Sunteți pe pagina 1din 16

Available online at www.sciencedirect.

com

Applied
Geochemistry
Applied Geochemistry 23 (2008) 11861201
www.elsevier.com/locate/apgeochem

Coupled thermo-hydro-chemical models of compacted


bentonite after FEBEX in situ test
Javier Samper a,*, Liange Zheng a,1, Luis Montenegro a, Ana Mara Fernandez b,
Pedro Rivas b
a
E.T.S. Ingenieros de Caminos, Canales y Puertos, Universidad de A Coruna, Campus de Elvina s/n, 15192 A Coruna, Spain
b
Departamento de Medio Ambiente, CIEMAT, Avda. Complutense 22, 28040 Madrid, Spain

Received 9 February 2007; accepted 8 November 2007


Editorial handling by R. Fuge
Available online 26 January 2008

Abstract

FEBEX (Full-scale Engineered Barrier EXperiment) is a demonstration and research project for the engineered barrier
of a high-level radioactive waste (HLW) repository. FEBEX is based on the Spanish reference concept for radioactive
waste disposal in crystalline rock according to which canisters are emplaced in horizontal drifts and surrounded by a
compacted bentonite clay barrier. The project includes two main large-scale tests which started in February 1997: an
in situ full-scale test performed at Grimsel, Switzerland, and a mock-up test operating at CIEMAT facilities in Madrid,
Spain. Coupled thermal, hydrodynamic and chemical (THC) models for the engineered barrier have been developed from
laboratory experiments and the mock-up test. Dismantling of section 1 of in situ test, which took place in the summer of
2002, provides a unique opportunity to test THC model predictions. Here a numerical THC model of FEBEX in situ test is
presented. Predictions performed with three conceptual geochemical models are compared to measured temperature, water
content and geochemical data obtained from aqueous extract tests of samples collected after dismantling of heater 1. These
data require the use of inverse hydrogeochemical models for their interpretation in order to account for the geochemical
processes suered by bentonite samples during aqueous extraction. Model predictions reproduce temperatures and capture
the trends of water content and most chemical species. However, there are some discrepancies near the heater which are

especially signicant for SO24 and HCO3 . The THC model could be improved by taking into account dierent types of
water in bentonite, bentonite swelling, protonation/deprotonation by surface complexation, and CO2 degassing and
dissolution processes.
2008 Elsevier Ltd. All rights reserved.

1. Introduction

FEBEX (Full-scale Engineered Barrier EXperi-


*
ment) is a demonstration and research project deal-
Corresponding author. Fax: +34 981 167 170.
ing with the bentonite engineered barrier designed
E-mail address: jsamper@udc.es (J. Samper).
1
Present address: Lawrence Berkeley National Lab, 1 Cyclo- for sealing and containment of a high-level radioac-
tron Road, Berkeley, CA 94720, USA. tive waste repository (ENRESA, 2000). FEBEX is

0883-2927/$ - see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.apgeochem.2007.11.010
J. Samper et al. / Applied Geochemistry 23 (2008) 11861201 1187

based on the Spanish reference concept for disposal compared to measured geochemical data obtained
of radioactive waste in crystalline rock. Canisters from aqueous extract tests. These data require the
are emplaced in horizontal drifts and surrounded use of inverse hydrogeochemical models for their
by a highly-compacted bentonite barrier (ENRESA, interpretation in order to account for the geochemi-
1994). Bentonite has the multiple purpose of provid- cal processes suered by bentonite samples during
ing canister mechanical stability by absorbing stres- aqueous extraction tests. The paper starts by describ-
ses and deformations, sealing discontinuities in the ing the FEBEX in situ test. Then, THC conceptual
adjacent rock and retarding radionuclide migration. and numerical models are presented. Interpretation
FEBEX includes two main large-scale experi- of aqueous extract data with three conceptual geo-
ments operating since February 1997: an in situ chemical models is described. Inferred chemical data
full-scale test performed at Grimsel (Switzerland), are compared to model predictions. Finally, main
and a mock-up test in Madrid, Spain (ENRESA, uncertainties of THC model are discussed.
2000).
One of the objectives of the FEBEX project is the 2. FEBEX in situ test
development and testing of conceptual and numeri-
cal thermal, hydrodynamic and chemical (THC) The FEBEX in situ test is being performed in a
models for the engineered bentonite barrier. The gallery excavated in granite in the northern zone
ability of these models to reproduce the observed of the Grimsel underground laboratory operated
THC patterns in a wide range of conditions should by NAGRA in Switzerland. The test includes the
enhance the condence in their predictive capabili- heating system, the clay barrier (bentonite) and
ties. Conceptual and numerical THC models of instrumentation, monitoring and control systems.
heating and hydration experiments performed on The drift is 70.4 m long and 2.28 m in diameter.
small-scale lab cells within the FEBEX project pro- The test zone is located in the last 17.4 m of the drift
vided excellent results for temperatures, water where heaters, bentonite and instrumentation were
inow and nal water content (ENRESA, 2000, installed. The test zone was sealed with a concrete
2006d; Samper et al., 2005a,b,c). These models have plug. The main elements of the heating system are
been used to simulate the evolution of FEBEX two heaters (1 and 2), separated horizontally by
in situ test. These predictions were compared to 1 m, which simulate full-sized canisters. Heaters
measured and inferred geochemical data collected were placed inside cylindrical steel liners. Heaters
after dismantling heater 1 in the summer of 2002. are 4.54 m long, 0.9 m diameter and have a thick-
Geochemical simulations of FEBEX in situ and ness of 0.1 m. Heaters are designed to maintain a
mock-up tests are presented by Gens et al. (2004). maximum temperature of 100 C at the steel liner/
Other in situ and mock-up tests include the LOT bentonite interface. The bentonite barrier is made
experiment at Aspo underground research labora- of blocks of highly-compacted bentonite. A layout
tory (URL) in Sweden (Karnland et al., 2000; Arcos of the in situ test is shown in Fig. 1. Weighted aver-
et al., 2003) and Ophelie mock-up test performed at age values of dry density and water content of ben-
HADES URL in Belgium (Van Humbeeck et al., tonite blocks are 1.70 g/cm3 and 14.4%, respectively
2006; Verstricht et al., 2007). (ENRESA, 2000). Mineralogical composition of
Here, a THC numerical model for the FEBEX FEBEX bentonite is listed in Table 1 (Fernandez
in situ test is presented. Model predictions performed et al., 2004). The main mineral phase (9092 wt.%)
with several conceptual geochemical models are is montmorillonite.

Sections 7 12 19 28 29

Access Concrete
gallery Heater 1 Heater 2 0.9 m 2.28 m

Fig. 1. Layout of the FEBEX in situ test. Vertical lines show the location of the dierent sampling sections.
1188 J. Samper et al. / Applied Geochemistry 23 (2008) 11861201

Table 1
Mineralogical composition (in weight %) of FEBEX bentonite (Fernandez et al., 2004)
Main minerals Accessory minerals Poorly ordered minerals
Smectite 92 3 Organic Matter (as CO2) 0.35 0.05 SiO2 0.038 0.005
Quartz 21 Carbonates (calcite, dolomite) 0.60 0.13
Plagioclase 21 Soluble sulfates (gypsum) 0.14 0.01 Al2O3 0.035 0.005
Cristobalite 21 Low soluble sulfates (barite, celestite) 0.02 0.00
K-Feldspars Traces Sulphides (pyrite) 0.02 0.01 Fe2O3 0.105 0.009
Tridymite Traces Chlorides (halite) 0.13 0.02
Calcite Traces

