Electrochemistry Communications 51 (2015) 1–5

Contents lists available at ScienceDirect

Electrochemistry Communications
journal homepage: www.elsevier.com/locate/elecom

Effect of electroconvection during pulsed electric field electrodialysis.

Numerical experiments
Aminat M. Uzdenova a, Anna V. Kovalenko b, Mahamet K. Urtenov b, Victor V. Nikonenko b,⁎
a
Karachaevo-Cherkessky State University named after U.D. Aliev, Karachaevsk, Russia
b
Kuban State University, Stavropolskaya st. 149, 350040 Krasnodar, Russia

a r t i c l e i n f o a b s t r a c t

Article history: One of the ways of improving electrodialysis (ED) and some microfluidic processes is the optimization of current
Received 5 September 2014 regime. It is recognised now that the use of pulsed electric fields (PEF) in ED allows enhancement of mass transfer
Received in revised form 20 November 2014 and mitigation of fouling. To explain these effects, Mishchuk et al. [1] suggested that due to inertial forces elec-
Accepted 25 November 2014
troosmotic mixing can remain during the pause. By solving fully coupled Nernst–Planck–Poisson–Navier–Stokes
Available online 3 December 2014
equations, we compute the distribution of velocity, concentration and electric fields in an ED channel. We simu-
Keywords:
late the situation where electroconvective vortices occur or not in the pause, and show that the total vortex at-
Ion-exchange membrane tenuation takes tenths of a second. However, inertial forces are effective only first 0.01 s, then the vortices are
Pulsed current fed by the chemical energy of non-uniform concentration field, the relaxation time of which is several seconds.
Electroconvection The remnant vortices contribute to earlier onset of electroconvection. However, more important is the formation
Mathematical modelling of new vortices after voltage re-application. For the first time, we show that nonuniformity of concentration field
produces an effect similar to the action of electrically or geometrically non-uniform surface. It causes formation of
spatially non-uniform electric body force, which hastens electroconvection.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction
Electrodialysis is a separation process widely used in water treat-
ment, juices and wine conditioning, recovery of valuable or harmful
components from solutions and other [2]. Three main ways to improve
this process are considered: improvement of (i) stack design and hydro-
dynamic conditions; (ii) membranes and ion-conducting fillers; and
(iii) current mode. For a long time the use of intensive overlimiting cur-
rents has been avoided because of water splitting at membrane/solution
interface. The latter leads to a loss in current efficiency and undesirable
shift in pH causing in particular precipitation of metal hydroxides. How-
ever, recent research has shown that at severe ion concentration polar-
ization where the current density essentially exceeds its so-called
limiting value, convective fluid flows occur at ion-selective surface
such as membranes [3,4], beads [5], nano-channels [6,7] or electrodes
[8]. While gravitational convection may be important at relatively
high concentrations (N0.1 M [9]), the main mechanism of the current-
induced convection is electroosmosis [9–11]. Electroconvection is a
phenomenon intensively studied now in connection with the improve-
ment of not only membrane separation processes [12], but nano-and
microfluidic devices as well [13–15]. Note also that water splitting
[11], the loss of membrane selectivity and current induced membrane
discharge [16] may contribute to overlimiting transfer.
⁎ Corresponding author. Tel.: +7 8612199 573.
E-mail address: nikon@chem.kubsu.ru (V.V. Nikonenko).
The PEF mode especially applied in overlimiting current range seems
a very promising way of developing electromembrane processes. The
first application of PEF in ED was apparently made by Karlin and
Kropotov [17] who have observed the variation in mass transfer rate
when passing from steady state to transient regimes. Mishchuk et al.
[1,18] have found that the rate of ED desalination can be higher in PEF
mode than in steady state when applying the same average voltage.
Similar results were obtained by other authors [19–21], who have
found also antifouling and antiscaling effects of PEF. Recently, Malek
et al. [22] have shown that the PEF effects are stronger at overlimiting
current densities. Moreover, application of PEF helps to minimize
water splitting and current efficiency losses. It becomes hence possible
to operate at higher voltages while avoiding the problems related to
pH changes [21–23].
In pioneer theoretical papers by Mishchuk et al. [1,18], electro-
diffusion time scales in PEF mode were examined, and the advantages
of the potentiostatic regime compared to the galvanostatic one were
established. To explain higher effect of PEF mode in overlimiting current
range, the authors [1,18] assumed that due to inertial properties,
electroconvective vortices can be conserved in a pause between the
pulses thus contributing to a more rapid restoration of concentration
at the membrane surface. However, the theory [1,18] was limited by
use of electroneutrality assumption and did not take into account
electroconvection. In this paper we present the results of mathematical
modelling using a recently developed 2D model involving the fully
coupled Nernst–Planck–Poisson and Navier–Stokes equations [24].

