Electrochimica Acta 55 (2010) 3845–3856

Contents lists available at ScienceDirect

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

Critical review article

Capacitive deionization as an electrochemical means of saving energy and

delivering clean water. Comparison to present desalination practices:
Will it compete?
Marc A. Anderson a,b,∗,1 , Ana L. Cudero b,1 , Jesus Palma b,1
a
Environmental Chemistry and Technology Program, University of Wisconsin – Madison, Madison, WI 53706, USA
b
Electrochemical Processes Unit, Madrid Institute for Advanced Studies in Energy (IMDEA Energy), E28933 Mostoles, Madrid, Spain

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

Article history: Potable water as well as water for agriculture and industry is critical to human habitation on this planet.
Received 9 November 2009 We have been squandering and polluting this precious resource and are now in need of finding cost com-
Received in revised form 3 February 2010 petitive newer technologies for reclaiming this valuable life-sustaining liquid. Capacitive deionization
Accepted 4 February 2010
(CDI) is an electrochemical water treatment process that holds the promise of not only being a commer-
Available online 12 February 2010
cially viable alternative for treating water but for saving energy as well. CDI works by sequestering ions,
or other charged species, in the electrical double layer of ultracapacitors. While removing these ions,
Keywords:
one actually stores capacitive energy. If one recovers this energy efficiently, this process likely consumes
Electrodialysis
Capacitive deionization
less power than any competing technology. This paper reviews current methods for treating water in
Double layer capacitance comparison to the state of art of the CDI process.
Zeta potential © 2010 Elsevier Ltd. All rights reserved.
Asymmetric electrodes

Contents

1. Introduction—water and energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3846

2. Water from the sea—current methods of desalination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3846
2.1. Energy required for desalination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3847
2.2. Capital, operational and maintenance costs for desalination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3847
3. Electrodialysis for desalinating brackish waters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3847
3.1. Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3847
3.2. The electrodialysis process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3847
3.2.1. Electrode reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3848
3.2.2. Energy consumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3848
3.3. Limitations of electrodialysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3849
4. Capacitive deionization systems—a competitive energy efficient water treatment technology? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3849
4.1. Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3849
4.1.1. Double layer models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3849
4.1.2. Zeta potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3849
4.2. The CDI process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3850
4.2.1. CDI history . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3850
4.2.2. CDI energy considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3850
4.3. Ultracapacitors in CDI systems for water treatment and energy storage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3851
4.3.1. Ultracapacitors for energy vs. CDI systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3851
4.3.2. Importance of electrolyte concentration and composition in CDI systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3851

∗ Corresponding author at: Environmental Chemistry and Technology Program, University of Wisconsin – Madison, 660 N. Park Street, Madison, WI 53706, USA.
Tel.: +1 608 262 2674; fax: +1 608 262 0454.
E-mail address: nanopor@wisc.edu (M.A. Anderson).
1
ISE member.

0013-4686/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.electacta.2010.02.012
3846 M.A. Anderson et al. / Electrochimica Acta 55 (2010) 3845–3856

4.3.3. Carbon based CDI systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3852

4.3.4. Other CD materials and our asymmetric system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3852
4.4. Devices and operational aspects of CDI systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3853
4.4.1. Two units in parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3853
4.4.2. Recycle loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3854
5. Conclusions and future trends for CDI systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3855
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3855

1. Introduction—water and energy It stands to reason that if most of the 98% of our waters are either
sea or brackish waters, we must find newer, more efficient, and cost
Hall and Day [1] have written a recent and compelling article effective means of removing salts from these waters. Indeed, major
related to the plethora of environmental problems facing present cities such as Sydney, Perth, Singapore, Los Angeles, Johannesburg,
and future generations of human inhabitants on the face of the Jubail, Ras Laffan (Qatar) and Miami are either building or design-
Earth. Hall and Day suggest that instead of thinking about “peak ing huge new desalination plants. We, the authors of this article
oil” we have now arrived at “peak everything”. However, they believe, as some select others, that CDI may ultimately provide a
admit that scientists and engineers may be able to avoid imme- competitive means of delivering potable waters and energy savings
diate calamity through technological solutions to these shortages at the same time. However, to be competitive, CDI must compare
(this review article on capacitive deionization (CDI) is related to favorably with more established methods both in terms of capital
one of these technological solutions). Thus, our near term priorities and operational costs.
should be focused on solving global warming, providing a stable and
sufficient energy supply, and delivering clean and potable water
2. Water from the sea—current methods of desalination
to the World’s people. These later three priorities are directly con-
nected. Global warming is tied to our prodigious demand for energy
Typical concentration of dissolved salts in seawater and brack-
and the consequent burning of fossil fuels. Furthermore, energy
ish water are 35,000 and 1000 mg/L, respectively. The most widely
and water are also related. Energy is needed to deliver water and
used processes for desalination include membrane separation sys-
water is needed to generate energy [2,3]. Indeed, one does not site
tems: reverse osmosis (RO), and electrodialysis (ED); and thermal
a nuclear power plant in the middle of the desert but rather near
separations including: multistage flash distillation (MSF), multi-
a major water body such as a river, lake or the sea. By the same
effect distillation (MED) and mechanical vapor compression (MVC).
token, we are likely not able to produce energy such as bio-fuels
Among these processes, RO and MSF methods are employed in the
without a sufficient supply of water. On the other hand, we need
bulk of the plants (90%) to desalinate seawater worldwide [6].
large quantities of energy to desalinate seawater for potable use. In
A visual summary of all these processes and their impact on
this review, we first examine the relationships between energy and
the desalination market is presented in Figs. 3 and 4 adopted from
water particularly with respect to the major methods of desalina-
the review of Chaudhry [7]. Among membrane based plants, 86%
tion focusing finally on capacitive deionization as a potential means
belong to Revere Osmosis plants while electrodialysis represents
of solving the energy–water problems simultaneously.
only 14%. As also illustrated, while there are more plants using
At the turn of this 21st century, Shawn Tully wrote an arti-
RO membranes than thermal methods, the total amount of water
cle in Fortune magazine indicating that water would be the “oil
processed by both is almost equal. Electrodialysis has the particu-
of this century” [4]. Indeed, major companies marketing drinking
larity that, while being a membrane process, the driving force is a
water like France’s Suez and Vivendi have been betting their future
potential applied between two electrodes; the same driving force
revenues on the scarcity and higher price of water. Fig. 1 shows dis-
as the CDI process. Thus, we will further analyze the electrodialysis
tributed rainfalls worldwide as well as the per capita availability of
technology below and compare it with CDI [7].
water on each continent [5]. It can readily be noticed that the con-
Reverse osmosis is the fastest growing method of desalination
tinents having the majority of people, Asia and Africa, also have the
but it would not have been possible without the seminal discovery
least amount of water. Furthermore, as illustrated in Fig. 2, much of
by, Loeb and Sourirajan [8] that transformed membrane separa-
the water in developing areas is being used for agriculture leaving
tion from a laboratory to an industrial process. The flux of the first
very little for human consumption [5].
Loeb–Sourirajan reverse osmosis membrane was 10 times higher
than that of any membrane then available and made reverse osmo-
sis a potentially practical method of desalting water. In this review
article, we hope to show that new materials and electrode con-
figurations may also allow for a similar break-through in the CDI
process.