The in situ test began on February 27th, 1997. dry air, (4) Advective ow of air dissolved in
Heater 1 was switched-o in February 2002. Dis- water, (5) Heat conduction through solid, liquid
mantling of heater 1 was performed from May to and gas phases; and (6) Heat convection through
September 2002. A comprehensive post-mortem liquid and gaseous phases. General mass balance
bentonite sampling and analysis program was equations for water, air and heat can be found in
designed to characterize solid and liquid phases, Samper et al. (2007c).
measure physical and chemical changes induced by Solute transport processes include: advection,
the combined eect of heating and hydration; and molecular diusion and mechanical dispersion. Sol-
test THM and THC models (ENRESA, ute transport equations are stated for each chemical
2006a,b,c,d). component (primary species) in the system.
Dierent conceptual geochemical models (CGM)
3. THC model have been proposed for FEBEX bentonite. They
dier depending on the type of test used to obtain
3.1. Conceptual model chemical data: squeezing and aqueous extracts tests
(AET). Conceptual model CGM-0 was derived
The thermo-hydrodynamic (TH) model assumes from a squeezing test performed on a bentonite
that the porous medium contains a liquid phase sample having a gravimetric water content of
consisting of water, aqueous species and dissolved 26.5% (ENRESA, 2000). Geochemical models were
gases, a gaseous phase which is made of water vapor used to derive the chemical composition at 14%
and dry air (i.e., a ctitious species which encom- water content. Later, Samper et al. (2005b) pre-
passes all the gaseous species except water vapor), sented a model derived from aqueous extract tests
and a solid phase which includes several minerals (model CGM-1) which assumes gypsum at equilib-
and the exchange complex. rium, calcite and dolomite with kinetic mineral dis-
Water may ow in liquid phase through com- solution, and a prescribed pressure of 103.5 bar for
pacted bentonite in response to a hydraulic gradient CO2(g). This model reproduces adequately mea-
according to Darcys law. It can also ow in gaseous sured aqueous extract data performed for dierent
phase as vapor due to air convection and in durations and S/L ratios. However, it fails to repro-
response to a moisture gradient according to Ficks duce HCO 3 and pH data, possibly due to changes in
law. Water vapor ux is several orders of magnitude CO2(g) pressure which are not considered in the
smaller than liquid water ux in most cases and model. In general, calculated results with model
therefore can be neglected. However, near the hea- CGM-1 could not t SO2 4 squeezing data. Samper
ter of the in situ test vapor ux can be important. et al. (2005b) presented a modied model (CGM-
Liquid water and water vapor uxes are related 2) based on data from both aqueous extract and
through evaporation and condensation. The energy squeezing tests. This model does not consider an ini-
transfer due to liquid water/vapor transition is very tial amount of gypsum in the system but allows for
important due to the large enthalpy increment of its precipitation. Models CGM-1 and CGM-2 t
evaporation/condensation (585 cal/g at 20 C). The simultaneously both squeezing and aqueous extract
TH conceptual model takes into account the follow- data for most chemical species, although they pres-

ing multiphase ow processes: (1) Advective ow ent discrepancies for SO2
4 and HCO3 data.
of liquid water, (2) Advective and diusive ow of Models CGM-0, CGM-1 and CGM-2 account
water vapor, (3) Advective and diusive ow of for aqueous complexation, acidbase, mineral disso-
J. Samper et al. / Applied Geochemistry 23 (2008) 11861201 1189

Table 2
Chemical species considered in models CGM-0, CGM-1 and CGM-2
Geochemical CGM-0 CGM-1 CGM-2
models

Primary species H2O, H+, Ca2+, Mg2+, Na+, K+, Cl, SO24 ; HCO3 , SiO2(aq),
Aqueous complexes OH, CaSO4(aq), CaCl+, MgCl+, NaCl(aq), MgHCO
3 , NaHCO3(aq), CaHCO3 , MgCO3(aq), CaCO3(aq), CO2(aq),
2   2 
CO3 , KSO4 , MgSO4(aq), NaSO4 ; H2 SiO4 ; HSiO3
Minerals Calcite (equilibrium) Calcite (kinetic) Calcite (kinetic)
Chalcedony (equilibrium) Chalcedony (equilibrium) Chalcedony (equilibrium)
Gypsum/anhydrite (equilibrium, only Gypsum/anhydrite Gypsum/anhydrite
precipitation) (equilibrium) (equilibrium,
only precipitation)
Dis-dolomite (kinetic) Dis-dolomite (kinetic)
Exchangeable Ca2+, Mg2+, Na+, K+ Ca2+, Mg2+, Na+, K+, H+ Ca2+, Mg2+, Na+, K+, H+
cations
Gases No condition on CO2(g) Fixed CO2(g) pressure Fixed CO2(g) pressure

lution/precipitation, gas dissolution/exsolution and which may dissolve or precipitate under kinetic con-
cation exchange. These processes are assumed at ditions in models CGM-1 and CGM-2 (see Table 2).
local equilibrium except for calcite and dolomite Relevant aqueous complexes were identied from
speciation runs performed with EQ3/6 (Wolery,
Table 3
Equilibrium constants for aqueous complexes, minerals and gases 1979). Equilibrium constants at 25 C for aqueous
and selectivity coecients at 25 C (Wolery, 1979) complexes, minerals and gases were taken from
Log K the EQ3/6 database (Wolery, 1979). They are listed
in Table 3. The GainesThomas convention is used
Aqueous complexes
CaCl+ , Ca2+ + Cl 0.6956 for cation exchange. Kinetic laws and rate constants
CaCO3 aq H () Ca2 HCO 3 7.0017 for calcite were taken from Plummer et al. (1979)
CaHCO 3 () Ca
2
HCO 3 1.0467 while those of disordered dolomite were derived
2 2
CaSO4 aq () Ca SO4 2.1111 from Ayora et al. (1994). Values of specic surface
CO2 aq H2 O () H HCO 6.3447

3 for calcite and dolomite were derived by Samper
CO2
3 H () HCO3 10.3288
H2 SiO2 et al. (2005b) from calibration of AET (see Table
4 2H () 2H2 O SiO2 aq 22.9116
HSiO 3 H
( ) H2 O SiO2 aq 9.9525 4). Bentonite chemical composition data at the ini-
KSO
4 () K SO4
2
0.8796 tial water content of bentonite buer (14.4%) for
MgCl+ , Mg2+ + Cl 0.1349 models CGM-0, CGM-1 and CGM-2 are listed in
MgCO3 aq () Mg2 CO2 2.9789
3 Table 5.
MgHCO 3 () Mg
2
HCO 3 1.0357
MgSO4 aq () Mg2 SO2 4 2.4117
NaCl(aq) , Na + Cl+ 
0.7770 3.2. Numerical model
NaHCO3 aq () Na HCO 3 0.1541
NaSO
4 () Na SO4
2
0.8200 The model assumes axial symmetry. Spatial
OH + H+ , H2O 13.9951
domain was discretized with a 1D mesh of rectangu-
Minerals lar nite elements which contains 614 nodes and 307
Calcite H () Ca2 HCO 3 1.8487 elements (Fig. 2). Two material zones were consid-
Dis  dolomite 2H 4.0579
() Ca2 Mg2 2HCO
ered in the mesh: bentonite and granite. The rst
3
Anhydrite () Ca2 SO2 4.3064 two nodes (1 and 2) are located at the heaterben-
4
Gypsum () Ca2 SO2 4 2H2 O 4.4823 tonite interface (r = 0.45 m). Bentonite which lies
Chalcedony , SiO2(aq) 3.7281 at the interval 0.45 m < r < 1.135 m is discretized
Gases with 220 nodes. Granite is simulated with 392 nodes
CO2 g H2 O () H HCO
3 7.8136 and extends from 1.135 m to 50 m. The mesh was
Cation exchange KNa-cation rened near the bentonitegranite interface because
Na+ + X-K ,K+ + X-Na 0.1380 large concentration gradients are expected to take
Na+ + X-H ,H+ + X-Na 8.4  106 place in this zone. The model was run for 1930 days,
Na+ + 0.5 X2-Ca , 0.5 Ca2+ + X-Na 0.2924 which corresponds to the time of dismantling of
Na+ + 0.5 X2-Mg , 0.5 Mg2+ + X-Na 0.2881
heater 1.
1190 J. Samper et al. / Applied Geochemistry 23 (2008) 11861201

Table 4
Kinetic dissolution/precipitation rates, R, for calcite (Plummer et al., 1979) and disordered dolomite (Ayora et al., 1994)
Mineral Kinetic rate R (mol/m2 s) Specic surface (m2/m3)
5 8 11
Calcite 8:91  10  aH 5:01  10  aH2 CO3 6:46  10  aH2 O 0.1
1:86  102  aCa2  aHCO3 1  X
Disordered dolomite 2.20  108 (1-X) 0.5
Specic surfaces derived from calibration of aqueous extract tests (Samper et al., 2005b). Note: a is the activity of chemical species and X is
saturation ratio.

