Sorption and diffusion of selenium oxyanions in granitic rock

Journal of Contaminant Hydrology 192 (2016) 203–211

Contents lists available at ScienceDirect

Journal of Contaminant Hydrology journal homepage: www.elsevier.com/locate/jconhyd

Sorption and diffusion of selenium oxyanions in granitic rock Jussi Ikonen a,⁎, Mikko Voutilainen a, Mervi Söderlund a, Lalli Jokelainen a, Marja Siitari-Kauppi a, Andrew Martin b a b

University of Helsinki, Laboratory of Radiochemistry, Department of Chemistry, P.O. Box 55, FIN-00014 Helsinki, Finland Nagra (National Cooperative for the Disposal of Radioactive Waste), Wettingen, Switzerland

a r t i c l e

i n f o

Article history: Received 22 March 2016 Received in revised form 16 July 2016 Accepted 4 August 2016 Available online 5 August 2016 Keywords: Selenium Selenite Selenate Diffusion Sorption TDD-modelling

a b s t r a c t The processes controlling diffusion and sorption of radionuclides have been studied extensively in the laboratory, whereas, only a few in-situ experiments have been carried out in order to study in-situ diffusion over the longterm (several years). This is largely due to the fact that in-situ experiments are typically time consuming and cost intensive, and it is commonly accepted that laboratory scale tests are well-established approaches to characterizing the properties of geological media. In order to assess the relevance of laboratory experiments, the Swiss National Cooperative for Disposal of Radioactive Waste (Nagra) have been conducting extensive experiments in the Underground Rock Laboratory (URL) at the Grimsel Test Site (GTS) in order to study radionuclide transport and retention in-situ. One of the elements used in these experiments is non-radioactive selenium, as an analog for the radiotoxic isotope Se-79, which is present in radioactive waste. In this work, two laboratory through-diffusion experiments using selenium as a tracer were carried out in block (decimeter) scale rock specimens to support one of the ongoing radionuclide transport and retention in-situ experiment at the GTS mentioned above. The though-diffusion tests of selenium were performed under atmospheric conditions in both Kuru grey granite (KGG) and Grimsel granodiorite (GG). The decrease of selenium concentration in an inlet hole drilled into each of the rock samples and the breakthrough of selenium into sampling holes drilled around the inlet were analyzed using Inductively Coupled Plasma Mass Spectrometry (ICPMS). The effective diffusion (De) and distribution coefficients (Kd) of selenium were then determined from the changes of selenium concentration in the inlet and sampling holes using a Time-Domain Diffusion (TDD) simulations. In addition, Kd of selenium was measured by batch sorption experiments as a function of pH and Se concentration in atmospheric conditions and nitrogen atmosphere. The speciation of selenium was studied by HPLCICP-MS in simulated ground waters of each of the rock types. The Kd of selenium was found to be in the range of (6.2–7.0 ± 2.0) × 10−3 m3/kg in crushed rock whereas the Kd obtained from block scale through diffusion experiment varied between (1.5 ± 0.3) × 10−3 m3/kg and (1.0 ± 0.6) × 10−4 m3/kg. The De of selenium was significantly higher for GG; De = (2.5 ± 1.5) × 10−12 m2/s than for KGG; De = (7 ± 2) × 10−13 m2/s due to the higher permeability of GG compared with KGG. © 2016 Elsevier B.V. All rights reserved.

1. Introduction The geological disposal of high-level radioactive waste and spent nuclear fuel has been studied in many countries, e.g. in Finland, Sweden, Switzerland, France etc. Host rocks considered viable have included deep-lying crystalline rock, clay formation sand salt (Witherspoon, 1996, Posiva 2013-01). The main mechanism responsible for radionuclide transport in crystalline rock is groundwater (GW) flow through water conducting fractures. The most significant processes that delay radionuclides flowing through fractures are diffusion into the rock matrix (i.e. matrix diffusion) and sorption onto the mineral surfaces (e.g.,

⁎ Corresponding author. E-mail address: jussi.ikonen@helsinki.fi (J. Ikonen).

http://dx.doi.org/10.1016/j.jconhyd.2016.08.003 0169-7722/© 2016 Elsevier B.V. All rights reserved.

Dai et al. 2007, Grisak and Pickens, 1980, Neretnieks, 1980, Séby et al. 1998). One radionuclide found in radioactive waste and spent fuel that is of particular concern for the long-term safety of radioactive waste repositories is Se-79 due to its long half-life (3.7 × 105 years) and its high mobility in the geological environment, especially in crystalline rocks (Atwood 2010; Lehto and Huo 2011). Other isotopes of selenium found in radioactive waste as a result of neutron activation of selenium includes Se-75, however, because Se-79 is produced from selenium bearing materials found in nuclear fuel and reactor construction materials, it is by far the most ubiquitous radioisotope of selenium found in radioactive waste. The mobility of Se-79 depends largely on the oxidation state of selenium; there are five possible oxidation states. Under oxic conditions, oxidation state VI dominates and selenium exists in the form of selenate