http://dx.doi.org/10.1016/j.elecom.2014.11.021
1388-2481/© 2014 Elsevier B.V. All rights reserved.
2 A.M. Uzdenova et al. / Electrochemistry Communications 51 (2015) 1–5
2. Mathematical model
The model used here differs from the “basic” model described in [24]
only by the boundary conditions at the inlet (y = 0):
 
∂φðx; 0; t Þ RT ∂c ðx; 0; t Þ ∂c ðx; 0; t Þ
¼−
z1 D1 1 þ z2 D 2 2
∂y F z1 D1 þ z2 D2 c0
2 2 ∂y ∂y
and at the outlet (y = L) of the desalination channel:
 
∂сi ∂φ
þ zi c i ðx; L; t Þ ¼ 0; i ¼ 1; 2
∂y ∂y
Eq. (1) follows from the condition that the tangential current is set
zero at y = 0. Eq. (2) shows that the sum of diffusion and migration tan-
gential components of fluxes is zero at y = L for cation (i = 1) and anion
(i = 2). All the notations are the same as used in [24].
3. Results and discussion
The calculations are made for a desalination channel of length L =
4 mm, intermembrane space H = 1 mm, formed by an anion-
exchange membrane (AEM) and a cation-exchange membrane (CEM)
both assumed ideally permselective. A 0.01 mol m− 3 NaCl solution
enters the channel and flows between the membranes with an average
velocity of V0 = 0.8 mm s−1. The diffusion coefficients are taken as
D1 = 1.33 × 10−9 m2s−1, D2 = 2.05 × 10−9 m2s−1. A potential differ-
ence Δφ across the channel (involving both membranes, the space
between them and two half-cells of the concentration compartment)
is set as a known function of time. Then the distribution of cation and
anion concentrations, potential, velocity and current density is calculat-
ed as functions of time. Note that only a diffusion layer (including the
electrical double layer (EDL) on the membrane surface) was the subject
of study in Refs [4,25].
The concentration used in computations is rather small compared to
the concentrations used in ED practice. The first reason for this choice is
that the electrokinetic effects are more pronounced in dilute solutions;
the second one, the computational difficulties arise essentially with in-
creasing concentration as EDL becomes thinner and a denser grid is
needed. An analysis of the concentration effect on electroconvection is
the subject of further work.
Fig. 1b shows the response of current density on variation of Δφ
(Fig. 1a) in our numerical experiments. Before t1 = 201 s and after
t2 = 201.5 s, Δφ is equal to − 2 V; between t1 and t2, Δφ = 0 V,
hence, the pause duration is 0.5 s as in the experimental study [23]. In
the pause, the current density changes its sign: the cell works as a
source of electric current similarly as in reverse electrodialysis [26].
Just after setting Δφ = 0, there is a jump of the current (with negative
sign), which is due to partial discharge of the space charge region
(SCR). Then the current is stabilized; generally it tends to zero as the
concentration gradients are vanishing. However, since this process is
rather long (several minutes), it cannot be seen in Fig. 1b. After
restarting Δφ = − 2 V, there is a current surge, which is favoured by
the decrease in resistance caused by concentration restoration. Then
concentration polarization increases, and the current density declines.
It is remarkable that the fast current decline lasts for a very short time,
about 0.05 s. Then a little jump in current occurs, and the decline
slows. 0.5 s after restarting voltage, the current density slightly grows
and becomes higher than its steady-state value, ist, before the pause. It
attains ist only 7 s after restarting voltage (not shown in Fig. 1b). Note
that there is no real steady state: when Δφ is maintained constant, the
current density oscillates [24] due to electroconvection occurring in
the Rubinstein–Zaltzman unstable mode [27]. The slow return to the
state before the pulse is a type of hysteresis similar to that described
by Pham et al. [25].
(a)
ð1Þ (b)
ð2Þ
Fig. 1. (a) Pulse mode of electric potential; (b) response of the current density: 1—viscosity
ν = 0.01 cm2 s−1 for the entire time; 2—ν = 0.01 cm2 s−1 for t b t1 = 201 s and t N t2 =
201.5 s, ν = 1.0 cm2 s−1 for t1 ≤ t ≤ t2; and 3—ν = 1.0 cm2 s−1 for the entire time. The
insertion shows the response just after voltage cutoff.
Fluid flow and electric current streamlines as well as the concentra-
tion distribution are shown in Fig. 2. In the conditions used, the calculat-
ed limiting current density ilim = 0.0185 A m−2. This value is in a good
agreement with that calculated by the Lévêque equation [9].
At Δφ = − 2 V, electroconvective vortices are developed at both
membranes (Fig. 2a). They are stronger near the CEM, as in the case of
NaCl ilim at this membrane (0.0185 A m− 2) is about 1.5 times lower
than that at the AEM (0.028 A m−2). It can be seen that after setting
Δφ = 0, the vortices do not decay immediately: after 0.01 s they are
clearly seen near both membranes (Figs. 2b and 3a). Even after 0.5 s a
small vortex of about 1 μm remains near the CEM (Fig. 3b).
In order to study the effect of remnant vortices on further develop-
ment of electroconvection, we examine the case where within the
pause the viscosity is set 1.0 cm2 s−1, hence 100 times higher than its
normal value, 0.01 cm2 s−1. After re-applying voltage, the viscosity
takes its normal value. Fig. 2c shows that really the increase in viscosity
suppresses the vortices. However, it gives only a small effect on the con-
centration field: the remnant vortices are too small (Fig. 3) to enhance
considerably mass transfer. The relaxation of concentration field
(which is weak during 0.5 s of the pause, Fig. 2d and e) is mainly due
to diffusion. The diffusion relaxation time, trel = δ 2/D, for characteristic
length δ = 100 μm of the concentration field heterogeneity is about 10 s.
After restarting voltage, the vortices reappear in 0.01 s near the CEM
and in 0.4 s near the AEM. After 0.6 s they are well developed near the
AEM in the case of normal viscosity (Fig. 2f) and not yet seen near this
membrane if ν = 1.0 cm2 s−1 within the pause (Fig. 2g). Namely the ap-
pearance of vortices near the AEM gives rise to the current density at
t = 202 s (Fig. 1b).
The third calculation is made for an elevated viscosity (ν =
1.0 cm2 s−1) during the entire time (line 3, Fig. 1). In this case no vorti-
ces are formed. In steady state i = 0.020 A m−2 that is slightly higher
than ilim = 0.0185 A m−2 given by the Lévêque equation. The deviation
is due to reduction in the diffusion layer thickness produced by the
extended SCR [28].
As Fig. 3c shows, the decay of the vortex after setting Δφ = 0 occurs
initially rather fast. However, 0.02 s after the voltage cutoff, the vortex
decay slows.
The relaxation time of vortex decay may be evaluated from the
 
