Chemical Engineering Journal Advances 8 (2021) 100169

Contents lists available at ScienceDirect

Chemical Engineering Journal Advances

journal homepage: www.sciencedirect.com/journal/chemical-engineering-journal-advances

Electrodiffusion of ions in ion exchange membranes: Finite element

simulations and experiments
Kuldeep a, Pertti Kauranen a, Heikki Pajari b, Risto Pajarre b, Lasse Murtomäki a, *
a
Department of Chemistry and Materials Science, School of Chemical Engineering, Aalto University, P.O. Box 16100, 00076 AALTO, Finland
b
VTT Technical Research Center of Finland, Tietotie 4E, 02150 Espoo, Finland

A R T I C L E I N F O A B S T R A C T

Keywords: Electrodiffusion of ions in both cation (CEM) and anion exchange membranes (AEM) has been studied with
Nernst-Planck equation theoretical calculations and experimental studies. Calculations are based on the Finite Element Method (FEM)
Multi-ionic transport using COMSOL Multiphysics® software. Nernst-Planck equations are solved in multi-ionic systems where no
Ion-exchange membranes
closed form solutions are available. Simulations are compared with laboratory-scale experiments in terms of
Finite element method
Simulation
current efficiency and membrane selectivity. Simulations revealed unexpected features in transport, due to
coupling of ionic fluxes when the local electroneutrality condition is assumed. Transport of weak electrolytes
showed the importance of involving ionic equilibria along the concentration profiles in both solutions and
membranes, compelling to consider ionic constituents. The advantage of the COMSOL simulations is the ease to
find concentration and potential profiles across the entire system, and to split fluxes to diffusion and migration
contributions, showing their coupling even in the absence of electric current.

Introduction thermodynamics of irreversible processes that takes explicitly into ac­

count the coupling of component fluxes [19]. Yet, the number of un­
Ion transport has a crucial role in water purification and desalina­ known parameters makes applying it rather unpractical. Therefore, the
tion, membrane separation processes, Chlor-alkali industry, electrodi­ almost exclusively applied approach to ionic transport is the
alysis (ED), fuel cells and batteries, to name the most common Nernst-Planck equation that formally ignores the coupling of fluxes, but
applications [1–3]. ED is a well-proven electrochemical separation via the electroneutrality condition and the definition of electric current,
method for water treatment by using ion-selective membranes and the coupling remains very strong [20–23]. Coupling of electroosmosis
electric current [4]. ED has also contributed to other fields, which with ionic fluxes can formally be taken into account via a convective
cannot be ignored, including industrial waste treatment [5], chemical term that must be separately determined [24]. Regardless of its apparent
process industries, food industries [6,7], pharmaceutical industries [7], simplicity, the Nernst-Planck equation is not solvable in closed form in a
galvanic industries [8], and salt production from seawater [9,10]. The charged membrane because of the fixed charge of the membrane that
essential aspects in this field have been summarised by, e.g. Tanaka leads to transcendental equations already in the simplest cases, i.e. ho­
[11]. With the recent development of membranes, it can be anticipated mogeneous membrane charge, binary system, 1D geometry.
that applications of membrane processes are going to become increas­ The theory of ion transport has naturally been subject to various
ingly important. This is one reason for the intensive research to solve the studies and books [25–27]. A detailed review of the various ion trans­
critical aspects of this process. port theories inside the membranes was already published by Buck in
To understand the charge transfer in the membranes as a function of 1983 [28]. Concerning electrodialysis of strong electrolytes, Bouzek
current density, a suitable mathematical approach is required [12–18]. et al. [29,30] investigated ion transport in the membrane with and
Simulation work on membranes has drawn attention of the researchers without convection in their models. They presented the transfer of Na+,
for decades. Nowadays, mathematical modelling is a complementary OH− , and Cl− in the membrane in the current density range of 0 to 250
approach to verify the exact mechanism of mass transfer across the mA cm− 2 for a very high concentration of the electrolytes. They dis­
membranes. The most rigorous approach to transport is via cussed the performance of an ion exchange membrane (IEM) under

* Corresponding author.
E-mail address: lasse.murtomaki@aalto.fi (L. Murtomäki).

https://doi.org/10.1016/j.ceja.2021.100169
Received 29 March 2021; Received in revised form 16 August 2021; Accepted 16 August 2021
Available online 20 August 2021
2666-8211/© 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Kuldeep et al. Chemical Engineering Journal Advances 8 (2021) 100169

