Electrochimica Acta 106 (2013) 91100

Contents lists available at SciVerse ScienceDirect

Electrochimica Acta
journal homepage: www.elsevier.com/locate/electacta

Long term stability of capacitive de-ionization processes for water

desalination: The challenge of positive electrodes corrosion
Izaak Cohen , Eran Avraham, Yaniv Bouhadana, Abraham Soffer, Doron Aurbach
Department of Chemistry, Bar-Ilan University, Ramat-Gan 52900, Israel

a r t i c l e

i n f o

Article history:
Received 16 January 2013
Received in revised form 4 May 2013
Accepted 11 May 2013
Available online 18 May 2013
Keywords:
Double layer
Capacitive deionization (CDI)
Water desalination
Electro-adsorption
Corrosion of electrodes
Activated carbon electrodes

a b s t r a c t
Corrosion of the positive electrodes, in capacitive deionization (CDI) cells for water desalination processes,
is a major problem that may prevent them from becoming practically important. This paper deals with the
consequence of the corrosion of the positive electrodes in CDI processes on the desalination performance,
in terms of capacity and the ratio between adsorption of counter-ions and desorption of co-ions. The
detrimental effect of the positive electrodes oxidation on the de-ionization efciency is demonstrated and
discussed. The role of the potential difference applied to CDI cells on the electrodes stability was explored
as well. We used for this study CDI cells comprising several pairs of activated carbon electrodes and 3
electrodes cells containing reference electrode. The interrelated parameters measured included potential,
current, concentration (translated from conductivity measurements) and pH vs. time. The present study
and the understanding gained herein, will enable the development of durable, long term and effective
CDI processes.
2013 Elsevier Ltd. All rights reserved.

1. Introduction
Due to the increasing need for potable water, the search for
energy-efcient water purication is widespread, with the most
widely investigated methods being electrodialysis [13], electrostatic shielding [46] and ion exchange [79]. The most frequently
examined and used methods for water desalination are reverse
osmosis [7,10,11] (RO) and direct distillation (DD) [12], which
require extensive amounts of energy either for maintaining high
pressure (RO), or to evaporate water from the saline mixture and
condensate it (DD).
Capacitive deionization (CDI) is an energy-efcient water
desalination method for brackish water (BW) that contains
5005000 ppm of salt [13]. In this method, the solution ows
through cells comprising sets of two high-surface area electrodes
which are periodically charged by applying constant potential differences between the electrodes. Thus, the salt is extracted from
the saline water in a capacitive way, so part of the charge consumed during the electrodes charging can be exploited back upon
discharge of the cell. The electrodes used in CDI are usually carbonaceous materials with high surface area of hundreds of m2 /g,
e.g., activated carbon bers (ACF) and carbon aerogel [1416]. In
our work we use ACF electrodes. Activated carbon electrodes in
CDI cells are expected to undergo very prolonged charge/discharge

Corresponding author. Tel.: +972 35317441.

E-mail address: izaakco@gmail.com (I. Cohen).
0013-4686/$ see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.electacta.2013.05.029

cycling, because the process mainly involves electrostatic, nonchemical interactions [17,18].
Desalination of water by CDI can be operated in two modes
related to the ow of solution in the cell: ow-by [1933], in which
the solution ows in parallel to, between the electrodes and owthrough [3441], in which the solution ows in perpendicular to,
through the electrodes. Using the ow-through regime, the solution has to ow through macro-porous electrodes. Their porous
structure has to allow a smooth ow of solution through them.
A very suitable electrodes material for ow through CDI cells is
carbon cloth comprising activated carbon bers.
Our previous work [39] showed a very interesting phenomenon
occurring during prolong CDI processes in ow through cells
(repeated chargedischarge cycling), which was referred to therein
as the inversion effect. In the experiments described therein [39],
the changes in solution concentration upon periodic operation of
the CDI cells were monitored by continuous (on line) ionic conductivity measurements at the outlet of the cell. When the cell is
polarized, electro-adsorption and hence removal of salt (reected
by low conductivity) are expected, while shortening the cell should
result in an increase in salt concentration (discharge) and hence an
increase in ionic conductivity. The inversion effect was reected
by desorption of ions (a rise in conductivity) while the cell was
being polarized and charged. In turn, ions were depleted from
the solution (decrease in conductivity) when the cell was discharged. The inversion effect was indirectly attributed to the
continuous corrosion of the positively polarized electrodes, during
charging.

92

I. Cohen et al. / Electrochimica Acta 106 (2013) 91100

Fig. 1. (A and B) Schemes of the ow-through cells used in this work. All the components are explained in the gures and in Section 2.1.

Another effect that should be taken under consideration, and

that can contribute to the inversion effect is the prevention of the
complete desorption of the adsorbed ions due to the complex pores
structure of the activated carbon electrodes. Thus as the electrodes
are polarized, after the potential is removed, there are remaining
counter-ions trapped in pores, which charge can hold some co-ions
[20].
The work described in this paper was devoted to understand
the inversion effect, how it depends on the parameters of the CDI
process, especially the applied potential. Corrosion inhibition of the
positive electrodes in capacitive deionization (CDI) processes can
help to make them an effective water desalination technology. Several long term experiments with different cells were conducted in
order to understand what happens to CDI cells during prolonged
operation. The impact of the applied potential was explored. The
use of chemically oxidized electrodes in these processes was examined as well. Possible corrosion processes of the carbon electrodes
were investigated. The parameters measured included potential,
current, salt concentration at the exit of the cells and the extent
of salt removal (translated from conductivity measurements) and
changes in pH of the solution that ows through the electrodes.
The present study is highly important, because the understanding gained herein, can help to promote CDI further as a promising
method for desalination of brackish water.