Fig. 1. Worldwide water resources per capita and level of rainfall as a function of
continent. Fig. 2. The type of water use by sector for developing and high-income countries.
 M.A. Anderson et al. / Electrochimica Acta 55 (2010) 3845–3856 3847

Table 1
Types of cost associated with operation and maintenance processes for the major
types of desalination plants worldwide.

Type of cost MSF (KUS$) MED (KUS$) RO (KUS$)

Chemicals for cleaning 1500 1000 2500

Operational chemicals 50 100 250
People for operations 400 600 500
People for maintenance 100 175 100
Membranes NA NA 2000
Other 300 200 250
Fig. 3. Percentage of membrane and thermal desalination plants worldwide.
Total 2350 2075 5600

2.1. Energy required for desalination

[6] compared the costs associated with operation and mainte-
nance, as well as other costs such as energy, capital cost, etc.
RO competes favorably with MSF in regard to energy required:
for a 170 × 103 m3 /day (45 MGD) [170 × 106 L/day] plant deliv-
2.9–3.7 kWh/m3 of water treated for RO vs. 4 kWh/m3 for MSF pro-
ering potable water located in the Middle East (Gulf Water).
cesses. The latter figure of 4 kWh/m3 of seawater treated is true
Tables 1 and 2 are adopted from their paper. A couple of points
only if MSF is cited with a working power plant to reduce the heat
can be noted. Energy costs are slightly higher for MSF and MED
required to drive this thermal separation process. Furthermore,
compared to RO plants. Major operational cost for RO is related to
these values do not consider the recovery of the energy in the pres-
membrane replacement. A recent review analyzing all the informa-
surized concentrated reject by means such as a Pelton turbine or
tion in the literature regarding water desalination cost as a function
an isobaric chamber. In this case, it is suggested that energy con-
of the type and size of the plant can be found in Ref. [11].
sumption is reduced to 1 kWh/m3 [7]. However, as Semiat exposes
in his recent review [2], a wide variety of energy consumption val-
3. Electrodialysis for desalinating brackish waters
ues are given for desalting processes and some confusion arises
when comparisons are tried.
3.1. Background
Regardless of the process (RO, MFS, ED or CDI) being used to
remove salts from sea or brackish waters, thermodynamically we
Electrodialysis is the closest cousin of capacitive deionization
can calculate the energy required. The minimum energy (work)
systems and has been successfully used for the desalination of
needed to separate ions from a solution is around 1.1 kWh/m3 for
brackish waters. Due to its similarity to CDI processes, it is instruc-
seawater (35,000 ppm) and 0.12 kWh/m3 and for brackish water
tive to examine this technology from two perspectives. Firstly, it
(4000 ppm) in both cases depending on the recovery [2,9,10].
is a commercially accepted technology for water treatment. Sec-
However, these minimum theoretical values depend upon the
ondly, there are commonalities between CDI and electrodialysis
input–output flow concentration ratio [9]. For example, if such
processes particularly with respect to ion transport in solution as
ratio is 0.5, energy consumption increases to 1.6 and 0.17 kWh/m3 ,
well as through membranes that are worth noting and compar-
respectively, and if the ratio increases to 0.7, consumption will be
ing. A critical review of electrodialysis separation technologies was
2.0 and 0.21, respectively. Biesheuvel has shown [9] that reversible
written by Xu and Huang in 2008 [12]. In addition, Davis has writ-
work needed per unit volume to produce a dilute stream scales with
ten an excellent chapter concerning the electrodialysis process in
the volume ratio (relationship between inlet and outlet volumes
Handbook of Industrial Membrane Technology [13].
and thus to salt concentration differences), and as a consequence,
the minimum work required increases at high water recovery
3.2. The electrodialysis process
values. In fact, if the concentrated stream becomes infinitely con-
centrated the required work will be infinite [9].
Electrodialysis involves moving ions in a potential field across
polymeric anion and cation-exchange membranes. Shown in Fig. 5,
2.2. Capital, operational and maintenance costs for desalination
is a pictorial illustration of the process. Cation- and anion-exchange
membranes are placed alternatively between the cathode and
As stated above, cost of desalination is tied greatly to both,
the anode. When a potential difference is applied between both
the desalination method and the size of the plant with smaller
electrodes, the cations are drawn towards the cathode (negative
plants (<1 million gallons per day (MGD), 3790 m3 /day) costing
electrode) and anions towards the anode (positive electrode). The
over 1.3 dollars/m3 (5 $/kgal) [0.13 cents/L] of seawater treated as
cations migrate through the cation-exchange membranes, but are
opposed to larger plants (>10 MGD) being able to deliver potable
retained by the anion-exchange membranes. The opposite occurs
water for as low as 0.4 dollars/m3 (1.5 $/kgal) [0.04 cents/L]. In the
with the anions that migrate through the anion-exchange mem-
case of brackish water, the costs vary between 0.10–1 dollars/m3
branes but not through the cation-exchange membranes. This
(0.4–4 $/kgal) [0.01–0.1 cents/L] [7,11]. Borsani and Rebagliati
movement produces a rise in the concentration of ions in some
compartments (brine streams) and the decrease in the adjacent
ones (dilute streams), from which purified water exits. As a result

Table 2
Total water cost for the major types of desalination plants worldwide.

Type of cost MSF MED RO

Thermal energy (M$) 105 105 0

Electric power (M$) 92 76 114
Operation and maintenance (M$) 25 22 60
Plant investment (M$) 180 195 170
Total cost (M$) 402 398 344
Fig. 4. Percentage of membrane and thermal desalination plants worldwide and
Water cost ($/m3 ) 0.52 0.52 0.45
water processed by these plants. MGD: million gallons per day.
3848 M.A. Anderson et al. / Electrochimica Acta 55 (2010) 3845–3856

Fig. 6. Electrical work to produce 1 m3 of a solution containing 0.3 g/L of NaCl as

a function of the ion concentration at the inlet using electrodialysis. Operating cell
voltage = 1.2 V, charge efficiency = 100%.