different conditions with the effect of the diffusion boundary layers on proportions of each ionic species in carrying electric current are evalu­
ion transfer. Moon et al. [31,32] studied ion transfer in a continuous and ated, as described below. The reason for choosing these particular sys­
batch mode electrodialysis (ED) process with 1D and 2D models. In 2D, tems is an effort to convert waste Na2SO4 to acid and base via
they clarified the effect of the flow rate and the current density on ionic electrodialysis using bipolar membranes. A particular focus is to identify
concentrations in the membrane and the electrolyte compartments. the charge carrying species in an anion exchange membrane soaked in
Fidaleo and Moresi [33] and Tanaka [32] have also modeled mass sulfuric acid solutions.
transfer and limiting current density in continuous ED of NaCl [33] or
seawater [32]. In the previous simulations [33], the maximum product Theory
concentration reached ca. 1.7 mol dm− 3, which can be utilized in scaling
up from a lab-scale or pilot scale to an industrial-scale. Ion transport is described with the Nernst-Plank equation that reads
The transport of weak electrolytes poses an extra degree of difficulty
→ →diff usion →migration →convection
because, in the presence of concentration gradients, dissociation equi­ jk= jk + jk + jk
libria shift along the concentration profile. The fluxes of weak electro­ (1)
lyte ions are consequently no more spatially constant, and ionic = − Dk ∇ck − Dk zk f ck ∇ϕ + →
v ck ; f = F/RT
constituent fluxes must be considered instead. The transport of phos­
phoric and carbonic acid [34], carbonates [35] and phosphates [36] has where jk is the flux density (mol/cm2s), Dk is the diffusion coefficient of
been studied both theoretically and experimentally, taking dissociation ion ‘k’, zk its valence and ck its molar concentration, and →v is convection
equilibria into account. Concerning the sulfate-bisulfate system that is velocity; ϕ is the Galvani potential of the solution (‘electrolyte potential’
also our interest, rather confusing results are reported in the literature. in COMSOL). In the membrane, counter-ions, i.e. ions with the opposite
Pourcelly et al. [37] explored the leakage of protons through anion ex­ sign to that of the membrane charge are the main charge carriers via
change membranes (AEMs) in-depth in HCl and H2SO4 systems. They migration, while the co-ions with the like charge are minor charge
suggested that although bisulfate balances the membrane charge, it does carriers, moving mainly by diffusion in the interstitial space between the
not carry the charge across an AEM but sulfate is the main charge car­ fixed charges [25].
rier. Koter et. al [38] also studied the transport of H2SO4 through The electroneutrality in the membrane is maintained by the fixed-
different AEMs in the presence of electric current. The results were charge sites as well as the counter-ions and co-ions in the membrane:
found in accordance with Pourcelly et al. [32,34]. Furthermore, Koter ∑
and Kultys studied the mixture of phosphoric and sulfuric acids in an zk ck + zM cM = 0 (2)
anion-exchange membrane [35] and found that the permeability of the k

former was much lower due to strong interaction with the membrane;
where zM and cM are the valence of the fixed charge group and their
agreement with the model, an extended Nernst-Planck equation, was
amount expressed as molar concentration, respectively. The current
quite good. →
Ionic mobilities in ion exchangers is yet another issue that compli­ density ( I ) is defined as:
cates the analysis of transport. Kamcev et al. [39] calculated the mo­ → ∑
I →
bilities of ionic species across the IEMs from experimental ion sorption, = zi j i (3)
F
salt permeability, and the ionic conductivity data. They concluded that i

counter-ions can exist either as a ‘condensed’ phase, i.e. tightly bound to In systems including sulfuric acid, the dissociation equilibrium be­
the fixed charges of the membrane, between which they move with a tween bisulfate and sulfate is also taken into account. The first dissoci­
hopping mechanism, or as a ‘free’ phase where ions move with Brow­ ation is always complete:
nian motion in the interstitial space. As a consequence, the apparent
mobility of counter-ions in the condensed phase was 2-2.5 fold to that in H2 SO4 →H+ + HSO−4 (4)
the free aqueous phase of the membrane, whereas the mobility of co-ions The dissociation of bisulfate
was predicted equivalent to its aqueous value. [39]
Purely geometrical effects have also an effect on the mobility. In HSO−4 ⇌H+ + SO2−4 (5)
narrow pores found in ion-exchange membranes, Brownian motion is
hindered, which can be taken into account via the Renkin correction has the pKa value of ca. 2. As pH can vary significantly across the
[40]. For example, with the solute to pore radii ratio of 0.4 (the upper membrane, also the ionic product of water needs to be taken into
limit of the Renkin correction validity), the effective diffusion coefficient account:
is only 10% of its value in a solution. K w = [H+ ][OH− ] (6)
The fast development of the computational power of personal com­
puters has made numerical solutions of all kinds of scientific problems In the case of weak electrolytes, ionic fluxes are necessarily not
accessible to researchers without the need to resort to mainframe constant due to dissociation equilibria. Therefore, ionic constituents
computers and software. Also, in modern software no actual coding is need to be considered instead, here, the hydrogen and sulfate
any more needed, but a user can define the problem with an interactive constituents:
graphical interface; certain knowledge of physics and chemistry is of → → →
J H = j H+ + j HSO−4
course needed to be able to set-up a problem. COMSOL Multiphysics® (7)
→ → →
that is based on the Finite Element Method (FEM) is a fine example of J S = j HSO−4 + j SO2−4
such a software. Using COMSOL Multiphysics®, we have solved the
problem of multi-ionic electrodiffusion in charged membranes in 1D, Simulations show that the constituent fluxes are, indeed, constant,
taking into account all above mentioned issues, and compared simula­ and it is possible to express the electric current density with the con­
tions to experimental results. The objective of this work is to understand stituent fluxes.
the relative proportions of the different transport mechanisms of ions in The integral transport number [25] (current efficiency) is calculated
ion-exchange membranes as a function of the bathing solution concen­ as
trations, the amount of fixed membrane charge and electric current →
zk F j k
density. Two different systems are studied: (i) a cation exchange mem­ Tk = → (8)
brane flanked between Na2SO4 and NaOH solutions, and (ii) an anion I
exchange membrane flanked between Na2SO4 and H2SO4 solutions. The

2
Kuldeep et al. Chemical Engineering Journal Advances 8 (2021) 100169

Fig. 1. Schematic of lab-scale experimental setup and 1D simulation setup with all domains and mesh a) CEM, and b) AEM.

