Anda di halaman 1dari 12

Journal of Colloid and Interface Science 297 (2006) 353–364

www.elsevier.com/locate/jcis

Electrokinetic phenomena in saturated compact clays


M. Rosanne, M. Paszkuta, P.M. Adler ∗
Institut de Physique du Globe de Paris (IPGP), Laboratoire des Milieux Poreux et Fracturés, Boîte 89, 4 place Jussieu, 75252 Paris cedex 05, France
Received 9 September 2005; accepted 11 October 2005
Available online 5 December 2005

Abstract
The membrane potential, the pressure difference, and the concentration difference induced by an applied concentration gradient through samples
of compact clay were measured as functions of sodium chloride concentration and of porosity. The results of some previous numerical predictions
are recalled to be functions of a characteristic length scale which can be derived from conductivity and permeability. Generally, the experimental
data were in agreement with these numerical predictions. Therefore, nondiagonal coupling coefficients can be derived with acceptable precision
from the diagonal coefficients.
 2005 Elsevier Inc. All rights reserved.

Keywords: Porous media; Coupled transports; Clay; Experimental

1. Introduction write

The ability of clay soils to act as “perfect” or “ideal” mem- I = γ11 · E + γ12 · (−∇P ) + γ13 · (−∇n), (1a)
branes that inhibit the passage of electrolytes is of great value U = γ21 · E + γ22 · (−∇P ) + γ23 · (−∇n), (1b)
for some possible storage of toxic wastes. Clays exhibit mem-
brane properties when charged ionic species are excluded from Jd = γ31 · E + γ32 · (−∇P ) + γ33 · (−∇n), (1c)
the pores, while uncharged species have a relatively free move- where k is the Boltzmann constant (=1.38 × 10−23 m2 kg
ment. s−2 K−1 ). A distinction is generally made between the diago-
The major purpose of this work was to experimentally de- nal phenomenological coefficients γij with i = j , and the non-
termine the coefficients which characterize coupled transports diagonal ones with i = j . Because of the Onsager symmetry
in compact clays and to compare them to some numerical pre- properties, the nondiagonal coefficients are not all independent.
dictions derived by [1,2] for various kinds of porous media. A literature search shows that measurements of these co-
More precisely, consider a porous medium filled by an elec- efficients for clays (and especially for compact clays) were
trolyte solution with a concentration n (defined as the number not frequently performed. Interesting contributions concerning
of molecules per m3 ). When a pressure gradient ∇P , an electric electrokinetic coefficients were made by Kemper and Quirk [4],
field E and a concentration gradient ∇n are simultaneously ap- Elrick et al. [5], Groenvelt et al. [6], and Sherwood and Cras-
plied to the porous medium, three fluxes are generated, namely, ter [7]. A recent review of all of these coefficients was written
a flow characterized by the seepage velocity U (m s−1 ), a solute by Horseman et al. [8].
flux Jd which is the number of molecules per unit surface and This paper is organized as follows. Section 2 summarizes the
per unit time (m−2 s−1 ) and a current density I (A m−2 ). Close theoretical determination of the coupling coefficients. The lo-
to equilibrium, when gradients are not too large, the problem is cal transport equations corresponding to the electrical potential,
linear and fluxes are proportional to these gradients (see [3]). the ionic concentrations, and the velocities are the Poisson–
Then, with respect to the solvent-fixed reference plane, one can Boltzmann, the convection–diffusion, and the Stokes equations,
respectively. Numerical results for various porous media corre-
* Corresponding author. spond to the analytical solutions valid for circular Poiseuille
E-mail address: adler@ipgp.jussieu.fr (P.M. Adler). flows when the capillary radius is replaced by a suitable length
0021-9797/$ – see front matter  2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcis.2005.10.026
354 M. Rosanne et al. / Journal of Colloid and Interface Science 297 (2006) 353–364

scale. The material, the experimental procedure and the data are Marino et al. [2] (and references therein),
described in Section 3. Two samples of argilite extracted from
 γ11 µe2
a Callovo–Oxfordian formation were characterized. Section 4 γ11 =
is devoted to the analysis of the effective diffusion coefficient, (el kT κ)2
osmotic coefficient and coefficient related to “membrane poten- 1   ζ   
= D1 + D2 − D1 − D2 I1 (κRc )
tial.” Results are considered as functions of solute concentration 2 κRc I0 (κRc )
and porosity in isothermal conditions. Data relative to the elec-   
λ
trokinetic coefficients are compared to numerical and analytical + 2ζ  3
− I1 (κRc ) , (6a)
κRc I0 (κRc )
results derived by [1,2]. The paper ends with some concluding  
 γ21 µ   2I1 (κRc )
remarks. γ21 = = γ 12 = ζ − 1 , (6b)
el ζ Kκ 2 κRc I0 (κRc )
 γ13 µe     I1 (κRc )
2. Theoretical background γ13 = 2
= D1 − D2 − 2ζ  D1 + D2 ,
el (kT ) κRc I0 (κRc )
2.1. Analysis on the pore scale (6c)
 
