistry of clay minerals: the coagulation of montmorillonite dispersions. Adv. Colloid Interface Sci. 100–102, 105–128. Laird, D.A., 2006. Influence of layer charge on swelling of smectites. App. Clay Sci. 34, 74–87. Michot, L., Bihannic, I., Porsch, K., Maddi, S., Baravian, C., Mougel, J., Levitz, P., 2004. Phase diagrams of Wyoming Na-montmorillonite clay. Influence of particle anisotropy. Langmuir 20, 10829–10837. Norrish, K., 1954. The swelling of montmorillonite. Discuss. Faraday Soc. 18, 120– 134. Ploehn, H.J., Liu, C., 2006. Quantitative analysis of montmorillonite platelet size by atomic force microscopy. Ind. Eng. Chem. Res. 45, 7025–7034. Shainberg, I., Kaiserman, A., 1969. Kinetics of the formation and breakdown of Camontmorillonite tactoids. Soil Sci. Soc. Am. Proc. 33, 547–551. Turesson, M., Forsman, J., Åkesson, T., Jönsson, B., 2004. Simulation of phase equilibria in lamellar surfactant systems. Langmuir 20, 5123–5126. Verburg, K., Baveye, P., 1994. Hysteresis in the binary exchange of cations on 2:1 clay minerals: a critical review. Clays Clay Miner. 42, 207–220.