2. Experimental
2.1. The CDI Cells structure
The CDI cell was designed for the ow-through mode, as
described in Fig. 1. The cell has a ange-type design. The solution is introduced through plates in order to ensure homogeneous
ow of the solution through the whole circular cross section of the
cell. The electrodes were discs (54 mm wide, 0.5 mm thick) made
of commercial carbon cloth (ACC-5092-15) from Nippon Kynol,
Japan, possessing high surface area (1440 m2 /g BET) originating
from phenolformaldehyde polymeric bers that underwent carbonization and activation. The CDI cell, shown in Fig. 1A contains 20
pairs of ACF disk electrodes (positive and negative); the CDI stuck
contains 10 cells and each cell contains 2 pairs of ACF electrodes.

Sheets of porous polypropylene cloth served as separators between

the electrodes and exhibited a fairly low resistance to the solution
ow. Silicon glue was soaked into the rims (at the perimeters) of
the separator discs, thus forming soft and elastic gaskets. These
separator sheets with the gaskets at their perimeters provided the
necessary mechanical and electrical separation between the electrodes (thus preventing short circuits). The cases for the electrodes
were ring spacers, made of poly-tetra-uoro-ethylene (PTFE), possessing 0.5 mm deep grooves that hold the carbon electrodes. The
current collectors are made of graphite paper discs (Grafoil Inc.)
that are attached to the electrodes in the cell. These graphite sheets
current collectors are perforated to allow a smooth ow of solution through them. When all these components: electrodes in their
plastic cases, graphite sheet current collectors, separator sheets
with polymeric gaskets at their perimeter are pressed together, in
the right order, they form hermetically sealed ow through multi
electrodes electrochemical cell. We used such cells containing a
reference electrode (RE). The RE was Ag mesh covered by AgCl (by
anodization of the silver mesh in NaCl solution) which was placed
at the middle of the cell (Fig. 1A). This RE could be also used as auxiliary electrode. The content of its AgCl layer was calculated [34] to
be equivalent to more than 3 times the charge that can be stored in
the electrical double layer of the electrodes in the cell (assuming 1 V
was applied between the electrodes). A simpler cell which included
three-electrode cell containing two sheets of ACF electrodes as the
working electrode (WE) and two Ag/AgCl electrodes as the counter
and reference electrodes, was also used (Fig. 1B). All the CDI experiments were carried out using the cell described in Fig. 1A except for
one experiment indicated by Fig. 11, which was carried out using
the cell described in Fig. 1B.

2.2. The system set-up

A graphical explanation of the layout of the entire setup used
herein, is presented by Fig. 2. A 20 l round-bottom ask was used as
the solution reservoir for the experimental system. It contained 18 l
of 1000 ppm of NaCl (>99% pure, Frutarom, Israel) solution in highly
puried water (18.2 M) that was circulated in a closed system
which included the electrochemical cell, the pumps and the various detectors. Two conductometers (Metrohm, 712 model) were

I. Cohen et al. / Electrochimica Acta 106 (2013) 91100

93

Laboratory (OCL)) until the pH was 7. The carbon electrodes were

then rinsed with puried water (18.2 M) and dried in a furnace
under air at 70 C for 2 h.
2.4. Resistivity measurement
4 points resistance measurements of ACF electrodes were carried by using two thin gold-coated plates between which ACF
electrodes were held and clamped. The diameter of the ACF electrodes in these measurements (1 cm) was in the same as the width
of gold-coated plates. The power source used forced constant current to ow and the resistance was calculated from the voltage
falling on the electrodes. We conducted a series of measurements
with different thickness of the electrodes samples measured (different stacks of electrodes held between the gold plates). Pristine
and cycled electrodes (both positive and negative) were measured.
We conrmed a strict linear dependence between the voltage measured (hence, the resistance calculated) and the thickness of the
electrodes samples.
3. Results and discussion
3.1. A general description of the behavior of CDI processes
measured herein
For the presentation of the results and discussion, several key
parameters of the CDI processes have to be dened:

Fig. 2. Layout of the setup used; all the components of the layout of the setup are
described in the gure.

used simultaneously in order to measure on line the conductivity

of the solution which ows out of the CDI cell. A probe connected
to a pH-meter (Metrohm, 827 model) measured the pH changes
in order to correct the conductivity measurements by subtracting
pH effects (usually negligible). In order to control the solution ow,
an electrical pump (Fluid Metering Inc.) was used and the solution
ow was xed to 50 ml/min. The potentials were applied to the
CDI cell by a Power Supply (Kiethley, 2440 5A SourceMeter) and
the potentials between the negative electrodes and the RE were
measured by a Multimeter (TES 2730). All the electrical equipment
described above was connected to a computer and controlled by a
LabView (National Instrument (NI)) Programming software using
two NI data acquisitions (NI USB-6008) and RS-232 cables in order
to control and monitor the long-term experiments. It is important
to note that the pH changes measured in this work were significant for the mechanistic studies but were minor enough not to
affect too much the conductivity measurements. Hence, changes in
conductivity could be always translated to changes in NaCl due to
adsorption and desorption processes.
2.3. Surface treatment
For one of the CDI cells, chemically oxidized carbon cloth sheets
(made of ACF) were used as positive electrodes. Oxidation was carried out with concentrated HNO3 solution (70%, Bio Lab) at room
temperature for 25 h. After oxidation, the carbon samples were
washed with a sodium bicarbonate solution (>99.5%, Ofer Chemical

a. CP (=positive) and CN (=negative) the capacity of each electrode

in the cell.
b. E the potential applied to the cell.
c. EPZC the electrodes potential of zero charge (PZC). This important parameter is usually the immersion potential measured
(vs. a reference electrode) when the dry and fresh electrode is
being submerged in solution. EPZC depends on the state of the
electrodes surface (e.g. the presence of surface groups) and the
solution. It can change during the electrochemical processes due
to changes in the electrodes surface [35,39,42] (e.g. oxidation).
d. E the difference between the actual electrodes potential and
EPZC , when a potential E is applied to the cell. We can dene
E+ and E for the positive and negative electrodes in the cell.
|E+ | + |E | = E, the potential applied to the cell. Depending of
the electrodes morphology, pore size, state of the surface and the
ratios between ions and pores sizes, there may be symmetrical
situation in which |E+ | = |E |or unsymmetrical situation in
which |E+ | =
/ |E |.
+