Table 5
Chemical, mineralogical and cation exchange initial composition at 14% water content of FEBEX bentonite and granite for dierent CGM
Conceptual geochemical models for bentonite Granite
CGM-0 CGM-1 CGM-2
T (C) 12 12 12 12
pH 7.86 7.62 7.62 9.25
Na+ (mol/L) 1.88  101 1.28  101 1.28  101 3.76  104
K+ (mol/L) 1.68  103 2.00  103 2.30  103 7.80  106
Ca2+ (mol/L) 1.13  102 2.10  102 2.5  102 2.16  104
Mg2+ (mol/L) 1.44  102 2.20  102 3.2  102 1.32  106
HCO 3 mol=L 5.80  104 3.90  104 4.1  104 3.97  104
SO2
4 mol=L 2.12  102 3.10  102 2.2  102 7.86  105
Cl (mol/L) 1.87  101 1.55  101 1.85  101 1.31  105
SiO2(aq) (mol/L) 1.05  104 2.8  104 2.8  104 6.07  104
Calcite (%) 1 1 1 5
Chalcedony (%) 4.5 4.5 4.5 20
Disordered dolomite (%) n.c.a 2 2 0
Anhydrite (%) 0 n.c.a n.c.a 0
Gypsum (%) n.c.a 5  104 0 0
Na+ (meq/100 g) 31.18 29.35 29.35 n.c.a
K+ (meq/100 g) 1.94 2.56 2.56 n.c.a
H+ (meq/100 g) n.c.a 1 1 n.c.a
Ca2+ (meq/100 g) 34.62 35.71 35.71 n.c.a
Mg2+ (meq/100 g) 34.01 32.29 32.29 n.c.a
a
n.c. means not considered in the model.

Axis Prescribed T = 97 C Prescribed T = 12 C


Water flux = 0 Prescribed liquid pressure = 700 kPa

Bentonite Granite
0 0.45 m 1.135 m 50 m
Heater 1

Fig. 2. Scheme of nite element rectangular grid and thermal and ow boundary conditions used for THC model of FEBEX in situ test.

The initial temperature is uniform and equal to tion induced by heaters does not extend to such a
12 C. A constant temperature of 97 C is prescribed boundary. Boundary conditions for ow include:
at the heaterbentonite interface (r = 0.45 m) while (1) No ow at r = 0.45 m and (2) A prescribed liquid
temperature is assumed to remain constant and pressure of 700 kPa at r = 50 m (see Fig. 2).
equal to its initial value of 12 C at the external Since the FEBEX in situ test was not designed to
boundary (r = 50 m) because the thermal perturba- be air-tight, nothing prevented gas from owing in
J. Samper et al. / Applied Geochemistry 23 (2008) 11861201 1191

or out of the test section at early stages of the test. is equal to 6.1  1011 m2/s for bentonite and
However, it is believed that bentonite swelling 4.38  1013 m2/s for granite (see Table 6).
sealed most of buer gaps and voids, making it dif- There are no dispersivity data for granite. Results
cult for gas to ow across the concrete plug. It is of sensitivity analysis to granite dispersivity indicate
assumed in the model that the system is closed for that dispersivity of granite mostly aects solute back
gaseous phases. This means that there are neither diusion from bentonite into granite (ENRESA,
gas uxes across external boundaries nor prescribed 2000, 2006d). Given the small value of the Peclet
gas pressures. Gas in granite and bentonite is simu- number in bentonite, model results are not sensitive
lated as a trapped phase overlapping the liquid to bentonite dispersivity. In order to prevent numer-
phase. ical oscillations near the bentonitegranite interface,
Table 5 lists initial concentrations of FEBEX a dispersivity of 0.03 m was used for both bentonite
bentonite and granite. Concentrations for the initial and granite. This ensures that the Peclet number is
gravimetric water content of bentonite of 14% are much smaller than 2.
listed for CGM-0, CGM-1 and CGM-2.
Thermal and hydrodynamic parameters for ben- 3.3. Code description
tonite and granite are summarized in Table 6. The
porewater diusion coecient, D0, is assumed to THC predictions of the in situ test have been per-
be equal to 2  1010 m2/s for all chemical species. formed with INVERSEFADESCORE (Zheng
This value was derived from calibration of heating and Samper, 2004), a code which solves both
and hydration laboratory experiments (ENRESA, forward and inverse multiphase ow and multicom-
2000, 2006d). The eective diusion coecient, De, ponent reactive transport problems in 1, 2 and 3D

Table 6
Thermal, hydrodynamic and solute transport parameters of FEBEX bentonite in THC model of in situ test
Parameter Bentonite Granite
2 21
Liquid intrinsic permeability K (m ) as a function K0 = 2.75  10 for
of ionic strength I (mol/L) I0=0.2674 mol/L
 1=2
K I
K0 a I0 with a = 1.097 5  1017
p 0:5 2
Liquid relative permeability krl as a function k rl S 3l k rl S l 1  1  S 0:5
l 
of liquid saturation Sl 1:1
19:1107 W
Retention curve: liquid saturation Sl as a function Sl 0:18 Sl = [(1 + 4.76104W)3.33]0.7
15105 W1:22 
of suction W
6 6
Solid density (kg/m3) 2780  e[2  10 (TTref )]
2700  e[2  10 (TTref )]

Thermal conductivity of solids (W/m C) 0.8 1.5


Specic heat of solids (J/kg C) 835.5 1029
Liquid viscosity (kg/m s) as a function of temperature T 0.6612(T  229)1.562
7
Liquid density (kg/m3) as a function of liquid 998.2  e[5  10 (P1  100)2.1  104(T  Tref )]

pressure Pl and temperature T


Intrinsic permeability of gas krg (m2) 5  1010
Gas relative permeability krg krg = (1  Sl)3
Vapor tortuosity 0.5
Gas viscosity (kg/m s) 1.76  105
Liquid thermal conductivity (W/m C) 1.5
Liquid specic heat (J/kg C) 4202
Reference temperature, Tref (C) 12
Air thermal conductivity (W/m C) 2.6  102
Air specic heat (J/kg C) 1000
Vapor thermal conductivity (W/m C) 4.2  102
Vapor specic heat (J/kg C) 1620
Vaporization enthalpy (J/kg) 2.454  106
Porosity 0.41 0.01
Porewater diusion coecient (m2/s) 2  1010
Dispersivity (m) 3  102
Eective diusion coecients (m2/s) 6.1  1011 4.38  1013
1192 J. Samper et al. / Applied Geochemistry 23 (2008) 11861201

axy-symmetric porous and fractured media. This


code is the result of integrating the capabilities of
FADES (Navarro and Alonso, 2000), CORE2D
(Samper et al., 2000) and INVERSECORE (Dai
and Samper, 2004). State variables of the forward
model are liquid pressure, gas pressure, and temper-
ature. They are solved by a NewtonRaphson GF

method. Reactive transport equations are solved


with a sequential iteration method to obtain concen-
trations of primary species. The inverse problem is
solved by minimizing a generalized least-squares
criterion with a GaussNewtonLevenbergMarqu- BB29-9
ardt method (Dai and Samper, 2004). Forward rou- BB29-13
tines of INVERSEFADESCORE have been BB29-11
widely veried with analytical solutions and used BB29-12 BB29-8

to simulate THC experiments (Samper et al., BB29-10


2005c; ENRESA, 2006d) and THMC problems BB29-5 BB29-7
BB29-6
(Zheng and Samper, 2004). Codes of the CORE ser-
ies have been extensively used to model laboratory
and in situ experiments including the CERBERUS Fig. 3. Location of bentonite blocks BB29-5BB-29-13 used for
chemical analyses after dismantling section 29 of FEBEX in situ
experiment in Boom clay in Belgium (Samper
test.
et al., 2006; Zheng et al., 2008), interpretation of
the Redox Zone experiment in a fracture zone of
Aspo site at Sweden (Molinero and Samper, 2004, were preserved immediately after their extraction
2006; Molinero et al., 2004), inverse reactive trans- in plastic lm, two layers of aluminized PET-sheets
port of laboratory experiments and coastal aquifers and vacuum-sealed plastic bags. The rst PET-sheet
(Dai and Samper, 2006; Dai et al., 2006), O2 con- was vacuum-sealed after ushing with N2. Protec-
sumption in a HLW repository in granite (Yang tion against mechanical actions was used to ensure
et al., 2007a), long-term geochemical evolution of the integrity of the material (Fernandez and Rivas,
a HLW repository in clay (Yang et al., 2007b), 2003).
and stochastic cation exchange and reactive trans-
port in aquifers (Samper and Yang, 2006). 4.2. Bentonite porewater chemistry