δ2
Navier–Stokes equation [29]: t ¼ 12ν ln 1− V , where V is the steady-
V
 A.M. Uzdenova et al. / Electrochemistry Communications 51 (2015) 1–5 3

(a)

(b)

(c)

(d)

(e)

(f)

(g)

Fig. 2. Distribution of salt concentration (shown with different colours), electric current (black lines) and fluid velocity streamlines (white lines) in an ED desalination compartment of
length L = 4 mm and intermembrane space H = 1 mm. (a) Steady state at Δφ = −2 V; (b) and (c) 0.01 s after setting Δφ = 0; (d) and (e) 0.01 s after restarting Δφ = −2 V; (f) and
(g) 0.6 s after restarting Δφtot = −2 V; (b), (d) and (f) are calculated at normal viscosity ν = 0.01 cm2 s−1; (c), (e) and (g), at viscosity ν = 0.01 cm2 s−1 in the pulses, and ν =
1.0 cm2 s−1 in the pause.

state value of velocity V, which varies under a constant body force, F. The
time needed for 100-fold decrease of V after setting F = 0 is 0.5 10−2 s, if
δ = 200 μm (the size of a vortex). Hence, if only inertial and viscosity
forces are taken into account, the vortex lifetime is essentially lower
than the 2D simulation predicts.
The reason of the vortex long lifetime may be in the fact that the
body forces feeding vortices are not really zero after setting Δφ = 0. It
might be due to non-uniform distribution of concentration in the
membrane vicinity.
The situation after voltage cutoff is quite similar to that occurring in
reverse electrodialysis [26]. There is a desalted solution in one
compartment and a concentrated one in the neighbour one. If the circuit
is open (zero current), a potential difference is generated; if the circuit is
short (zero voltage), a current is generated. In both cases, electric power
is produced at the expense of chemical energy. The remnant membrane
potential can be calculated by using an approximate equation [30]
RT
Cr
Δφmb ¼ − t 1 −t 2 ln l ð3Þ
F C
where Cr and Cl are the (1:1) electrolyte concentrations at the right and
the left-hand solutions; t i is the transport number in the membrane. As
4 A.M. Uzdenova et al. / Electrochemistry Communications 51 (2015) 1–5

Fig. 3. Evolution of a vortex with time, fluid flow and electric current streamlines 0.01 s (a) and 0.5 s (b) after setting Δφ = 0, respectively. (с) Vortex diameter vs. time after setting Δφ = 0.