 γ23 µκ 2 2I (κR )
= 4ζ 
1 c
γ23 = −1 , (6d)
The coefficients γij defined by (1) can be derived by ana- kT κRc I0 (κRc )
lyzing the phenomena on the pore level. Such an analysis was
 γ33 µe2 1    I1 (κRc )
performed in [2]. If no temperature gradient is imposed on γ33 = 2
= D1 + D2 − ζ  D1 − D2 ,
el (kT ) 2 κRc I0 (κRc )
the medium, the transport is governed by three types of cou-
(6e)
pled equations. First, the concentration ni (R, t) of each ith ion
species is the solution of a convection–diffusion equation where κ −1 is the Debye–Hückel length. The dimensionless val-
  ues γij (i, j = 1, . . . , 3) are defined by the first equalities and
∂ni Di marked with a prime.
+ ∇ · −Di ∇ni − ezi ni ∇ψ + ni u = 0,
∂t kT It is important to note that most numerical results obtained
i = 1, 2, . . . , N, (2) for various types of porous media were shown to be very close
to the analytical results (6) when the radius of the tube is re-
where t is the time, Di the diffusion coefficient of ion i, e the
placed by the characteristic length Λ , defined by
absolute value of the electron charge, zi the valency, ψ the elec-
 
tric field, and u the fluid velocity. N is the number of ionic  12.5Kσ ∞ 1/2
species; it is equal to 2 in this paper. Λ = , (7)
σ
Since the Reynolds number is in most cases much less than
unity, u and the pressure P (R, t) are governed by modified where σ ∞ is the fluid conductivity, σ the macroscopic conduc-
Stokes equations, tivity of the medium filled by the electrolyte, and K the perme-
ability of the porous medium. The definition of this length scale
∇ · u = 0, µ∇ 2 u = ∇P + ρ∇ψ, (3) was inspired by the length scale Λ introduced by [9]. Note that
Λ is homogeneous to a length. The use of Λ was already ex-
where µ is the fluid viscosity and ρ the charge density. perimentally justified for electroosmotic phenomena in [10].
Finally, ψ(R, t) is the solution of the Poisson–Boltzmann Therefore, the major purpose of this work is to check this
equation property with data obtained for compact clays.
e 
N
ρ
∇ 2ψ = − =− ni zi (4) 2.2. Analysis on the macroscopic scale
el el
i=1

where el is the fluid permittivity. Consider now an experiment performed on a porous sam-
ple of thickness 2l and cross-sectional area S separating two
To be solved, the set of Eqs. (2)–(4) is subject to the bound-
compartments each of volume V filled by aqueous solutions at
ary conditions on the solid/fluid interface Si
different concentrations n10 and n20 of the same electrolyte.
ν · ji = 0, u = 0, and ψ = ζ, (5) The system of Eqs. (1a)–(1c) must be completed by the one
dimensional continuity equations
where ν is the outward normal to the solid surface, and ζ the
zeta potential. ∂I
= 0, (8a)
These equations were linearized around equilibrium, made ∂x
dimensionless, and solved by [2] (and the references therein) ∂
· U = 0, (8b)
for spatially periodic porous media. The coefficients γij (i, j = ∂x
1, . . . , 3) can be derived by integrating the local fields over the ∂n ∂
+ · JL = 0. (8c)
unit cell. ∂t ∂x
The only cases which can be solved analytically, are the ones The initial concentrations are
of a plane channel or of a circular tube of radius Rc . Let us recall
the coefficients given in the latter case by Coelho et al. [1] and n(x = −l, t = 0) = n10 , n(x = l, t = 0) = n20 . (9)
M. Rosanne et al. / Journal of Colloid and Interface Science 297 (2006) 353–364 355

In addition, we used the boundary conditions at x = ±l Neither ω nor τ depends on the formation factor F . However,
∂n it is important to note that ω is a nondimensional parameter,
±a = JL , (10a) whereas τ is expressed in A−1 s−1 m3 .
∂t
We interpreted ω by considering the flow of solvent to be a
P = Pa , (10b)
consequence of the interplay of the applied concentration gra-
where Pa is the atmospheric pressure and the parameter a is dient of the solute and the interaction forces between the solute
defined as and the charged pore surfaces, which causes both the pressure
V gradient and the electrical field [11]. When the clay is a “per-
a= . (11) fect” or “ideal” membrane, i.e., totally impermeable to the
S
solute, ω is unity and the solute flux is equal to zero. For a clay
Let V be the common volume of the two reservoirs and S
that is completely permeable to the solute, ω = 0. However, in
the cross section of the porous medium.
most real systems, the solute permeates the clay to some ex-
tent and ω falls within the range 0 < ω < 1. Thus, the clay is
2.2.1. General case where U (t) = 0
considered as a “nonideal” membrane [12].
Let us assume that the processus is mostly one-dimensional,
The degree to which the clay prevents ions from entering
along the x-axis.
into the pore space is termed its “efficiency.” Hence, if a clay is
Since no external electric field is applied and since both solu-
an ideal membrane, its efficiency is 100%.
tions in the reservoirs are insulated from each other, the current
Moreover, substitution of Eqs. (16) and (17), after rearrange-
density derived from (8a) is equal to zero,
ment with relations (18a) and (18b), into Eqs. (1b) and (1c)
I (t) = 0. (12) yields the set of coupled equations
Therefore, Eq. (1a) can be written as ∂P ∂n
U = −K + 2kT ωK , (19a)
∂ψ γ12 ∂P γ13 ∂n  ∂x  ∂x
=− − . (13) ∂n
∂x γ11 ∂x γ11 ∂x Jd = D̄ − − nωU, (19b)
∂x
Integrating Eqs. (13) and (1b) along the x-direction and as-
where K is given by
suming as a first approximation that the coefficients γij do not
depend on x yields γ12 γ21
K = γ22 − . (19c)
γ12 γ13 γ11
ψ = − P − n, (14a)
γ11 γ11 D̄, which was called the dispersion coefficient in [6], is defined
lU (t) = −γ21 ψ − γ23 n. (14b) by
With the help of Eqs. (1a) and (1b), one obtains D̄ = (γ33 − nτ γ13 ) + nωγ23
∂ψ ∂P ∂n = (γ33 − 2kT τ γ31 ) + 2kT ωγ32 . (19d)
γ11 + γ12 = −γ13 , (15a)
∂x ∂x ∂x K contains the contribution of self-induced electroosmosis
∂ψ ∂P ∂n
γ21 + γ22 = −U (t) − γ23 . (15b) to the water flow. D̄ contains the effect of salt sieving formu-
∂x ∂x ∂x lated by ωγ32 , due to the presence of a hydrostatic pressure
Rearrangement of Eqs. (15a) and (15b) yields the follow- difference and the contribution to Jd of self-induced electrod-
ing expressions for the electrical potential gradient ∂ψ
∂x and the iffusion, formulated by τ γ31 . Recall that an ideal membrane
pressure gradient ∂P
∂x in terms of the concentration gradient and completely prohibits the passage of solutes. Thus, by definition,
seepage velocity, D̄ must approach zero for an ideal membrane.
The governing one-dimensional partial differential equation
∂ψ (−γ13 γ22 + γ12 γ23 ) ∂x
∂n
+ γ12 U is obtained by substituting Eq. (19b) into the mass balance ex-
= , (16)
∂x γ11 γ22 − γ12 γ21 pression (8c),
∂P (−γ13 γ21 + γ11 γ23 ) ∂x
∂n
+ γ11 U   
=− . (17) ∂n ∂ ∂n
∂x γ11 γ22 − γ12 γ21 + D̄ − − nωU = 0. (20)
∂x ∂x ∂x
The membrane behavior of the clay, i.e., its ability to be par-
tially or totally impermeable to the solutes, is generally quan- 2.2.2. Special case where U (t) = 0
tified by two parameters ω and τ , commonly referred to as the Since the electric current through the pore must be zero in
“reflection coefficient” and the “charge fluidity,” respectively. the absence of any externally applied electrical field and since
They are defined as (see [6]) U (t) is assumed to be zero, Eqs. (15a) and (15b) reduce to
   