4. Geochemical data after dismantling FEBEX in situ Obtaining reliable data for clay porewater chem-
test istry is a dicult task. A great deal of eort has been
made during recent years to improve water extrac-
4.1. Bentonite samples tion methods, develop numerical interpretation
methods and achieve consistency of experimental
Bentonite samples were taken from vertical sec- data and geochemical modelling (Sacchi and Mich-
tions normal to the axis of the tunnel. In each sec- elot, 2000; Sacchi et al., 2001; Bradbury and Baey-
tion several samples were taken along dierent ens, 2002, 2003; Pearson and Scholtis, 2003;
radii. Bentonite samples were taken for THC analy- Fernandez, 2004).
ses from 3 sections in the heater zone (S29, S28 and Preserving in situ conditions of core samples is
S19), a section in the zone between the concrete seal dicult. A general discussion of perturbations
and the heater 1 (S12) and a section in contact with caused by sampling, storage, preparation and
the concrete seal (S7) (see Fig. 1). Each sampling extraction techniques is given by Sacchi and Miche-
section consisted of an outer, central and inner layer lot (2000), Sacchi et al. (2001) and Pearson and
of bentonite blocks. Fig. 3 shows the location of Scholtis (2003). Aqueous extract tests (AET) have
bentonite blocks in section 29 collected after dis- been used to obtain porewater chemistry for com-
mantling of heater 1. A total of 9 bentonite blocks pacted FEBEX bentonite. AET is a method to
were taken (BB29-5 to BB2913). Samples were ana- quantify the total content of soluble salts in a clay
lyzed at dierent radial distances. Bentonite blocks sample. A 1:R AET consists of adding to a mass
J. Samper et al. / Applied Geochemistry 23 (2008) 11861201 1193

Ms of powdered clay sample a mass of distilled 4.3. Inference of porewater chemistry by inverse
water equal to R times Ms. Clay sample and water geochemical modelling
are stirred during a period of time usually of two
days during which equilibration of water and clay The methodology to infer bentonite porewater
sample is allowed. Chemical analyses are performed chemical composition from AET data is based on
on the supernatant solution after phase separation the denition of a conceptual geochemical model
by centrifugation (Sacchi et al., 2001). The solid- (CGM) for the claywater system. A CGM for a clay
to-liquid ratio, S/L, is related to the aqueous extract sample is dened in terms of relevant chemical pro-
ratio R through cesses taking place during aqueous extraction. Iden-
S 1 tication of the CGM requires knowing: (1) The list
1 of relevant aqueous complexes, (2) relevant mineral
L wi R1 wi
phases, their initial volume fractions and equilibrium
where wi is the gravimetric water content of the clay constants, (3) cation exchange reactions, CEC and
sample. It should be noted that S/L coincides with cation selectivities, and (4) relevant gas phases, pres-
1/R only when the clay sample is dry (wi = 0). sures and conditions (open or closed). The method
In addition to dilution, several chemical pro- starts from an initial CGM and a guess of sample
cesses take place during porewater extraction from porewater concentrations, ci. Inverse modelling
clay samples, including dissolution of soluble miner- accounts for perturbations caused by aqueous
als (halite, sulfates and carbonates mostly), dissolu- extraction and computes concentrations of aqueous
tion and ex-solution of gases and cation exchange. extracts. Optimum estimates of ci are those which
All these processes change the concentrations of dis- minimize the dierences between measured and cal-
solved species in a complex nonlinear manner which culated aqueous extract concentrations. Large devi-
makes it dicult to derive the chemical composition ations of model results from measured data may
of the original (before aqueous extraction) clay indicate the need to modify or update the CGM.
porewater from aqueous extract data (Bradbury Estimated parameters include pH and concentra-
and Baeyens, 1998; Sacchi and Michelot, 2000; Sac- tions of major anions and cations of the original clay
chi et al., 2001). The inference of dissolved concen- sample. Additional details of inverse AET interpre-
tration for reactive species requires geochemical tation can be found in Zheng et al. (2008).
modelling based on mineralogical data (Baeyens
and Bradbury, 1991; Pearson and Scholtis, 1995; 4.4. Inferred chemical compositions
Baeyens and Bradbury, 1997, 1998, 2003; Pearson
and Scholtis, 2003; Fernandez and Rivas, 2005a; Table 7 shows measured water content and AET
Zheng et al., 2008). data for bentonite samples at dierent radial dis-

Table 7
Chemical composition (in mg/L) of bentonite aqueous extracts samples (1:4 solid-to-liquid ratio) at dierent radial distances
Sample Radial distance (cm) Water content (w. %) pH Cl SO2
4 HCO
3 SiO2 Na+ K+ Ca2+ Mg2+
BB29-7/2-5 106.0 26.8 8.4 22 77 207 24.0 155 31 42 144
BB29-7/2-6 106.0 27.4 8.9 21 70 216 56.4 168 44 65 35
BB29-7/2-7 101.5 26.6 8.3 26 132 176 31.4 143 8.1 10 49
BB29-7/2-8 101.5 25.5 8.4 27 118 184 37.6 148 16 21 70
BB29-7/1-9 97.5 26.5 8.1 24 259 136 23.2 185 2.9 <1 1.1
BB29-7/1-10 97.5 26.5 8.2 29 156 162 28.2 148 3.2 1.1 2
BB29-7/1-11 93.5 25.8 8.0 38 260 141 26.1 185 2.9 <1 1.1
BB29-7/1-12 93.5 26.0 8.1 33 264 136 25.7 185 3.0 <1 1.1
BB29-12/2-1 87.5 21.0 8.1 59 270 143 26.3 215 3.1 1.1 1.6
BB29-12/2-2 82.5 22.0 8.1 76 293 139 27.6 219 3.6 1.9 2.1
BB29-12/2-3 77.5 20.3 8.0 164 255 121 27.1 250 4.1 2.8 2.6
BB29-12/2-4 72.5 19.9 8.0 280 207 108 26.8 280 5.4 6.2 4.3
BB29-11/2-1 66.0 16.3 8.1 348 137 119 31.5 300 6.4 7.9 5.1
BB29-11/2-2 61.0 15.2 7.9 370 171 111 32.4 310 6.9 9.1 5.6
BB29-11/2-3 56.0 14.2 8.1 348 192 123 34.6 310 7.0 8.1 5.4
BB29-11/2-4 51.0 14.2 8.2 315 209 153 35.7 280 6.7 5.1 3.4
1194 J. Samper et al. / Applied Geochemistry 23 (2008) 11861201

Table 8
Inferred, calculated and measured 1:4 aqueous extract data for bentonite sample BB29-7/1-12 having a gravimetric water content of 26%
(concentrations in mol/L)
CGM-0 CGM-1 CGM-2 Measured
Inferred Calculated Inferred Calculated Inferred Calculated
Ca2+ 8.3  103 2.5  105 1.3  103 2.7  105 6.3  103 2.4  105 2.5  105
Cl 1.9  102 9.2  104 1.9  102 9.7  104 1.9  102 9.2  104 9.3  104
HCO 3 4.6  102 2.2  103 1.5  104 7.2  104 1.7  104 1.2  103 2.2  103
K+ 1.1  103 7.6  105 5.3  104 8.0  105 1.0  103 7.6  105 7.6  105
Mg 2+
1.5  102 4.4  105 2.4  103 4.9  105 1.3  102 4.6  105 4.5  105
Na+ 1.2  101 8.0  103 5.5  102 8.4  103 1.0  101 8.1  103 8.0  103
SO2
4 5.7  102 2.7  103 9.6  103 2.9  103 5.7  102 2.7  103 2.7  103