(a) (b)
1.5 30
0.02 s 0.02 s
0.05 s 0.05 s
0.10 s 20 0.10 s
1 0.20 s 0.20 s
0.30 s 10 0.30 s
C2, 10-3 mol m-3

0.40 s Fy, 103 N m-3 0.40 s

0
0.5
-10

0 -20
1750 1850 1950 2050 1750 1850 1950 2050
y, μm
y, μm

Fig. 4. Longitudinal distribution of the anion concentration, C2 (a), and the tangential body force, Fy (b), at distance 0.05 μm from the CEM surface, while the EDL thickness is about 0.3 μm.
The time elapsed after setting Δφ = 0 is shown in figures.

we can see, to generate electricity, a difference in transport numbers is by small stable “seed” vortices [25] (of Dukhin's type). The earlier onset
needed. of instability is facilitated also by surface hydrophobicity [24,33].
However, together with a concentration gradient across the mem- ED system in PEF mode shows a hysteretic electrokinetic behaviour
brane, there are lateral gradients of concentration (Figs. 2, 4a). More- firstly studied by Pham et al. [25]. During a pause, concentration resto-
over, as the electrolyte concentration at the depleted membrane ration occurs mainly near the membrane surface thus contributing to
surface (C2) is not uniform along the membrane, the EDL thickness is fast decrease of system resistance. Similarly, after restarting voltage,
a function of the lateral coordinate, as well as the space charge located the concentration decreases firstly near the surface. As a result, the
in solution. The remnant lateral electric field is, according to Eq. (3) in SCR thickness grows rapidly and reaches its threshold (corresponding
differential form: to transition in instable electroconvection) when the concentration far
from the surface remains relatively high providing low system resis-
  tance. Optimization of PEF mode might go through exploitation of this
∂φ RT ∂ ln c1 ∂ lnc2
¼− t1 −t 2 ð4Þ hysteretic behaviour.
∂y F ∂y ∂y