1 P 1 γ21 γ13 − γ23 γ11 ∂ψ ∂P ∂n
ω= = , (18a) γ11 + γ12 = −γ13 , (21a)
2kT n I =0,U =0 2kT γ22 γ11 − γ12 γ21 ∂x ∂x ∂x
   
1 ψ 1 γ22 γ13 − γ23 γ12 ∂ψ
+ γ22
∂P ∂n
= −γ23 .
τ =− = . (18b) γ21 (21b)
2kT n I =0,U =0 2kT γ22 γ11 − γ12 γ21 ∂x ∂x ∂x
356 M. Rosanne et al. / Journal of Colloid and Interface Science 297 (2006) 353–364

The solute flux Jd is expressed by


 
∂n
Jd = D̄ − . (22)
∂x
Thus, the mass balance expression is a classical diffusion
equation of the form
  
∂n ∂ ∂n
+ D̄ − = 0, (23)
∂x ∂x ∂x

where D̄ is the same as in Eq. (19d).


In order to determine D̄, a simple one-dimensional model
was used [13] with the boundary conditions
∂n ∂n
±a = D̄ for i = 1, 2 and x = ±l. (24) Fig. 1. The zeta potential ζ as a function of NaCl concentration C. Data are for
∂t ∂x
EST104/2364 (2) and EST104/2487 (").
Equation (23), together with the boundary and initial con-
ditions (24) and (9), is solved numerically by a finite volume
The solute was sodium chloride supplied by SIGMA (pu-
technique. D̄ was calculated from the experimentally measured
rity 99.5%). The solvent was pure water filtered by an HPCL
concentration difference C by means of an elementary pro-
Maxima unit. The concentration C ranged between 10−4 and
gram. Generally speaking, D̄ corresponds to the best fit be-
10−2 mol l−1 .
tween the experimental slope and the slope of the numerical
The zeta potential ζ was estimated by measuring the elec-
solution.
trophoretic mobility of clay particles in electrolyte solutions.
Since the ratio between κ −1 and the particle dimension is small,
3. Experimental
the Smoluchowski formula [14] can be used for all particle
shapes with an estimated precision of 10%,
The phenomenological coefficients γij were measured in a
µue
series of one-dimensional experiments performed on porous ζ= , (27)
samples of thickness 2l and cross-sectional area S separating el
two compartments, each of volume V , filled by aqueous so- where ue is the electrophoretic mobility.
lutions at different concentrations of the same electrolyte. The Results for ζ in various NaCl solutions are displayed in
concentration n expressed in m−3 is related to the molar con- Fig. 1. |ζ | is an increasing function of the concentration; it
centration C expressed in mol m−3 by changes by only 7 mV (−18 to −25 mV) over a concentra-
tion range of 10−4 to 10−2 mol l−1 at a constant pH value equal
C = n/Na , (25) to 5.
where Na is the Avogadro number (=6.02 × 1023 mol−1 ).
The relationships between the coefficients γij and the mea- 3.2. Measurements
surable clay parameters can be summarized by
The salt concentration was determined by electrical conduc-