Fig. 5. Schematic diagram of the electrodialysis process showing the migration of

charged ions towards two charged electrodes. where z = charge of the ion; F = Faraday constant
(96,485 Amp s/mol); Qf = diluent flow-rate, L/s; C = change of
of the anion and cation migration, electric current flows between ion concentration from inlet to outlet, mol/L.
the cathode and anode with equal charge equivalents transferred Combining Eqs. (1) and (2) the net energy consumption is
so that charge balance is maintained in each stream. zFQf Ct
WED = V (3)
d
3.2.1. Electrode reactions
Using Eq. (3) and taking d = 1, we have calculated the minimum
In electrodialysis, anode and cathode reactions may occur at
theoretical work to produce 1 m3 (0.27 kgal) of a solution contain-
each electrode depending upon the pH of the water and the poten-
ing 300 mg/L of NaCl from solutions of different concentrations.
tial applied.
We have selected an operating cell voltage of 1.2 V. Results, plotted
At the cathode:
in Fig. 6, show that from a thermodynamic point of view, and even
2e− + 2 H2 O → H2 (g) + 2 OH− Reaction 1 assuming efficiency of 100%, electrodialysis is not competitive with
Reverse Osmosis when applied to solutions with ion concentrations
At the anode: beyond 2000–3000 mg/L.
Unfortunately, this is an ideal situation. Real current efficiency
H2 O → 2 H + + ½ O2 (g) + 2e− Reaction 2
is always lower than 100%. However, it is important to note that
current efficiencies higher than 80% are desirable in order to mini-
or 2 Cl− → Cl2 (g) + 2e− Reaction 3 mize energy costs. Low current efficiencies indicate: water splitting
in the diluent or concentrate streams, shunt currents between the
In this process, hydrogen gas may be generated at the cathode electrodes, or back-diffusion of ions from the concentrate to the
and either oxygen or chlorine gas (depending upon the concentra- diluent. Again, combining Eqs. (1) and (2), the following expression
tion of the electrode stream and the end ion exchange membrane can be derived for current efficiency [14]:
arrangement) at the anode. The amount of gas evolved depends as
well on the potential applied. These gases are subsequently dissi- Im zFQf C
d = = (4)
pated as effluents from each electrode compartment and may be I I
either combined to maintain a neutral pH or discharged. Alterna- where I is the actual applied current, in Amperes.
tively, hydrogen gas may be used for other applications. It can be seen in Eq. (4) that the efficiency of the process is a
direct function of the inlet feed concentration of ions. This can easily
3.2.2. Energy consumption be understood as there are more charged species in more concen-
In electrodialysis, the net energy consumption is the work trated systems to carry current. However, from a practical point of
applied to remove the dissolved ions from the solution: view, one should also take into account that, due to concentration
polarization at the membrane surfaces, a limiting current is reached
Edeionization Im t
WED = Wdeionization = =V (1) [13]. In the brine compartments, the concentration is higher at
d d
the membrane surfaces than in the compartment bulk and in the
where W is work, E is energy, in Joules, V is cell voltage, in V, Im is the diluted compartments the situation is reversed. The resistance to
minimum theoretical current necessary to remove a given amount the flow of electric current increases as this interfacial concentra-
of ions, in Amperes, t is time, in s, and d is the deionization or cur- tion decreases. Net charge transport is related to the charge and the
rent efficiency, a measure of ion transport across the ion exchange equivalent ionic conductance of charged species, anions and cations
membranes for a given applied current. not necessarily being equivalent. Furthermore, charge transport in
Additionally, the minimum theoretical current, Im , is the product solution is different than charge transport through the exchange
of the flow-rate times the change in concentration from the inlet membranes (the later numbers are not generally available). For a
to the outlet: complete description of this problem one should see the description
offered by Davis [13]. Ortiz et al. have developed a mathemati-
Im = zFQf C (2) cal model of a conventional electrodialysis process applied to the
 M.A. Anderson et al. / Electrochimica Acta 55 (2010) 3845–3856 3849

desalination of brackish water [15]. The model allows one to predict 4.1.1. Double layer models
the system’s behavior under various operating conditions and to Due to factors such as surface polarization, adsorption of ions
calculate the electrical energy consumption of commercial equip- and orientation of polar molecules, the region between two dif-
ment. ferent phases has a complex distribution of charge. This interface
is known as double layer. Historically, there have been three
3.3. Limitations of electrodialysis models describing the double layer. The first model was pro-
posed by Helmholtz in 1883 [22], and describes the distribution
Electrodialysis (ED) works best for removing low molecular of charges at the double layer as in the case of a capacitor: sur-
weight charged species. The energy consumption of the ED sys- face accumulates charge of one sign while at the solution side the
tem is proportional to the water salinity; thus, ED is more feasible opposite sign charges are accumulated. A second model devel-
when the salinity of the water is not higher than 6000 mg/L of oped by Guoy–Chapman in 1913 already took into account the
dissolved solids. Also, the process is not suitable for water with gradient of electron density at a charged interface, the so-called
a value of dissolved solids below 400 mg/L due to the low con- Thomas–Fermi screening distance, and its variation with distance
ductivity. Furthermore, comparatively larger membrane areas are from the surface. The consequence of this screening is that there
required to satisfy capacity requirements for low concentration exists a distribution of electric charge in the double layer region
(and sparingly conductive) feed solutions. As with RO, electrodialy- depending on the potential at the surface. In this model, other fac-
sis systems require feed pre-treatment to remove species that coat, tors such as the Boltzmann distribution due to thermal effects (ions
precipitate onto, or otherwise “foul” the surface of the ion exchange are not static) were also included. However, this model described
membranes. However, electrodialysis reversal can minimize scal- ions as point charges. As a result, it predicted unrealistic high capac-
ing by periodically reversing the polarity of the electrodes and/or itance values due to extremely short distances. Lastly in 1924,
the flows of the diluent and concentrate streams. Stern completed the model by assuming that the double layer can
be divided in an “inner” region where ion distribution followed
4. Capacitive deionization systems—a competitive energy Langmuir’s adsorption isotherm, while the region further from the
efficient water treatment technology? surface could be roughly described with the Gouy–Chapman model.
Thus, the total capacitance can be calculated like a series union
4.1. Background of both, inner double layer (or Helmholtz’s) and diffuse layer. For
more detailed information the reader is addressed to Chapter 6 in
As mentioned before, capacitive deionization (CDI), or as some- Ref. [23].
times referred to as electro-sorption, operates using an applied
potential, like its closest cousin, electrodialysis, to drive charged
species (ions) to the electrodes. However, CDI does not involve 4.1.2. Zeta potential
membranes. It is therefore a low pressure process of deionization Molecules, particles or ions in a solution form hydration shells
that has the possibility of directly competing with reverse osmosis that can also be described by means of double layer principles. Ions
or distillation as a means of delivering waters free of ions at reduced with a charge contrary to that of the surface will be highly attached
cost and operating expense [16–21]. CDI is an electrochemical pro- to that surface, forming the inner or Stern layer, while those fur-
cess that operates by adsorbing ions in the double layer formed at ther away will form a diffuse layer. As a consequence, a potential
the electrodes by the application of a potential difference. The prin- difference between the surface of the particle, molecule or ion and
ciples of the process can be traced to the work of Helmholtz and to that in the solution bulk is established. That parameter is known as
the modeling of the electrical double layer by Guoy–Chapman, as electrokinetic (zeta) potential [24–27].
explained further below. Zeta potential is an important and useful indicator of interfacial
In a classic parallel plate capacitor, charge separation is electro- electrochemical character. For instance, it can be used to predict
static. Capacitance scales directly with the area of the plates and the stability of colloidal suspensions or emulsions, and its variation
the inverse distance of separation as shown in Eq. (5): with parameters such as: conductivity, concentration, etc. How-
ever, the most important parameter affecting zeta potential for
A materials subsequently described in this paper is pH. The greater
C = εr ε0 (5)
D the zeta potential, the more likely the suspension is to be stable
where C, is the capacitance in farads F; A is the area of each plate in a because two charged particles will repel each other and thus over-
traditional capacitor (usually metal) in square meters; εr is the rel- come the natural tendency to aggregate. A typical value of ±30 mV
ative static permittivity (sometimes called the dielectric constant) is considered to be the transition between stable and unstable
of the material between the plates, (vacuum =1 F/m), ε0 is the per- suspensions, e.g. a suspension of particles with a zeta potential
mittivity of free space (8.854 × 10−12 F/m) and D is the separation within those values would probably be unstable. The pH at which
between the plates, in meters. the zeta potential value is zero is called isoelectric pH, and would
Capacitors can be connected either in parallel or in series, being correspond to the least stable pH value for that suspension. The
the values of equivalent capacitance obtained by the following measurement of zeta potential is often the key to understanding
expressions: dispersion and aggregation processes in applications as diverse as
water purification, emulsions, paints, cosmetics, etc. We will show
Parallel : Ceq = ˙Cn (6) it to also be important to the CDI process.
The zeta potential is measured indirectly. When an electric field
1 1
Series : =˙ (7) is applied, molecules, particles or ions will move through the solu-
Ceq Cn
tion together with all those ions (and polar molecules such as
The energy stored can be calculated by water) contained within its diffuse layer. The velocity of the parti-
cles moving towards the electrode when an electric field is applied
1 2
Estored = CV (8) is known as electrophoretic mobility, and constitutes the basis of a
2 well known separation technique: electrophoresis. That velocity is
where E = energy in Joules; C = capacitance in Farads, V = potential influenced by factors such as the strength of the field, the viscosity
difference in V. and dielectric constant of the medium and the zeta potential, and
3850 M.A. Anderson et al. / Electrochimica Acta 55 (2010) 3845–3856