e. N /N the ratio between the amounts of cations and anions

that can be electro-adsorbed to the electrodes pores at the same
E. In a symmetrical situation (can be dened also as a symmetrical CDI cell) N+ /N = 1. An obvious condition for symmetric
situation is large enough pores that allow a similar transport and
adsorption of cations and anions at the same E.
f. EP and EN the potentials of the positive and negative electrodes
(respectively) in the CDI cell vs. a reference electrode.
g. E0 the potential of the electrodes vs. a reference electrode,
when the CDI cell is shorted (EP = EN = E0 ).
h. Ew the electrolysis potential of water, 1.299 V.
It is important to note that polarization of any CDI cell (i.e. applying a potential E to the cell) leads always to two parallel processes:
adsorption of counter-ions and desorption of co-ions (i.e. ions with
the same sign of charge as that of the electrode). The capacity of
the electrodes is an intrinsic property the system, depending at a
rst approximation on their surface area, the dielectric constant of
the solution and the ratio between the electrodes pores size and

I. Cohen et al. / Electrochimica Acta 106 (2013) 91100

/- (mol)

/- (mol)
+

E0

PZC

E0

E0

/- (mol)

E (V) vs.
s. Ref

94

PZC

E0

E (V) vs. Ref

Negative
Electrodes

E (V) vs. Ref

Positive
Electrodes

PZC

Working potential domains

Working potential domains

E0

PZC

E0

E (V) vs. Ref

Working potential domains

Negative Electrodes

Positive Electrodes

/- (mol)

PZC

/- (mol)

Working potential domains

Negative Electrodes

E (V) vs. Ref

Working potential domains

C

Positive Electrodes

Fig. 3. (A) Illustration of the operation of a symmetric CDI cell whose electrodes are shorted to 0 V upon discharge. (B) The inversion point illustration of the operation
of unsymmetrical CDI cells, which positive electrodes are oxidized and their EPZC were positively shifted. (C) Illustration of the operation of unsymmetrical CDI cells, which
positive electrodes are oxidized and have a high resistivity and a low surface area because of detrimental oxidation processes. The electrodes EPZC and the potential of cell
shorting, E0 , vs. a virtual reference electrode, are marked therein. A graphical explanation of how to read the Fig. 3 illustrations is given by Fig. 4.

the size of the counter-ions (the ions that are adsorbed to the electrode when it is polarized). In fact, for a given system (CDI cell and
solution), the only controllable parameter is E.
Before the presentation of the experimental data, it is important to present important introductory remarks in light of which,
the results thus obtained can be analyzed. As mentioned in Section
1, upon prolonged operation of symmetrical CDI cells comprising initially identical pairs of activated carbon electrodes, they
undergo changes that take the cells away from symmetrical behavior. We observe inversion in the cells behavior: the solution owing
through the cell becomes more concentrated in ions upon charge
and ions are depleted from solution upon discharge (when the
cell is shorted). Fig. 3 describes schematically the behavior of the
electrodes (both positive and negative ones) in terms of potential
domains of operation, adsorbed and desorbed amounts of ions as
a function of potential, the changes in EPZC and E0 at the 3 stages
of a prolonged periodic CDI experiment. Graphical explanations of
how to read easily the illustrations in this gure are given by Fig. 4.
Fig. 3 illustrations are based on previously published experimental
results [27,3436].
The electro-adsorption/electro-desorption behavior of a typical
symmetrical cell is illustrated in Fig. 3A (stage 1), which shows a
schematic dependence of the amount of counter-ions adsorbed and
co-ions desorbed onto/from the positive and negative electrodes as
a function of the electrodes potential.
The electrodes used in this work have a well-developed pores
system that allows equal adsorption of Na+ and Cl ions at similar E values (see typical voltammetry response in Fig. 5). Hence,
the CDI cells used herein are initially fully symmetrical (unless the
electrodes were treated before the experiments or damaged in the
course of them). Hence, these cells can be modeled by a simple
equivalent circuit analog of two identical capacitors in series (electrical double-layer; EDL capacitors) that have the same capacity
CP and CN , with a solution resistance in between (Rs). When these
cells are initially polarized with E < Ew , it can be expect that only

electrostatic interactions take place. However, there are situations

in which the positive electrodes are not ideally polarized because
a parasitic oxidation process takes place in parallel to the capacitive/electrostatic interactions. As a result, 3 intrinsic parameters of
the positive electrodes are directly affected: EPZC , CP and the resistance of the positive electrode (RP ). Along the continues charging
and discharging cycles of the CDI operation, the positive electrodes
surface may be oxidized, resulting in a positive shift of the EPZC ,
and the movement of the working potential domains of the positive and negative electrodes, yet with no degradation in CP and
no signicant high RP . [36,40] This situation is illustrated in Fig. 3B
(stage 2): the EPZC of the positive electrodes is shifted positively
due to a gradual oxidation, while the EPZC of the negative electrodes
remains unchanged. Consequently the potential domains of operation of both electrodes are changed compared to the initial stage
(Fig. 3A). In fact, Fig. 3B presents the inversion point: the shift in EPZC
and the consequent change in the electrodes potential domains of
operation reaches a situation in which for both electrodes, when
the cell is charged, adsorption of counter-ions is nearly equal to
desorption of co-ions. Fig. 3C presents a more advanced stage of
the oxidation of the positive electrode. The signicant increase of
RP , based on Ohms law leads to the increase of the potential which
is applied to the positive electrodes, that leads to an electrochemical reactions (pH change). In addition the degradation of the surface
area of the positive electrode affects badly its capacity, CP , which is
signicantly reduced, assuming we can neglect parasitic Faradaic
side reactions, (e.g. by working with potentials which are lower
than the water electrolysis potential). In any event, no matter what
are the difference in the capacitance of the electrodes, the charge
on them should be always equal when the cell is polarized: QP = QN .
Thus Eq. (1) can be derived:
Q N = E N C N(=negative) = E P C P(=positive) = Q P