or in terms of the constituent fluxes (see Results). It should be noted that fixed charge density in the membrane; we checked that both options
ionic fluxes consist of diffusion and migration and, therefore, Tk cannot gave identical results. The membrane phase is assumed homogeneous, i.
be considered as the pure migration transport number. e. the porosity of the membrane or its microstructure are not seen in the
The dissociation constant of reaction (5) has the thermodynamic pKa Donnan equilibrium, and it is assumed to be aqueous because no
of ca. 2, i.e. it is expressed in terms of activities, not concentrations. In chemical partition coefficient, i.e. difference in the standard chemical
transport equations, however, concentrations are considered. Therefore, potentials between the aqueous and membrane phases exists in the
we have to correct for the thermodynamic dissociation constant with the expression of the Donnan potential:
activity coefficients. Fortunately, this issue has already been discussed in ( w)
RT c
literature [41,42], and we are able to write down the concentration ΔϕD = ln k ⇔ ck = cwk e− zk f ΔϕD (10)
zk F ck
equilibrium constant, using the data of ref. [41], as
[ ] In the above equation that applies to all ions in the system separately,
[H+ ] SO2−4 γHSO−4
Kc = [ ] = Ka (9) ck is the concentration of ion k in the aqueous (w) and membrane phases,
HSO−4 γSO2−4 γH+
respectively, and zk is its charge. The Donnan potential is calculated
As anticipated, Kc depends on the concentration of sulfuric acid. For from the electroneutrality condition, Eq. (2), and it applies also when
example, in 0.5 M sulfuric acid, Kc ≈ 0.24 M, hence 24-fold to Ka. In current is flowing. Flux continuity in the entire simulation domain is
Supporting Information, the dependence of Kc on the acid concentration assumed, if not specifically otherwise defined. Hence, Neumann condi­
is presented, as well as the concentrations of the ions. tion is valid at the solution-membrane boundaries. The advantage of
There still remains a couple of issues to be considered: First, the value Comsol simulations is that no explicit definitions of the boundary con­
of Kc within a membrane is probably different from that in an aqueous ditions are needed from a user, just selecting appropriate ones from
solution due to the existence of fixed charge groups. We present these menus.
calculations in Supporting Information; we chose Kc = 0.25 M in the
membrane. Second, during electrolysis, concentration polarization takes Experiments
place at the membrane-solution interface so that, strictly speaking, Kc
varies also in the flanking aqueous solutions. Taking into account con­ A single electrodialysis cell (Micro Flow Cell, Electrocell, Denmark)
centration polarization would require an iterative process that is not is assembled either with two AEMs and one CEMs or two CEMs and one
easily doable in COMSOL. In this work, however, concentration polari­ AEM, as well as Pt coated Ti anode and cathode, Figure 1. The active
zation is quite moderate and we calculated that in the worst case – membrane and electrode area is 10 cm2 and the membrane spacing is 3
lowest concentration, highest current density – the deviation of Kc from mm. The CEM and AEM characteristics are given in Table 1.
the given parameter value is only ca. 5%. Hence, we ignored this issue. The flow rate was 100 mL/min for each liquid channel (width 3 mm)
In COMSOL, the Donnan equilibrium can be chosen by activating this and was controlled by peristaltic pumps (Watson Marlow 323). Sodium
option in the menu defining the ion-exchange membrane properties. sulfate (Fluka, purity >99.0%. (Germany)), sulfuric acid (Merck KGaA
Alternatively, the Poisson-Boltzmann equation can be solved assigning with MQ 100 (Germany)), NaOH (Merck KGaA pellets p.a. grade

Table 1
Membranes characteristics [43].
Membranes Membranes Type Area Thickness Water Content (% Ion exchange capacity Membrane fixed Membrane fixed charge in
(from SUEZ) (cm2) (mm) wet resin) (meq/dry g resin) charge (M)* pore water (M)†

CEM CR61P 10 0.58 44 2.20 1.23 3.075

AEM AR103P 10 0.57 39 2.37 1.45 3.625
*
,† The formula used to calculate membrane fixed charge is shown in Supporting Information.

3
Kuldeep et al. Chemical Engineering Journal Advances 8 (2021) 100169

Table 2
Experimental sets for CEM at 25◦ C.
Experiment Set Anode Compartment Cathode Compartment Dialysate Compartment Permeate Compartment I (mA/cm2)

1 Na2SO4 (0.1 M) NaOH (0.2 M) Na2SO4 (0.1 M) NaOH (0.2 M) 30

2 Na2SO4 (0.25 M) NaOH (0.5 M) Na2SO4 (0.25 M) NaOH (0.5 M) 50
3 Na2SO4 (0.5 M) NaOH (1.0 M) Na2SO4 (0.5 M) NaOH (1.0 M) 100

Table 3
Experimental sets for AEM at 25◦ C.
Experiment Set Anode Compartment Cathode Compartment Dialysate Compartment Permeate Compartment I (mA/cm2)

1 H2SO4 (0.1 M) H2SO4 (0.1 M) H2SO4 (0.1 M) H2SO4 (0.1 M) 0, 30

2 H2SO4 (0.25 M) H2SO4 (0.25 M) H2SO4 (0.25 M) H2SO4 (0.25 M) 0, 50
3 H2SO4 (0.5 M) H2SO4 (0.5 M) H2SO4 (0.5 M) H2SO4 (0.5 M) 0, 100
4 H2SO4 (0.1 M) Na2SO4 (0.1 M) Na2SO4 (0.1 M) H2SO4 (0.1 M) 30
5 H2SO4 (0.25 M) Na2SO4 (0.25 M) Na2SO4 (0.25 M) H2SO4 (0.25 M) 50
6 H2SO4 (0.5 M) Na2SO4 (0.5 M) Na2SO4 (0.5 M) H2SO4 (0.5 M) 100

EMSURE® (Germany) are used as received. The concentrations of

dialysate compartment is varied from 0.1 M to 0.5 M in CEMs and AEMs Table 4
List of simulation parameters.
experiments, as shown in Tables 2 and 3.
The purpose of experiments 1-3 in Table 3 is to identify the mobile Parameter Value and unit
species inside the AEM. Conductivities of permeate solutions are Thickness of the diffusion boundary layers 50 μm
measured by WTW TetraCon 925 conductometer. Conductivities are Membrane thickness 0.6 mm
then converted to concentrations with the data from CRC Handbook of Membrane charge Range (0-3 M).
Diff. coeff. of Na+ 1.3 × 10− 9 m2 s− 1
Chemistry and Physics [44]. The concentration of feed and products are Diff. coeff. of H+ 9.3 × 10− 9 m2 s− 1
also verified by titration with Titrette® 25mL class A precision. Diff. coeff. of HSO4− 1.3 × 10− 9 m2 s− 1
The theoretical rate of the permeate concentration increase was Diff. coeff. of SO42− 1.0 × 10− 9 m2 s− 1
calculated from: Diff. coeff. of OH− 5.2 × 10− 9 m2 s− 1
Current density Range (0-100 mA cm− 2)
→ Temperature 25 ◦ C
Δnk Δ(Vck ) Tk i
= = (11) Electrolyte volume fraction in the membrane 0.4
Δt Δt zk F Tortuosity of the diffusion pathway 1.5

