Diffusion and retention behaviour of Cs in illite-added compacted montmorillonite

Abstract Compacted bentonite is to be used as a component of an engineered barrier system to retard the migration of radionuclides in the geological disposal of radioactive waste. In such an environment, montmorillonite in compacted bentonite might be altered to illite due to the hydrothermal reactions caused by the decay heat of radionuclides. In the present study, the diffusion and retention behaviour of Cs in compacted montmorillonite containing illitewas investigated using through-diffusion experiments. The experimental results showed that the flux of Cs attributed to the surface diffusion was independent of the sorption of Cs on illite, indicating that the Cs sorbed on illite was immobile or considerably less mobile than the Cs sorbed on montmorillonite. Consequently, the illite content in compacted bentonite is expected to enhance the sorption capacity of Cs without causing surface diffusion.

1 Japan Atomic Energy Agency (JAEA), 4-33, Muramatsu, Tokai-mura, Naka-gun, Ibaraki, 319-1194, Japan 2 Inspection Development Corporation, 4-33, Muramatsu, Tokai-mura, Naka-gun, Ibaraki, 319-1112, Japan A B S T R AC T : Compacted bentonite is to be used as a component of an engineered barrier system to retard the migration of radionuclides in the geological disposal of radioactive waste. In such an environment, montmorillonite in compacted bentonite might be altered to illite due to the hydrothermal reactions caused by the decay heat of radionuclides. In the present study, the diffusion and retention behaviour of Cs in compacted montmorillonite containing illite was investigated using through-diffusion experiments. The experimental results showed that the flux of Cs attributed to the surface diffusion was independent of the sorption of Cs on illite, indicating that the Cs sorbed on illite was immobile or considerably less mobile than the Cs sorbed on montmorillonite. Consequently, the illite content in compacted bentonite is expected to enhance the sorption capacity of Cs without causing surface diffusion.
In the geological disposal of radioactive waste, compacted bentonite will be used as a component of the engineered barrier system to retard the migration of radionuclides from waste packages. The migration of radionuclides is governed by diffusion in compacted bentonite due to the low permeability caused by the swelling of montmorillonite, the major component of bentonite. The large sorption capacity of montmorillonite also contributes to the retardation of radionuclides diffusing in compacted bentonite as cationic species. Concern exists, however, that montmorillonite is altered to other minerals because of the hydrothermal reactions induced by the decay heat of radionuclides contained in the waste. Illite is considered to be the most likely alteration product of montmorillonite (e.g. Japan Nuclear Development Institute (JNC), 2000;NAGRA, 2002;SKB, 2011). The difference in diffusion and sorption behaviours of radionuclides due to the illite generated in compacted bentonite is, therefore, one of the issues to be investigated in terms of the safety assessment (Ohnuki et al., 1994;Ahn et al., 1995).
Illite is known to have a large sorption capacity for Cs due to the presence of the so-called Frayed Edge Sites (FES), which are considered to be located at the edge of the illite particle (e.g. Sawhney, 1971). The affinity of the FES to Cs is known to be considerably greater than that of the sorption sites on montmorillonite. In the safety performance analysis of a geological disposal, 135 Cs is one of the dominant radionuclides in the biosphere (JNC, 2000). The increase in illite content can, therefore, be expected to have a positive impact on the performance of the EBS, unless the extent of alteration of montmorillonite is large enough to affect significantly the permeability and sorption capacity of the compacted bentonite. On the other hand, the diffusivities of Na + , Sr 2+ , Co 2+ and Zn 2+ have been reported to be enhanced in compacted illite (Glaus et al., 2010(Glaus et al., , 2015a. For Cs, enhanced diffusivities have been obtained in clay rocks (Van Loon et al., 2004;Melkior et al., 2005;Wersin et al., 2008) and in compacted bentonite (Muurinen et al., 1985;Kim et al., 1993;Suzuki et al., 2007;Sawaguchi et al., 2013). This phenomenon is often referred to as the surface diffusion effect, which has been explained by the increase in mobile cations concentration due to the sorption of cations on the sites to which sorbed cations are not fixed. It is possible, therefore, that the diffusivity of Cs increases with increasing sorption of Cs in compacted bentonite. When taking illite content into account in the safety assessment, the possibility of the surface diffusion of Cs sorbed on illite needs to be investigated.