(1)

We can assume that the electrodes capacity is proportional to

their effective surface area and that the oxidation of the positive

I. Cohen et al. / Electrochimica Acta 106 (2013) 91100

Y-axis: The amount of ion adsorbed

/desorbed on/from the
electrodes respectively

+
/- (mol)

E0

E0

X-axis: The differential potential

applied to the electrodes in
respect to a reference
electrode (in this study we
used Ag/AgCl)

E (V) vs.. Ref

PZC

95

Point of Zero Charge (PZC): The

immersion potential, measured
versus reference electrode.
E0: The mutual potential of the
electrodes when the CDI cell is
discharged, measured versus
reference electrode .In this case it
equals to the PZC value.

Working potential domains

Negative polarized electrodes

operation (cathode)

E0

E0
E (V) vs. Ref

Desorption of anion (co-ion)

While polarizing the negative
electrodes, the solution anion
next to the positive electrodes
are rejected to the bulk of the
solution and it takes energy

PZC

/- (mol)

/- (mol)

Adsorption of cation (counter-ion)

The higher the differential potential
is the higher the adsorption of
cation is.

Positive polarized electrodes

operation (anode)

E (V) vs. Ref

PZC

Working potential domain

Adsorption of anion (counter-ion)

The higher the differential
potential is the higher the
adsorption of anion is.

Working potential domain

Desorption of cation (co-ion)

While polarizing the positive
electrodes, the solution cation
next to the positive electrodes
are rejected to the bulk of the
solution and it takes energy

Fig. 4. Illustration of typical ion adsorption/desorption behavior of symmetric identical electrodes in the CDI cell. Explanation about how to read Fig. 3.

electrode reduce its capacity due to reduction in its surface area.

Therefore, we can set the following equation:
E N AN(=negative) = E P AP(=positive)

(2)

We can assume that as the potential on the positive electrode

is higher, its surface oxidation and the reduction in its surface area
is higher. Also the positive shift of its EPZC and RP is higher as well.
This may bring the CDI cell to the situation illustrated in Fig. 3C
(stage 3). At this the specic surface area of the positive electrode is
lower than that of the negative electrode and the RP is signicantly
higher, thereby; most of the potential applied to the cell falls on
the positive side. Upon charging, desorption of co-ions dominates
the charge balance at the solution side, while upon discharge coions are depleted from the solutions in order to balance charge of
the counter-ions in the pores. This scenario is the inversion behavior observed. The adsorption of co-ions in stage 3 is much lower
than the adsorption of counter-ions in stage 1, due to the positive

Fig. 5. A typical cyclic voltammograms of the ACF electrodes used in this work,
measured in a three electrodes cell (vs. Ag/AgCl RE), that determines well their electrochemical potential window. The solution of NaCl concentration was 0.1 M, Scan
rate of 1 mV/s and voltage domain as indicated. Note the typical capacitive behavior (the buttery shape) of symmetrical activated carbon electrodes, at potential
domains where no electrolysis occurs.

asymmetric high potential and the relatively low surface area of the
positive electrode. Thereby, after inversion, the overall capacity of
the processes measured is much lower compared to the adsorption/desorption capacities measured initially. Consequently, it is
critically important to avoid the above described detrimental processes of the positive side in CDI processes. This work intends to
address this challenge of CDI processes.
3.2. Long-term stability of the CDI cells comprising symmetrical
activated carbon electrodes upon cycling in the potential range
00.9 V
20 pairs of pristine (without any pre-treatment) ACF electrodes were assembled in CDI cells for long term testing. Fig. 6
presents typical continuous measurements of the concentration
of NaCl (ppm) in saline solution (translated from conductivity
measurements) that was owing through a cell during repeated
chargedischarge cycling. The concentration values were translated from continuous conductivity measurements (with an
appropriate calibrated translation). The voltage applied was 0.9 V
upon charge and 0 V (short) upon discharge, while each cycle lasted
1 h. The 3 stages in the behavior of these systems (Fig. 3), are clearly
reected in the concentration/time plot in Fig. 6 (marked therein).
The system reaches stage 2, namely the inversion point, after 5
days of cycling. After then, the system demonstrates stage 3 of it
cycling behavior (depletion of salt upon discharge and an increase
in salt concentration upon charging). The four plots shown in Fig. 6
are enlargements of four time domains, taken from continuous
plots measured during typical prolonged cycling experiments. In
a normal plot, we expect to see only two peaks: a negative, going
down peak, when the cell is charged (adsorption of counter-ions
to the electrodes, depletion of salt) and a positive, going up peak,
when the cell is discharged (shorted). This positive peak marks
desorption of the counter-ions from the electrodes, to the solution
stream. The four plots presented in Fig. 6 relates to the development

96

I. Cohen et al. / Electrochimica Acta 106 (2013) 91100

Fig. 6. Plots of salinity in ppm of NaCl vs. time (in minutes) measured during prolonged cycling of CDI cells: 18 days of experiment, in which the potentials applied to the cell
were 00.9 V and the electrodes use comprised pristine, untreated ACF. The 4 plots relate to selected time domains. The relevant potential application (charge/discharge) vs.
time is marked.