204

J. Ikonen et al. / Journal of Contaminant Hydrology 192 (2016) 203–211

(SeO2− 4 ). At lower redox potentials, oxidation state IV is found and selenium exists in the forms of selenite (SeO23 −) and biselenite (HSeO− 3 ) (Lehto and Huo 2011). The reduction of selenate to selenite occurs slowly when redox potential is lower than + 400 mV under neutral pH. Under intermediate redox potentials (approximately + 200 mV, pH ) selenium is reduced to elemental selenium. Under low redox potentials selenium forms selenide ions Se2 − at oxidations state –II (Lehto and Huo 2011). Selenate and selenite are the most common forms found in natural waters. In the soil, selenate, selenite and elemental selenium are the most prevalent forms (Atwood 2010). After being released to the host rock from a geological repository, selenium is assumed to be at oxidation states IV and VI. However thermodynamic calculations show that selenium exists also at lower oxidation states 0 and –II due to contact with the iron canisters that the waste is contained in (Altmann 2008). Selenium reduction is slow and kinetically hindered because it involves the transfer of multiple electrons along with multiple oxygen atoms between its various oxidation states (VI, IV, 0, –I, –II) (De Cannière et al. 2010, Grambow 2008, Savoye et al. 2012, Séby et al. 1998). Sorption of anionic species typically decreases as a function of pH because the mineral surfaces become negatively charged which leads to the higher mobility of selenium oxyanions in neutral and basic pH (De Cannière et al. 2010). Sorption of selenium oxyanions is known to be strong only on aluminium and iron minerals, such as goethite, hematite and pyrite, through surface complexation (Boult et al. 1998, Duc et al. 2003, Parida et al. 1997, Rovira et al. 2008). Selenate forms weakly bound outer sphere complexes while selenite forms stronger inner sphere complexes with hydrous aluminium oxides and iron oxides (De Cannière et al. 2010). Adsorption capacity of selenium on aluminium oxides is higher than oniron oxides (Chan et al. 2009, Peak 2006) and retention of selenium is controlled by reduction of selenium to lower oxidation states or by surface complexation. However, selenate ions were not reduced along their migration path in Boom Clay even though IV and –II were thermodynamically stable oxidations states. Indications of selenite reduction to 0 and –II was observed in diffusion experiments in Callovo-Oxfordian clay-stones (Beauwens et al. 2005, Descostes et al. 2008, Savoye et al. 2012). Contrary to clays the content of aluminium and iron bearing minerals is low in crystalline rock leading to slight sorption. Presence of the iron bearing minerals increase the retention significantly because they act as selective sorbent for selenium species (Videnská et al. 2013; 2015, Yllera de Llano et al. 1996). The effective diffusion coefficient (De) of highly mobile selenium oxyanions decrease with decreasing salinity due to anion exclusion that is caused by the negative surface charge of minerals (Iida et al. 2011). Distribution coefficients (Kd) of radionuclides can be determined by batch sorption and diffusion experiments. Most of them have been determined using simple crushed rock batch sorption method (Posiva 2012-41). These values are typically slightly higher than the ones from diffusion experiments (Bradbury and Stephen 1985, Tachi et al. 1998). To be able to compare results obtained from the batch and in-situ diffusion experiments reliably, block scale diffusion experiments are a useful step for upscaling Kd derived from batch sorption experiments. In the block scale diffusion experiments a measurement geometry similar to in-situ experiments is used and the samples contain less disturbances than crushed rock. However, results of block scale diffusion experiments are difficult to analyze using analytical or traditional tools since either the initial or boundary conditions are challenging to handle due to the geometry of the experiment or complex sampling procedure. Such challenges could be solved using, e.g., commercial numerical solvers of partial differential equations or continuous time random walkers (Noetinger and Estebenet, 2000, Noetinger et al. 2001, Glaus et al. 2015). One possibility to handle the challenges is to apply the Time Domain Diffusion (TDD) method which has been developed for simulating diffusion in heterogeneous media (McCarthy 1993, Delay et al. 2002, Delay and

Porel 2003).TDD is a rapid particle-tracking method that makes it possible to simulate diffusion in heterogeneous media when the local porosities and diffusion coefficients are known. In this method a particle is forced to jump to a neighboring point during a certain random transition time, which makes it faster than the traditional particle-tracking methods (Sardini et al. 2003). The TDD method has been used to study the effect of structural heterogeneities on diffusion in various cases (Sardini et al. 2007, Robinet et al. 2008, Robinet et al. 2012, Voutilainen et al. 2013). Dentz et al. (2012) have also modified the method by including the sorption of migrating element in heterogeneous media. They included a trapping frequency parameter which may be used to mimic chemical sorption and desorption processes. The idea from Dentz et al. (2012) can be applied so that the distribution and diffusion coefficients, which are typically used for geological materials, can be utilized. The method has also been found to be a powerful tool for analyzing results of both in-situ and laboratory experiments especially when initial or boundary conditions are complicated (Voutilainen et al. 2013, Soler et al. 2014, Ikonen et al. 2016). Recently the method has been developed further by including flow in fractured media (Gjetvaj et al. 2015, Noetinger et al. n.d.). The Swiss National Cooperative for Disposal of Radioactive Waste (Nagra) has been conducting extensive in-situ experiments at the Grimsel test site (GTS) in the field of radionuclide migration and retention in the rock matrix. The second Long Term Diffusion (LTD) experiment was started in spring 2014 using radionuclides H-3, Na-22, Cs134, Cl-36 and Ba-133 as well as nonradioactive selenium. The aim of this work was to study the sorption and diffusion of selenium in granitic rock at laboratory to provide data for modelling and to compare the results with the ones from the in-situ experiment. Furthermore, sorption of selenium was studied with batch sorption experiments on crushed rock in atmospheric conditions and in nitrogen atmosphere. The Kd values were measured as a function of pH and selenium concentration. Speciation of selenium was followed using HPLC-ICP-MS. The data from the block scale diffusion experiment was analyzed using TDD method that utilizes the Kd and De of selenium and the results were compared to the ones from batch sorption experiments. 2. Experimental 2.1. Materials KGG is a homogeneous, non-oriented fine grained granitic rock quarried from Kuru in the central part of Finland. The main minerals present are plagioclase (21%), quartz (35%), potassium feldspar (36%) and hornblende + biotite (8%). Minor assemblages are muscovite, chlorite, sericite, zircon, apatite and opaque minerals. The porosity of KGG is 0.47% (Jokelainen et al. 2009) and permeability (3 ± 1) × 10− 18 m2 (Kuva 2016).GG is a homogeneous, medium grained and slightly foliated granodiorite. The main minerals are plagioclase (37%), quartz (33%), potassium feldspar (17%) and biotite (6%). The remaining minerals, which do not exceed 5% in volume, are green amphibole (hornblende), muscovite, epidote, titanite and opaque minerals; most probably iron sulphides like pyrite (Jokelainen et al. 2013). The porosity of GG is 0.65% (Kelokaski et al., 2006) and permeability (1.3 ± 0.3) × 10−17 m2 (Kuva 2016). The selenium used as a tracer in all experiments was stable Se (Romil) in oxidation state + IV with a concentration of 1000 mg/L in 0.5 M nitric acid. This solution was added into the two different GW simulants so that the concentrations in the inlet hole in the blocks were the same as the initial concentration of selenium in the in-situ experiment at 50 mg/L. This high concentration was chosen to make sure that detectable levels of the tracer reached the sampling holes. Selenium concentration in the GW from the Grimsel LTD experiment site was measured to be below 15 μg/L (Giroud 2016). Table 1 presents the elemental composition of the two GW simulants that were used in the experiments. In the case of GG

J. Ikonen et al. / Journal of Contaminant Hydrology 192 (2016) 203–211

205

Table 1 The elemental composition of GW simulants used in the block diffusion and batch sorption experiments.

Grimsel Allard OX

pH⁎

pH ⁎⁎

Na+ (mg/L)

K+ (mg/L)

Ca2+ (mg/L)

Mg2+ (mg/L)

F− (mg/L)

Cl− (mg/L)

NO− 3 (mg/L)

HCO–3 (mg/L)

SO2− (mg/L) 4

⁎⁎⁎ (mg/L) SiO23

9.6 8.4

7.2 8.2

25.7 57.5

0.3 3.9

5.6 18.0

– 4.4

7.0 –

6.0 42.5

17.4 –

27.5 122.0

5.8 9.6

19.0 18.3

⁎ At the beginning of the experiment. ⁎⁎ At the end of the experiment. ⁎⁎⁎ SiO3 was added in form of Na2SiO3·H2O.

experiments, the simulant was prepared according to Mäder et al. (2006) which represents the fracture water of the Grimsel test site. Modified Allard GW (Allard and Beal 1979) was used in the case of KGG experiments. Allard GW simulant is average GW of some Swedish moderately saline GWs. It has been commonly used in the experiments in Sweden and Finland. Diffusion experiments were carried out under atmospheric conditions. Dissolving of carbon dioxide from air caused the pH to stabilize to levels presented in Table 1.