In the present study, the diffusion and retention behaviour of Cs in compacted montmorillonite containing illite was investigated by through-diffusion experiments. The dominant sorption sites of Cs in compacted montmorillonite samples were controlled by adjusting the amount of illite added and the concentration of Cs in the compacted montmorillonite samples. From the experimental results, the possibility of surface diffusion of Cs sorbed on illite contained in compacted montmorillonite was discussed.

Materials
Kunipia F (supplied by Kunimine Industry Co. Ltd., Japan), a commercially produced Na-montmorillonite purified from a crude bentonite Kunigel V1, was used as the sample montmorillonite. Kunipia F consists of 98% montmorillonite and small amounts of quartz and calcite. Mica minerals were not detected in either Kunipia F or Kunigel V1 (Ito et al., 1993 EF.110-2) was used as the illite sample in the present study. The purified illite was obtained by elutriation from a shale rock containing 85% illite (Rochester, New York, USA) and its grain size was adjusted to <250 μm. The illite was added to the montmorillonite sample at the ratios listed in Table 1. The montmorillonite samples containing illite were compacted in the diffusion cell to a dry density of 1.2 mg/m 3 . Quartz sand was added to obtain the same dry density without adjusting the montmorillonite ratio, because the dry density affected significantly the sorption and diffusion behaviour of Cs in compacted montmorillonite (Sato et al., 1992;Molera & Eriksen, 2002). The grain size of the quartz sand was <800 μm.

Experimental procedure
The diffusion cell used in the present study is shown in Fig. 1a (Suzuki et al., 2004). The montmorillonite sample was compacted into the cylindrical space in a 5 mm-thick sample plate. The plate was held by two filter holders with the hole where filters were placed. A porous filter with a pore size of 70 μm and a thickness of 2 mm made from polypropylene filter plate (Flon Filters, EMD Millipore), was placed between the porous filter and the compacted montmorillonite sample to prevent leaching of montmorillonite. Tubes with 1 mm internal diameter made from polytetrafluoroethylene (PTFE) were connected to the filter holder. The solution in a reservoir was circulated through the porous filter via the tubes. After the montmorillonite sample was compacted into the diffusion cell, the cell was immersed in a 0.5 mol/dm 3 NaCl solution for 1 day under low pressure without connection of the PTFE tubes. The diffusion cell was then connected to a reservoir as illustrated in Fig. 1b by the aforementioned PTFE tubes and a peristaltic pump. A flexible tube (PharMed® BPT) was used as the liquid feeding part of the peristaltic pump. To complete the saturation of compacted montmorillonite sample, 0.5 mol/dm 3 of NaCl solution was circulated from the reservoir through the diffusion cell for >1 month. The volume of 0.5 mol/dm 3 of NaCl in the reservoir was 100 mL.
In through-diffusion experiments, the diffusion cell was connected to two reservoirs (Fig. 1c). A tracer was spiked in the solution in the larger reservoir (denoted as high-concentration solution). The solution in the smaller reservoir (denoted as low-concentration solution), into which the tracer diffuses from the high-concentration solution through the compacted montmorillonite sample, was replaced periodically to keep the tracer concentration sufficiently low compared to that in the high-concentration solution. The volume of the high-concentration solution was 1 dm 3 . The volume of the low-concentration solution varied from 30 to 250 mL depending on the increasing rate of the tracer concentration in the solution.