of abnormal behavior of the cell that can be explained in terms of

irreversible processes the electrodes may undergo. Looking at the
four plots in this gure from A to D, the rst plot (Fig. 6A) shows the
behavior of the CDI cell before inversion occurs: the high peak (values higher than 1000 ppm of NaCl) indicates the desorption of the
counter-ions upon discharge, while the small positive peak (values
higher than 1000 ppm) indicates the beginning of the inversion
effect, where the desorption of the co-ions is clearly seen when
the cell is polarized charged. The adsorption of the counter-ions
(indicated by the going down peak, to values less than 1000 ppm of
NaCl) appears after the small, going up peak, which marks the desorption of the co-ions. The second enlarged domain (Fig. 6B) shows
the inversion point, where we can see the equal concentration
peaks of the adsorption of co-ions upon discharge and their desorption during the charging step, vs. the normal (expected) adsorption
of counter-ions at the charging step and their desorption upon
discharge (i.e., in contrast to the normal behavior where we have
negative change of concentration on charge and positive change in
concentration on shorting, we see both negative and positive peaks
during both charge and discharge steps). After the system reaches
the inversion point, we observe gradual changes (Fig. 6C) until the
system reaches a steady state behavior (Fig. 6D) in which a positive
concentration peak appears upon charge (desorption of co-ions
dominates) and a negative concentration peak appears upon discharge (shorting), which belongs to adsorption. The main trend is
the increasing dominance of desorption of co-ions over adsorption
of counter-ions during charging (up to the steady state behavior).
Fig. 7 shows results of XPS measurements of electrodes after
prolonged cycling. On the qualitative level, positive electrodes from
CDI cells that underwent prolonged processes as those described
in Fig. 6, exhibit high level of oxygen content, compared to negative electrodes. Detailed surface studies of electrodes from CDI cells
after cycling by SEM (to be published separately, beyond the scope
of this paper) clearly showed that the positive electrodes morphology was considerably changed while that of the negative electrodes
remained similar to that of pristine activated carbon electrodes.
Additionally the surface area of the positive electrodes which relate
to Fig. 6 became smaller by 15% (BET), compared to the negative
electrode that remained with the same surface area as the pristine

ACF electrodes (BET). The surface area of the positive electrodes

in these processes became smaller as the experiments were longer,
until the positive electrodes reach a point where no adsorption and
desorption of ions can occurs.
The resistivity () of the positive and negative electrodes after
the long term experiments (which Fig. 6 reects) and that of the
pristine electrodes, was calculated from the electrodes resistance
measurements, by the following simple equation:
R=

l
A

(3)

where R is the resistance (), l, A and  are the length, area

and specic conductivity of the sample (respectively). The average
resistivity values obtained for the positive electrodes were around
150 ( cm) while that of the negative and pristine electrodes were
the same, around 5 ( cm). These results clearly show an increase of
two orders of magnitude in the resistivity of the positive electrodes
during prolonged cycling. This result indicates the asymmetric distribution of the potential in the cells as prolonged cycling proceeds.
In a follow up experiment to Fig. 6, new pristine ACF electrodes
were assembled in the CDI cell and the cell was polarized and held
at a constant potential of 0.9 V, after 3 chargedischarge cycles (in
order to bring the CDI cell to a dynamic steady state [38]). The cell
was sampled several times (a few dischargecharge cycles between
0.9 and 0 V each time) in intervals of hundreds of minutes, in order
to measure the effect of the application of constant polarization
on its cycling behavior. Fig. 8 shows the response of the cell in 3
selected sampling periods, in terms of NaCl concentration (ppm)
uctuation upon cycling. Fig. 8A and B shows two periods of times
before the inversion point. As can be seen from Fig. 8C the cell
reached stage 2 (Fig. 3B the inversion point) after 2.5 days, about
half the time it takes the cell to reach this stage upon continuous periodic cycling between 0 and 0.9 V. Hence, we can conclude
that an overall application of 0.9 V to the CDI cell comprising the
ACF electrodes used herein during 2.5 days (either constantly or
periodically) changes the electrodes in terms of their specic surface area and the location of their EPZC compared to the potential
they reach upon charge and discharge and the distribution of the

I. Cohen et al. / Electrochimica Acta 106 (2013) 91100

97

Fig. 7. XPS oxygen (A) (O 1s) and carbon (B) (C 1s) response of pristine and cycled (00.9 V) positive and negative electrodes (stage 3 in Fig. 6). The measurements were
identical in terms of the X-ray beam power and cross section of the electrode samples measured. The comparison is of course on the qualitative level, however it is clear
from these surface studies that the positive electrode exhibits the highest level of oxygen content and the lowest level of carbon content.

Fig. 8. Plot of salinity in ppm of NaCl vs. time in minutes; in which a continuous direct potential of 0.9 V was applied to the pristine electrodes in the CDI cell. (A and B) Shows
two periods of times before the inversion point. As can be seen from Fig. 8C the cell reached stage 2 (Fig. 3B, the inversion point) after 2.5 days, about half the time it takes
the cell to reach this stage upon continuous periodic cycling between 0 and 0.9 V.

potential between the positive and the negative electrodes upon

polarization.
3.3. The behavior of CDI cells with pre-oxidized positive electrodes
The oxidation of the carbon electrodes surface can be done
either chemically (using oxidation agents such as nitric acid
solutions [36,40]) or electrochemically by polarization to positive
potentials in the aqueous solutions. A new CDI cell (same as in
Section 3.1) was constructed where pristine ACF electrodes served
as the negative ones, while surface oxidized ACF electrodes (etched
with a concentrated HNO3 solution at room temperature for 25 h)
served as the positive electrodes. If our understanding regarding
the situation in these cells upon cycling, as described above (discussions related to Figs. 37) is correct, we expect that cells with
pre-oxidized positive electrodes should behave from the beginning
of cycling as regular cells in stage 2 or 3 (Fig. 3) after prolonged
cycling (which brings them to inversion, as demonstrated and
discussed above). We anticipate that this should occur because oxidation changes the resistivity and the surface area of the positive
electrodes and alters their EPZC . Fig. 9 shows the response of such a
cell in terms of current and salt concentration as a function of time,
during periodic polarization between 0.9 and 0 V. Indeed, the cell
exhibit from the beginning a stage 2 type behavior: the desorption
of co-ions upon polarization is of a similar magnitude as desorption

Fig. 9. Plot of salinity in ppm of NaCl vs. time (minutes) and a parallel plot of current
in amperes vs. time; in experiment with a CDI cell which positive electrodes were
pretreated by a concentrated nitric acid for 25 h and the negative electrodes were
not treated. The potentials applied to the cell were 00.9 V.