scale and in-situ diffusion experiments. These concentrations were chosen because this range was estimated to being the matrix pore water in the sampling holes after several years of diffusion. In the Kuru block the selenium concentration in the sampling hole at a depth of 1 cm from the inlet hole rose to the lower end and in Grimsel to the higher end of the chosen range. Distribution coefficient Kd of selenium was determined as follows:

2.2. Batch sorption experiments

Kd ¼

Batch sorption experiments were conducted using crushed rock of grain sizes below 0.3 mm. The ratio of solid to solution was 100 g/L mixed in 20 mL PolyEthylene bottles. The initial pH of solution was adjusted with HCl and NaOH. The crushed rock samples were left to stabilize for 14 days. Grimsel rock samples were saturated with Grimsel and Kuru rock samples with Allard GW simulant. Duplicate samples of each experiment were produced for both rock types. After stabilization selenium was added and samples were left in the mixer for 14 days. The samples were then centrifuged and aliquots of sample solutions were filtered with 0.45 μm polypropylene membrane filters for analysis. Selenium concentration in 0.5 M HNO3 solution was measured with ICP-MS Agilent 7500ce with Octopole reaction system using isotope Se-82. Quantification of selenium was done using external calibration. During the measurements an internal germanium standard was introduced through an autosampler for stability correction. The detection limit was 0.6 ppb from pure solution with helium as reaction gas. Sorption experiments were carried out under atmospheric conditions and in a glove box under nitrogen atmosphere to study the effect of low oxic environment on selenium sorption. Eh measured with silver chloride electrode (Orion Redox/ORP Electrode 9778BNWP) was +150 mV. Batch experiments were performed with three different selenium concentrations (0.04, 0.4 and 4 mg/L) and in three different pH (7, 8 and 9), and measured with Hannah Instruments HI1331 pH electrode. The pH range covered the possible pH changes in the block-

c0 −c V  c m

ð1Þ

where c0 is initial concentration of selenium in the solution and c the final concentration of selenium in the solution. V is the volume of solution and m the mass of the solid phase. With increasing distribution coefficient a larger portion of nuclide is sorbed on the solid phase. 2.3. Diffusion experiments Diffusion experiments were carried out in two rock samples. The KGG rock sample was a 30 × 30 × 30 cm3 cube with inlet hole in the centre and nine sampling holes around it (see Fig. 1). Sampling holes A1, A2 and A3 were 4 cm, 3.6 cm and 1 cm distance from the inlet hole respectively. Sampling holes T1–T6 were about 6.6–6.8 cm distance from the inlet hole (see Table 2). The GG rock sample was 30 cm diameter core 20 cm in length with two inlet holes (A and B) and five sampling holes (See Fig. 1). Sampling holes A1 and A2 were 4 cm and 3.6 cm distance from inlet hole A and B1, B2 and B3 1 cm, 2 cm and 3 cm distance from inlet hole B (see Table 2). B1 was to the direction of the rock foliation and A1 against it. In Table 2 volumes of each inlet and outlet holes as well as the volumes of water samples taken from them are shown. The water volumes in the inlet and sampling holes was monitored and adjusted during the diffusion experiment. This was necessary especially in the GG where the saturation of the whole rock block was difficult to maintain due to its high permeability.

Fig. 1. The rock samples from KGG (left) and GG (right) used in the block-scale diffusion experiments with boreholes drilled for the though diffusion tests.

206

J. Ikonen et al. / Journal of Contaminant Hydrology 192 (2016) 203–211

Table 2 Total volumes of inlet and sampling holes of KGG and GG blocks and volumes of water samples taken from the holes in both diffusion experiments. KGG sample

GG sample

Hole

Volume of hole (mL)

Volume of sample (mL)

Distance from inlet hole (cm)

Hole

Volume of hole (mL)

Volume of sample (mL)

Distance from inlet hole (cm)

A A1 A2 A3 T1 T2 T3 T4 T5 T6

105 15 15 15 42 42 42 42 42 42

0.1 10 10 10 10 10 10 10 10 10

– 4 3.6 1 6.6 6.8 6.9 6.9 6.6 6.6

A A1 A2 B B1 B2 B3

35 7 7 35 7 7 7

0.1 7 7 0.1 7 7 7

– 1 2 – 1 2 3

Block scale diffusion experiments were carried out under atmospheric conditions. The rock samples were placed in the plastic containers with about 5 cm of deionised water at the bottom to keep the rocks saturated. Before selenium injection, inlet and sampling holes were saturated for three months. Selenium concentration in the inlet holes was 50 mg/L. The inlet hole of KGG was filled with 105 mL of modified Allard GW simulant. Sampling holes of KGG were filled with of GW simulant; 15 mL into A1–A3 and 42 mL into T1–T6 (see Table 2). In GG, 35 mL of Grimsel GW simulant was placed into inlet holes A and B and 7 mL into sampling holes A1–A2 and B1–B3. Sodium azide (5%) was added into inlet holes (525 μL in KGG and 175 μL in GG) to prevent microbial growth. Sampling and analysis of selenium concentration in the inlet holes and sampling holes was performed every four weeks in Kuru and three weeks in Grimsel. From the inlet holes, 100 μL samples were taken and diluted (ratio 1:100) into 0.5 M suprapure Nitric acid (HNO3). In KGG, 10 mL samples from the sampling holes were taken. In GG, all the water in sampling holes was changed. After the sampling the solutions were filtered with 0.45 μm polypropylene filters and strong suprapure HNO3 (Romil) was added until a nitric acid concentration of 0.5 M was reached before performing the ICP-MS analysis. 2.4. Speciation of selenium High-performance liquid chromatography connected to an inductively coupled plasma mass spectrometry (HPLC-ICP-MS; Agilent 1260 Infinity and Agilent 7500 ce) was used in the speciation analysis of selenium. Dionex AG11 guard column and AS11 anion exchange column were used in the separation between the anionic selenium species and were attached to the HPLC instrument. ICP-MS was used as a concentration detector and the isotope of selenium quantified was Se-82. The total measurement time applied was 8.00 min, during which peaks of selenite and selenite were detected at 3.47 min and 5.55 min respectively. 10 mM sodium hydroxide (NaOH) was used as an eluent with a flow rate of 0.8 mL/min. The eluent was bubbled with argon gas throughout the measurements to remove dissolved carbon dioxide from the solution. 2.5. Time Domain Diffusion (TDD) simulations It is not possible to analyze the results of the diffusion experiments using analytical tools due to complicated geometries of the experimental systems and large sample volumes taken from sampling holes. We decided to simulate the diffusion of selenium using the TDD method as this approach can handle heterogeneous geometries and concentrations. The approach is also generally convenient when dealing with large systems since computation times do not considerably depend on sample size, and it is faster than more traditional simulation methods (Sardini et al. 2003). A more detailed description of the theory is given