where V is the volume of the permeate, ck the concentration of the

permeate, Tk the integral transport number, and zk the valence of
→ where c2 = [proton], c3 = [bisulfate] and c4 = [sulfate]; plus sign for
permeating ion (H+ or OH− ); i is the current crossing the membrane proton and sulfate and minus sign for bisulfate; species 1 is Na+ that does
and F the Faraday constant. not participate in the reaction. The reaction is brought at equilibrium by
assuming an arbitrarily high value for kd = 100 s− 1 and then ka = kd /Kc.
Simulations In calculating the boundary concentrations, this equilibrium is naturally
included. It must be emphasized that the calculation of three transport
The cell is divided into three simulation domains. The center domain equations with the reaction term is quite challenging, while in Comsol it
represents the ion selective membrane. The domain on the left is the salt does not pose a problem.
or dialysate compartment. Third domain on the right is for either for All the initial values including physical and chemical properties of
base or acid, named as the permeate or acid and base compartment the feed solutions as well as geometric characteristics of the membrane
where the ion concentration increases during the electrodialysis Fig. 1. are imported from the experimental laboratory setup. All the diffusion
The dialysate domain is fed with electrolyte consisting of 0.5 M Na2SO4, coefficients, electrolyte volume fraction and electrolyte tortuosity
whereas the concentration and the electrolyte of the permeate values are taken from the reported literature [45–47], Table 4. Note that
compartment is varied. the diffusion coefficients are those in an aqueous solution at infinite
The following assumptions are made in the simulations: (a) a sta­ dilution, which will be discussed later on. Membrane charge was varied
tionary state; (b) Donnan equilibrium between the membrane and the in order to see its effect on transport characteristics. The thickness of the
bathing electrolyte solutions; (c) homogeneous membranes in terms of diffusion boundary was estimated from similar system in the literature
its charge density and porosity; (d) no convection. The model is solved [48,49] these calculations are presented in Supporting Information.
increasing the membrane fixed charge gradually to facilitate the Meshing is a crucial issue in FEM. Mesh points were inserted in the
convergence of the simulation, with varying values of current density form of a geometric series (element ratio 25) with the smallest elements
and concentrations. The direction of current is set from right to left, close to phase boundaries. The total number of mesh elements was 250.
Fig. 1. The default solver, Direct PARDISO was used to solve the problem.
Simulations are carried out with COMSOL Multiphysics® 5.6 soft­
ware applying Tertiary Current Distribution physics. The free variables are Result and discussion
the ionic concentrations and the electrolyte potential; one of the ions is
solved from the electroneutrality condition. The dissociation of bisulfate In this section we describe the results of the COMSOL simulations
is taken into account with a reaction term in the transport equations. At and discuss them in the light of experimental results. During the simu­
steady state, lations, the concentrations of the electrolytes are fixed at the limits of the
∂ci [ ] simulation domain. The electrolyte potential is fixed to zero at the left
∂t
= Di ∇2 ci + zk f ∇ck ⋅∇ϕ ± R = 0
(12) limit, and current density is defined at the right limit, hence we simulate
a galvanostatic experiment. Since the model is 1D, the current density is
R = kd c3 − ka c2 c4 ; i = 2, 3, 4

4
Kuldeep et al. Chemical Engineering Journal Advances 8 (2021) 100169

2 2
Fig. 2. CEM simulations: (a) Concentration profiles in system Na2SO4 (0.5 M) |CEM| NaOH (0.5 M) at 100 mA cm− and (b) Na+ flux components at 100 mA cm− in
system Na2SO4 (0.5 M) |CEM| NaOH (0.5 M) . Membrane charge is 3 M.

continuous across the entire system. The amount of fixed charge in the is not straightforward.
membrane, the electrolyte concentration at the limit of the permeate We made several diffusion experiments that will be published later
compartment (right limit) and the current density are systematically because, for example, the question of how to interpret the electrolyte
varied in the simulations. Nernst-Planck equation is solved under the diffusion coefficient in terms of three species, proton, bisulfate and
electroneutrality condition. As a result, the concentration profiles and sulfate is unclear. Yet, we found, in accordance of Kamcev et al. [39],
fluxes of all species, as well as the electrolyte potential profile are ob­ that the salt diffusion coefficients were ca. 10% of their aqueous values.
tained. Furthermore, ionic fluxes are split to diffusion, migration and Hence, all the simulations were done with the values of diffusion co­
convection contributions; the last contribution is zero in our simula­ efficients of 10% of their aqueous values in Table 4, except 60% for
tions. The continuity of the fluxes and electric current at the membrane- proton, to account for its high mobility.
solution interfaces are implicitly taken care of. At these interfaces, the
Donnan potential boundary condition is applied, which is seen as an
CEM simulations
abrupt jump of ionic concentrations.
The values of the ionic diffusion coefficients within an ion-exchange
As mentioned in Introduction, simulations with a cation exchange
membrane are known to be lower than in the free aqueous solutions, as
membrane were carried out in the system where the dialysate
we discussed above. In the very rigorous paper by Bui et al. [50], explicit
compartment was initially a 0.5 M Na2SO4 solution and the permeate
formulae are given to calculate diffusion coefficients in an ion-exchange
compartment NaOH at variable concentrations. The system thus con­
membrane; we estimated that they were ca. 60% lower than in an
tains three ionic species, Na+, OH− and SO42− . The thickness of the
aqueous solution. The disadvantage is that their model does not account
diffusion boundary layers flanking the membrane are set to 50 µm
for the membrane charge and do not discern counter-ions from co-ions.
(Table 4). The concentrations at their outer boundaries are kept
The model by Kamcev [39] lowers the diffusion coefficients by ca. 90%,
constant.
and it also addresses the membrane charge as well as counter-ions and
Fig. 2a shows the concentration profiles of all species throughout the
co-ions. But both these models concern a binary system with strong
cell. The concentration is decreased in the dialysate compartment and
electrolytes. Therefore, their use in our systems with a weak electrolyte
increased in permeate compartments due to concentration polarization,