98

I. Cohen et al. / Electrochimica Acta 106 (2013) 91100

Fig. 10. Plots of salinity in ppm of NaCl vs. time (in minutes) during prolonged cycling of CDI cells: 18 days of experiment, in which the potentials applied to the cell were
00.7 V and the electrodes used were comprised from pristine, untreated ACF. The four plots relate to selected signicant time domains. The relevant potential application
(charge/discharge) vs. time is marked therein.

of counter-ions upon discharge. In general it can be seen that when

the cell is charged, desorption of ions dominates while upon discharging (shorting), the adsorption of ions is more signicant than
their desorption (what we dene herein as inversion behavior).
Also it can be seen that the current is almost 0.1 A at the end of
the charging step. Such as high current can be explained by a side
reaction (leading to Faradaic currents) which takes place when a
potential of 0.9 V is applied to a CDI cell which is not symmetrical
anymore (due to the changes in the positive electrode (corrosion)).
Due to changes in the positive electrode, the potential is not equally
distributed between the electrodes. Hence, the positive electrodes
may be polarized to potentials that accelerate their corrosion.
Since the conclusion from these experiments is that oxidation
of the positive electrodes brings to the inverted behavior described
herein, it is interesting to examine the effect of the applied potential
on the behavior of these cells.
3.4. Long-term stability of CDI cells cycled in the potential range
00.7 V
Fig. 10 presents the uctuation of NaCl concentration (ppm)
upon periodic polarization of the cell between 0.7 and 0 V. The
experiment discussed herein is quite similar to the prolonged
cycling experiment described in Fig. 6. Fig. 10 also presents (as
Fig. 6) enlargement of the response for 4 selected time domains
(AD) measured during the prolonged cycling experiment. The
rst 2 enlarged time domains (A and B) in Fig. 10, reect the
normal/initial behavior of CDI cells: adsorption of counter-ions
is dominant upon charging (depletion of NaCl concentration) and
desorption of counter-ions is dominant upon shorting (discharge),
reected by an increase in the NaCl concentration. The enlarged
time domain shown in Fig. 10C, exhibits the inversion point, after
14 days of cycling (compared to 5 days that it takes the CDI cells
to reach this stage when cycled between 0.9 and 0 V, Fig. 6). As
expected, after the cell reaches stage 2 (the inversion point) further
cycling exhibit the inverted behavior, as indeed well reected

in the response presented in the fourth (enlarged) time domain

in Fig. 10D. Hence, reducing the potential applied to the cells
by 200 mV, increases their stability by almost 300% (inversion is
reached after 14 days instead of 5 days of cycling).
3.5. The relation between the ACF electrodes potential in CDI cells
(vs. a reference electrode) and possible changes in the solution pH
It is expected that electrochemical oxidation of carbon electrodes in aqueous solutions should be connected to local (surface)
pH changes [43]. Monitoring pH changes at the ACF electrodes used
herein upon polarization, may serve as a good qualitative indication of their possible surface oxidation or reduction, as a function
of the potential applied to the electrode [44]. Eqs. (4)(7) below
provide examples of possible surface reactions of carbon surfaces
in aqueous solutions, in which pH changes are involved.
C| + H2 O C| O + 2H+ + 2e

(4)

C| + H2 O C| O + H+ + e

(5)

C| + 2H2 O CO2 + 4H+ + 4e

(6)

C| + e + H2 O C  H + OH

(7)

Each proton released to the solution should be accompanied

either by hydroxide generation in the negative electrodes or by
cation electro-adsorption (proton or sodium).
For this purpose, we have elaborated a specially designed owthrough CDI cell (Fig. 1b), which is a three-electrode cell containing
two sheets of ACF electrodes as the working electrode (WE) and
two Ag/AgCl electrodes as the counter and reference electrodes.
A pH meter was placed at the exit of the cell. The solution contained 1000 ppm NaCl. The use of Ag/AgCl as the counter electrode
enabled the measurement of pH change arising only from a single
ACF electrode [34]. The ow rate was adjusted to 50 ml/min and
the measurements were taken under Argon environment. Fig. 11

I. Cohen et al. / Electrochimica Acta 106 (2013) 91100

99

Fig. 11. Plot of the dependence of pH and charge on the potential applied. The electrodes used in this work, measured in the 3 electrodes cell (ow through mode, see Fig. 1B).

Fig. 12. Overall changes of pH vs. time (minutes) of the solution, during the prolonged cycling experiments in which the potentials applied to the cells were 00.9 V (blueplot,
that has the highest negative slope) and 00.7 V (red plot). (For interpretation of the references to color in this gure legend, the reader is referred to the web version of the
article.).