by Delay et al. (2002); Delay and Porel (2003), Sardini et al. (2003), and Voutilainen (2012) so only a brief outline is given here. A diffusing molecule is defined as a particle which is forced to jump during a “transition time” to one of its neighboring voxel on a gridded mesh defining the diffusion space. This is not the case in the traditional methods in which a particle may do numerous tries before a successful jump. In the TDD method a particle in voxel i jumps to neighboring voxel j with the transition probability P ij ¼

bij ∑ j bij

ð2Þ

where bij ¼


 Aij εDap ij Lij

ð3Þ

with Aij the total area between voxels i and j, (ɛDp)ij the harmonic mean of the product of porosity (ɛ) and pore diffusion coefficient (Dp) at voxels i and j, and Lij the distance between the centers of voxels i and j. Note that for an impermeable voxel (ɛ = 0 of Dj = 0) bij becomes to zero, which leads to vanishing transition probability. The transition probability of a jump by a particle at a site i is taken into account such that, reaching the target site takes certain time, the transition time, which depends on the properties of the two voxels. The transition time for a jump from voxel i to voxel j is given by t i→ j ¼ −

εi V i logðu01 Þ; ∑ j bij

ð4Þ

where Vi is the total volume of voxel i and u01 a random number from a uniform distribution between 0 and 1. Eqs. (2) and (3) show that a particle jumps more likely to a voxel whose porosity and diffusion coefficient are high. Also, the mean transition time of a particle is shorter to a target voxel whose porosity and diffusion coefficient are higher (see Eq. 4). Dentz et al. (2012) modified the TDD method so that sorption and desorption of particles can be taken into account. They proposed that transition time between two positions for sorbing element (Ti → j) is simply the sum of times during which the particle is mobile (ti → j) and sorbed (ts). They also introduced a trapping frequency and distributions for trapping time and number of trappings during ti → j which could be used to mimic chemical sorption and desorption. However, the trapping frequencies and distributions are difficult to measure for natural materials. Thus we propose a form for sorption time during the transition: ts ¼ −

" # εi V i εV logðu01 Þ− − i i logðu01 Þ ; ∑ j wij ∑ j bij

ð5Þ

where wij ¼

Aij ðεDa Þij Lij

ð6Þ

with Da the apparent diffusion coefficient. In Da the effect of sorption is taken into account Da ¼

εDp De ¼ ; ε þ ρK d ε þ ρK d

ð7Þ

where ρ is the density of rock and De the effective diffusion coefficient. Note that Da is equal to Dp when Kd is equal to zero and thus the first term in Eq. 5) is equal to the second term if Kd is zero in voxel i and all of its neighboring voxels. Note also that wij cannot be used instead of bij when determining the transition probabilities since it leads

J. Ikonen et al. / Journal of Contaminant Hydrology 192 (2016) 203–211

to particles avoiding voxels where they can be sorbed to. In reality it has been seen that in rock with heterogeneous sorption properties the elements are found more likely from places where the sorption takes place (Jokelainen et al., 2013). In the simulations, 3D grey scale images were used as the simulation geometries, in which voxels formed a simulation grid. These grids were created according to actual dimensions of the rock samples so that one voxel represents (1 × 1 × 1) mm3 sized cube. This lead led to simulation grids whose sizes were 300 × 300 × 183 (KGG) and 295 × 295 × 184 (GG) voxels (see Fig. 2). In the 3D image each component (rock, inlet holes and sampling holes) were colored with a unique grey-scale value which was linked to its porosity and diffusion coefficient. Unique grey values were also used to assist the sampling during the simulation by checking the grey value of voxel in which the particle is located after experimental sampling time has been reached (see steps 3 and 4 in simulation procedure below). The simulations were done using 500,000 particles for GG and 2,000,000 in KGG because of the different Kd and effective diffusion coefficient. The initial position for each particle was chosen randomly in the inlet holes (KGG: hole A, GG: holes A and B). The porosity of rock was set to 0.47% in the KGG sample and 0.65% in the GG sample according to porosity measurements performed previously. The porosity, diffusion coefficient and distribution coefficient in the inlet and sampling holes were set to 100%, 9.46 × 10− 10 m2/s (Augustithis 1983), and 0 m3/kg, respectively. The simulations were performed using effective diffusion coefficients from 1.0 × 10−13 to 8.0 × 10−12 m2/s and distribution coefficients from 6.0 × 10−5 to 4.0 × 10−3 m3/kg. During the simulation following procedure was repeated until a preset total diffusion time was elapsed: 1. Determine direction of jump according transition probabilities (see Eq. 1) and a selected random number. 2. Determine transition time according Eq. 3 and a second random number. 3. Check if experimental sampling time has passed during the jump. If not, return to point 1. 4. Check if the particle is in one of the holes. If not, return to point 1. 5. Record the location of particle and remove part of the particles according third random number. The ratio of removed particles is set according to the sample volume and total volume of the hole in which the particle is currently located. 6. Check if total diffusion time is elapsed. If not return to point 1. The fitting of simulated curves to experimental curves was performed by determining the sum of squared deviation for each parameter pair and by finding a minimum of the sum (least squared approach). The confidence limits for simulated curves were defined so that the measured curves remained within the limits and that concentration changes in inlet and outlet holes could be explained using similar values for De and Kd.

207

3. Results and discussion 3.1. Speciation results The initial chemical form of the added selenium was selenite based on HPLC-ICP-MS analysis. At the end of the five years diffusion experiment, 3.7 ± 0.2% of selenite was oxidized to selenite in the water in the inlet hole of KGG. However significantly higher amounts of selenite was oxidized to selenate in the GG rock block during three years; 16.3 ± 0.1% of selenium was determined to be in form of selenate in the inlet hole water of GG. Eh typically ranges between 500 and 600 mV in waters which are in equilibrium with atmospheric oxygen (Mohamed and Antia 1998) and it is expected that selenate is a dominating form (Lehto and Huo 2011). The difference between the amount of selenate in GG and KGG at the end of the diffusion experiments cannot be explained with the mineral composition because they are quite similar. The differences of the GW simulants are more significant (pH, salinity, carbonate concentration). In the GG block the basic conditions at the beginning of the experiment favor selenate which is observed as the predominate species in the inlet hole. The carbonate concentration is three times higher in Allard GW than in Grimsel GW. In Grimsel GW the carbonate concentration is not in equilibrium with the atmospheric CO2 at the beginning of the experiment and the equilibration leads to pH decrease from 9.6 into 7.2 during three years whereas in KGG the pH remains at about 8. Speciation was measured only in samples from hole A3 in KGG where 28.6 ± 1.7% of selenium was found to be in the form of selenate which is about eight times more than in the inlet hole. Sorption of selenite is known to be higher than that of selenate (De Cannière et al. 2010) which leads to faster migration of selenate in the rock matrix. This leads to accumulation of selenate into the sampling hole A3.