electrochemical parameter pumping [33,34]. During the late 1970s

until the mid-1980s, Oren was the single largest advocate of the
CDI process publishing papers on new carbon materials as well
as fundamental electrochemical double layer aspects of electrodes
[35–39]. However, a renewed interest in the CDI concept came in
the mid-1990s when Farmer et al. [17–19] at Lawrence Livermore
National Labs (LLNL), working on high surface area conducting car-
bon aerogels, developed their own version of a CDI device. The
crucial aspect of these new materials was their large increase in
surface area. As already stated, capacitance scales with surface area
(Eq. (5)); thus high surface area carbon materials used as electrodes
improve capacitance and have better performance with respect
Fig. 7. Schematic diagram of capacitive deionization showing the removal of to electro-sorption. These high surface area materials have been
charged ions or species by two charged electrodes. essential in the development of both CDI devices and also electri-
cal double layer (EDL) capacitors (sometimes called super or ultra
is defined by the Henry Eq. (9): capacitors). Both processes (EDL capacitors and CDI systems) oper-
 2εf (k )  ate in an identical fashion by adsorbing ions at a charged interface.
a
Ue = (9) Therefore, these aerogels and other similar high surface area con-
3 ducting carbon materials can be used in both devices.
Being Ue = electrophoretic mobility; ε = dielectric constant;
 = viscosity, and  = zeta potential and f(ka ) = Henry’s function.
Thus, by measuring the velocity of the particles, by means of 4.2.2. CDI energy considerations
techniques such as laser Doppler electrophoresis, the value of zeta In the case of capacitive deionization, provided that the energy
potential can be obtained. stored in the capacitor can be easily recovered in the regenera-
tion cycle, the net energy consumption is the difference between
4.2. The CDI process the energy supplied during deionization (charging of the capaci-
tor) and the energy recovered during regeneration (discharging of
CDI makes use of the above-mentioned basic capacitor princi- capacitor):
ples to remove dissolved ions from an electrolyte (water) stream.
In this case, the plates happen to be the electrochemical surfaces WCDI = Wdeionization − Wregeneration = Wcharge − Wdischarge (11)
where ions adsorb following the principles of the double layer
[9,28].
Note that this equation is similar to that for electrodialysis (Eq. (1))
A representation of how the electro-sorption process works in
in which an additional subtractive term is included due to the fact
the CDI system is shown in Fig. 7. Essentially, a solution of ions flows
that CDI systems actually store energy during the ion removal pro-
through a pair of electrodes and anions (or other negatively charged
cess; that energy can be recovered during the regeneration cycle.
species) are retained at the anode (positive electrode) while the
In principle, this would make the entire process more energet-
cations (or positively charged species) are separated from solution
ically favorable. However, some aspects such as efficiencies and
at the cathode. Ideally, no redox process occurs, and as a conse-
irreversibility during the charging/discharging cycle must also be
quence the process is reversible and the electrochemical response
considered [9,28,40]. Unfortunately, until now, little effort has been
is purely capacitive without any faradaic contribution.
expended upon developing electronic and electrochemical meth-
Electrochemically, the capacitance of a CDI system can be eval-
ods for performing regeneration effectively [29]. Since this task is an
uated by means of cyclic voltammetry. The potential window is
important and essential part of the CDI process, this energy savings
constrained by the water redox potential values, typical operation
strategy should be a major component of any working system.
values being 1.2–1.5 V. In absence of faradaic contributions, capaci-
On the other hand, some limitations of regeneration with
tance can be obtained directly from the voltammogram and sweep
respect to energy efficiency should be noted. Efficiency is defined
rate:
  [9] as the ratio between the amount of salt molecules removed
dq dt I from the solution and the amount of electronic charge transferred
C= =I = (10)
dV dV  between the electrodes, in the charge–discharge cycle, for adsorbed
ions into a highly porous surface. Biesheuvel [9] have described a
where q = charge in Coulombs, V = potential difference in V, t = time
thermodynamic model for the CDI process based in the GCS model
in s, I = intensity (in Amperes), and  = sweep rate in V s−1 .
including ion size constraints according to Carnahan-Starling equa-
The behavior of new electrode materials towards electro-
tion of state. They obtain an analytical solution for charge efficiency
adsorption could be directly obtained from resulting voltammo-
concluding that efficiency only approaches unity by increasing cell
grams. The more featureless and rectangular the curve the better,
voltage, Stern capacity or decreasing the ionic strength of the solu-
meaning that only capacitive processes take place without any con-
tion being treated. In this respect, attempts are being made to
tribution of faradaic (redox) reactions.
further evaluate the effect of these variables [40].
To understand the importance of the charge efficiency in CDI
4.2.1. CDI history
systems, we propose a macroscopic analysis based on the idea that
A recent review of the CDI process directed specifically to
the energy for deionization is equivalent to the energy stored in a
the desalination community has been presented by Oren [29].
capacitor, and the energy released during regeneration corresponds
In this article, all of the fundamental aspects concerning the
to the discharge of such capacitor. This energy is related to the work
electro-adsorption process were exposed and therefore will not be
of charging and discharging the capacitor through the expressions:
examined in detail here. Historically, CDI dates back to the pio-
neering electrochemical demineralization work of the Caudle and
Wdischarge
Johnson groups in the late 1960s and early 1970s [30–32]. Oren Edeionization = Estored = Wcharge charge = (12)
also worked on his own version of CDI, which he referred to as discharge
 M.A. Anderson et al. / Electrochimica Acta 55 (2010) 3845–3856 3851

Fig. 8. Electrical work to produce 1 m3 of a solution containing 0.3 g/L of NaCl from
solutions of different concentrations using capacitive deionization (CDI). The oper-
ating cell voltage = 1.2 V, round trip efficiencies 70–95%.