Table 9
Inferred, calculated and measured 1:4 aqueous extract data for bentonite sample BB29-11/2-3 having a gravimetric water content of 14.2%
(concentrations in mol/L)
CGM-0 CGM-1 CGM-2 Measured
Inferred Calculated Inferred Calculated Inferred Calculated
Ca2+ 6.2  102 1.9  104 4.3  102 2.0  104 6.2  102 2.1  104 2.0  104
Cl 3.2  101 9.8  103 3.2  101 9.8  103 3.2  101 9.8  103 9.8  103
HCO 3 6.7  102 2.1  103 1.4  104 8.2  104 1.5  104 9.1  104 2.0  103
K+ 2.8  103 1.8  104 2.4  103 1.8  104 2.8  103 1.8  104 1.8  104
Mg2+ 6.5  102 2.1  104 4.0  102 2.2  104 6.0  102 2.1  104 2.2  104
Na+ 1.9  101 1.3  102 1.6  101 1.3  102 1.9  101 1.3  102 1.3  102
SO2
4 6.7  102 1.9  103 2.0  102 2.0  103 6.7  102 1.9  103 2.0  103

tances of section 29 (Fernandez and Rivas, 2003). Bicarbonate inferred concentrations obtained
Inverse geochemical modeling was performed for with model CGM-0 are almost two orders of magni-
all AET data of section 29. Detailed results for sam- tude greater than those of models CGM-1 and
ples BB29-7/1-12 and BB29-11/2-3 which have CGM-2. This large dierence is caused by the
gravimetric water contents of 26% and 14.2%, hypothesis of constant CO2(g) pressure of models
respectively, are listed in Tables 8 and 9. These CGM-1 and CGM-2. These two models assume that
tables include measured AET data as well as most of the measured HCO 3 in aqueous extract
inferred and calculated compositions with CGM-0, tests is supplied by a constant CO2(g) pressure. In
CGM-1 and CGM-2. Inferred concentrations are other words, the assumption of constant CO2(g)
much smaller than calculated concentrations due pressure provides the main source of dissolved
to dilution caused by aqueous extraction. It can be HCO 3 . Then, inferred concentrations are much
seen by comparing calculated and measured col- smaller than they would otherwise be.
umns in Tables 8 and 9 that models generally repro-
duce measured data. Models CGM-1 and CGM-2, 5. Testing THC model predictions
however, do not reproduce HCO 3 concentrations
accurately. This is an indication of their limitation THC model reproduces properly the radial distri-
to represent adequately the chemistry of FEBEX bution of computed temperatures in bentonite and
bentonite. granite just when heater 1 was switched o at
It should be noted that all models lead to similar t = 1827 days (Fig. 4). The t to measured temper-
inferred results for all chemical species except for atures at dismantling time (t = 1930 days) is worse

SO2 2
4 and especially for HCO3 . Inferred SO4 con- because dismantling of concrete plug and bentonite
centrations with CGM-1 are smaller by a factor of buer induced a 3-dimensional heat re-distribution
36 than those inferred with CGM-0 and CGM-2 which is not taken into account in the 2D axy-sym-
because in model CGM-1 part of the measured metric THC model (Fig. 5).
SO24 in aqueous extract tests is assumed to be The model captures the general trend of mea-
released by gypsum dissolution. sured gravimetric water contents at the bentonite
J. Samper et al. / Applied Geochemistry 23 (2008) 11861201 1195

100 0.6

0.5 Bentonite
Granite
80

heater
1827 days
Temperature (C)

measured data 0.4 recalculated data CGM0


model results

Cl- (mol/L)
60 recalculated data CGM1
recalculated data CGM2
0.3 model results CGM0
model results CGM1
40 model results CGM2
0.2

20
0.1

0 0.0
0 4 8 12 16 0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.35 1.50

Radial distance (m) Radial distance (m)

Fig. 4. Radial distribution of computed (line) and measured Fig. 7. Comparison of predicted Cl concentrations with models
temperatures after 1827 days of heating, just the day in which CGM-0, CGM-1 and CGM-2 (lines) with inferred aqueous
heater 1 was switched-o. extract data (symbols) at section 29 after 1930 days of heating.

40 0.16

1930 DAYS HEATER1 0.14 Bentonite


35 Granite
measured data

heater
0.12
model results
Temperature (C)

30 recalculated data CGM0


Ca2+ (mol/L)

0.1 recalculated data CGM1


recalculated data CGM2
0.08 model results CGM0
25
model results CGM1
model results CGM2
0.06
20
0.04

15
0.02

10 0
0 4 8 12 16 0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.35 1.50
Radial distance (m) Radial distance (m)

Fig. 5. Radial distribution of computed (line) and measured Fig. 8. Comparison of predicted Ca2+ concentrations with
temperatures after 1930 days of heating. models CGM-0, CGM-1 and CGM-2 (lines) with inferred
aqueous extract data (symbols) at section 29 after 1930 days of
heating.

28

bentonite granite
gravimetric water content (%)

heater (Fig. 6). It should be noted that measured water


24 contents near the granitebentonite interface are
larger than the water contents corresponding to full
20
saturation at the original dry density of the buer
(w = 25%) due to bentonite swelling, a process not
taken into account in the model.
16 data Overprediction of measured water contents near
model results the heater is partly due to the fact that the model
12
considers a xed temperature of 97 C at the heater
0.3 0.45 0.6 0.75 0.9 1.05 1.2 1.35 1.5 which is equal to the average of measured tempera-
Radial distance (m)
tures at the heater surface which range from 75 to
Fig. 6. Radial distribution of computed and measured gravimet- 100 C. Model t to measured water contents near
ric water content after 1930 days of heating. the heater would improve if a temperature larger
than 97 C had been prescribed at the heaterben-
barrier after 1930 days. However, it underpredicts tonite interface. Another reason for overpredicting
measured water contents near the granitebentonite measured water contents near the heater is benton-
interface and overpredicts them near the heater ite shrinkage which is not considered in the model.
1196 J. Samper et al. / Applied Geochemistry 23 (2008) 11861201

0.12 0.08
Bentonite Granite
recalculated data CGM0
0.10 Bentonite
Granite recalculated data CGM1

heater
recalculated data CGM2
heater

0.06 model results CGM0


recalculated data CGM0
0.08 recalculated data CGM1 model results CGM1

SO4 2- (mol/L)
Mg2+ (mol/L)

recalculated data CGM2 model results CGM2

model results CGM0


0.06 model results CGM1 0.04
model results CGM2

0.04

0.02
0.02

0.00
0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.35 1.50 0.00
0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.35 1.50
Radial distance (m)
Radial distance (m)
Fig. 9. Comparison of predicted concentrations of Mg2+ (lines)
Fig. 12. Comparison of predicted concentrations of SO24 (lines)
with inferred aqueous extract data (symbols) at section 29 after
with inferred aqueous extract data (symbols) at section 29 after
1930 days of heating using models CGM-0, CGM-1 and CGM-2.
1930 days of heating using models CGM-0, CGM-1 and CGM-2.

0.5
Bentonite Granite
1x10-1
0.4 recalculated data CGM0
heater

recalculated data CGM1


Granite
recalculated data CGM2 1x10-2

heater
model results CGM0
Na+ (mol/L)

0.3 model results CGM1


HCO3- (mol/L)

model results CGM2


1x10-3

0.2

1x10-4
recalculated data CGM0
0.1
recalculated data CGM1
-5 recalculated data CGM1
1x10
model results CGM0
model results CGM1
0.0
model results CGM2
0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.35 1.50 Bentonite
1x10-6
Radial distance (m) 0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.35 1.50
Radial distance (m)
+
Fig. 10. Comparison of predicted concentrations of Na (lines)
with inferred aqueous extract data (symbols) at section 29 after Fig. 13. Comparison of predicted concentrations of HCO 3
1930 days of heating using models CGM-0, CGM-1 and CGM-2. (lines) with inferred aqueous extract data (symbols) at section
29 after 1930 days of heating using models CGM-0, CGM-1 and
CGM-2.
0.006
Bentonite Granite
0.005
at t = 1930 days (see Figs. 713). Such a comparison
heater

recalculated data CGM0

0.004
recalculated data CGM1
recalculated data CGM2
provides a unique opportunity to test the THC
K+ (mol/L)

model results CGM0


model results CGM1
model of the engineered barrier system.
0.003 model results CGM2
Spatial distribution of the concentration of a
0.002
conservative species such as Cl is the result of the
combined eect of the following processes: (1) Ben-
0.001 tonite porewater evaporation near the heater, (2)
Water condensation some distance away from hea-
0
0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.35 1.50 ter, (3) Dilution due to inow of granitic groundwa-
Radial distance (m) ter, (4) Advective displacement of the hydration
Fig. 11. Comparison of predicted concentrations of K+ (lines) front, and (5) Solute diusion which tends to dissi-
with inferred aqueous extract data (symbols) at section 29 after pate solute gradients. All of these processes lead to
1930 days of heating using models CGM-0, CGM-1 and CGM-2. a pattern of decreasing Cl concentration with
increasing radial distance (Fig. 7). THC predictions
THC predictions of the in situ test performed for CGM-0 and CGM-2 are slightly larger than
with CGM-0, CGM-1 and CGM-2 are compared those of CGM-1 because initial Cl concentration
to inferred AET geochemical data from section 29 of CGM-0 and CGM-2 (0.187 M) was derived from
J. Samper et al. / Applied Geochemistry 23 (2008) 11861201 1197