Fig. 3. AEM simulations at at 100 mA cm− 2 a) Concentration profiles in the system Na2SO4 (0.5 M) |AEM| H2SO4 (0.5 M), b) Total flux variation in all domains in
system Na2SO4 (0.5 M) |AEM| H2SO4 (0.5 M). Membrane charge is 3 M.

5
Kuldeep et al. Chemical Engineering Journal Advances 8 (2021) 100169

2
Fig. 4. Current efficiency at at 100 mA cm− a) CEM b) AEM. Inset a) is hydroxide and inset b) proton leakage current.

which comes from the ion-selectivity of the membrane, i.e. enhanced →

rate of the counter-ions transport in the membrane. The high concen­ TS = −
2F J S
(14)
→
tration of the counter-ions in the membrane is naturally due to the I
Donnan equilibrium. The non-zero concentrations of co-ions in the Fig. 3b proves that the fluxes of the ionic constituents are, indeed,
membrane reflect the Donnan failure and, thus, the leakage of OH− constant. The proton leakage is also calculated from the total flux of the
through a CEM Fig. 2 a); sulfate anions are expelled almost completely proton constituent:
from the membrane.
Fig. 2b shows the diffusion and migration contributions of the Na+ →
FJH
flux. The total flux is ca. 7.4×10− 7 mol cm− 2 s− 1 that corresponds to ca. TH = → (15)
I
70 mA cm− 2. Since the total current density was 100 mA cm− 2, the OH−
flux is ca. 30 mA cm− 2. Hence, the current efficiency of Na+ is ca. 70%, The behavior is found very similar to hydroxyl ion leakage in CEM.
which is quite plausible at this high electrolyte concentrations. As can be However, the proton constituent leakage is somewhat higher than the
anticipated, in the membrane, the Na+ flux is almost entirely migra­ hydroxyl ion leakage because of the smaller size and high mobility of the
tional although it has a concentration difference of 0.5 M over the proton, insets in Fig. 3.
membrane. The total flux of Na+ (and other ions as well) is constant
across the system as it naturally must be at steady state.
Current efficiency
AEM simulations
Fig. 4a shows the effect of membrane charge on the current efficiency
In the case of AEM simulations, the dialysate compartment was the of Na+ (see Eq. 8) at different concentrations of the permeate
same as earlier, 0.5 Na2SO4, but the permeate compartment was sulfuric compartment. The current efficiency is found to be higher at higher
acid at variable concentrations; boundary conditions are the same as membrane charge and lower concentrations of the permeate. Hydroxyl
with CEMs. The bisulfate-sulfate equilibrium (5) has been taken into ion leakage is also calculated from its total flux that is converted to
account with the equilibrium constant Kc, Eq. (9) that has different current density and plotted against various concentrations of permeate
values in the solution and membrane (see Supporting Information). Now compartments, inset Fig. 4a. The leakages is observed to increase line­
the system consists of four ionic species, Na+, H+, HSO4− and SO42− . arly as the concentration of the permeate increases. At high NaOH
Concentration profiles are shown with current density of 50 mA cm− 2 in concentrations, the share of OH− transport of the total current seems
Fig. 3a. Although in 0.5 M sulfuric acid the majority (74%) of anions is in surprisingly high, i.e. the selectivity of the CEM would be very low, but it
the bisulfate form, the major component in the membrane is sulfate. must be realized that Na+ resides on the both sides of the membrane,
Fig. 3b shows that the ionic fluxes are no more constants. In the right making its diffusion contribution quite small, while there is a steep
hand side compartment, bisulfate carries most of the current but in the concentration gradient of OH− , facilitating its transport. Hence, this
membrane and in the Na2SO4 containing compartment, sulfate is the behavior is a result of the definition of the current density, Eq. 8.
charge carrier, which agrees with ref. [37]. Since ionic fluxes are not Similarly Fig. 4b is the current efficiency of the sulfate constituent in
spatially constant we must to resort to ion constituent fluxes, Eq. (7). AEM. The current efficiency is found lower in AEMs than in CEMs in
Combining Eqs. (3) and (7), it is obtained that1 similar conditions of concentrations and membrane charges. The reason
is the higher proton leakage rate in AEM than hydroxy leakage in CEM,
→
I → → → inset Fig. 4b, due to the smaller size and high mobility of the protons.
= j Na+ + J H − 2 J S (13)
F
Hence, the integral transport number of the sulfate constituent is
Zero current simulations in AEM

If the two solutions flushing a membrane are not identical, diffusion

potential, ϕdiff, is developed across the system, Fig. 5b. Therefore, even
in the absence of net electric current, migration of ions takes place,
1
Note that ionic fluxes are denoted with lower case j and constituent fluxes although the sum of total ionic fluxes via Eq. (3) remains zero. The
with upper case J. diffusion potential is given as [25]

6
Kuldeep et al. Chemical Engineering Journal Advances 8 (2021) 100169

Fig. 5. a) Electrolyte potential at zero-current in the system H2SO4 (0.5 M) |AEM| H2SO4 (0.5 M); b) Electrolyte potential at zero-current in the system H2SO4 (0.1 M)
|AEM| H2SO4 (0.5 M).