 σ γ12 γ13

γ11 γ12 γ13 tivity measurements made with a pair of platinum electrodes
γ21 γ22 γ23 =  γ12 K
µ γ23  . (26) related to a Consort P407 conductimeter by means of a cali-
γ31 γ32 γ33 n n
2kT γ13 2kT γ23 γ33 bration curve. The voltage across the reversible Ag/AgCl elec-
trodes was measured using a Keitley 175A multimeter (from 1
3.1. Materials to 200 mV full scale). We observed an asymmetry potential of
about 0.5 mV between the electrodes when zero concentration
The material, supplied to us by ANDRA, is a compact was imposed across the clay. We treated this potential as an ex-
argilite extracted in east France from a Callovo–Oxfordian for- perimental error. The hydrostatic pressure difference between
mation. The wellbore, referred to as EST104, was sampled at both sides of the clay sample was measured continuously with
two different levels, 483.6 m (EST104/2364) and 505.15 m a Kimo 25 differential pressure transducer (from 0 to 600 Pa
(EST104/2487). Both cores were used to obtain powder or solid full scale). Solute conductivity σf , voltage differences ψmeas ,
cylinders of thickness 3 mm and diameter 12 mm. A SEM and pressure difference P were measured once a day.
analysis of the argilite powder shows a heterogeneous compo-
sition; owing to the careful crushing process to obtain powder 3.3. Ag/AgCl electrodes
from the original block, the average grain radius ranges from
1 to 10 µm. The cation exchange capacity (CEC) of argilite 3.3.1. Electrode potential
was found (data from ANDRA) to be 25 meq 100 g−1 for The absolute potential ψm/s of an electrode is defined as
EST104/2364 and 17 meq 100 g−1 for EST104/2487. the potential difference between the metal (m), when immersed
M. Rosanne et al. / Journal of Colloid and Interface Science 297 (2006) 353–364 357

in the solution, and the solution (s) [15],

ψm/s = ψm − ψs . (28)
One cannot measure the equilibrium potential of a metal
with respect to a given solution, but only the potential difference
between the metals (m1) and (m2) of two electrodes immersed
in the same solution,

ψm1/m2 = ψm1 − ψm2 = ψm1/s − ψm2/s . (29)


The electrode potential, denoted by ψN , is the potential dif-
ference between the electrode and a second electrode taken as
a reference,

ψN = ψm1/s − ψref/s . (30)


Fig. 2. Example of calibration of an Ag/AgCl membrane.
This electrode potential can also be expressed by means of
the Nernst law, which is presented for simplicity as a function 3.4. Membrane potential determination
of the concentration C (mol l−1 ),
kT The total measured potential difference ψmeas induced by a
ψN = ψm
0
− ln C, (31) concentration gradient was measured by means of two Ag/AgCl
e
electrodes as described previously. These electrodes are im-
0 is the standard potential of the electrode, k the Boltz-
where ψm mersed in two solutions of concentrations C1 and C2 . Hence,
mann constant (=1.38 × 10−23 J K−1 ), e the elementary charge
(=1.6 × 10−19 C), and T the absolute temperature (K). ψmeas = ψm1 − ψm2 = ψm1 − ψs1 + ψs1 − (ψm2 − ψs2 )
− ψs2
3.3.2. Electrode preparation and test = ψN1 − ψN2 + (ψs1 − ψs2 ) = ψN + ψ. (34a)
Silver membranes made by SPI (SPI-PoreTM Silver Mem-
brane Filters) were coated by electrolysis with silver chloride Thus, ψmeas consists of the electrode potential difference
(AgCl) in a 10−1 mol l−1 KCl solution using an external direct ψN referred to as the “Nernst potential,” and the so-called
current voltage of 1 V. membrane potential difference ψ = ψs1 − ψs2 .
The Ag/AgCl electrodes obtained by this method have been The electrode potential difference ψN is given by the
systematically compared to a calomel electrode. According to Nernst equation (31); therefore, when the two electrodes have
the electrochemistry convention, the positive pole was put at identical properties,
the calomel electrode. The theoretical electrode potential can kT C1
ψN = − ln . (35)
be expressed as e C2
kT Hence, the membrane potential difference ψ can be ex-
ψN = ψcal
0
− ψAgCl
0
+ ln C, (32) pressed as
e
0 = 0.244 V and ψ 0 kT C1
where ψcal AgCl = 0.220 V are the standard ψ = ψmeas + ln . (36)
potentials of the calomel and AgCl electrodes, respectively. e C2
The electrodes which were used to measure the potentials
4. Results and discussion
were thoroughly checked. The results were systematically fitted
by the relation
4.1. Experimental results
ψN = A + B ln C. (33)
Fig. 3 shows typical results obtained when a concentration
Selected membranes give values of A and B close to Ath = gradient is imposed across the clay sample; the evolutions of
22.34 mV and Bth = 25.7 mV, respectively, for a tempera- C = C1 − C2 , P , and ψmeas were recorded. After the tran-
ture equal to 298 K. An example of the response obtained by sient period observed during the first 100 h, these evolutions are
such an Ag/AgCl electrode is displayed in Fig. 2. The least- linear with time and decrease very slightly.
squares fit of the data by Eq. (33) gives A = 18.57 ± 3 mV C, P , and ψ decrease at the same rate and the ratios
ψ
and B = 25.1 ± 1 mV. The correlation coefficient r 2 is equal P
C and C appear to be constant, as one can observe in Fig. 4.
to 0.998. The agreement between the experimental and theoret- This is in agreement with the theoretical predictions [4].
ical values is good. However, it should be noted that this good The membrane potential plays an important role in the abil-
agreement required some care; for example, electrodes were ity of a medium to restrict solute transport. To explain its phys-
fabricated with reduced current intensity in order to obtain a ical meaning, we can consider the following two limiting con-
uniform deposit. ditions [11].
358 M. Rosanne et al. / Journal of Colloid and Interface Science 297 (2006) 353–364

(a) (b)

(c)

Fig. 3. A typical example of the evolution with time of the measured parameters: (a) C = C1 − C2 , (b) P , (c) ψ(= ψmeas − ψN ). (!) obtained by
subtracting from the measured potential difference ψmeas (∗) the electrode potential difference ψN (1).

(a) (b)

Fig. 4. Evolution of P / C (a) and ψ/ C (b) with time.