squeezing tests while initial Cl concentration for concentrations generally decrease with increasing
CGM-1 was derived from AET and is equal to radial distance and show local minima at
0.155 M (Table 5). The THC model overpredicts r = 0.66 m. The best t to inferred SO2 4 data is
Cl concentrations near the granitebentonite inter- achieved with CGM-1 (Fig. 12).
face (0.75 < r < 1.1 m) while it underpredicts them CGM-1 and CGM-2 lead to similar inferred
for r < 0.75 m. While computed concentrations HCO 3 concentrations. They are more than two
decrease monotonically with distance in a convex orders of magnitude smaller than those inferred
manner, measured values show a maximum at with CGM-0. Such a discrepancy in HCO 3 data is
r = 0.6 m. related to the dierence in CO2(g) conditions (see
Inferred aqueous extract Ca concentrations Table 2). THC predictions with CGM-0 deviate lar-
from section 29 are largest near the heater and gely from the cloud of inferred data (Fig. 13). Pre-
decrease with increasing radial distance. Near the dictions of CGM-1 and CGM-2 capture the trend
bentonitegranite interface they increase again of inferred data, although such data are highly
due to a sampling artefact caused by colloidal par- uncertain.
ticles (Fernandez and Rivas, 2003). Similar to Cl, Calcite which in model CGM-0 is assumed at
inferred Ca data show a maximum at r = 0.6 m. equilibrium (Table 2) precipitates at the node closest
Calcium inferred with CGM-0 and CGM-2 are to the heater and dissolves everywhere in the buer
similar and generally larger than those inferred (Fig. 14). The maximum cumulative precipitation at
with CGM-1. Such a discrepancy in Ca data is the heater interface is 1.6 103 mol/L. Cumulative
related to the fact that CGM-1 assumes that ben- calcite dissolution/precipitation computed with
tonite contains gypsum which may dissolve/precip- models CGM-1 and CGM-2 is extremely small
itate at local equilibrium (see Table 2). THC due to the slow kinetic rate adopted for calcite
predictions with CGM-0 and CGM-1 are similar (see Table 4).
and deviate from the cloud of inferred data. The Dolomite is only considered in models CGM-1
best t to inferred Ca data is achieved with and CGM-2. This mineral dissolves at the benton-
CGM-2 (Fig. 8). itegranite interface with cumulative values of
Inferred Mg concentrations show a maximum at about 103 mol/L and precipitates in the rest of
r = 0.5 m. CGM-0 and CGM-2 lead to similar the buer (not shown here).
inferred Mg concentrations which are generally lar- All models have in common that anhydrite pre-
ger than those inferred with CGM-1. The three cipitates only at the node closest to heater. Cumula-
models lead to dierent predictions for Mg. The tive precipitation is about 0.1 mol/L. In model
best t for this cation is obtained with CGM-1 CGM-1 anhydrite dissolves everywhere in the
(Fig. 9).
In contrast to dissolved Ca and Mg, inferred Na
and K concentrations do not show a maximum at
r = 0.6 m. Sodium data inferred with CGM-0 and 0.0002
CGM-2 are similar and generally larger than those Bentonite Granite
heater

inferred with CGM-1. Predictions, however, are


0.0001
similar for CGM-1 and CGM-2 which are smaller
Calcite (mol/L)

than those obtained with CGM-0. Computed Na


concentrations reproduce the overall trend of 0.0000
inferred data, but largely overpredict them near
the heater (Fig. 10). The best t to inferred Na data
-0.0001 model results CGM0
is achieved with CGM-0 (Fig. 10). model results CGM1
Inferred K concentrations are similar for all model results CGM2

models. Predictions, however, are similar for -0.0002


0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.35 1.50
CGM-1 and CGM-2 and are larger than those Radial distance (m)
obtained with CGM-0. The best t to inferred K
data is obtained with CGM-2 (Fig. 11). Fig. 14. Spatial distribution of computed cumulative calcite
dissolution/precipitation (negative values for dissolution and
Inferred SO24 concentrations with CGM-0 and positive for precipitation) with model CGM-0 after 1930 days of
CGM-2 are similar and generally larger by a factor heating. Precipitation at the heater interface is much larger than
of 23 than those inferred with CGM-1. Inferred dissolution at other places and is not shown here.
1198 J. Samper et al. / Applied Geochemistry 23 (2008) 11861201

40
Bentonite
Concentration of exchanged Ca, X2-Ca,
decreases about 6 meq/L near the heater and
Exchangeable Ca2+ (meq/100 g)

36
Granite increases slightly at the bentoniteconcrete inter-
heater

32 face due to calcite dissolution (Fig. 15). Exchanged


model results CGM0 Mg shows a pattern similar to that of X2-Ca
model results CGM1
28
model results CGM2 (Fig. 16). Exchanged Na, on the other hand,
increases 9 meq/L near the heater and decreases
24
slightly (1 meq/L) at the bentoniteconcrete
20 interface (Fig. 17). Exchanged K does not change
signicantly.
16
0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.35 1.50
Patterns of reactive species show some similari-
Radial distance (m) ties with those of Cl, although their data show
more scatter. Reactive species are aected by ther-
Fig. 15. Spatial distribution of computed concentrations of
interlayer Ca with models CGM-0, CGM-1 and CGM-2 after
mohydrodynamic processes as well as by chemical
1930 days of heating. reactions. Aqueous concentrations increase near
the heater due to evaporation. This leads to precip-
40
itation of carbonate and sulphate minerals. Precipi-
tation of calcite causes a depletion of dissolved Ca
Exchangeable Mg2+ (meq/100 g)

Granite
which is compensated by a release of exchanged
heater

Bentonite Ca, X2-Ca. To ll the sites left by the release of


36
exchanged Ca, dissolved Na gets into the exchanger
and consequently X-Na increases. Since total cation
exchange concentration, CEC, is constant, an
32 increase in X-Na induces also a decrease in X2-Mg.
Mixing of bentonite and granite porewaters at
model results CGM0
model results CGM1 the bentonitegranite interface induces calcite disso-
model results CGM2

28
lution (Fig. 14) and cation exchange. Divalent cat-
0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.35 1.50 ions have larger selectivities than those of
Radial distance (m)
monovalent cations and therefore both exchanged
Fig. 16. Spatial distribution of computed concentrations of Ca and Mg increase while X-Na decreases slightly
interlayer Mg with models CGM-0, CGM-1 and CGM-2 after (see Figs. 1517).
1930 days of heating.
6. Conclusions and uncertainties
50
model results CGM0
model results CGM1
THC predictions of the in situ test performed
Exchangeable Na+ (meq/100 g)

Bentonite
model results CGM2
with three conceptual geochemical models have
heater