shows the pH develops and the charge consumed in the cell at certain applied potentials (constant polarization). Two potential scales
are provided: vs. the reference Ag/AgCl electrode and vs. EPZC of the
WE (i.e. the immersion potential of these ACF electrodes measured
vs. Ag/AgCl RE). There is a correlation between signicant changes
in pH and increase in the residual current, at potentials higher than
250 mV vs. EPZC . This can be logically explained in terms of oxidation
of the electrodes surface as the potential applied is higher, according to Eqs. (3)(5), leading to pronounced charge consumption and
decrease of pH. These ndings correlate well with the pronounced
difference in the CDI cells stability when cycled between 0.7 and
0 V or 0.9 and 0 V (Figs. 6 and 3, respectively). (Corresponding to
initial application of 350 mV or 450 mV vs. EPZC respectively, on the
positive electrode.)
An interesting phenomenon can be seen in Fig. 12. This gure
shows the pH changes of the solution during prolonged cycling of
the CDI cells between 0.70 V and 0.90 V (indicated). The pH probe
was placed at the exit of the cell and since the solution is circulated
in the system, cycling can change irreversibly the content of the
solution (e.g. due to continuous surface reactions of the electrodes
and consequent pH changes, Eqs. (4)(7)). As expected, the pH of the
solution uctuates in accordance with cycling. However, there is a
pronounced negative drift in pH and an increase in pH uctuations
during cycling in the 0.90 V range, while the pH drift upon cycling
in the 0.70 V range is very small and the uctuations in the pH
are small and steady during cycling. Although the plot shows that
when cycling in the 00.9 V range, we reach the pH of 4.5, its inuence on the conductivity is in two orders of magnitude smaller than
the salt concentration used. These results show that application of
too high potential to the positive electrodes (e.g. 450 vs. 350 mV initially), leads to their continuous surface reactions, which affect very
strongly the desalination behavior of the CDI cells during prolonged
cycling.

4. Conclusions
This work addresses highly important problem that can
be developed in CDI cells during periodic water desalination
processes. Cells comprising identical pairs of activated carbon electrodes, which are initially symmetric, loose their symmetry upon
cycling. The positive electrodes are gradually oxidized and their
pore systems are damaged. Consequently, the potential applied
to the cell is not divided equally. Most of the potential difference
falls more and more on the positive electrodes and their oxidation process accelerates, leading to water oxidation and decrease
in pH upon the charging steps. The change from the symmetric
to the asymmetric behavior of the electrodes is explained and
illustrated in Figs. 3 and 4. These changes in the electrodes and
the way how the potential is distributed in the cell, leads to an
inversion phenomenon in which desorption of co-ions becomes
a dominant phenomenon during charging (potential application)
and their re-adsorption is observed upon discharge (shorting). The
trivial way to avoid these negative phenomena is to apply to cells
low enough potential, so, the positive electrode should never reach
oxidation stages. However, low applied potential means low desalination capacity. The direction has to be development of oxidation
protected positive electrodes. A future work will explore how
the behavior of prolonged CDI processes described in this paper
depends on the ow mode.
References
[1] Y. Tanaka, Ion-exchange membrane electrodialysis program and its application
to multi-stage continuous saline water desalination, Desalination 301 (2012)
1025.
[2] T. Chakrabarty, A.M. Rajesh, A. Jasti, A.K. Thakur, A.K. Singh, S. Prakash, V. Kulshrestha, V.K. Shahi, Stable ion-exchange membranes for water desalination
by electrodialysis, Desalination 282 (2011) 28.

100

I. Cohen et al. / Electrochimica Acta 106 (2013) 91100

[3] L.J. Banasiak, T.W. Kruttschnitt, A.I. Schfer, Desalination using electrodialysis
as a function of voltage and salt concentration, Desalination 205 (2007) 3846.
[4] K. Dermentzis, Removal of nickel from electroplating rinse waters using electrostatic shielding electrodialysis/electrodeionization, Journal of Hazardous
Materials 173 (2010) 647652.
[5] K. Dermentzis, Continuous electrodeionization through electrostatic shielding,
Electrochimica Acta 53 (2008) 29532962.
[6] K. Dermentzis, K. Ouzounis, Continuous capacitive deionizationelectrodialysis
reversal through electrostatic shielding for desalination and deionization of
water, Electrochimica Acta 53 (2008) 71237130.
[7] A. Venkatesan, P.C. Wankat, Simulation of ion exchange water softening pretreatment for reverse osmosis desalination of brackish water, Desalination 271
(2011) 122131.
[8] A. Morel, K. Zuo, X. Xia, J. Wei, X. Luo, P. Liang, X. Huang, Microbial desalination cells packed with ion-exchange resin to enhance water desalination rate,
Bioresource Technology 118 (2012) 4348.
[9] C. Jacob, Seawater desalination: boron removal by ion exchange technology,
Desalination 205 (2007) 4752.

[10] B. Penate,
L. Garca-Rodrguez, Current trends and future prospects in the design
of seawater reverse osmosis desalination technology, Desalination 284 (2012)
18.
[11] R. Abejn, A. Garea, A. Irabien, Analysis, modelling and simulation of hydrogen
peroxide ultrapurication by multistage reverse osmosis, Chemical Engineering Research and Design 90 (2012) 442452.
[12] M. Shari, B.J. Zebarth, D.L. Burton, C.A. Grant, M.A. Hajabbasi, G. Abbassi-Kalo,
Sodium hydroxide direct distillation: a method for estimating total nitrogen in soil, Communications in Soil Science and Plant Analysis 40 (2009)
25052520.
[13] J.C. Farmer, D.V. Fix, G.V. Mack, J.F. Poco, J.K. Nielsen, R.W. Pekala, J.H. Richardson, The Seventh National Technology Transfer and Conference and Exposition,
Anaheim, CA, October, 1996, pp. 2931.
[14] G. Rasines, P. Lavela, C. Macas, M. Haro, C.O. Ania, J.L. Tirado, Electrochemical
response of carbon aerogel electrodes in saline water, Journal of Electroanalytical Chemistry 671 (2012) 9298.
[15] D. Wu, X. Chen, S. Lu, Y. Liang, F. Xu, R. Fu, Study on synergistic effect
of ordered mesoporous carbon and carbon aerogel during electrochemical
chargedischarge process, Microporous and Mesoporous Materials 131 (2010)
261264.
[16] Y.J. Lee, H.W. Park, S. Park, I.K. Song, Electrochemical properties of Mn-doped
activated carbon aerogel as electrode material for supercapacitor, Current
Applied Physics 12 (2012) 233237.
[17] P.M. Biesheuvel, Thermodynamic cycle analysis for capacitive deionization,
Journal of Colloid and Interface Science 332 (2009) 258264.
[18] P.L. Taberna, P. Simon, J.F. Fauvarque, Electrochemical characteristics and
impedance spectroscopy studies of carboncarbon supercapacitors, Journal of
the Electrochemical Society 150 (2003) A292.
[19] J.-B. Lee, K.-K. Park, H.-M. Eum, C.-W. Lee, Desalination of a thermal power plant
wastewater by membrane capacitive deionization, Desalination 196 (2006)
125134.
[20] Y.-J. Kim, J.-H. Choi, Improvement of desalination efciency in capacitive deionization using a carbon electrode coated with an ion-exchange polymer, Water
Research 44 (2010) 990996.
[21] P.M. Biesheuvel, R. Zhao, S. Porada, A. van der Wal, Theory of membrane capacitive deionization including the effect of the electrode pore space, Journal of
Colloid and Interface Science 360 (2011) 239248.
[22] P.M. Biesheuvel, A. van der Wal, Membrane capacitive deionization, Journal of
Membrane Science 346 (2010) 256262.
[23] S. Porada, M. Bryjak, A. van Der Wal, P.M. Biesheuvel, Effect of electrode thickness variation on operation of capacitive deionization, Electrochimica Acta 75
(2012) 148156.