Within the EDL near the CEM, the cation concentration C1 is high;
hence the space charge is high and t1 is rather close to 1 where C2 is
small, and they are low where C2 is elevated. Hence, ∂φ/∂y and tangen-
tial body force, Fy, are functions of y (Fig. 4b). In the condition of our nu-
merical experiments, this produces lateral electric currents occurring
from the zones with high space charge to the zones with low space
charge. These currents generate electroosmotic flow resulting in the
remnant vortices after voltage cutoff.
The small difference between curves 1 and 2 in Fig. 1b suggests that
the main cause of the fast restart of electroconvection after a pause is the
nonuniformity of concentration field. As Figs. 2 and 3 show, the electric
current streamlines are not homogeneous: they are concentrated in re-
gions with relatively high concentration. This causes spatially non-
uniform electric body force with a significant tangential component.
The situation is similar to that where membranes with electrically or
geometrically heterogeneous surface are used [9,10,25,27,31,32]
where earlier onset of electrokinetic Rubinstein's instability is facilitated
4. Conclusion
Inertial forces decay in b0.01 s after voltage cutoff. However,
electroconvective vortices remain during several tenths of a second
fed by non-uniform concentration field. After reapplication of voltage,
the remaining nonuniformity of concentration field, like heterogeneity
of membrane surface, causes non-uniform distribution of electric body
force, which rapidly restarts electroconvection after a pause.
Conflict of interest
None.
Acknowledgments
This research was financially supported by the Russian Science
Foundation (grant Nb.°14-19-00401).
 A.M. Uzdenova et al. / Electrochemistry Communications 51 (2015) 1–5
References
[1] N.A. Mishchuk, L.K. Koopal, F. Gonzalez-Caballero, Colloid Surf. A 176 (2001) 195.
[2] H. Strathmann, Desalination 264 (2010) 268.
[3] S.M. Rubinstein, G. Manukyan, A. Staicu, I. Rubinstein, B. Zaltzman, R.G.H.
Lammertink, F. Mugele, M. Wessling, Phys. Rev. Lett. 101 (2008) 236101.
[4] R. Kwak, V.S. Pham, K.M. Lim, J. Han, Phys. Rev. Lett. 110 (2013) 114501.
[5] N. Mishchuk, F. Gonzalez-Caballero, P. Takhistov, Colloid Surf. A 181 (2001) 131.
[6] T.A. Zangle, A. Mani, J.G. Santiago, Chem. Soc. Rev. 39 (2010) 1014.
[7] S.J. Kim, Y.-C. Wang, J.H. Lee, H. Jang, J. Han, Phys. Rev. Lett. 99 (2007) 044501.
[8] V. Fleury, J.N. Chazalviel, M. Rosso, Phys. Rev. Lett. 68 (1992) 2492.
[9] V. Nikonenko, N. Pismenskaya, E. Belova, Ph. Sistat, Ch. Larchet, G. Pourcelly, Adv.
Colloid Interf. Sci. 160 (2010) 101.
[10] S.S. Dukhin, N.A. Mishchuk, J. Membr. Sci. 79 (1993) 199.
[11] J.J. Krol, M. Wessling, H. Strathmann, J. Membr. Sci. 162 (1999) 155.
[12] M. Wessling, L. Garrigós Morcillo, S. Abdu, Sci. Rep. 4 (2014) 4294.
[13] J. De Jong, R.G.H. Lammertink, M. Wessling, Lab Chip 6 (2006) 1125.
[14] A. Höltzel, U. Tallarek, J. Sep. Sci. 30 (2007) 1398.
[15] M.E. Suss, A. Mani, T.A. Zangle, J.G. Santiago, Sensors Actuators A 165 (2011) 310.
[16] M.B. Andersen, M. van Soestbergen, A. Mani, H. Bruus, P.M. Biesheuvel, M.Z. Bazant,
Phys. Rev. Lett. 109 (2012) 108301.
[17] J.V. Karlin, V.I. Kropotov, Russ. J. Electrochem. 24 (1989) 1654.
[18] N.A. Mishchuk, S.V. Verbich, F. Gonzalez-Caballero, Colloid J. 63 (2001) 643.
[19] H.J. Lee, S.H. Moon, S.P. Tsai, Sep. Purif. Technol. 27 (2002) 89.
5
[20] B. Ruiz, P. Sistat, P. Huguet, G. Pourcelly, M. Araya-Farias, L. Bazinet, J. Membr. Sci.
287 (2007) 41–50.
[21] N. Cifuentes-Araya, G. Pourcelly, L. Bazinet, J. Colloid Interface Sci. 361 (2011) 79.
[22] P. Malek, J.M. Otriz, B.S. Richards, A.I. Schafer, J. Membr. Sci. 435 (2013) 99.
[23] S. Mikhaylin, V. Nikonenko, G. Pourcelly, L. Bazinet, J. Membr. Sci. 468 (2014) 389.
[24] M.K. Urtenov, A.M. Uzdenova, A.V. Kovalenko, V.V. Nikonenko, N.D. Pismenskaya,
V.I. Vasil'eva, P. Sistat, G. Pourcelly, J. Membr. Sci. 447 (2013) 190.
[25] V.S. Pham, Z. Li, K.M. Lim, J.K. White, J. Han, Direct numerical simulation of
electroconvective instability and hysteretic current-voltage response of a
permselective membrane, Phys. Rev. E. 86 (2012) 046310.
[26] D.A. Vermaas, M. Saakes, K. Nijmeijer, J. Membr. Sci. 385–386 (2011) 234.
[27] I. Rubinstein, B. Zaltzman, Phys. Rev. E. 62 (2000) 2238.
[28] I. Rubinstein, L. Shtilman, J. Chem. Soc. Faraday Trans. 75 (1979) 231.
[29] C. Larchet, S. Nouri, B. Auclair, L. Dammak, V. Nikonenko, Adv. Colloid Interf. Sci. 139
(2008) 45.
[30] B. Auclair, C. Larchet, L. Dammak, V. Nikonenko, M. Métayer, J. Membr. Sci. 195
(2002) 89.
[31] J. Balster, M.H. Yildirim, D.F. Stamatialis, R. Ibanez, R.G.H. Lammertink, V. Jordan, M.
Wessling, J. Phys. Chem. B 111 (2007) 2152.
[32] H.-C. Chang, E.A. Demekhin, V.S. Shelistov, Phys. Rev. E. 86 (2012) 046319.
[33] V.S. Shelistov, E.A. Demekhin, G.S. Ganchenko, Electrokinetic instability near charge-
selective hydrophobic surfaces, Phys. Rev. E. 90 (2014) 013001.