In the limit of extremely low concentration, the co-ions are almost the entire pore volume is electrically neutral. Thus, the
almost completely excluded from the pores, because of the large potential that arises across the medium is nothing but the “free
double-layer repulsion. Therefore, assuming ideal behavior, diffusional potential” given by
kT  
ψmax = − ln(C1 /C2 ) = −2.57 × 10−2 ln(C1 /C2 ). (37) kT D1 − D2
e ψmin = − ln(C1 /C2 )
e D1 + D2
On the other hand, in the limit of large concentrations, the
electrical double layers formed in the pore are compressed and = 5.18 × 10−3 ln(C1 /C2 ), (38)
M. Rosanne et al. / Journal of Colloid and Interface Science 297 (2006) 353–364 359

where D1 and D2 are the diffusion coefficients in a free medium To take into account the dependence of ψ on C1 /C2 , C̄,
of cations and anions, respectively. and , ψ/ ln(C1 /C2 ) is plotted in Fig. 6 as a function of κΛ
In the intermediate concentration range, the induced poten- for both samples EST104/2487 and EST104/2364. The accu-
tial has values between the two former limits. racy on ψ/ ln(C1 /C2 ) is about 10–30%.
Generally, the membrane potential ψ can be expressed in Results seem to be independent of the clay sample, but de-
terms of the transference number t± of the cations (t+ ) and the pend on C1 /C2 and C̄. | ψ| increases as C1 /C2 increases.
anions (t− ) as follows: ψ/ ln(C1 /C2 ) displays asymptotic behavior at large and
kT small κΛ , while it varies quite rapidly in the intermediate range
ψ = − (t+ − t− ) ln(C1 /C2 ). (39) of κΛ . With increasing κΛ , ψ/ ln(C1 /C2 ) increases and be-
e
The transference number t+ (=1 − t− ) can be deduced from comes positive for κΛ ≈ 7.
the quotient of ψm and ψN [15], The two former limits are plotted in Fig. 6. The membrane
1 ψm potential tends to the free diffusional potential at large κΛ ,
t+ = . (40) indicating that ψ is only related to the diffusion difference be-
2 ψN
tween anions and cations. The surface effects do not influence
The transference number t + gives information about the solute diffusion and the clay efficiency is close to zero. More-
ability of a medium to restrict solute transport. For a nonse- over, for κΛ < 7, the negative sign of ψ indicates that the
lective medium, t + is equal to 0.4 for NaCl. However, if there double layer effects play a role. The clay efficiency increases as
is a total exclusion of ions from the pores, t + is equal to 1. κΛ decreases. The lowest value of κΛ obtained experimen-
In Fig. 5, we display the induced potential ψ of samples tally is equal to 0.51 with an associated value of ln(C ψ equal
EST104/2487 and EST104/2364 as functions of the initial con- 1 /C2 )
−2
to −1.35 × 10 V, which corresponds to a clay efficiency of
centration ratio C1 /C2 and the porosity  for different average
concentrations C̄. Results seem to be independent of the clay
sample, but depend on C1 /C2 , , and C̄. | ψ| increases as
C1 /C2 increases.
For C̄ > 10−3 mol l−1 , the dependence of ψ on C1 /C2 can
be approached by a logarithmic relation (see Fig. 5a),
ψ = 5.07 × 10−3 ln(C1 /C2 ). (41)
This relation is very close to relation (38), indicating that ψ is
only related to the diffusion differences of anions and cations.
Therefore, there is no anionic exclusion.
For C̄ < 10−3 mol l−1 , the negative sign of ψ indicates
that the double-layer effects play a role. The dependence of ψ
on C1 /C2 can also be approximated by a logarithmic relation,
ψ = −6.3 × 10−3 ln(C1 /C2 ), (42)
with a correlation coefficient r equal to 0.7. Such a bad value of Fig. 6. ψ/ ln(C1 /C2 ) as a function of κΛ . Data are for EST104/2487 (")
r can be explained by experimental errors; moreover, porosity and EST104/2364 (2). Solid line: Eq. (44). The broken lines are the two limits
is varying slightly in these experiments, as shown by Fig. 5b. corresponding to Eqs. (37) and (38).

(a) (b)

Fig. 5. The induced potential ψ as a function of C1 /C2 (a) and of  (b) for various C̄. Data are for EST104/2487 for C̄ > 10−3 mol l−1 (") and C̄ < 10−3
mol l−1 (!); EST104/2364 for C̄ < 10−3 mol l−1 (1) and C̄ > 10−3 mol l−1 (2).
360 M. Rosanne et al. / Journal of Colloid and Interface Science 297 (2006) 353–364