[24] H. Li, L. Zou, Ion-exchange membrane capacitive deionization: a new strategy

for brackish water desalination, Desalination 275 (2011) 6266.
[25] M.-W. Ryoo, G. Seo, Improvement in capacitive deionization function of
activated carbon cloth by titania modication, Water Research 37 (2003)
15271534.
[26] J. Yang, L. Zou, H. Song, Z. Hao, Development of novel MnO2 /nanoporous carbon
composite electrodes in capacitive deionization technology, Desalination 276
(2011) 199206.
[27] Y. Oren, Capacitive deionization (CDI) for desalination and water treatment
past, present and future (a review), Desalination 228 (2008) 1029.
[28] S. Porada, L. Weinstein, R. Dash, A. van der Wal, M. Bryjak, Y. Gogotsi, P.M.
Biesheuvel, Water desalination using capacitive deionization with microporous
carbon electrodes, ACS Applied Materials and Interfaces 4 (2012) 11941199.
[29] Y. Bouhadana, E. Avraham, A. Soffer, D. Aurbach, Several basic and practical
aspects related to electrochemical deionization of water, AIChE Journal 56
(2010) 779789.
[30] J.-H. Lee, J.-H. Choi, The production of ultrapure water by membrane capacitive
deionization (MCDI) technology, Journal of Membrane Science 409/410 (2012)
251256.
[31] L. Yuan, X. Yang, P. Liang, L. Wang, Z.H. Huang, J. Wei, X. Huang, Capacitive
deionization coupled with microbial fuel cells to desalinate low-concentration
salt water, Bioresource Technology 110 (2012) 735738.
[32] T.J. Welgemoed, C.F. Schutte, Capacitive deionization technologyTM : an alternative desalination solution, Desalination 183 (2005) 327340.
[33] M.A. Anderson, A.L. Cudero, J. Palma, Capacitive deionization as an electrochemical means of saving energy and delivering clean water. Comparison to
present desalination practices: will it compete? Electrochimica Acta 55 (2010)
38453856.
[34] E. Avraham, Y. Bouhadana, A. Soffer, D. Aurbach, Limitation of charge efciency
in capacitive deionization, Journal of the Electrochemical Society 156 (2009)
P95.
[35] E. Avraham, M. Noked, Y. Bouhadana, A. Soffer, D. Aurbach, Limitations of charge
efciency in capacitive deionization, Journal of the Electrochemical Society 156
(2009) P157.
[36] E. Avraham, M. Noked, Y. Bouhadana, A. Soffer, D. Aurbach, Limitations of charge
efciency in capacitive deionization processes. III. The behavior of surface oxidized activated carbon electrodes, Electrochimica Acta 56 (2010) 441447.
[37] E. Avraham, M. Noked, I. Cohen, A. Soffer, D. Aurbach, The dependence of the
desalination performance in capacitive deionization processes on the electrodes PZC, Journal of the Electrochemical Society 158 (2011) P168.
[38] E. Avraham, M. Noked, A. Soffer, D. Aurbach, The feasibility of boron
removal from water by capacitive deionization, Electrochimica Acta 56 (2011)
63126317.
[39] Y. Bouhadana, E. Avraham, M. Noked, M. Ben-Tzion, A. Soffer, D. Aurbach, Capacitive deionization of NaCl solutions at non-steady-state conditions: inversion
functionality of the carbon electrodes, Journal of Physical Chemistry C 115
(2011) 1656716573.
[40] I. Cohen, E. Avraham, M. Noked, A. Soffer, D. Aurbach, Enhanced charge efciency in capacitive deionization achieved by surface-treated electrodes and
by means of a third electrode, Journal of Physical Chemistry C 115 (2011)
1985619863.
[41] A.M. Johnson, J. Newman, Desalting by means of porous carbon electrodes,
Journal of the Electrochemical Society 118 (1971) 510517.
[42] H. Tobias, A. Soffer, The immersion potential of high surface electrodes, Journal
of Electroanalytical Chemistry 148 (1983) 221232.
[43] K. Kinoshita, in: K. Kinoshita (Ed.), Carbon: Electrochemical and Physicochemical Properties, John Wiley & Sons, New York, 1988, p. 257.
[44] B. Kastening, M. Hahn, J. Kremeskotter, The double layer of activated carbon
electrodes part 2. Charge carriers in the solid material, Journal of Electroanalytical Chemistry 374 (1994) 159166.