3.2. Sorption results Kd values of selenium varied from (1.6 ± 0.3) × 10− 2 m3/kg to (3.0 ± 1.0) × 10−3 m3/kg for KGG and from (1.3 ± 0.3) × 10−2 m3/kg to (3.0 ± 1.0) × 10−4 m3/kg for GG. There was no significant difference, probably because of the minerology of these two rocks are similar, even though Allard GW is more saline than Grimsel GW. However, there was a slight decrease in Kd values of selenium over the studied pH range (Fig. 3). pH can have an influence on selenium sorption especially under basic conditions where the mineral surfaces might become negatively charged resulting to less sorption (Bray et al. 2014). All in all the sorption remained relatively low. As can be seen in Fig. 3, selenium concentration affects slightly to the Kd results. Highest concentration (4 mg/L) used showed slightly higher values for Kd than lower concentrations. This difference is more significant with a neutral pH, but under basic conditions the effect of selenium concentration becomes almost negligible. The mechanism involved here is as yet unknown. In the experiments done under nitrogen

Fig. 2. Schematic drawing of experimental setup used in simulations for KGG (left panel) and GG (right panel). In the simulations a 300 × 300 × 183 voxel simulation grid was used for KGG sample and 295 × 295 × 184 voxel for GG sample. In both cases 1 mm3 sized voxels were used.

208

J. Ikonen et al. / Journal of Contaminant Hydrology 192 (2016) 203–211

Fig. 3. Kd of selenium as a function of pH in three different concentrations determined by batch sorption experiments in the atmospheric conditions. GG shown in left and KGG right graph.

atmosphere Eh decreases to + 150 mV which is expected to decrease oxidation of selenite to selenate leading to increase in sorption(De Cannière et al. 2010). The Kd for GG varied between (1.5 ± 0.3) × 10−2 m3/kg and (2.0 ± 0.3) × 10−2 m3/kg. In neutral pH the Kd was similar under atmospheric and nitrogen conditions but it was notably higher in basic solutions with a slight tendency to decrease along decrease of selenium concentration. Kd for KGG varied between (2.5 ± 0.3) × 10−2 m3/kg and (3.5 ± 0.4) × 10−2 m3/kg. Kd of KGG was slightly higher in nitrogen atmosphere at pH 7. Under basic conditions Kd of selenium was one decade higher than under atmospheric conditions. When comparing the Kd values under atmospheric conditions and nitrogen atmosphere it was seen that the values under nitrogen atmosphere are notably higher in basic solutions and they do not decrease with increasing pH. 3.3. Diffusion experiment results The decrease of selenium in the inlet holes and increase in the sampling holes in diffusion experiments described in Section 2.3 were analyzed using TDD simulations. In case of KGG we were able to measure a notable decrease in selenium concentration from the inlet hole A and the breakthrough into sampling hole at a distance of 1 cm (A3). The

concentration in other sampling holes (T1–T2, A1 and A2) remained at or below the detection limit of ICP-MS (0.6 μg/L). The measured and modelled curves for in-diffusion of selenium from inlet hole A and breakthrough of selenium into the sampling hole A3 are shown in Fig. 4. The best agreement between the measured and modelled in-diffusion curves were found when the values of (7 ± 2) × 10−13 m2/s and (1.5 ± 0.3) × 10−3 m3/kg were used for De and Kd, respectively. In the case of breakthrough, the best agreement was found when values of (7 ± 1) × 10−13 m2/s and (1.6 ± 0.2) × 10−3 m3/kg were used (see Table 3). From these results, values of (1.7 ± 0.6) × 10−13 m2/s and (1.6 ± 0.6) × 10− 13 m2/s were determined for Da for inlet (A) and breakthrough (A3), respectively. The confidence bands for the modelled indiffusion curve and error bars for the modelled through diffusion curve are also shown. De has previously been measured in KGG for HTO and Cl-36 using a similar experimental system as described in this paper (Jokelainen et al. 2009). For HTO they reported values from 5.3 × 10− 13 m2/s to 13 × 10− 13 m2/s with average of 8 × 10−13 m2/s and for Cl-36 from 1.3 × 10−13 m2/s to 2.8 × 10−13 m2/s with average of 2 × 10−13 m2/s. As far as we are aware Kd values for selenium in KGG have not been measured so we can only compare Kd values determined in this work by batch sorption experiments (see Section 3.4).

Fig. 4. Measured and modelled curves for selenium in-diffusion from inlet hole A (left) and breakthrough into sampling hole A3 (right) together with the confidence bands of the model in KGG. For the breakthrough measurements, the detection limit of selenium concentration is also indicated by the blue line.

J. Ikonen et al. / Journal of Contaminant Hydrology 192 (2016) 203–211

209

Table 3 The results for De and Kd of selenium in KGG granite and GG analyzed from the diffusion experiment using TDD simulations. The values for Da have been determined using Eq. (7). KGG

De (m2/s) Kd (m3/kg) Da (m2/s)

GG

Decrease from A

Increase into A3

Decrease from A

Decrease from B

Increase into B1

(7 ± 2) × 10−13 (1.5 ± 0.3) × 10−3 (1.7 ± 0.6) × 10−13

(7 ± 1) × 10−13 (1.6 ± 0.2) × 10−3 (1.6 ± 0.4) × 10−13

(3 ± 1) × 10−12 (1.0 ± 0.3) × 10−4 (11 ± 5) × 10−12

(2.5 ± 1) × 10−12 (0.9 ± 0.3) × 10−4 (10 ± 6) × 10−12

(2.5 ± 1.5) × 10−12 (1.2 ± 0.6) × 10−4 (8 ± 5) × 10−12

In case of GG we were able to model measured in-diffusion curves of selenium from both inlet holes A and B as well as the breakthrough curve for one of the sampling holes B, the results of which are shown in Fig. 5. The best agreement between the measured and modelled indiffusion curves were found when the values of (3.0 ± 1.0) × 10−12 m2/s and (1.0 ± 0.3) × 10− 4 m3/kg (inlet hole A) and (2.5 ± 1.0) × 10−12 m2/s and (0.9 ± 0.3) × 10−4 m3/kg(inlet hole B) were used for De and Kd, respectively. For the breakthrough, the best agreement was found when values of (2.5 ± 1.5) × 10−12 m2/s and (1.2 ± 0.6) × 10−4 m3/kg were used (see Table 3). From these results, values of (11 ± 5) × 10− 12 m2/s, (10 ± 6) × 10− 12 and (8 ± 5) × 10−12 m2/s were determined for Da for the inlet holes (A and B) and sampling hole (B1), respectively. The confidence bands for modelled in-diffusion curves and error bars for modelled through diffusion curve are also shown. Here, the uncertainties of modelled curves are higher than in KGG even though the measured values were well over the detection limit of selenium concentration. The higher uncertainties are caused by the higher variations in the experimental data which might be caused by the slight changes in volume of water in sampling holes during the long time duration of the experiment (see Section 2.3) and structural heterogeneities of the sample. The chemical effects which cannot be taken account using linear Kd concept might have impact as well. Due to higher diffusivity of selenium in GG than in KGG, such effects might be more pronounced in GG. However, based on previous studies, the effect of heterogeneity on De is expected to be minor in these rock samples (Jokelainen et al. 2009, Ikonen et al. 2016). The breakthrough of selenium was detected in all sampling holes in GG. However, the breakthrough curves fluctuate heavily and thus accurate modelling of them was not feasible. It was observed during the experiment that in some of the sampling holes the water level decreased randomly between the samplings. Such fluctuation was more extensive in other sampling holes than in A, B, and B1 holes. The features with high transmissivity are likely to have chased the water loss. The results of the in-situ diffusion experiment that lasted 2.5 years in Grimsel URL were reported by Soler et al. (2015). The De of HTO in GG