60%. The clay efficiency E is defined by The agreement between predictions and experiments is good
for large κΛ , except for some experimental errors. The analyt-
( ψ − ψmin )
E= × 100. (43) ical results display asymptotic behaviors both at very low and
( ψmax − ψmin ) very large κΛ . The minimum close to 0.4 indicates no surface
As it is very difficult experimentally to obtain values of κΛ effects, and thus E = 0. The maximum close to 0.88 indicates
lower than 1, we cannot conclude that this value corresponds to a maximum efficiency E = 80%.
the maximum efficiency.
In order to estimate the maximum efficiency, we have also 4.2. Transport coefficients
displayed in Fig. 6 the analytical values of ψ/ ln(C1 /C2 ).
Analytical values were obtained by integrating Eq. (13) and In the present experimental setup, the seepage velocity u(t)
writing it in a dimensionless form, is forced to be zero. Moreover, since there is no externally ap-
plied electrical field and since the electrical circuit is open, the
 current density I (t) is equal to zero.
ψ kT 1 γ13
=−  . (44) In this section, attention is focused on the coefficients related
ln(C1 /C2 ) e 2 γ11
to the concentration gradients, namely γ13 (γ31 ), γ23 (γ32 ), and
The method for deriving the coefficients γ13 and γ11 will be γ33 .
explained in the next section. The values of the γ13 and γ23 coefficients were calcu-
Fig. 6 indicates qualitative agreement between analytical ψ
lated from the experimental constant values P C and C using
predictions and experiments. When compared with the numer- Eqs. (21a) and (21b),
ical value for κΛ = 0.51, for example, the experimental value
P ψ
of ψ/ ln(C1 /C2 ) is underestimated by about 20%. However, γ13 = −γ12 − γ11 , (46a)
the minimum and the maximum seem to correspond to the ex- n n
perimental values. Therefore, the analytical results help us to P ψ
γ23 = −γ22 − γ21 . (46b)
conclude that the maximum value of ψ/ ln(C1 /C2 ) reached n n
for κΛ ≈ 0.4 is close to −1.84 × 10−2 V, which corresponds to The coefficients corresponding to electrokinetic phenomena,
a maximum efficiency of 77%. Argilite only partially restricts namely γ11 , γ22 , and γ21 (or γ12 ), were measured separately in
the passage of NaCl and it cannot be considered as an “ideal another series of experiments described in [17].
membrane,” as can be expected for natural clays. Next, γ33 was derived from Eq. (23) when D̄ is known. The
The cation transference number t + in both clay samples is measured concentration difference C(t) yields D̄ by means
plotted as a function of κΛ in Fig. 7. The same comments as of an elementary program described in [13].
for ψ can be made. t+ decreases from 0.75 to 0.4 when κΛ
increases from 1 to 30. For large κΛ , t+ is equal to 0.4, i.e, 4.2.1. Analysis of γ13 and γ31
essentially the same as in the bulk solution, indicating that E The coefficient γ13 , related to the membrane potential ψ,
is equal to zero. For the small experimental values of κΛ , i.e., can be analyzed more precisely with the help of (6) in order to
1, t+ reaches a value of 0.75. Again, the experimental values eliminate the bulk effects; it may be written in dimensionless
were compared to the values obtained by combining Eqs. (39) form as


and (44), 
γ13 γ13 (ζ  = 0)
 1
  
α13 =  

1 1 γ13 D̄1 − D̄2 D̄1 − D̄2 ζ 
 
t+ = 1 +  . (45)
2 2 γ11 (D1 + D2 ) I1 (κΛ )
= −2 (47a)
(D1 − D2 ) κΛ I0 (κΛ )
with
Di
D̄i = , i = 1, 2. (47b)
F
 by (D  − D  ) and by the electric for-
Hence, division of γ13 1 2
mation factor F cancels the influence of the molecular diffusion
coefficients and of the restricted geometry, respectively.
 as a function of κΛ and compares exper-
Fig. 8 displays α13
imental results to the analytical solution for circular Poiseuille
flows. The experimental data are seen to cluster around the an-
alytical expression. By standard techniques, the experimental
errors on κΛ and α13  are estimated to be equal to 46 and 60%.

Fig. 8 shows that α13  decreases as κΛ increases. When κΛ

is larger than 10, α13 is close to zero; this means that γ


13 is
Fig. 7. t+ as a function of κΛ . Data are for EST104/2487 (") and close to the value with no surface effects. As κΛ decreases,
EST104/2364 (2). Solid line: Eq. (45).  increases, which implies salt exclusion due to electrostatic
α13
M. Rosanne et al. / Journal of Colloid and Interface Science 297 (2006) 353–364 361

 as a function of κΛ . Data


Fig. 8. The dimensionless coupling coefficient α13
are for EST104/2487 (") and EST104/2364 (2). Solid line: Eq. (47a). Fig. 10. γ33 as a function of κΛ . Data are for EST104/2487 (") and
EST104/2364 (2). Comparison with D̄1 /2 (E), D̄2 /2 (P) and (D̄1 +
D̄2 )/2 (∗).

a pressure gradient, the solute concentration increases on the


higher pressure side of the porous medium and decreases on
the other side. Since γ32 is related to γ23 (see Eq. (26)), it is not
useful to plot it.
It is interesting to discuss the reflection coefficient ω. As,
experimentally, it was shown that γ11 γ22  γ12 γ21 , Eq. (18a)
can be rewritten as follows:
 
1 γ21 γ13 γ23
ω= − . (48)
2kT γ22 γ11 γ22
ω was found to be close to zero since the generated pressure
gradients are small in both clay samples. Consequently, γ23 is
 /γ  as a function of κΛ .
correlated to γ21 and γ13 by
Fig. 9. The ratio of the dimensionless coefficients γ23 22
Data are for EST104/2487 (") and EST104/2364 (2). Solid line: Eq. (6d). γ21 γ13
γ23 = . (49)
γ11
repulsion of the anions by electrical fields associated with the
surface effects. 4.2.3. Analysis of γ33
Since the coefficient γ31 that gives the magnitude of the In a first approximation, γ33 can be considered as the dif-
solute flux induced by the application of an electric field (elec- fusion coefficient of NaCl through the clay. However, it is
trodialysis) is proportional to γ13 (see Eq. (26)), it is not useful important to note that it is not exactly the diffusion coeffi-
to plot it. cient. Indeed, as we can observe in Eq. (6) when ζ = 0, γ33 =
(D1 + D2 )/2 = 1.69 × 10−9 m2 s−1 , but the diffusion coeffi-
4.2.2. Analysis of γ23 and γ32 cient of NaCl in a free medium is equal to 2D1 D2 /(D1 + D2 ) =
In order to compare γ23 to permeability, the ratio of the di- 1.62 × 10−9 m2 s−1 .
mensionless coefficients (see Eq. (6)) (γ23  /γ  ) is plotted in The surface effects on γ33 can be more precisely analyzed
22
  /γ  ) has been re- by introducing the coefficient α33  where the bulk effects have
Fig. 9 as a function of κΛ . The ratio (γ23 22
ferred to as the chemicoosmotic efficiency. (γ23  /γ  ) is a de- been subtracted:
22