As a consequence, the net energy consumption is related to the

charge and discharge efficiencies by Eq. (13):
 
1
WCDI = Edeionization − discharge (13)
charge
Fig. 9. Number of publications in supercapacitors and capacitive deionization along
the last years.
Similarly to capacitors, the round trip efficiency can be calcu-
lated as the ratio between the work retrieved during the discharge
of the capacitor to the work applied for charging: (2.03 kWh/m3 ) for desalination of brackish waters (AWWA M46,
1999 cited in [21]). In our theoretical analysis above, Farmer’s
Wdischarge
roundtrip = charge discharge = (14) results correspond to 90% roundtrip efficiency, while Welgemoed
Wcharge results are consistent with an efficiency of just 50%. Nevertheless,
Supercapacitors can show round trip efficiencies over 95% [41]. If Welgemoed and Schutte [21] predicted that by including energy
a CDI device were to reach similar values, the practical application recovery during regeneration and by optimizing internal electrical
of CDI could be extended to a range of concentrations much wider connections, future industrial units could approach the laboratory
than electrodialysis. We have made theoretical calculations for the scale energy consumptions.
work required to produce 1 m3 of a solution containing 300 mg/L of
NaCl from solutions of different concentrations (the same analysis 4.3. Ultracapacitors in CDI systems for water treatment and
as shown in Fig. 6), with an operating cell voltage of 1.2 V and round energy storage
trip efficiencies ranging from 70 to 95%. Results are plotted in Fig. 8,
where they are compared to the minimum thermodynamic work to 4.3.1. Ultracapacitors for energy vs. CDI systems
remove such salts producing 1 m3 of diluted solution and 0.25 m3 In these days of water and energy shortage, it is interesting
of concentrated solution, calculated from the model described by to remark that the same principle of charge sequestering at an
Biesheuvel [9]. This model serves as a reference for calculating the interface, may be used to help solve both problems: delivering
maximum theoretical roundtrip efficiency, which in this case is clean water and storing energy. Both fields are rather young.
around 96%. As seen in Fig. 9, interest in supercapacitor systems started to
Fig. 8 indicates that, under the selected conditions and at increase around the late 1980s but it was ten years later that
concentrations below 5000 mg/L, CDI could be a competitive tech- more people started examining capacitive deionization systems.
nology even if moderate efficiencies, from 60 to 70%, are attained. Unfortunately, it would seem that energy shortage is receiving
Additionally, if efficiencies over 85% can be reached, CDI could the bulk of attention, as the number of papers being published
become a serious competitor with RO, not only for brackish water, per year in supercapacitor research is around 200 but only around
but for seawater desalination as well. Indeed Kötz and Carlen have 5–6 in the area of capacitive deionization. This situation may be
reported round trip efficiencies of 92% for these type of superca- changing since this past spring (2009) seven new papers have been
pacitor electrodes [42], while Miller and Burke reported round trip devoted to CDI or electro-sorption processes [9,43–48]. Many of
efficiencies even higher than 95% [41]. these papers are related to the use of new materials such as carbon
In the original experiments of Farmer et al. [17,18], an energy nanotubes in the CDI process.
consumption value of 0.1 kWh/m3 was obtained for brackish
waters. This number serves as a figure of merit or preliminary 4.3.2. Importance of electrolyte concentration and composition in
benchmark for this technology. However, Welgemoed and Schutte CDI systems
[21], using a larger pilot plant unit based on similar carbon aero- The role of the electrolytes with respect to ion size, level of
gel CD materials, obtained a value of 0.6 kWh/m3 to produce water hydration and mobility in connection with pore size and pore con-
with 500 mg/L of total dissolved salts from a 2000 mg/L synthetic nectivity is of high importance and many studies have been devoted
solution. This value is considerably higher than the laboratory to these issues [49–54]. Gabelich et al. [55] noted that, in spite of
scale value obtained by Farmer et al. [17,18] although smaller high surface area due to the presence of small pores (4–9 nm), only
than that obtained in the case of electrodialysis reverse systems about 40 m2 /g was accessible to the ions, depending upon hydrated
3852 M.A. Anderson et al. / Electrochimica Acta 55 (2010) 3845–3856

ion radius. Li et al. [44] and Zou et al. [56] have more recently taken
a fresh look at the subject of ion size using mesoporous carbon elec-
trodes that are fabricated much in the same fashion as the famous
ZSM-5 heterogeneous catalyst first patented by Mobil Oil in 1978
[57,58]. These materials have very small (<5 nm) regular pores. The
authors suggest that not only pore size but pore regularity greatly
influences electro-adsorption in these materials.
In brackish and seawaters, there are single valent cations
present such as sodium, double valent ions such as calcium, and
even triple valent ions such as iron. Anions are as well single and
multivalent species having different levels of hydration. Gabelich et
al. [55] remarked that CDI systems employing aerogel carbon elec-
trodes present problems regarding multivalent ions removal due
to size limitations. In the case of our own CDI system, we have not
observed that problem [59].
It should be noted that the CDI process is not only dependent
upon the composition of the ions present but also their con-
centration. Similar to that of ED systems, we can expect that as
ionic strength increases, double layer distances collapse and sur-
face potentials generated by the applied fields are shielded. This Fig. 10. Typical electrode pairs of carbon used in capacitive deionization.