Fig. 6. a) The flux components of the sulfate constituent in system H2SO4 (range 0.1-1.0 M) |AEM| H2SO4 (0.75 M); b) concentration polarisation of sulfur species at
zero current in H2SO4 (0.01 M) |AEM| H2SO4 (0.5 M). Membrane charge is 3 M.

∫ ∑
RT tk our result, showing that sulfate prevails in the membrane when the
ϕdiff = − dlnck (16)
F z
k k
concentration-based equilibrium constant, Kc is used. Although their
statement is based on experiments, it is hard to understand why bisulfate
where tk is the transport number and ck the concentration of species k. would not carry electric current, unless its mobility in the membrane is
While in the binary aqueous solutions the transport numbers are con­ very low. If we used the thermodynamic dissociation constant Ka = 0.01
stants, the concentrations in the membrane (see Figs. 2a and 3a) and M, bisulfate would prevail in the membrane, but also carry current.
thus the transport numbers vary along it. This is why the integral in Eq. Using instead the concentration dependent Kc, no contradictions appear.
(16) above cannot be solved in closed form. Fortunately, the potential
profiles (Fig. 5), the flux components (Fig. 6a) and concentrations Validation of simulation
(Fig. 6b) can be evaluated with Comsol across the system.
The trend in the total potential drop, Fig. 5b, is a bit surprising. The According to Eq. (11), bisulfate (z = − 1) should convey sulfur twice
highest diffusion potential is found in the case of a neutral membrane. as fast as sulfate (z = − 2). When H2SO4 is used as the dialysate, bisulfate
This is, however, understandable realizing that in an AEM the concen­ is the main anion in liquid phase, while sulfate is the only anion in
tration of proton, that has the highest mobility, is the lower the higher is Na2SO4. Therefore, the rate of increase of the permeate concentration in
the membrane charge. Therefore, the difference in the transport the experiments with H2SO4 feed (Experiments 1-3 in Table 3) should be
numbers, which is the origin of the diffusion potential, gets gradually ca. twice as high as with the Na2SO4 feed (Experiments 4-6 in Table 3) if
smaller. bisulfate would carry the charge in the AEM. However, according to our
As can be seen in the Fig. 6a above, migration does take place even in experiment, the rate of the permeate concentration increase is equal
the absence of electric current because diffusion potential is its driving within experimental error when equal concentrations of H2SO4 or
force. The flux of the proton constituent cancels out this flux, making the Na2SO4 are used as the dialysate (Supporting information Fig. 2).
total current zero. Moreover, the experiment shows that also water is transferred from
Fig. 6b replies to the dilemma in the paper of Pourcelly et al. [37] dialysate to permeate by electroosmosis (Supporting information
who stated that in the absence of current bisulfate is the major species in Table 1 and Table 2) which cannot be accounted for in our 1D model.
the membrane, although sulfate is the charge carrier. This disagrees with However, the experimental current efficiency calculation is corrected

7
Kuldeep et al. Chemical Engineering Journal Advances 8 (2021) 100169

Fig. 7. The comparison of simulated current efficiency with experiments a) CEM (experiment set 1-3) and simulations (CEM range 1-3M; dialysate, permeate and
current density are kept the same as experimental sets Table 2). b) AEM (experiments set 1-6) and simulations (AEM range 1-3M; dialysate, permeate and current
density are kept the same as experimental sets Table 3) and equilibrium is maintained inside the membrane with concentration equal to the membrane fixed charge.

Fig. 8. Zero current experiments: a) The flux of sulfate constituent as a function of dialysate H2SO4 concentration; permeate H2SO4 concentration 0.01 M. b) Sulfate
constituent flux components in 3 M membrane. D(SO42− )M = 0.1 × D(SO42− )w, D(HSO4− )M = 0.1 × D(HSO4− )w and D(H+)M = 0.6 × D(H+)w.

for the volume change, Eq. (17): Conclusions

( ) ( )
Δnpermeate = Vpermeate cpermeate t=2h − Vpermeate cpermeate t=0 (17)
The transport of ions across ion exchange membranes is studied with
the Finite Element Method, using COMSOL Multiphysics and compared
and sulfate is assumed to be the main charge carrier in the AEM, as
with laboratory-scale experiments. The aim of the study was to under­
simulated and also confirmed in the literatute [32,34].
stand the effect of the membrane fixed charge density, electric current
The experimental results of the CEM system corresponds to the
density and the solution concentrations on the efficiency of the elec­
simulation results with a 3 M fixed charge of the membrane, Fig. 7a,
trodialysis process where Na2SO4 is converted to acid and alkaline. In a
although in Table 1 the membrane fixed charge is given as 1.23 M, but if
multi-ionic case, despite of the simplicity of the system (1D), it is not
we take into account that the free volume (porosity) if the membrane is
possible to have a closed form solution of the Nernst-Planck equation
40% the charge that an ion feels actually is ca. 3 M.
that forms the basis of the study. With COMSOL, however, the concen­
Transport under zero current conditions was studied in a two-
trations profiles and fluxes of all species, as well as the potential profile
compartment cell setup where the permeate and dialysate compart­
across the system can be achieved with relative ease. Furthermore, it is
ment are separated by an AEM. The permeate concentration is set to be
possible to separately analyze the diffusion and migration contributions
fixed at 0.01 M H2SO4, whereas dialysate concentration is varied from
to the total ionic fluxes. Donnan equilibrium was assumed at the
0.1 M to 0.6 M H2SO4. The same setup is simulated with COMSOL, and
solution-membrane interfaces, and electroneutrality throughout the
the simulated and experimental results are shown in Fig. 8. It appears,
system.
quite surprisingly, that the membrane charge does not have that great an
It was found that, in order to reach an agreement with experiments,
effect.
the thermodynamic dissociation constant of bisulfate had to be replaced
with one expressed in terms of ionic concentrations, viz taking the ac­
tivity coefficients into account. In the case of weak electrolyte transport,