varied between 1.7 × 10−13 and 1.1 × 10−12 m2/s where the lowest values hypothesize kind of retardation for HTO in this matrix. However, a particle tracking approach to analyze the diffusion of non-sorbing radionuclides has shown that the De is more likely closer to 3 × 10−11 m2/s than 3 × 10−13 m2/s in GG (Ikonen et al. 2016). Soler et al. (2015) have also reported De values from laboratory experiments ranging from 3.3 × 10−12 to 2.9 × 10−11 m2/s for HTO which are in the same range with the in-situ De values. Moreover, values measured in the laboratory using helium as a gas phase indicated a De of 2 × 10−12 m2/s (the values are converted to water phase; Kuva et al. 2015). Note that these values have not been reported for selenium. Generally, it can be said that the results for De of selenium oxyanions determined in this work are in fair agreement with the ones for HTO and Cl-36 determined from the in-situ and block scale diffusion experiments (Soler et al. 2015, Ikonen et al. 2016 and Jokelainen et al. 2009). 3.4. Comparison of Kd values When comparing the Kd values of selenium determined by batch sorption and through diffusion experiments, we point out the differences of pH in different experiments. For example the pH has changed strongly in GG during the 3 years diffusion experiment; at the beginning it was 9.6 and at the end 7.2. This might be due to the equilibration with atmospheric CO2.·For KGG batch sorption Kd was (6.0 ± 2.0) × 10−3 m3/kg in pH 8. This was about four times higher than the Kd value (1.5 ± 0.3) × 10−3 m3/kg obtained from the through diffusion experiment. For GG the Kd value from the batch experiments at pH was (7.0 ± 2.0) × 10−3 m3/kg which is about one order of magnitude higher than the Kd value (1.0 ± 0.6) × 10−4 m3/kg from the through diffusion experiment. However we have to take into account that at the beginning of the block experiment, the pH was 9.6 which favors selenate species. In all experiments there might be present some selenate. Thus in this work the Kd values determined incorporate both species. In this work, sorption of selenium is in agreement with the work by Fujikawa and Fukui (1997) who reported Kd values in granodiorite in low ionic

Fig. 5. Measured and modelled curves for selenium in-diffusion from inlet holes A and B (left) and breakthrough into sampling hole B1 (right) together with confidence of bands of the model in GG.

210

J. Ikonen et al. / Journal of Contaminant Hydrology 192 (2016) 203–211

strength solution ranging from 1.8 × 10−1 m3/kg to 3.2 × 10−3 m3/kg. Kd values are also in agreement with the ones in granite ranging from 5.7 to 8.6 × 10−3 m3/kg reported by Jan et al. (2007). The difference between the Kd results with crushed and intact rock has also been observed with other studies (e.g., Bradbury and Stephen 1985, Tachi et al. 1998). Kd is known to be strongly dependent on specific surface area (SSA). The SSA of crushed KGG rock material of grain size b0.3 mm is 0.428 ± 0.003 m2/g (Nuclear Chemistry Chalmers 2015 BET/ krypton) and that of GG 0.330 ± 0.002 m2/g. However, it is assumed that the SSA in the intact rock blocks is low compared to the crushed material. For example the SSA of intact granitic rock cores varied between 0.02 m2/g and 0.14 m2/g when measured by BET/Kr method (Hellmuth et al. 1995). The difference between Kd values obtained from batch and diffusion experiments is much higher in GG than in KGG even though the SSA values are not very far from each other for crushed material. From mercury prosimetry measurements, the average pore opening in GG is 400 nm, implying a smaller SSA in intact rock compared to KGG where the average pore opening is 200 nm. As seen from batch sorption experiments, the minerology and water composition does not seem to have significant effect on Kd values. However there is one decade difference in Kd between these two rock types in the block experiments, where the Kd in KGG was greater than in GG. It seems that this is due to the structural effects. Moreover, pH 9.6 in the beginning of GG experiment favors selenate which is known to be less sorbing than selenite.

4. Conclusions The results suggest that selenium is poorly retained in the granitic rock environment under atmospheric conditions when selenium is present as an oxyanion. In the pH range from 7 to 9.6, Kd values of selenium varied from (1.6 ± 0.3) × 10−2 to (3.0 ± 1.0) × 10−3 m3/kg in KGG and from (1.3 ± 0.3) × 10−2 to (3.0 ± 1.0) × 10−3 m3/kg in GG. The batch Kd values were similar for the two rock types used due to only minor differences in mineral and GW compositions. Kd values were slightly higher under a nitrogen atmosphere when Eh was + 150 mV. In the block experiments some selenite (IV) was observed to oxidise to selenate (VI). However, under a nitrogen atmosphere selenium stays in the form of selenite which increases slightly the Kd of selenium in these conditions as concluded by Lehto and Huo (2011). The TDD modelling tool was found to be efficient for analyzing the block scale diffusion experiments, especially when handling the challenging boundary conditions due to the geometry of the experimental setup and irregular sampling procedures. As a result, De values of (2.5 ± 1.5) × 10− 12 m2/s and (7.0 ± 2.0) × 10− 13 m2/s for selenium in GG and KGG were determined from the modelling. These results are in fairly good agreement with the earlier reported Devalues for non-sorbing elements in the same rock types. More importantly, the block scale experiment seems to produce similar results for De values as derived from the in-situ experiment. In addition, chemical sorption using linear Kd was adapted to a model developed from the work by Dentz et al. (2012). From this modelling, Kd values of (1.0 ± 0.2) × 10− 4 m3/kg and (1.6 ± 0.2) × 10−3 m3/kg for selenium in GG and KGG, respectively, were determined. In general, the Kd values obtained from the batch sorption experiment were larger than the ones obtained from the block scale diffusion experiments for both KGG and GG. In the case of KGG, the Kd values from the batch experiment were about four times higher than from the diffusion experiment while in GG the difference was about one decade. This can be explained by the increase in the specific surface area due to crushing of the rock. Laboratory studies conducted with the crushed rock was seen to overestimate the sorption of selenium compared to the intact rock specimens. Block scale experiments where sample geometry is closer to the in-situ experiments provide valuable data for upscaling. By comparing the results from the laboratory block-scale