creasing function of κΛ ; this is physically expected since γ23 


γ33  (ζ  = 0)
γ33
    1
is close to zero when κΛ is large and close to γ22 when κΛ is α33 = −
small. Since (γ23 /γ  ) essentially expresses surface effects, the D̄1 + D̄2 D̄1 + D̄2 ζ
22
specific surface and the related effects decrease when κΛ in- (D1 − D2 ) I1 (κΛ )
creases. For small κΛ , ζ  has a significant influence on γ23  . =− . (50)
(D1 + D2 ) κΛ I0 (κΛ )
The comparison with analytical results is acceptable and in-
dicates that the maximum value of (γ23  /γ  ) is equal to 0.7; Before the experimental values of γ33 are compared with
22
the deviation from 1 indicates that argilite is not an ideal mem- the numerical ones (see Eq. (50)), Fig. 10 displays γ33 , D̄1 /2,
brane. D̄2 /2, and (D̄1 + D̄2 )/2 as functions of κΛ . γ33 is almost equal
The hyperfiltration (or inverse osmosis) coefficient γ32 cor- to D̄1 /2 for large κΛ . This indicates that γ33 must be com-
responds to a filtration process by which, upon application of pared to the diffusion coefficient of cations (i.e., Na+ ) when
362 M. Rosanne et al. / Journal of Colloid and Interface Science 297 (2006) 353–364

 as a function of κΛ . Data


Fig. 11. The dimensionless diagonal coefficient α33 Fig. 12. D33 (=D̄/γ33 − 1) as a function of κΛ . Data are for EST104/2487
are for EST104/2487 (") and EST104/2364 (2). Solid line: Eq. (51). (") and EST104/2364 (2). Solid line: Eq. (54).

only geometric effects are taken into account. Moreover, in the surface effects influences D̄ by generating electroosmosis and
free medium D2 > D1 . Therefore, the best comparison of γ33 electrodiffusion. Moreover, as shown in Fig. 12, the comparison
with analytical results is obtained when Eq. (50) is modified as with analytical data is qualitatively successful.
    
 γ33 γ33 
 1 I1 (κΛ ) 4.3.2. The conversion efficiency
α33 = − ζ = 0 = . (51)
D̄1 D̄1 ζ  κΛ I0 (κΛ ) Let us follow the development of Gross and Osterle [16] con-
The experimental values of α33  and the analytical expression cerning the conversion efficiency of the clay. The maximum
(51) are displayed in Fig. 11 as functions of κΛ . For large κΛ , value of the conversion efficiency ηij is related to the “figure
 approaches zero indicating that the anionic mobility (i.e.,
α33 of merit” βij by
 increases
Cl− ) has a negligible effect on γ33 . For small κΛ , α33 (1 + βij )1/2 − 1
up to 1, indicating that surface effects delay the cation diffusion. (ηij )max = , (55)
(1 + βij )1/2 + 1
4.3. Discussion where βij is given by the relationship
 −1
γii γjj
4.3.1. The dispersion coefficient βij = −1 . (56)
γij γj i
Relation (19d) can be made dimensionless as
  Equations (1a)–(1c) imply that there are three dissipative
D̄ γ32 γ31 processes coupled together in clay. They are osmosis, elec-
D33 = − 1 = 2kT ω −τ . (52)
γ33 γ33 γ33 troosmosis, and electrodialysis, described by β23 , β21 , and β31 ,
Since ω is close to zero, the first term on the right-hand side respectively. The results are displayed in Fig. 13 as functions of
of (52) is negligible; therefore, κΛ together with the analytical data.
In osmosis, the maximum value of β23 is close to 0.006, cor-
γ31
D33 = −2kT τ . (53) responding to a maximum efficiency of 0.2% for κΛ ≈ 1.7.
γ33 Then, β23 decreases from this value down to zero for larger and
Hence, by using the condition γ11 γ22  γ12 γ21 and by sub- smaller values of κΛ .
stituting Eq. (49) into Eq. (18b), D33 can be expressed as β21 contains the contribution to the water flow of self-
  induced electroosmosis. The maximum value of β21 is about
γ31 γ13 γ21 γ12
D33 = − 1− . (54) 0.04 at κΛ ≈ 3–4, which corresponds to a maximum efficiency
γ33 γ11 γ22 γ11
of the clay with respect to electroosmosis about 1%.
Thus, D33 contains contributions of the self-induced elec- β31 contains the contribution of electrodialysis to the solute
troosmosis term (γ12 γ21 /γ11 ) and of the selfinduced electro- flux. β31 decreases monotonically as κΛ decreases from a
diffusion term (−γ13 γ31 /γ11 ) to the water flow and the solute maximum value of approximatively 5.9 corresponding to an ef-
flux, respectively. ficiency of 45%. The efficiency is rather small, which is due
In Fig. 11, D33 is plotted against κΛ for both clays. One to the small surface charges. The data are in poor agreement
can observe that D33 is close to zero for large κΛ (>7), indi- with (56).
cating that D̄ approaches γ33 , which, for large κΛ , represents
the diffusion coefficient of cations (Na+ ) without any surface 4.3.3. Comparison with the analytical predictions
effects. Hence, D̄ is essentially dominated by geometrical ef- The experimental data were compared with the analytical
fects and is independent of the self-induced electroosmosis and predictions (6) obtained for a circular tube where the radius Rc
electrodiffusion. However, for small κΛ (<7), the presence of is replaced by the length Λ given by (7).
M. Rosanne et al. / Journal of Colloid and Interface Science 297 (2006) 353–364 363

(a) (b)

(c)

Fig. 13. Figures of merit of (a) osmosis, (b) electroosmosis, and (c) electrodialysis. Solid lines: Eq. (56).