For samples C00L and C05L (Table 1), the diffusion cell was connected to the reservoirs containing 0.5 mol/dm 3 NaCl (Fig. 1c). 137 Cs in the form of CsCl was then spiked into the high-concentration solution at a total Cs concentration of ∼7 × 10 -8 mol/ dm 3 including non-radioactive Cs used as a carrier. The low-concentration solution was replaced by a new 0.5 mol/dm 3 NaCl solution at intervals of 1 to 7 days. The activity of 137 Cs in the low-concentration solution replaced was measured to obtain the flux of Cs from the compacted montmorillonite sample. An aliquot of the high-concentration solution was taken once per week to monitor the Cs concentration in the high-concentration solution by measuring 137 Cs activity. The 137 Cs activity was obtained from the measurement of the activity of 137m Ba, the daughter radionuclide of 137 Cs, by γ-ray spectrometry. At the end of the experiment, the compacted montmorillonite sample was sliced into sections at a thickness of ∼0.4 mm to obtain the concentration profile of Cs in the sample. The sliced sections were wrapped in a sheet of paper and enclosed in a polyethylene bag. The 137m Ba γ-ray from the wrapped and enclosed sliced sections was measured directly by γ-ray spectrometry. For samples C00H, C05H and C50H, the reservoir connected for saturation with 0.5 mol/dm 3 NaCl ( Fig. 1b) was placed in a new reservoir filled with 100 mL of 0.5 mol/dm 3 NaCl solution containing 1 × 10 -4 mol/dm 3 non-radioactive CsCl. The solution was replaced once per week until the Cs concentration in the solution was constant, indicating that the sorption of Cs in the compacted montmorillonite sample had reached equilibrium. The Cs concentration was measured by inductively coupled plasma mass spectrometry (ICP-MS; Perkin Elmer, NexION300X). After the sorption equilibrium was reached, the diffusion cell was disconnected from the reservoir and connected to the reservoirs (Fig. 1c). The highconcentration and low-concentration reservoirs were filled with 0.5 mol/dm 3 NaCl solution containing 1 × 10 -4 mol/dm 3 non-radioactive CsCl. 137 Cs tracer was then spiked into the high-concentration solution. The increase in Cs concentration was ∼1 × 10 -8 mol/dm 3 through the addition of the 137 Cs tracer. The total Cs concentration in the high-concentration solution was considered to be constant. The low-concentration solution was replaced by a new 0.5 mol/dm 3 NaCl solution containing 1 × 10 -4 mol/dm 3 non-radioactive CsCl at intervals of 1 to 7 days. At the end of the experiment, the compacted montmorillonite sample was sliced to obtain the profile of 137 Cs in the sample. The same procedure for measurement of 137 Cs activity as for the samples C00L and C05L was applied to samples C00H, C05H and C50H.
As for samples D00, D05 and D50, the throughdiffusion experiment of deuterated water (HDO) was carried out after saturation with 0.5 mol/dm 3 NaCl solution. HDO tracer was spiked by adding 50 mL of 0.5 mol/dm 3 NaCl solution prepared in D 2 O into the high-concentration solution after the same amount of 0.5 mol/dm 3 NaCl solution was removed from the reservoir. The concentration of HDO was measured by a Fourier transform infrared spectrometer (FTIR: Thermo Fisher Scientific, Nicolet 6700) with attenuated total reflectance (ATR) spectroscopy (Suzuki et al., 2004). As the concentration of HDO in the compacted montmorillonite samples was insufficient for measurement by FTIR, the concentration profile was not obtained.

Analytical method
The effective diffusion coefficient (D e ) was calculated from the flux and the tracer concentration in the high-concentration solution at steady state based on Fick's first law by the following equations: where J is the tracer flux, L is the thickness of compacted montmorillonite sample, ΔC p is the difference of tracer concentration in pore water between each end of the compacted montmorillonite sample, C H is the tracer concentration in the high-concentration solution and C L is the tracer concentration in the lowconcentration solution.
In through-diffusion experiments, the tracer diffusion in filters may dominate the total tracer flux depending on the tracer diffusivity in the filters and the thickness of the filters (Glaus et al., 2008;Aertsens et al., 2012;Glaus et al., 2015b). In order to avoid the influence of filters, a porous filter and a membrane filter with high permeability were used in the present study. The solution from a reservoir was circulated through the porous filter. As the solution was flowing continuously through the porous filter, the tracer concentration was considered to be uniform. The solution is considered to be easily exchanged between the porous filter and the membrane filter due to the high permeability of the membrane filter. The tracer concentration in the pore water in the compacted montmorillonite sample at the boundary in contact with the membrane filter was, therefore, assumed to be the same as that in the solution in the reservoir.