and in-situ experiments, realistic data for De and Kd can be applied in safety case calculations. References Allard, B., Beal, J., 1979. Sorption of americium on geologic media. J. Environ. Sci. Health A 14 (6), 507–518. Altmann, S., 2008. Geo'chemical research: a key building block for nuclear waste disposal safety cases. J. Contam. Hydrol. 102, 174–179. Atwood, D.A., 2010. Radionuclides in the Environment. John Wiley & Sons, Chichester, pp. 121–130. Augustithis, S.S., 1983. Leaching and Diffusion in Rocks and their Weathering Products. Theophrastus publications, Athens, p. 6. Beauwens, T., De Canniere, P., Moors, H., Wang, L., Maes, N., 2005. Studying the migration behaviour of selenate in Boom Clay by electromigration. Eng. Geol. 77 (3-4), 285. Boult, K. A., Cowper, M. M., Heath, T. G., Sato, H., Shibutani, T., Yui, M., 1998. Towards an understanding of the sorption of U(VI) and Se(IV) on sodium bentonite. J. Contam. Hydrol. 35, 141—150 Bradbury, M.H., Stephen, I.G., 1985. Diffusion and permeability based sorption measurements in intact rock samples In: (eds. Werne, L.O.) Mater. Res. Soc. Symp. Proc. 50. Bray, A., Benning, L., Bonneville, S., Oelker, E., 2014. Biotite surface chemistry as a function of aqueous fluid composition. Geochim. Cosmochim. Acta 128, 58–70. Chan, Y.T., Kuan, W.H., Chen, T.Y., Wang, M.K., 2009. Adsorption mechanism of selenite and selenite on the binary oxide system. Water Res. 43, 4412–4420. Dai, Z., Wolfsberg, A., Lu, Z., Reimus, P., 2007. Upscaling matrix diffusion coefficients for heterogeneous fractured rocks. Geophys. Res. Lett. http://dx.doi.org/10.1029/ 2007GL029332. De Cannière, P., Maes, A., Williams, S., Bruggeman, C., Beauwens, T., Maes, N., Cowper, M., 2010. Behaviour of Selenium in Boom Clay SCK·CEN ref:CO 90 01 1467.01 1467 RP·W&D.037 – NIROND ref: CCHO2004/00/00 DS251-A44/2.1. Boeretang. Descostes, M., Blin, V., Bazer-Bachi, F., Meier, P., Grenut, B., Radwan, J., Schlegel, M.L., Buschaert, S., Coelho, D., Tevissen, E., 2008. Diffusion of anionic species in Callovo-Oxfordian argillites and Oxfordian limestones (Meuse/haute-Marne, France). Appl. Geochem. 23, 655–677. Delay, F., Porel, G., Sardini, P., 2002. Modelling diffusion in a heterogeneous rock matrix with a time-domain Lagrangian method and an inversion procedure. Compt. Rendus Geosci. 334 (13), 967–973. Delay, F., Porel, G., 2003. Inverse modeling in the time domain for solving diffusion in a heterogeneous rock matrix. Geophys. Res. Lett. 30, 1147–1150. Dentz, M., Gouze, P., Russian, A., Dweik, J., Delay, F., 2012. Diffusion and trapping in heterogeneous media: an inhomogeneous continuous time random walk approach. Water Res. 49, 13–22. Duc, M., Lefevre, G., Fedoroff, M., Jeanjean, J., Rouchaud, J.C., Monteil-Rivera, F., Dumonceau, J., Milonjic, S., 2003. Sorption of selenium anionic species on apatites and iron oxides from aqueous solutions. J. Environ. Radioact. 70, 61–72. Fujikawa, Y., Fukui, M., 1997. Radionuclide sorption to rocks and minerals: effects of pH and inorganic anions. Part 2. Sorption and speciation of selenium. Radiochim. Acta 76, 163–172. Giroud, N., 2016. Personal Communication after PSI Se Analyses. Gjetvaj, F., Russian, A., Gouze, P., Dentz, M., 2015. Dual control of flow field heterogeneity and immobile porosity on non-Fickian transport in Berea sandstone, water Resour. Restor. Dent. 51, WR017645. Glaus, M.A., Aertsens, M., Maes, N., Van Laer, L., Van Loon, L.R., 2015. Treatment of boundary conditions in through-diffusion: a case study of 85Sr2+ diffusion in compacted illite. J. Contam. Hydrol. 177-178, 239–248. Grambow, P., 2008. Mobile fission and actinide products in nuclear waste disposal. J. Contam. Hydrol. 102, 180–186. Grisak, G.E., Pickens, J.F., 1980. Solute transport through fractured media: 1. The effect of matrix diffusion. Water Resour. Res. 16 (4), 719–730. Hellmuth, K.-H., Klobes, P., Meyer, K., Röhl-Kuhn, B., Siitari-Kauppi, M., Hartikainen, J., Hartikainen, K., Timonen, J., 1995. Matrix retardation studies: size and structure of accessible pore space in fresh and altered crystalline rock. Z. Geol. Wiss. 23 (5/6), 691–706. Iida, Y., Yamaguchi, T., Tanaka, T., 2011. Experimental and modeling study on diffusion of selenium under variable bentonite content and porewater salinity. J. Nucl. Sci. Technol. 48 (8), 1170–1183. Ikonen, J., Sardini, P., Jokelainen, L., Siitari-Kauppi, M., Martin, A., Eikenberg, J., 2016. The tritiated water and iodine migration in situ in Grimselgranodiorite. Part I: determination of the diffusion profiles. J. Radioanal. Nucl. Chem. http://dx.doi.org/10.1007/ s10967-016-4890-6. Jan, S.-C., Wang, T.-H., Li, M.-H., Tsai, S.-C., Wei, Y.-Y., Hsu, C.-N., Teng, S.-P., 2007. Evaluating adsorption ability of granite to radioselenium by chemical sequential extraction. J. Radioanal. Nucl. Chem. 273 (2), 299–306. Jokelainen, L., Meski, T., Lindberg, A., Soler, J.M., Siitari-Kauppi, M., Martin, A., Eikenberg, J., 2013. The determination of 134Cs and 22Na diffusion profiles in granodiorite using gamma spectroscopy. J. Radioanal. Nucl. Chem. 295, 2153–2161. Jokelainen, L., Ikonen, J., Read, D., Hellmuth, K.-H., Siitari-Kauppi, M., 2009. The diffusion of tritiated water, chloride and uranium through granite. Mater. Res. Soc. Symp. Proc. 1193, 461–468. Kelokaski, M., Siitari-Kauppi, M., Sardini, P., Möri, A., Hellmuth, K.-H., 2006. Characterisation of pore space geometry by 14C-PMMA impregnation-development work for in situ studies. J. Geochem. Explor. 90 (1–2), 45–52. Kuva, J., Voutilainen, M., Kekäläinen, P., Siitari-Kauppi, M., Timonen, J., Koskinen, L., 2015. Gas phase measurements of porosity, diffusion coefficient, and permeability in rock samples from Olkiluoto bedrock. Finl. Transp. Porous. Med. 107 (1), 187–204.