In most cases, the nondiagonal coefficients (see Figs. 6–9) Such estimations can be very useful since the non diagonal
are in agreement with the analytical formulae. Generally speak- coefficients are quite difficult to measure.
ing, the agreement is better for large κΛ , i.e., for large con-
centrations. For these values, the discrepancies are seen to be 5. Conclusion
random in character and the average is very close to the ana-
lytical predictions. This provides an order of magnitude for the The objective of this study was to measure the coupling
experimental errors. transport coefficients of NaCl solutions through a compacted
However, in most cases, for small κΛ , the data deviate from natural clay submitted to concentration gradients and to com-
pare the experimental coupling coefficients with the analytical
the analytical predictions. This is probably due to the fact that
solutions obtained for circular Poiseuille flows. There are three
the experiments corresponding to these values had to be per-
processes coupled together: osmosis, described by γ23 , elec-
formed with very low concentrations and thus were more diffi-
troosmosis, described by γ21 , and electrodialysis, described by
cult and uncertain than the other ones.
γ31 .
This property is also valid for diagonal coefficients such as The coefficients were measured as functions of C and  and
γ33 where surface effects can be important. When bulk effects represented in a dimensionless form as functions of κΛ .
are removed from γ33 , the resulting dimensionless coefficient The maximum efficiencies related to osmosis and electroos-
 is seen to verify the previous property in Fig. 11; the same
α33 mosis were found to be lower than 1%. The maximum effi-
comments apply for the variations of the quality of the compar- ciency related to the self-induced electrical field (electrodial-
ison with κΛ . ysis) was found to be 45% for κΛ < 1.
This property is practically very important, since it can be Thus, an important result is that the major contribution to
used to estimate all these coefficients when diagonal properties solute transport is due to the concentration gradient and the self-
such as K and σ/σ ∞ are known. (7) implies Λ . When the elec- induced electrical field. Moreover, convection, osmosis, and
trolyte concentration and ζ are known, all the coefficients can electroosmosis do not have a strong effect on solute and water
be estimated from (6). transports. The efficiency, delaying the passage of ionic species
364 M. Rosanne et al. / Journal of Colloid and Interface Science 297 (2006) 353–364

by the charged clay pores increases with decreasing κΛ and [5] D.E. Elrick, D.E. Smiles, N. Baumgartner, P.H. Gronevelt, Soil Sci. Soc.
the maximum efficiency was found for κΛ < 1. Since the abil- Am. J. 40 (1976) 490–491.
[6] P.H. Gronevelt, D.E. Elrick, T.J.M. Blom, Soil Sci. Soc. Am. J. 42 (1978)
ity to delay the passage of NaCl is only about 45%, we can
671–674.
conclude that the argilite is not a perfect membrane. [7] J.D. Sherwood, B. Craster, J. Colloid Interface Sci. 230 (2000) 349–358.
Finally, experimental and analytical results were found to [8] S.T. Horseman, J.J.W. Higgo, J. Alexander, J.F. Harrington, Water, Gas
be in agreement, especially for large κΛ , indicating that the and Solute Movement through Argillaceous Media, Report CC-96/1, Nu-
coupling coefficients can be estimated from permeability, elec- clear Energy Agency, 1996.
[9] D.L. Johnson, J. Koplik, L. Schwartz, Phys. Rev. Lett. 57 (1986) 2564–
trical formation factor, zeta potential, and diffusion coefficients
2567.
of cations and anions in the free medium. [10] M. Rosanne, M. Paszkuta, J.-F. Thovert, P.M. Adler, Geophys. Rev.
Lett. 31 (2004) L18614, 1-5.
References [11] V. Sasidhar, E. Ruckeinstein, J. Colloid Interface Sci. 85 (1981) 332–
362.
[12] M.A. Malusis, C.D. Shackelford, H.W. Olsen, Eng. Geol. 70 (2003) 235–
[1] D. Coelho, M. Shapiro, J.-F. Thovert, P.M. Adler, J. Colloid Interface 248.
Sci. 181 (1996) 169–190. [13] M. Rosanne, N. Mammar, N. Koudina, B. Prunet-Foch, J.-F. Thovert,
[2] S. Marino, M. Shapiro, P.M. Adler, J. Colloid Interface Sci. 243 (2001) E. Tevissen, P.M. Adler, J. Colloid Interface Sci. 260 (2003) 195–203.
391–419. [14] R.J. Hunter, Zeta Potential in Colloid Science, Academic Press, New York,
[3] S.R. de Groot, P. Mazur, Non-Equilibrium Thermodynamics, North- 1988.
Holland, Amsterdam, 1962. [15] P.W. Hatkins, Physical Chemistry, Oxford Univ. Press, Oxford, 1998.
[4] W.D. Kemper, J.P. Quirk, Soil Sci. Soc. Am. Proc. 36 (1972) 426– [16] R.J. Gross, J.F. Osterle, J. Chem. Phys. 49 (1968) 228–234.
433. [17] R.F. Probstein, R.E. Hicks, Science 260 (1993) 498–503.

Anda mungkin juga menyukai