40 been compared to measured temperature, water


content and chemical data collected after disman-
tling section 29.
Granite
30
Raw aqueous extract data for section 29 have
been interpreted with an inverse geochemical model
which accounts for geochemical processes taking
place in bentonite samples during aqueous extrac-
20
0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.35 1.50 tion. To ensure consistency between inferred data
Radial distance (m) and model predictions, aqueous extract chemical
Fig. 17. Spatial distribution of computed concentrations of data have been inferred using the same conceptual
interlayer Na with models CGM-0, CGM-1 and CGM-2 at geochemical model used to perform THC
section 29 after 1930 days of heating. predictions.
Estimation errors for inferred chemical data are
buer, but its cumulative dissolution is four orders much smaller than errors associated with the identi-
of magnitude smaller than the amount precipitated cation of the conceptual geochemical model for
at the heater. FEBEX bentonite. For this reason, three dierent
J. Samper et al. / Applied Geochemistry 23 (2008) 11861201 1199

conceptual models have been used to interpret AET. is related to the fact that the model assumes that all
Models reproduce measured aqueous extract data water content is accessible to solutes for transport
except for CGM-1 and CGM-2 which cannot t and geochemical reactions. Compacted bentonite,
HCO 3 data accurately. They lead to similar inferred however, contains dierent types of water such as
results for all chemical species except for HCO3 and interlayer water located at the interlayer space
SO2
4 . Inferred aqueous extract concentrations of between montmorillonite units, double layer and
most species near the heater are larger than at the free water (Bradbury and Baeyens, 2002, 2003;
granitebentonite interface where dilution is much Fernandez et al., 2004; and Fernandez and Rivas,
more pronounced. 2005b). Dual porosity models based on the dual
Model predictions reproduce temperatures and continuum approach of Zheng and Samper (2005)
capture the trends of water content. The model provide a framework for considering free water,
underestimates measured water contents near the interlayer and double layer water.
bentonitegranite interface and overestimates mea- THC predicted concentrations systematically
sured values near the heater. These discrepancies overestimate measured values near the heater and
are caused by the clay swelling/shrinking phenom- show gradients sharper than those of measured
ena which are not taken into account in the current data. This discrepancy could be overcome by con-
THC model. sidering bentonite swelling/shrinking and improving
Predicted Cl concentrations near the granite the estimate of water evaporation at heaterbenton-
bentonite interface overestimate measured values. ite interface.
This discrepancy could be caused by changes in Laboratory experiments and model results
porosity induced by the swelling/shrinking phenom- performed on FEBEX bentonite have provided rm
ena and uncertainties in eective diusion coecient evidence for the relevance of proton surface
of Cl and vapor tortousity. complexation for buering pH and controlling
Computed results of Mg, Na and K with the HCO 3 concentrations (Fernandez, 2004; Samper
three models reproduce the general trends of aque- et al., 2007b). Future THC models should account
ous extract data. Models CGM-0 and CGM-1 t also for surface complexation protonation/
Ca data better than model CGM-2. Predicted deprotonation.
SO2
4 concentrations underestimate systematically THC models could be improved by incorporating
inferred aqueous extract data. Such a discrepancy mechanical and geochemical couplings to account
could be related to the role of gypsum/anhydrite. for porosity changes caused by swelling phenomena.
Bicarbonate inferred concentrations obtained with This leads to fully coupled THMC models (Zheng,
model CGM-0 are greater than those from models 2006). Preliminary results from a coupled THMC
CGM-1 and CGM-2. This discrepancy is related model of a heating and hydration experiment on
to the assumption on CO2(g). FEBEX bentonite indicate that geochemical results
Models CGM-1 and CGM-2 are less likely to improve when changes in porosity caused by swell-
represent FEBEX bentonite porewater due to their ing are considered (Zheng, 2006).
limitations in reproducing aqueous extract HCO 3 As indicated by an anonymous reviewer, high
data. On the other hand, models CGM-1 and espe- temperatures close to the heater may cause CO2
cially CGM-2 lead to the best t to inferred chemi- degassing from liquid water. Such CO2 will trans-
cal data of section 29. Therefore, since there is no port through the gas phase and eventually redissolve
single model which performs well for inferring aque- in liquid water again when it reaches the condensa-
ous extract data and reproduces in situ test data, it tion zone. Future models should account for CO2
cannot be clearly concluded which conceptual geo- degassing and dissolution processes which aect
chemical model is the best. Future studies should pH, dissolved HCO 3 and calcite dissolution and
improve CGM by identifying the role of sulfate precipitation.
minerals and ascertaining the role of gaseous The THC model has been tested with data from
phases. section 29 of the FEBEX in situ test. Available data
THC model predictions have been tested with indicate that there are chemical heterogeneities
aqueous extract data which in turn must be previ- along the buer which remain to be analyzed (Fern-
ously inferred with a hydrochemical model. This is andez and Rivas, 2003).
one of the main sources of uncertainty in the THC It should be pointed out that the THC model
model for bentonite. Another source of uncertainty presented here was aimed at evaluating the geo-
1200 J. Samper et al. / Applied Geochemistry 23 (2008) 11861201

chemical evolution of the clay barrier during the Dai, Z., Samper, J., Ritzi, R., 2006. Identifying geochemical
transient thermal and hydration stage. Performance processes by inverse modeling of multicomponent reactive
transport in Aquia aquifer. Geosphere 4, 210219.
assessment of the long-term geochemical evolution ENRESA, 1994. Almacenamiento geologico profundo de resid-
of a HLW repository requires models which uos radiactivos de alta actividad (AGP). Conceptos prelim-
account for processes not considered here such as inares de referencia. ENRESA Technical Publication 07-94.
redox reactions and interactions of canister corro- ENRESA, 2000. Full-scale engineered barriers experiment for a
sion products with bentonite (see Samper et al., deep geological repository in crystalline host rock (FEBEX
Project). European Commission. EUR 19147 EN.
2008). ENRESA, 2006a. FEBEX: Updated nal report. ENRESA
Technical Publication PT 05-0/2006.
ENRESA, 2006b. FEBEX: Post-mortem bentonite analysis.
Acknowledgements ENRESA Technical Publication PT 05-1/2006.
ENRESA, 2006c. FEBEX: Final report on thermo-hydro-
This research was performed within the frame- mechanical modelling. ENRESA Technical Publication PT
work of FEBEX project funded by ENRESA and 05-2/2006.
European Union through Contracts FI4W-CT95- ENRESA, 2006d. FEBEX: Final THG modelling report. ENRE-
SA Technical Publication PT 05-3/2006.
0006 and FIKW-CT-2000-00016 of the Nuclear Fis- Fernandez, A.M., 2004. Caracterizacion y modelizacion del agua
sion Program. We acknowledge contributions of intersticial de materiales arcillosos: Estudio de la bentonita de
UDC researchers who participated throughout the Cortijo de Archidona. Ph.D. Dissertation. Editorial CIE-
various stages of the FEBEX project. We thank MAT.
Wilfried Pngstein and a second anonymous re- Fernandez, A.M., Rivas, P., 2003. Task 141: post-mortem
bentonite analysis. Geochemical behaviour. CIEMAT Inter-
viewer for comments, suggestions and recommenda- nal Note 70-IMA-L-0-107 v0.
tions which have contributed to greatly improve the Fernandez, A.M., Rivas, P., 2005a. Pore water chemistry of
paper. saturated FEBEX bentonite compacted at dierent densities.
In: Alonso, E.E., Ledesma, A. (Eds.), Advances in Under-
standing Engineered Clay Barriers. A.A. Balkema Publishers,
References Leiden, The Netherlands, pp. 505514.
Fernandez, A.M., Rivas, P., 2005b. Analysis and distribution of
Arcos, D., Bruno, J., Karnland, O., 2003. Geochemical model of waters in the compacted FEBEX bentonite: pore water
the granitebentonite-groundwater interaction at Aspo HRL chemistry and adsorbed water properties. In: Alonso, E.E.,
(LOT experiment). Appl. Clay Sci. 23, 219228. Ledesma, A. (Eds.), Advances in Understanding Engineered
Ayora, C., Taberner, C., Samper, J., 1994. Modelizacion de Clay Barriers. A.A. Balkema Publishers, Leiden, The Neth-
transporte reactivo: Aplicacion a la dedolomitizacion. Estu- erlands, pp. 257275.
dios Geologicos 50, 397410. Fernandez, A.M., Baeyens, B., Bradbury, M., Rivas, P., 2004.
Baeyens, B., Bradbury, M.H., 1991. A physico-chemical charac- Analysis of the pore water chemical composition of a Spanish
terisation technique for determining the porewater chemistry compacted bentonite used in an engineered barrier. Phys.
in argillaceous rock. NAGRA Technical Report 90-40:61. Chem. Earth 29, 105118.
Bradbury, M.H., Baeyens, B., 1997. Derivation of in situ Opali- Gens, A., Guimarares, L., do, N., Olivella, S., Sanchez, M., 2004.
nus Clay porewater composition from experimental and Analysis of the THMC behaviour of compacted swelling clay
geochemical modelling studies. NAGRA Technical Report for radioactive waste isolation. In: Stephansson, O., Hudson,
97-07:50. J.A., Jing, L. (Eds.), Coupled Thermo-Hydro-Mechanical-
Bradbury, M.H., Baeyens, B., 1998. A physicochemical charac- Chemical Processes in Geo-Systems, Fundamentals, Model-
terisation and geochemical modelling approach for determin- ling, Experiments and Applications. Elsevier Geo-Engineer-
ing porewater chemistries in argillaceous rocks. Geochim. ing Book Series, vol. 2. Elsevier, pp. 317322.
Cosmochim. Acta 62, 783795. Karnland, O., Sanden, T., Johannesson, L.E., Eriksen, T.E.,
Bradbury, M.H., Baeyens, B., 2002. Porewater chemistry in Janson, M., Wold, S., Pedersen, K., Motamedi, M., Rosborg,
compacted re-saturated MX-80 bentonite: physico-chemical B., 2000. Long term test of buer material. Final report on the
characterisation and geochemical modelling. PSI Bericht Nr. pilot parcels. SKB Technical Report 00-22.
02-10. NAGRA Technical Report 01-08. Molinero, J., Samper, J., 2004. Groundwater ow and solute
Bradbury, M.H., Baeyens, B., 2003. Porewater chemistry in transport in fracture zones: an improved model for a large-
compacted resaturated MX-80 bentonite. J. Contam. Hydrol. scale eld experiment at Aspo (Sweden). J. Hydraulic Res. 42
61, 329338. (Extra Issue), 157172.
Dai, Z., Samper, J., 2004. Inverse problem of multicomponent Molinero, J., Samper, J., 2006. Modeling of reactive solute
reactive chemical transport in porous media: formulation and transport in fracture zones of granitic bedrocks. J. Contam.
applications. Water Resour. Res. 40, W07407. doi:10.1029/ Hydrol. 82, 293318.
2004WR003248. Molinero, J., Samper, J., Yang, C., Zhang, G., 2004. Biogeo-
Dai, Z., Samper, J., 2006. Inverse modeling of water ow and chemical reactive transport model of the Redox zone exper-
multicomponent reactive transport in coastal aquifer systems. iment of the Aspo hard rock laboratory (Sweden). Nucl.
J. Hydrol. 327 (34), 447461. Technol. 48, 151165.
J. Samper et al. / Applied Geochemistry 23 (2008) 11861201 1201