During the experiment, the concentration of Cs in the low-concentration solution was increased to ∼7% of the concentration in the high-concentration solution. The concentration of Cs in the low-concentration solution just after the replacement of solution was ∼0.5% of the concentration in the high-concentration solution, which was calculated from the amount of Cs left in the tubes and filters. In the calculation of D e , the intermediate value between the maximum concentration and the minimum concentration was used as the concentration of Cs in the low-concentration solution, C L .
The surface diffusion effect is often characterized by the greater diffusivity of cations than that of electrically neutral species caused by the diffusion of mobile cations sorbed on the surface of clay minerals (Van Loon et al., 2004;Melkior et al., 2005;Suzuki et al., 2007;Wersin et al., 2008;Sawaguchi et al., 2013;Glaus et al., 2015a). In the present study, the flux of Cs attributed to the surface diffusion, J Cs SD , was defined by the following equations: where J Cs PW is the flux of Cs in pore water, J Cs Total is the total flux of Cs diffusing into the low-concentration solution, D Cs 0 is the self-diffusion coefficient of Cs + in water (2.057 × 10 -9 m 2 /s), D Water 0 is the self-diffusion coefficient of water (2.236 × 10 -9 m 2 /s) and D HDO e is the effective diffusion coefficient of HDO obtained in the present study. The J Cs PW was defined as the fraction of flux derived from the diffusion of Cs under the assumption that the Cs tracer diffused in the compacted montmorillonite sample without sorption and interaction with solid phase. In the calculation of J Cs PW , the value of D HDO e obtained was converted to the D e of Cs in pore water using D Cs 0 and D Water 0 under the assumption that the geometric pathway was the same between the diffusion of HDO and Cs in the compacted montmorillonite sample. As the difference in grain size between quartz sand and illite might affect the diffusivity of HDO in the compacted montmorillonite sample, D HDO e was obtained for each sample with different illite content.
The distribution coefficient (K d ) was obtained from two measurements: the concentration profile obtained by slicing the compacted montmorillonite sample into sections at the end of the through-diffusion experiment and the amount of Cs retained in compacted montmorillonite sample in the saturation procedure with 0.5 mol/dm 3 NaCl solution containing 1 × 10 -4 mol/dm 3 non-radioactive CsCl. The K d from the concentration profile was calculated by the following equations: where K Pro d is the distribution coefficient obtained from the concentration profile, C p is the concentration of 137 Cs in pore water, C b is the concentration of 137 Cs per unit volume of the compacted montmorillonite sample, ρ is the dry density of the compacted montmorillonite sample, ε is the porosity of the compacted montmorillonite sample and x is the distance from the boundary between the compacted montmorillonite sample and the filter on the high-concentration side. ε was calculated from the equation ε = ρ/ρ clay , where ρ clay is the mineral density, 0.59 for the compacted montmorillonite sample containing 50% illite and 0.57 for the other compacted montmorillonite samples. ρ clay was assumed to be 2.9 mg/m 3 for montmorillonite and illite, and 2.7 mg/m 3 for quartz sand. In equation 6, the C p was assumed to decrease proportionally with increasing distance from the boundary in contact with the high-concentration solution. The C p at the boundaries in contact with the high-concentration solution (x = 0) and the low-concentration solution (x = L) were assumed to be the same for the C H and C L , respectively. The concentration in the low-concentration solution at the end of the experiment was used as the C L value in the calculation.