fact limits the total charge that can be stored at a given ionic
strength and potential, which essentially corresponds to the max-
imum capacitance of the inner layer of the electrochemical double
covers each electrode. One electrode is coated with an acidic SiO2
layer model [49]. This is thought to be one of the reasons why CDI
nanoporous film and the other with a basic Al2 O3 nanoporous film.
methods of desalination may be more effective for brackish than
We refer to these asymmetric systems as fourth generation devices.
for seawater scenarios [29]. However, by using asymmetric elec-
The main advantage of this sort of asymmetric system is the
trode materials having different pore size and differing charging
different intrinsic properties of each electrode. A distinct zeta
mechanisms, it may be possible to improve the performance of CDI
potential exists for each material (electrode) at a given pH. This
systems in waters having higher salt concentrations.
provides a superficial charge that is negative in the case of SiO2 but
positive in the Al2 O3 film. Typically, seawater and drinking waters
4.3.3. Carbon based CDI systems
have a pH value between 7 and 8.5. As illustrated in Fig. 12, coatings
Most current CDI or electro-adsorption systems typically utilize
on the carbon grid would be expected to be highly negative for the
high surface carbon in a variety of forms. Many studies on CDI or
SiO2 electrode and positive for an Al2 O3 or Mg-doped Al2 O3 elec-
electro-sorption utilize carbon aerogels [17,18,20,21,49,55,60–67],
trode. That superficial charge will favour the electro-adsorption of
others use either carbon cloths [68–74], carbon sheets [75], car-
cations or other positively charged species on the SiO2 electrode. In
bon nanotubes [47] or carbon nanofibers [76–80]. Some carbons
contrast, anions or other negatively charged species are most likely
are deposited using a chemical vapor deposition process to reduce
to deposit at the Mg-doped Al2 O3 electrode during the desalination
the size of the pores in carbon fiber electrodes [81]. Other carbon
process, but most importantly, it will avoid ions of the opposite
electrodes have been fabricated from carbon suspensions using
charge to adsorb during the regeneration step, as will be discussed
a wet phase inversion method [43]. A simpler electrode can be
further below. However, we firstly introduce some concepts for an
fabricated simply by using pressed activated carbon granules or
electrochemical analysis of deionization efficiency for this type of
ordered mesoporous carbon synthesized by a modified sol–gel pro-
CDI system.
cess [44,56,82]. A substantial review on carbon properties and their
role from the point of view of its use in supercapacitors has been
published in 2006 by Pandolfo and Hollenkamp [83].
A typical electrode pair configuration for carbon based CDI sys-
tems is shown in Fig. 10. As stated above, these systems became
practical as a means of deionizing water only with the intro-
duction of new materials—namely high surface area conducting
carbons [17]. Important characteristics of electrodes in a CDI sys-
tem together with capacitance and pore size distribution are range
of stability with respect to applied potential and the stability of
these electrodes in flowing aqueous media.

4.3.4. Other CD materials and our asymmetric system

It should be mentioned that other materials different from car-
bon can be used for CDI processes but examples are scarce. One
exception has been the work of Bladergroen and Linkov [84]. These
researchers studied electro-sorption using ceramic membranes.
In our research group, we have also employed the concept of a
ceramic membrane not as a mere separator between electrodes
nor as a primary material of the electrodes, but rather as a thin-film
nanoporous inorganic coating on the carbon fibers which changes
the physical-chemical properties of the carbon [59]. A schematic of
our CDI system is shown in Fig. 11. Indeed, as we shall discuss below, Fig. 11. Schematic diagram of our fourth generation CDI device showing the carbon
two asymmetric electrodes are used in this process. A different film fibers coated with different types of films.
 M.A. Anderson et al. / Electrochimica Acta 55 (2010) 3845–3856 3853

Fig. 12. Variation of zeta potential with pH for MgAl2 O4 and SiO2 .

Deionization efficiency, d , is defined as the ratio between the

charge injected to the electrodes and the amount of salt removed Fig. 13. Regeneration process of the 4th generation CD system using +1 V, pH 6.0
(black squares) and electrochemical regeneration using only diffusion (red trian-
upon the cell polarization:
gles).
dn
d = zF (15)
dq The alternative we have proposed is to modify the electrodes
Here n is the amount of salt in moles removed from the bulk solu- so that they spontaneously stay at zeta potentials far from the
tion, q is the amount of electrical charge (in Coulombs) added to the PZC. To obtain this condition, one of the electrodes must have a
CDI cell and F is the Faraday constant. This equation for capacitive spontaneous positive charge and the other negative. Under these
deionization is essentially the same as Eq. (4) for Electrodialysis, conditions, very few positive co-ions will be adsorbed on the pos-
provided that: dn = Qf dC dt and dq = I dt. itive electrode, nor negative co-ions on the negative electrode.
Efficiency of CDI systems depends on multiple variables; some of Coatings using oxides having differing PZC’s have been demon-
these variables are reviewed in the next paragraphs. One of them strated to modify zeta potential of the electrode, without damaging
is redox process occurring on the surface of the electrodes such their capacity. On the contrary, preliminary results show a moder-
as water electrolysis. Such faradaic reactions begin to become sig- ate increase in their capacity to store electrical charges. Depending
nificant only if a certain voltage level is exceeded, which can be on the coating material zeta potentials can be positive or negative
viewed as a practical limit for electro-sorption processes. Below at the usual pH of brackish and seawater [59], so with an appropri-
this threshold voltage, redox processes can be neglected unless ate selection of coating materials the detrimental effect of co-ions
electroactive dissolved species such as iron are present in the feed on charge efficiency can be reduced without any external voltage
solution. The most relevant effect arises from the obvious fact that regulation.
the electrolyte side of the electrical double layer is comprised of The contribution of our asymmetric system involving two mate-
two types of charge carriers. Thus, electrical charge added to an rials with distinct and opposite surface potentials at the pH of
electrode upon polarization is balanced not only by adsorbed coun- seawaters and drinking waters allows us to regenerate electrodes
terions, which are opposite in sign to that of the electrical charge, more efficiently. Each electrode material has a given zeta poten-
but also by the desorption of the co-ions having the same charge tial at the pH of the solution that repels co-ions that are generated
sign as that of the electrical charge of the electrode. This is quanti- during the rinse cycle.
tatively expressed by Eq. (16): In Fig. 13, the concentration of Ca2+ ions in the regenerating
solution is shown. Without applying a potential, ions simply dif-
dn = d − − d + (16) fuse into solution (red triangles). However, if a reverse potential
In which d − and d + are the infinitesimal changes in the surface of 1 V is applied, we can regenerate the system to 75% of the ini-
molar excesses of anions and cations, respectively [40]. tial Ca2+ concentration in ten minutes (black squares) as compared
Combining Eqs. (15) and (16) we obtain: to a 25% regeneration level relying on diffusion only. This increase
in regeneration rates by applying a reverse potential substantially
d − − d + reduces the amount of waste brines produced in this process.
d = zF (17)
dq
According to the above equation, charge efficiency depends on 4.4. Devices and operational aspects of CDI systems
adsorption of counterions and desorption of co-ions. This is partic-
ularly important if the electrode is at the potential of zero charge In the previous sections, we have shown the importance of
(PZC) where the amount of positive and negative charges adsorbed materials for building cells with stable and optimal performances.
on its surface is the same, so the difference d − − d + will be zero However, when real life devices are conceived, we also need to con-
and d = 0. When the electrode is polarized, more counterions are sider operational aspects of these systems. Appropriate designs for
adsorbed while co-ions are desorbed, so the efficiency will increase the complete CDI system as well as sensible process engineering are
proportionally. When the amount of desorbed co-ions is zero, d = 1. needed for the implementation of this technology for specific appli-
Some authors propose working at polarization voltages far from cations. Some of what we consider relevant operational aspects of
the PZC in order to maintain high charge efficiency [86]. However, the CDI technology are discussed below.
this has the disadvantage of reducing the capacity of the electrodes
to store charge. Furthermore, from a practical point of view, this 4.4.1. Two units in parallel
option requires a fine voltage control of the device during regener- In order to advantageously utilize energy stored in a CDI unit,
ation, so as to avoid every electrode returning to its natural open once its electrodes are fully charged, it is necessary to implement
circuit voltage. an electrical load that could allow discharging of such a unit. For our
3854 M.A. Anderson et al. / Electrochimica Acta 55 (2010) 3845–3856

Fig. 14. Schematic drawing of the charge/discharge configurations. In configuration 1, the CDI Unit A is in the deionization cycle producing a diluted solution and Unit B is
in the regeneration cycle discharging the adsorbed ions to the concentrate solution. In configuration 2, the situation is reversed.