8
Kuldeep et al. Chemical Engineering Journal Advances 8 (2021) 100169

ionic fluxes are no more constant, which requires the consideration of [12] Y. Tanaka, M. Reig, S. Casas, C. Aladjem, J.L. Cortina, Computer simulation of ion-
exchange membrane electrodialysis for salt concentration and reduction of RO
the hydrogen and sulfate constituent fluxes. In the previous literature, it
discharged brine for salt production and marine environment conservation,
has been suggested that during the sulfuric acid electrolysis across an Desalination 367 (2015) 76–89.
anion exchange membrane, sulfate ions are the major charge carriers in [13] G. Belfort, J.A. Daly, Optimization of an electrodialysis plant, Desalination 8
the membrane, although in the absence of current, mainly bisulfate (1970) 153–166.
[14] M. Avriel, N. Zelingher, A computer method for engineering and economic
balances the membrane fixed charge. In our simulations, however, the evaluation of electrodialysis plants, Desalination 10 (1972) 113–146.
concentration of bisulfate in the membrane is always quite low, which [15] H.-J. Lee, F. Sarfert, H. Strathmann, S.-H. Moon, Designing of an electrodialysis
explains the fact that sulfate carriers the majority of charges. desalination plant, Desalination 142 (2002) 267–286.
[16] M. Sadrzadeh, A. Kaviani, T. Mohammadi, Mathematical modeling of desalination
The mobilities of the ion within the ion-exchange membranes were by electrodialysis, Desalination 206 (2007) 538–546.
set to 10% of their aqueous values at infinite dilution, which agrees with [17] V.V. Nikonenko, N.D. Pismenskaya, A.G. Istoshin, V.I. Zabolotsky, A.A. Shudrenko,
literature. Only for the proton 60% of its aqueous mobility was used due Description of mass transfer characteristics of ED and EDI apparatuses by using the
similarity theory and compartmentation method, Chem. Eng. Process. 47 (2008)
to its small size and consequently high mobility.The current efficiencies 1118–1127.
of the counter-ions or ion constituents were evaluated in varying con­ [18] E. Brauns, W. de Wilde, B. van den Bosch, P. Lens, L. Pinoy, M. Empsten, On the
ditions mentioned above. Quite good an agreement with experiments experimental verification of an electrodialysis simulation model for optimal stack
configuration design through solver software, Desalination 249 (2009) 1030–1038.
were achieved, considering that atomic scale modelling, through which [19] S. Koter, W. Kujawski, I. Koter, Importance of the cross-effects in the transport
specific membrane properties could be implemented, is not possible through ion-exchange membranes, J. Membr. Sci. 297 (2007) 226–235.
with COMSOL. Also, the rate of the co-ion leakage (H+ in AEMs and OH− [20] F.S. Rohman, N. Aziz, Mathematical Model of Ion Transport in Electrodialysis
Process, Bull. Chem. React. Eng. Catal. 3 (2010).
in CEMs) was very close to experimental ones. We can conclude that
[21] S.T. Psaltis, T.W. Farrell, Comparing charge transport predictions for a ternary
COMSOL provides a modern means to study ion transport in a way that electrolyte using the Maxwell–Stefan and Nernst–Planck equations, J. Electrochem.
has not been possible earlier. In particular, splitting total flux to diffu­ Soc. 158 (2010) A33.
sion and migration contributions is very illustrative and clarifies their [22] E.E. Graham, J.S. Dranoff, Application of the Stefan-Maxwell equations to diffusion
in ion exchangers. 1, Theory, Ind. Eng. Chem. Fundam. 21 (1982) 360–365.
roles in the membrane selectivity. The study continues with 2D model­ [23] M.W. Verbrugge, R.F. Hill, Ion and Solvent Transport in Ion-Exchange Membranes:
ling which makes it possible to involve also convection via I. A Macrohomogeneous Mathematical Model, J. Electrochem. Soc. 137 (1990)
electroosmosis. 886.
[24] P. Meares, Coupling of ion and water fluxes in synthetic membranes*, J. Membr.
Sci. 8 (1981) 295–307.
[25] K. Kontturi, L. Murtomäki, J.A. Manzanares, Ionic transport processes.
Declaration of Competing Interest electrochemistry and membrane science, Oxford University Press, Oxford, 2015.
[26] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport phenomena, 2nd ed., John Wiley,
New York, 2007.
The authors declare no conflict of interest. [27] A. Moura Bernardes, M.A. Siqueira Rodrigues, J. Zoppas Ferreira (Eds.),
Electrodialysis and Water Reuse: Novel Approaches, Springer Berlin Heidelberg,
Acknowledgment Berlin, Heidelberg, 2014.
[28] R.P. Buck, Kinetics of bulk and interfacial ionic motion: microscopic bases and
limits for the Nernst - Planck equation applied to membrane systems, J. Membr.
This work has received funding from the European Institute of Sci. 17 (1984) 1–62.
Innovation and Technology (EIT), a body of the European Union, under [29] V. Fíla, K. Bouzek, The effect of convection in the external diffusion layer on the
results of a mathematical model of multiple ion transport across an ion-selective
the Horizon 2020, the EU Framework Programme for Research and membrane, J. Appl. Electrochem. 38 (2008) 1241–1252.
Innovation. Project Name and Number: 680652 Credit -18243. SUEZ [30] V. Fíla, K. Bouzek, A mathematical model of multiple ion transport across an ion-