For samples C00H, C05H and C50H, the K d was calculated from the amount of Cs retained in the compacted montmorillonite sample obtained in the saturation procedure with 0.5 mol/dm 3 NaCl solution containing 1 × 10 -4 mol/dm 3 non-radioactive CsCl in addition to the K d from the concentration profile. The K d from the saturation procedure was calculated from the following equation: where K Sat d is the distribution coefficient obtained from the saturation procedure, Q Sat Cs is the amount of Cs retained in the compacted montmorillonite sample, m is the dry weight of compacted montmorillonite sample and C Sat p is the concentration of Cs in the solution used for saturation (1 × 10 -4 mol/dm 3 ). In the saturation procedure, the solution for saturation was replaced repeatedly until the Cs concentration was constant.  137 Cs was used as the tracer for C00L and C05L (total Cs concentration <7 × 10 -8 mol/dm 3 ) and for C00H, C05H and C50H (under saturation with 1 × 10 -4 mol/dm 3 non-radioactive Cs), while HDO was used for D00, D05 and D50. All experiments were carried out with 0.5 mol/dm 3 NaCl.
in the experiments for a Cs tracer concentration of <7 × 10 -8 mol/dm 3 . On the other hand, steady state was considered to have been reached within 30 days for the samples saturated with 0.5 mol/dm 3 of NaCl solution containing 1 × 10 -4 mol/dm 3 non-radioactive CsCl (samples C00H, C05H and C50H). The fluxes of HDO reached steady state within 5 days for all samples (samples D00, D05 and D50). The D e values calculated from equation 1 are summarized in Table 2 together with the total fluxes of Cs and the fluxes of Cs attributed to the surface diffusion calculated from the values for D e of HDO obtained from equation 3. The values for D e of Cs obtained by throughdiffusion experiments were 7.2-8.0 × 10 -10 m 2 /s in simulated sea water (Suzuki et al., 2007) and 7.8 × 10 -10 m 2 /s at 0.5 mol/dm 3 NaCl (Sawaguchi et al., 2013). These experiments were carried out for a mixture of quartz sand and Kunigel V1 with a mass fraction of 3:7 compacted to a dry density of 1.6 mg/m 3 . For the D e of HDO and HTO, the dependence of D e on porosity was described as D e ¼ 0:471 4:3 Á D Water 0 for compacted Kunipia F when the direction of diffusion was normal to the direction of compaction (Suzuki et al., 2004). Based on this equation, the D e for HDO was calculated as ∼1.1 × 10 -10 m 2 /s under the sample conditions used in the present study. Considering the difference in terms of bentonite properties such as the dry density and smectite content, the D e values of Cs and HDO obtained here were comparable to the values reported.
The amount of Cs sorbed in the compacted montmorillonite samples obtained from the saturation procedure with 0.5 mol/dm 3 NaCl solution, containing 1 × 10 -4 mol/dm 3 non-radioactive CsCl is illustrated in Fig. 3. The sorption of Cs reached equilibrium after seven rounds of solution replacement for all samples. The K d values obtained from the saturation procedure, K Sat d , calculated by equation 7 are summarized in Table 2.
0.5 mol/dm 3 NaCl solution containing 1 × 10 -4 mol/dm 3 non-radioactive CsCl, the concentration profiles of 137 Cs showed almost the same trend except for the sample containing 50% illite. These results indicate that the illite content contributed to the high concentration of 137 Cs in the compacted montmorillonite samples.
The profiles of K d in compacted montmorillonite samples, K Pro d , calculated from equation 5 are shown in Fig. 5. The K Pro d increased slightly near both ends of compacted montmorillonite samples. As for the diffusion experiment using compacted montmorillonite, heterogeneous density distribution has been indicated (Glaus et al., 2011). The density of compacted montmorillonite near both ends has been reported to be less than that at the middle even if the saturation is completed. The K d of Cs in compacted bentonite has been reported to increase with decreasing dry density in some cases (Oscarson et al., 1994). The increases in K Pro d near both ends of the compacted montmorillonite samples can be interpreted as the increases in Cs sorption capacity caused by the decrease in the density of montmorillonite samples. The K Pro d values are summarized in Table 2.