CDI system we propose a second CDI unit working in parallel with operational drawbacks. A major issue is the fact that solution
the first, so when the first is regenerated, the energy stored can be concentrations vary with time. At the beginning of the batch deion-
applied to deionize the solution stream passing through the second ization cycle, the ion concentration is that of the feed solution, while
unit. Once the second unit is saturated, the system will be reverted at the end of the cycle the composition of the solution is mostly that
and the energy stored used to deionize a feed solution passing of a dilute solution. In most cases, there will be remarkable changes
through the first CDI unit. This procedure is schematically depicted in the physicochemical properties of the solution such as pH, con-
in Fig. 14. Energy stored in the CDI unit at the end of its deionization ductivity, density and viscosity. Such changes require continuous
cycle will be delivered exactly as in the discharge of a capacitor. adjustment of critical operating parameters such as current density
Such discharge will follow the typical V–I curve of any capacitor, and pumping or agitation power. In addition, the CDI cells, partic-
which means that there is more or less a linear decrease of volt- ularly the electrode materials, have to be stable over a wide range
age with current or current density. However, charging a capacitor, of working conditions.
in other words feeding the CDI unit during the deionization cycle, To overcome these problems one needs to design a continuous
shall preferably be performed at constant current or constant volt- operational system. This is based on recycling the outlet flux of the
age. In order to match two CDI units, one in regeneration and the CDI unit to its inlet and coupling this with a make up of feed solution
other in deionization cycle, a DC/DC converter shall be placed in and a bleed from the solution to be treated. A schematic diagram
between. Such converter will transform the variable V–I curve of of this system is shown in Fig. 15.
the CDI unit in regeneration cycle (discharge) into a constant cur- Although there is no doubt that this system is more complex
rent or constant voltage supply for the CDI unit in deionization cycle than that of a batch system, its operation is easier because it is
(charge). possible to adjust the concentration gap between inlet and outlet
In configuration 1, the CDI Unit A is in the deionization cycle streams to a selected value simply by changing the ratio between
using the energy stored in Unit B with the help of an external power the flow-rates of the recycle and bleed streams. At high recycle
source. Unit B is in the regeneration cycle discharging the adsorbed to feed ratio, the concentration in the unit remains almost con-
ions to the rejected solution while supplying DC power to Unit A. In stant. This allows one to operate under constant physicochemical
configuration 2, CDI Unit A is in the regeneration mode while Unit parameters.
B is in the deionization cycle and being fed with the energy stored Another important feature of this system is that the actual oper-
in Unit A in addition to the external power supply. ating concentration of salts within the CDI unit is very close to
the concentration of the deionized product and independent of
the concentration of the feed solution. Under these conditions,
4.4.2. Recycle loop the efficiency remains rather constant and, more importantly, will
Very often, water treatment systems are designed for batch always be kept near its maximum value for each specific appli-
operation because these systems are easier to construct and usu- cation; keeping in mind that lower concentrations provide higher
ally require a lower investment. However, batch systems have charge efficiencies [29].

Fig. 15. Block diagram showing a continuous mode of operation. Feed solution of 35 g/L NaCl, treated solution of 0.3 g/L and Qrecycle /Qbleed = 100.
 M.A. Anderson et al. / Electrochimica Acta 55 (2010) 3845–3856 3855