Navarro, V., Alonso, E.E., 2000. Modeling swelling soils for Samper, J., Zheng, L., Fernandez, A.M., Montenegro, L., 2007b.
disposal barriers. Comput. Geotech. 27, 1943. Inverse modeling of multicomponent reactive transport
Pearson, F.J., Scholtis, A., 1995. Controls on the chemistry of through single and dual porosity media. J. Contam. Hydrol.
pore water in a marl of very low permeability. In: Kharaka, (in review).
Y.K., Chudaev, O.V. (Eds.), WaterRock Interaction: Pro- Samper, J., Montenegro, L., Zheng, L., Yang, C., Alonso, J.,
ceedings of 8th International Symposium on WaterRock 2007c. Lessons learnt from the development and application
Interaction, Vladivostok, Russia. Balkema, Rotterdam, pp. of reactive solute transport and geochemical models of
3538. dierent levels of complexity for the EBS. In: Proceedings
Pearson, F.J., Arcos, D., Bath, A., Boisson, J.-Y., Fernandez, of Workshop Engineered Barrier Systems (EBS) in the Safety
A.M., Gabler, H.-E., Gaucher, E., Gautschi, A., Griault, L., Case: The Role of Modelling, Nuclear Energy Agency,
Hernan, P., Waber, H.N., 2003. Geochemistry of Water in the Radioactive Waste Management, OECD, pp. 6184.
Opalinus Clay Formation at the Mont Terri Rock Labora- Samper, J., Lu, C., Montenegro, L., 2008. Reactive transport
tory. Swiss Federal Oce for Water and Geology Series N 5. model of interactions of corrosion products and bentonite.
Plummer, N.L., Parkhurst, D.L., Wigley, T.M.L., 1979. Critical Phys. Chem. Earth, accepted for publication.
review of the kinetics of calcite dissolution and precipitation. Van Humbeeck, H., Verstricht, J., Li Xiang Ling, Bernier, F., De
In: Jenne, E.A. (Ed.), Chemical Modelling in Aqueous Canniere, P., 2006. The OPHELIE mock-up. Euridice News
Systems. Am. Chem. Soc. Symp. 93, 537573. Nr 3, April.
Sacchi, E., Michelot, J.L., 2000. Porewater extraction from Verstricht, J., Van Humbeeck, H., Li, X.L., De Canniere, P.,
argillaceous rocks for geochemical characterisation. Radio- 2007. Final report on the mock-up OPHELIE, EIG EURI-
act. Waste Manage. NEA. DICE.
Sacchi, E., Michelot, J.L., Pitsch, H., Lalieux, P., Aranyossy, Wolery, T.J., 1979. Calculation of chemical equilibrium between
J.F., 2001. Extraction of water and solutes from argillaceous aqueous solution and minerals: The EQ3/6 software package.
rocks for geochemical characterization: methods, processes, Lawrence Livermore National Laboratory Report UCRL-
and current understanding. Hydrogeol. J. 9, 1733. 52658.
Samper, J., Yang, C., 2006. Stochastic analysis of transport and Yang, C., Samper, J., Molinero, J., Bonilla, M., 2007a. Modelling
multicomponent competitive monovalent cation exchange in geochemical and microbial consumption of dissolved oxygen
aquifers. Geosphere 2, 102112. after backlling a high level radioactive waste repository. J.
Samper, J., Juncosa, R., Delgado, J., Montenegro, L., 2000. Contam. Hydrol. 93, 130148.
CORE2D: A code for non-isothermal water ow and reactive Yang, C., Samper, J., Montenegro, L., 2007b. A coupled non-
transport. Users manual version 2. ENRESA Technical isothermal reactive transport model for long-term geochem-
Report 6/2000. ical evolution of a HLW repository in clay. Environ. Geol.
Samper, J., Fernandez, A.M., Zheng, L., Montenegro, L., Rivas, doi:10.1007/s00254-007-0770-2.
P., 2005a. Testing coupled thermo-hydro-geochemical models Zheng, L., 2006. Coupled thermo-hydro-mechanical-geochemical
with geochemical data from FEBEX in situ test. In: Alonso, models for structured deformable porous media. Ph.D Dis-
E.E., Ledesma, A. (Eds.), Advances in Understanding Engi- sertation, Univ. de A Coruna, La Coruna, Spain.
neered Clay Barriers. A.A. Balkema Publishers, Leiden, The Zheng, G., Samper, J., Montenegro, L., 2008. Coupled thermo-
Netherlands, pp. 565575. hydro-bio-geochemical reactive transport model of the CER-
Samper, J., Vazquez, A., Montenegro, L., 2005b. Inverse BERUS heating and radiation experiment in Boom clay.
hydrochemical modelling of aqueous extracts experiments Appl. Geochem. in press. doi:10.1016/j.apgeochem.2007.09.
for the estimation of FEBEX bentonite pore water chemistry. 010.
In: Alonso, E.E., Ledesma, A. (Eds.), Advances in Under- Zheng, L., Samper, J., 2004. Formulation of the inverse problem
standing Engineered Clay Barriers. A.A. Balkema Publishers, of non-isothermal multiphase ow and reactive transport in
Leiden, The Netherlands, pp. 553563. porous media. In: Miller, C.T., Farthing, M.W., Gray, W.G.,
Samper, J., Zheng, L., Montenegro, L., Fernandez, A.M., Rivas, Pinder, G.F. (Eds.), Computational Methods in Water
P., Dai, Z., 2005c. Direct and inverse modelling of multi- Resources, vol. 2. Elsevier, pp. 13171327.
component reactive transport in single and dual porosity Zheng, L., Samper, J., 2005. A dual continuum coupled multi-
media. In: Alonso, E.E., Ledesma, A. (Eds.), Advances in phase ow model with mixed second order water transfer
Understanding Engineered Clay Barriers. A.A. Balkema term for structured soils: part II: testing with synthetic cases
Publishers, Leiden, The Netherlands, pp. 493503. and application to a real experiment. In: VII Jornadas de
Samper, J., Zhang, G., Montenegro, L., 2006. Coupled microbial Zona no Saturada, ZNS05, Coruna, Spain, pp. 301306.
and geochemical reactive transport models in porous media: Zheng, L., Samper, J., Montenegro, L., 2008. Inverse hydro-
formulation and application to synthetic and in situ experi- chemical models of aqueous extracts tests. Phys. Chem. Earth,
ments. J. Iberian Geol. 32, 211217. accepted for publication.

S-ar putea să vă placă și