D I S C U S S I O N
In the present study, ∼65% of the total flux of Cs, J Cs Total , was attributed to the flux caused by surface diffusion, J Cs SD , for the compacted montmorillonite samples C00L and C00H, which were illite-free. The J Cs SD /C H values for samples C00L and C00H were 7.3 × 10 -8 and 6.9 × 10 -8 m/s, respectively (Table 2). These values were almost equivalent to the J Cs SD /C H values for the samples containing illite (samples C05L, C05H and C50H). Although the J Cs SD /C H value of 5.0 × 10 -8 m/s for the sample C05L was less than the other values, the difference was considered to be small given the large experimental error. This result indicates that the capacity of the sorption sites on which Cs was sorbed as a mobile cation was the same in the compacted montmorillonite samples used in this study. The J Cs SD is, therefore, considered to be caused by the surface diffusion of Cs sorbed on the sorption sites on montmorillonite. The illite content in the compacted montmorillonite samples scarcely contributed to the J Cs SD . On the other hand, the K d values differed among the compacted montmorillonite samples depending on the illite content and the Cs concentration. As listed in Table 2, the K Pro d value for the sample containing 5% illite obtained at a Cs tracer concentration of <7 × 10 -8 mol/dm 3 (sample C05L) was about one order of magnitude greater than that for the sample without illite (sample C00L). The K Pro d and K Sat d values of 0.036-0.038 m 3 /kg were obtained for the samples without illite and containing 5% illite saturated with 0.5 mol/dm 3 NaCl solution containing 1 × 10 -4 mol/dm 3 non-radioactive CsCl (samples C00H and C05H). Compared to these values, the K Pro d and K Sat d values for the sample containing 50% illite (sample C50H) were greater. These increases in K d values might be caused by the contribution of sorption sites other than the cation exchange sites on montmorillonite. The Cs sorbed on these sites is considered to be immobile because the flux of Cs attributed to the surface diffusion was independent of the K d values.
In order to identify the sorption sites contributing to K d in the compacted montmorillonite samples, the contribution ratio of sorption sites to K d was calculated based on the previously proposed models of Cs sorption on montmorillonite and illite. For montmorillonite, most of the model studies have described the sorption of Cs as a one-site sorption on the cation exchange sites (e.g. Wanner et al., 1996). The sorption capacity of Cs on the cation exchange sites on montmorillonite can be calculated from the selectivity coefficient, Cs Na K, of cation exchange reaction, XNa + Cs + ⇔ XCs + Na + , where X indicates the cation exchange sites on montmorillonite. Only the cation exchange reaction between Na + and Cs + was considered for the calculation because the compacted montmorillonite samples were saturated with 0.5 mol/ dm 3 NaCl. For illite, two (Poinssot et al., 1999) or three (Bradbury & Baeyens, 2000) types of sorption sites have been assumed for the modelling of Cs sorption. In these studies, the FES with a high sorption affinity for Cs and a low capacity and the Type II sites with a low affinity for Cs and a high capacity were assumed as   Ito et al. (1993); b Wanner et al. (1996); c Bradbury & Baeyens (2000). sorption sites. The three-site sorption model proposed by Bradbury & Baeyens (2000) considers the planar sites in addition to the two sites above. The sorption affinity of the planar sites for Cs has been considered to be the lowest and the capacity to be the greatest. In this study, the three-site sorption model by Bradbury and Baeyens (2000) was applied to the calculation considering the possibility that the planar sites might be the dominant sorption sites contributing to K d under the experimental conditions used here. Only the exchange reaction between Na + and Cs + was considered for illite. The parameters used for the calculation are summarized in Table 3. The K d attributed to the sorption on each sorption site and the total K d calculated from the models are presented in Fig. 6 together with the K d obtained from the concentration profile, K Pro d , for each sample. The calculation results for the samples without illite, containing 5% illite and with 50% illite, respectively, are shown in Fig. 6a,b,c. In Fig. 6b,c the calculated K d values attributed to the sorption on the planar sites on illite are not shown because the K d values were smaller than the range of the vertical axis. The K d values obtained from the saturation procedure, K Sat d , are not shown in Fig. 6 because the K Sat d values were in good agreement with K Pro d . The K Pro d values were somewhat different from the calculated total K d (Fig. 6). The reason for the discrepancies was not clear. The trend of K Pro d values corresponded roughly to the curves of the total K d calculated. The aim of the model calculation is to identify the dominant sorption sites corresponding to the respective sample conditions used in this study. Although some differences were observed between the K Pro d and the K d values calculated, the model calculation is considered to be applicable.