water technologies and solutions is acknowledged for the supply of the selective membrane under current load conditions, J. Appl. Electrochem. 33 (2003)
675–684.
membranes. [31] P. Moon, G. Sandí, D. Stevens, R. Kizilel, Computational Modeling of Ionic
Transport in Continuous and Batch Electrodialysis, Sep Sci Technol 39 (2004)
2531–2555.
Supplementary materials
[32] Y. Tanaka, A computer simulation of ion exchange membrane electrodialysis for
concentration of seawater, Membr. Water Treat. 1 (2010) 13–37.
Supplementary material associated with this article can be found, in [33] M. Fidaleo, M. Moresi, Optimal strategy to model the electrodialytic recovery of a
the online version, at doi:10.1016/j.ceja.2021.100169. strong electrolyte, J. Membr. Sci. 260 (2005) 90–111.
[34] N. Pismenskaya, V. Nikonenko, B. Auclair, G. Pourcelly, Transport of weak-
electrolyte anions through anion exchange membranes, J. Membr. Sci. 189 (2001)
References 129–140.
[35] V. Nikonenko, K. Lebedev, J.A. Manzanares, G. Pourcelly, Modelling the transport
of carbonic acid anions through anion-exchange membranes, Electrochim. Acta 48
[1] W. Juda, W.A. McRae, Coherent Ion-Exchange Gels and Membranes, J. Am. Chem.
(2003) 3639–3650.
Soc. 72 (1950) 1044.
[36] E.D. Belashova, N.D. Pismenskaya, V.V. Nikonenko, P. Sistat, G. Pourcelly, Current-
[2] M.G. Buonomenna, Membrane processes for a sustainable industrial growth, RSC
voltage characteristic of anion-exchange membrane in monosodium phosphate
Adv 3 (2013) 5694.
solution. Modelling and experiment, J. Membr. Sci. 542 (2017) 177–185.
[3] D.H. Kim, A review of desalting process techniques and economic analysis of the
[37] Y. Lorrain, G. Pourcelly, C. Gavach, Transport mechanism of sulfuric acid through
recovery of salts from retentates, Desalination 270 (2011) 1–8.
an anion exchange membrane, Desalination 109 (1997) 231–239.
[4] H. Strathmann, Ion-exchange membrane separation processes, Elsevier,
[38] S. Koter, M. Kultys, Electric transport of sulfuric acid through anion-exchange
Amsterdam, Boston, 2004.
membranes in aqueous solutions, J. Membr. Sci. 318 (2008) 467–476.
[5] K. Asada, L. Gerdes, T. Kawahara, Electrodialysis of effluents from treatment of
[39] J. Kamcev, D.R. Paul, G.S. Manning, B.D. Freeman, Ion Diffusion Coefficients in Ion
metallic surfaces, in: Proceedings of 79th AESF annual technological conference,
Exchange Membranes: Significance of Counterion Condensation, Macromolecules
Atlanta, 1992.
51 (2018) 5519–5529.
[6] G. Pourcelly, L. Bazinet, Developments of Bipolar Membrane Technology in Food
[40] K.A. Johnson, G.B. Westermann-Clark, D.O. Shah, Diffusion of charged micelles
and Bio-Industries, in: A. Kumar Pabby, S. Rizvi, A. Maria Sastre (Eds.), Handbook
through charged microporous membranes, Langmuir 5 (1989) 932–938.
of Membrane Separations, CRC Press, 2008, pp. 581–657.
[41] E.B. Robertson, H.B. Dunford, The State of the Proton in Aqueous Sulfuric Acid,
[7] M. Greiter, S. Novalin, M. Wendland, K.-D. Kulbe, J. Fischer, Electrodialysis versus
J. Am. Chem. Soc. 86 (1964) 5080–5089.
ion exchange: comparison of the cumulative energy demand by means of two
[42] K.S. Pitzer, R.N. Roy, L.F. Silvester, Thermodynamics of electrolytes. 7. Sulfuric
applications, Journal of Membrane Science 233 (2004) 11–19.
acid, J. Am. Chem. Soc. 99 (1977) 4930–4936.
[8] H. Strathmann, Electrodialysis, a mature technology with a multitude of new
[43] SUEZ Water Technologies, https://my.suezwatertechnologies.com/WTSCustome
applications, Desalination 264 (2010) 268–288.
rPortal/s/content-download?DN=FSelIXMembranes.pdf.
[9] M.Y. Kariduraganavar, A.A. Kittur, S.S. Kulkarni, Ion exchange membranes:
[44] R.C. Weast, Handbook of chemistry and physics: A ready-reference book of
preparation, properties, and applications, in: Ion Exchange Technology I, Springer,
chemical and physical data, 56th ed., CRC Press, Cleveland, 1975.
2012, pp. 233–276.
[45] W.M. Haynes, D.R. Lide, T.J. Bruno (Eds.), CRC handbook of chemistry and
[10] H. Strathmann, Electrodialysis state of the art, membranes—proceedings of India-
physics: A ready-reference book of chemical and physical data 2012-2013, 93rd
EC workshop, Oxford & IBH, New Delhi, 1991.
ed., CRC Press, Boca Raton (Fla.), London, New York, op. 2012.
[11] Y. Tanaka, Bipolar Membrane Electrodialysis, in: Ion Exchange Membranes -
Fundamentals and Applications, Elsevier, 2007, pp. 405–436.

9
Kuldeep et al. Chemical Engineering Journal Advances 8 (2021) 100169

[46] V.M.M. Lobo, J.L. Quaresma, Handbook of electrolyte solutions, B, Physical [49] N. Ibl, O. Dossenbach, Convective Mass Transport, in: Yeager, Bockris et al. (Ed.) C
sciences data (1989) 41. [2013]– Comprehensive treatise of electrochemistry, pp. 133–237.
[47] R. Mills, V.M.M. Lobo, Self-diffusion in Electrolyte Solutions: A Critical [50] J.C. Bui, I. Digdaya, C. Xiang, A.T. Bell, A.Z. Weber, Understanding Multi-Ion
Examination of Data Compiled from the Literature, Elsevier Science, Amsterdam, Transport Mechanisms in Bipolar Membranes, ACS Appl. Mater. Interfaces 12
2014. (2020) 52509–52526.
[48] S.J.C. Weusten, Mass transfer in parallel plate electrolyzers. Master’s dissertation,
5612 AZ Eindhoven, Netherlands, 2017.

10