J. Ikonen et al. / Journal of Contaminant Hydrology 192 (2016) 203–211 Kuva, J., 2016. Personal communication. Lehto, J., Huo, X., 2011. Chemistry and Analysis of Radionuclides Laboratory Tehniques and Methodology. WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim, pp. 233–237. McCarthy, J.F., 1993. Continuous-time random walks on random media. J. Phys. A Math. Gen. 26, 2495–2503. Mohamed, A.M.O., Antia, H.E., 1998. Geoenvironmental Engineering. Elsevier, Amsterdam, pp. 172–173. Mäder, U.K., Fierx, T., Frieg, B., Eikenberg, J., Rüthi, M., Albinsson, Y., Möri, A., Ekberg, S., Stille, P., 2006. Interaction of hyperalkaline fluid with fractured rock: Field and laboratory experiments of the HPF project (Grimsel test site, Swirzerland). J. Geochem. Explor. 90, 68–94. Neretnieks, I., 1980. Diffusion in the rock matrix: an important factor in radionuclide retardation? J. Geophys. Res 85, 4379–4397. Noetinger, B., Estebenet, T., 2000. Up-scaling of double porosity fractured media using continuous-time random walks methods. Transp. Porous Media 39 (3), 315–337. Noetinger, B., Estebenet, T., Landereau, P., 2001. A direct determination of the transient exchange term of fractured media using a continuous time random walk method. Transp. Porous Media 44 (3), 539–557. Noetinger, B., Roubinet, D., Russian, A., Le Borgne, T., Delay, F., Dentz, M., de Dreuzy, J.-R and Gouze, P., Random walk methods for modeling hydrodynamic transport in porous and fractured media from pore to reservoir scale, Transp. Porous Media, in press DOI http://dx.doi.org/10.1007/s11242-016-0693-z Parida, K.M., Gorai, B., Das, N.N., Rao, S.R., 1997. Studies on Ferric Oxide HydroxidesIII. Adsorption of Selenite (SeO2− 3 ) on Different Forms of Iron Oxyhydroxides. J. Colloid Interface Sci. 185, 355–362. Peak, D., 2006. Adsorption mechanisms of selenium oxyanions at the aluminum oxide/ water interface. J. Colloid Interface Sci. 303, 337–345. Posiva 2013-01, 2013. Safety case for the disposal of spent nuclear fuel at Olkiluotomodels and data for the repository system 2012. posiva Oy, Eurajoki. Robinet, J.C., Sardini, P., Coelho, D., Parneix, J.C., Prêt, D., Sammartino, S., Boller, E., Altmann, S., 2012. Effects of mineral distribution at mesoscopic scale on solute diffusion in a clay-rich rock: Example of the Callovo-Oxfordian mudstone (Bure, France). Water Resour. Res. 48, W05554. Robinet, J.C., Sardini, P., Delay, F., Hellmuth, K.-H., 2008. The effect of rock matrix heterogeneities near fracture walls on the residence time distribution (RTD) of solutes. Transp. Porous Media 72 (3), 393–408. Rovira, M., Gim'enez, J., Mart'ınez, M., Mart'ınez-Llad'o, X., de Pablo, J.V., Mart'I, V., Duro, L., 2008. Sorption of selenium(IV) and selenium(VI) onto natural iron oxides: Goethite and hematite. J. Hazard. Mater. 150, 279–284. Sardini, P., Delay, F., Hellmuth, K.-H., Porel, G., Oila, E., 2003. Interpretation of out-diffusion experiments on crystalline rocks using random walk modeling. J. Contam. Hydrol. 61 (1–4), 339–350.

211

Sardini, P., Robinet, J.C., Siitari-Kauppi, M., Delay, F., Hellmuth, K.-H., 2007. Direct simulation of heterogeneous diffusion and inversion procedure applied to an out-diffusion experiment. Test case of Palmottu granite. J. Contam. Hydrol. 93 (1–4), 21–37. Savoye, S., Schlegel, M., Frasca, B., Fayette, A., Grenut, B., 2012. How mobile is selenite diffusion through the Callovo-Oxfordian claystones? Insights given by radiochemistry and X-ray absorption spectrometry. 4th Annual Workshop Proceedings of the 7th EC FP. Recosy CP, Karlsruhe. Séby, F., Potin-Gautier, M., Giffaut, E., Donard, O.F.X., 1998. Assessing the speciation and the biogeochemical processes affecting the mobility of selenium from a geological repository of radioactive wastes to the biosphere. Analusis 26 (5), 193–198. Soler, J.M., Landa, J., Havlova, V., Tachi, Y., Ebina, T., Sardini, P., Siitari-Kauppi, M., Martin, A., 2014. Modeling of an in-situ diffusion experiment in granite at the Grimsel test site. MRS Proc. 1665, 85–91. Soler, J.M., Landa, J., Havlova, V., Tachi, Y., Ebina, T., Sardini, P., Siitari-Kauppi, M., Eikenberg, J., Martin, A.J., 2015. Comparative modeling of an in situ diffusion experiment in granite at the Grimsel test site. J. Contam. Hydrol. 179, 89–101. Tachi, Y., Shibutani, T., Sato, H., Yui, M., 1998. Sorption and diffusion behavior of selenium in tuff. J. Contam. Hydrol. 35, 77–89. Videnská, K., Gondolli, J., Štamberg, K., Havlová, V., 2015. Retention of selenium and caesium on crystalline rock: the effect of redox conditions and mineralogical composition. J. Radioanal. Nucl. Chem. 304 (1), 417–423. Videnská, K., Palágyi, S., Štamberg, K., Vodičková, H., Havlová, V., 2013. Effect of grain size on the sorption and desorption of SeO4 2- and SeO3 2- in columns of crushed granite and fracture infill from granitic water under dynamic conditions. J. Radioanal. Nucl. Chem. 298 (1), 547–554. Voutilainen, M., 2012. Characterization of structure and diffusion in geological materials. PhD Thesis. Department of Physics, University of Jyväskylä Research, Jyväskylä Reports 12/2012, ISBN: 978-951-39-4996-9. Voutilainen, M., Sardini, P., Siitari-Kauppi, M., Kekäläinen, P., Aho, V., Myllys, M., Timonen, J., 2013. Diffusion of tracer in altered tonalite: experiments and simulations with heterogeneous distribution of porosity. Transp. Porous Media 96, 319–336. Witherspoon, P.A., 1996. Introduction to second world wide review of geological problems in radioactive waste isolation, geological problems in radioactive waste isolation, second worldwide review. Lawrence Berkeley National Laboratory Report, Berkeley. Yllera De Llano, A., Bidoglio, G., Avogadro, A., Gibson, P.N., Rivas Romero, P., 1996. Redox reactions and transport of selenium through fractured granite. J. Contam. Hydrol. 21 (1–4), 129–139.