The calculated total K d was derived entirely from the sorption on the cation exchange sites on montmorillonite as illite was not present in the sample (Fig. 6a). The dashed line of K d attributed to the sorption on montmorillonite lies behind the solid line of total K d in Fig. 6a. For the sample containing 5% illite, the dominant sorption sites for K d at a Cs tracer concentration of 7 × 10 -8 mol/dm 3 were calculated to be due to the FES which accounted for 76% of the sorption amount of Cs (Fig. 6b). For the samples saturated with 0.5 mol/dm 3 NaCl solution containing 1 × 10 -4 mol/dm 3 non-radioactive CsCl, the dominant sorption sites were the cation exchange sites on montmorillonite for the sample containing 5% illite (the condition for the sample C05H), while the type II sites for the sample containing 50% illite (the condition for the sample C50H) (Fig. 6c). The dominant sorption sites are, thus, identified as the cation exchange sites on montmorillonite for samples C00L, C00H and C05H, the FES for C05L and the Type II sites for C50H from the model calculations.
For the compacted montmorillonite samples containing 5% illite, the K Pro d of sample C05L, which was obtained at a Cs tracer concentration of <7 × 10 -8 mol/dm 3 , was more than ten times larger than that of the sample saturated with 1 × 10 -4 mol/dm 3 non-radioactive CsCl (sample C05H). This increase in K d can be interpreted as the contribution of the FES. For samples under saturation with 1 × 10 -4 mol/dm 3 FIG. 6. K d values attributed to the sorption at each sorption site and the total K d calculated from the models for the samples without illite (a), containing 5% illite (b), and containing 50% illite (c). Mt = montmorillonite.
non-radioactive CsCl, the greater K Pro d and K Sat d values in compacted montmorillonite samples containing 50% illite (sample C50H) was observed compared to the sample containing 5% illite (sample C05H). The greater K Pro d and K Sat d values can be interpreted as the contribution of the Type II sites. In this case, the K Pro d and K Sat d values of the sample C05H were equivalent to those of the sample without illite (sample C00H) because the increase in K d caused by the contribution of the Type II sites is considered to be small in sample C05H. The increase in K Pro d and K Sat d caused by the addition of 50% illite were calculated to be 0.024 and 0.027 m 3 /kg, respectively, comparing the K Pro d and K Sat d values between samples C50H and C00H. The increase in K d caused by the addition of 5% illite is expected to be 0.002-0.003 m 3 /kg for sample C05H. The increase in K d was, therefore, scarcely observable in sample C05H taking the experimental error into account.
As mentioned above, the flux values of Cs attributed to the surface diffusion, J Cs SD /C H , were almost the same among all of the compacted montmorillonite samples regardless of the illite content. From these results, it is concluded that Cs sorbed on the FES and Type II sites on illite was immobile or considerably less mobile than the Cs sorbed on the cation exchange sites on montmorillonite. As for the Type II sites, the K Pro d of sample C50H was equivalent to the K Sat d . The sorption of Cs on the Type II sites was, therefore, considered to be reversible.

C O N C L U S I O N S
In the present study, the surface diffusion of Cs sorbed on illite added to compacted montmorillonite was investigated by through-diffusion experiments. The experiments were carried out under conditions where the FES or the Type II sites on illite were the dominant sorption sites. As a result, the flux of Cs attributed to the surface diffusion was independent of the dominant sorption sites, indicating that the surface diffusion of Cs sorbed on the FES and the Type II sites on illite was negligible compared to the Cs sorbed on the cation exchange sites on montmorillonite. Consequently, the increase in illite content in compacted bentonite because of hydrothermal alteration of montmorillonite is expected to enhance the sorption capacity of Cs without increasing the diffusivity by the surface diffusion on illite surface, unless the decrease in montmorillonite content in compacted bentonite due to the alteration is so large as to affect the permeability and sorption capacity of compacted bentonite.