5. Conclusions and future trends for CDI systems important issues are now being investigated in some research and
development projects that are currently in progress both at IMDEA
In order for CDI systems to function at water treatment levels of Energy and at the University of Wisconsin in Madison.
millions of gallon per day (MGD) (hundreds of thousands of cubic The CDI technology is young and needs testing. As Oren [29]
meters per day), pairs of electrodes must be coupled into stacks reminds us, there are only a few companies trying to commercialize
and stacks into modules. To the best of our knowledge, the largest this technology. While some results of these early commercializa-
CD stack tested and reported in the open scientific literature has tion stories seem to look promising, no information can be obtained
been in the study of Welgemoed and Schutte [21]. He experienced on the length of field-testing, or how the electrodes behaved after
particular problems with the corrosion of stainless steel bus con- longer periods of operation. More work remains, however, we
nectors for carrying current from the potentiostat to the conducting should remember the pioneering work of Loeb and Sourirajan on
carbon electrodes. These problems were later resolved by using RO membranes [8]. Their work on new materials made present-
graphite bus connectors. The CDTTM patent of LLNL regarding the day reverse osmosis of seawater possible. Perhaps new electrode
Resorcinol Formaldehyde (RF) based carbon aerogel [17–20] has materials and better process control strategies will make this true
been licensed by CDT technologies of Addison, TX. They have also for CDI. One hopes that we can recover our water resources and
licensed a newer porous carbon obtained from carbohydrates (TDA save energy as well.
Research, Inc. (TDA), of Wheat Ridge, Colorado).
While problems of market penetration of CDI technologies may
relate to the youth of the process, there are some other factors as yet References
to be resolved. One of these issues is related to the cost of materials.
[1] C.A.S. Hall, J.W. Day, Am. Sci. 97 (2009) 230.
Much like supercapacitors, the cost of carbon needs to be reduced [2] R. Semiat, Environ. Sci. Technol. 42 (2008) 8193.
and the materials must prove to be stable over time and use. Several [3] M.E. Webber, Earth 3.0 (2008). http://www.sciam.com/article.cfm?id=the-
companies are playing in this sector but one company, Reticle [87], future-of-fuel&page=5.
[4] S. Tully, Fortune Magazine 141 (2000) 342.
claims to have superior carbon for CD applications at reduced cost. [5] M. Black, The No-nonsense Guide to Water, New Internationalist, Oxford, 2004.
There are additional problems related to regeneration rates, fouling [6] R. Borsani, S. Rebagliati, Desalination 182 (2005) 29.
and long-term stability that have not been adequately addressed [7] S. Chaudhry, CA Desalination Task Force Sausalito, CA (2003)
http://www.owue.water.ca.gov/recycle/desal/Docs/UnitCostDesalination.pdf.
until this time. To the best of our knowledge studies reported above
[8] S. Loeb, S. Sourirajan, U.S. Patent 3,133,132 (1964).
by Welgemoed and Schutte [21] and that of Gabelich et al. [55] have [9] P.M. Biesheuvel, J. Colloid Interface Sci. 332 (2009) 258.
begun to examine some of these problems in detail. [10] K.S. Spiegler, Y.M. El-Sayed, Desalination 134 (2001) 109.
[11] I.C. Karagiannis, P.G. Soldatos, Desalination 223 (2008) 448.
Fouling is of great concern for RO systems and as suggested
[12] T. Xu, C. Huang, AlChE J. 54 (2008) 3147.
by Oren [29], CDI systems may also suffer from this problem in [13] T.A. Davis, in: M.C. Porter (Ed.), Handbook of Industrial Membrane Technology,
actual operation. However, there is a significant advantage to the Noyes, Park Ridge, NJ, 1990.
CDI process in this aspect as it has been shown that, by switching [14] K.S. Spiegler, D.K. Laird, Principles of Desalination, 2nd ed., Academic Press,
New York, 1980.
potentials of electrodes in electrochemical systems, one drastically [15] J.M. Ortiz, J.A. Sotoca, E. Expósito, F. Gallud, V. García-García, V. Montiel, A.
reduces fouling [88]. Lastly, it is conceivable to apply a pulsed field Aldaz, J. Membr. Sci. 252 (2005) 65.
to the electrodes to reduce both inorganic and organic fouling [89]. [16] Q. Daoduo, Z. Linda, H. Eric, Res. J. Chem. Environ. 11 (2007) 92.
[17] J. Farmer, D. Fix, G. Mack, R. Pekala, J. Poco, J. Appl. Electrochem. 26 (1996) 1007.
Fouling is a problem in all water treatment scenarios and particu- [18] J.C. Farmer, D.V. Fix, G.V. Mack, R.W. Pekala, J.F. Poco, J. Electrochem. Soc. 143
larly in membrane based systems. In the case of CDI systems, some (1996) 159.
controversy arises regarding this aspect. While Andelman [90] has [19] J.C. Farmer, S.M. Bahowick, J.E. Harrar, D.V. Fix, R.E. Martinelli, A.K. Vu, K.L.
Carroll, Energy Fuels 11 (1997) 337.
theoretically provided a CDI system that claims to overcome these [20] R.W. Pekala, J.C. Farmer, C.T. Alviso, T.D. Tran, S.T. Mayer, J.M. Miller, B. Dunn,
fouling problems from a design perspective, Oren has suggested J. Non-Cryst. Solids 225 (1998) 74.
that there is little practical evidence in the real world that this in [21] T.J. Welgemoed, C.F. Schutte, Desalination 183 (2005) 327.
[22] Helmholtz, J. Franklin Inst. 115 (1883) 310.
fact works effectively [29]. On the other hand, Xu et al. [67], working
[23] J.O. Bockris, A.K.N. Reddy, M. Gamboa-Aldeco, Modern Electrochemistry, 2nd
on the treatment of brackish waters, have indicated that CDI sys- ed., Kluwer Academic/Plenum Publishers, 2000.
tems are relatively free of fouling problems. Likely, this will remain [24] M. Elimelech, J. Gregory, X. Jia, R.A. Williams, J. Gregory, X. Jia, R.A. Williams,
Particle Deposition and Aggregation, Butterworth-Heinemann, Woburn, 1995,
the subject of further and longer term studies.
Ch. 2.
Finally, a couple of remarks concerning the benefits of CDI sys- [25] R.J. Hunter, Zeta Potential in Colloid Science: Principles and Applications, Aca-
tems over that of RO and CDF desalination are worth mentioning. demic Press, London, 1981.
CDI is a low pressure process and is therefore expected, at least for [26] J.J. Lyklema (Ed.), Fundamentals of Interface and Colloid Science Solid–Liquid
Interfaces, Academic Press, 1995, Ch. 4.
brackish waters, to consume 1/3 less energy [91]. However, as we [27] L. Dongqing (Ed.), Interface Science and Technology Electrokinetics in Microflu-
noted above, these CDI systems are performing as supercapacitors idics, Elsevier, 2004, Ch. 10.
and therefore, while they are cleaning water they are also storing [28] P.M. Biesheuvel, B. van Limpt, A. van der Wal, J. Phys. Chem. C 113 (2009) 5636.
[29] Y. Oren, Desalination 228 (2008) 10.
energy. We can recover much of this energy so that figures of merit [30] D.D. Caudle, T.H. Tucker, J.L. Cooper, B.B. Arnold, A. Papastamataki, Research
mentioned above will improve. In addition, we may not always Report Oklahoma University Research Institute, 1966.
need to deliver the quality of water that RO and CDF provide. In [31] A.M. Johnson, J. Newman, J. Electrochem. Soc. 118 (1971) 510.
[32] A.M. Johnson, U.S. Patent, 3,755,135 (1973).
situations such as coal bed methane recovery or the treatment of [33] Y. Oren, A. Soffer, J. Electrochem. Soc. 124 (1977) C121.
other brackish waters where CDI systems could be used to obtain [34] Y. Oren, A. Soffer, J. Electrochem. Soc. 125 (1978) 869.
water suitable to irrigate croplands [91]. [35] Y. Oren, A. Soffer, J. Electroanal. Chem. 206 (1986) 101.
[36] Y. Oren, A. Soffer, J. Electrochem. Soc. 126 (1979) C330.
The reader is reminded that this is a review paper as it examines
[37] Y. Oren, H. Tobias, A. Soffer, Bioelectrochem. Bioenerg. 11 (1983) 347.
the present state of art and past history of the CDI process in com- [38] Y. Oren, H. Tobias, A. Soffer, J. Electroanal. Chem. 162 (1984) 87.
parison to what we presently know about competing desalination [39] Y. Oren, A. Soffer, J. Electroanal. Chem. 186 (1985) 63.
[40] E. Avraham, Y. Bouhadana, A. Soffer, D. Aurbach, J. Electrochem. Soc. 156 (2009)
systems. We are aware that in this review we have omitted some
95.
practical aspects with remarkable effects on capital and operational [41] J.M. Miller, A.F. Burke, Interface 17 (2008) 53.
costs of these CDI systems. At the present time, there is a paucity [42] R. Kötz, M. Carlen, Electrochim. Acta 45 (2000) 2483.
of data on these systems with respect to the design of flow condi- [43] J.A. Lim, N.S. Park, J.S. Park, J.H. Choi, Desalination 238 (2009) 37.
[44] L. Li, L. Zou, H. Song, G. Morris, Carbon 47 (2009) 775.
tions inside CDI units, as well as regarding the power electronics [45] J.B. Lee, K.K. Park, S.W. Yoon, P.Y. Park, K.I. Park, C.W. Lee, Desalination 237
or control logics needed to support these systems. Some of these (2009) 155.
3856 M.A. Anderson et al. / Electrochimica Acta 55 (2010) 3845–3856

[46] O. Kitous, A. Cheikh, H. Lounici, H. Grib, A. Pauss, N. Mameri, J. Hazard. Mater. [69] H.J. Oh, J.H. Lee, H.J. Ahn, Y. Jeong, Y.J. Kim, C.S. Chi, Thin Solid Films 515 (2006)
161 (2009) 1035. 220.
[47] Y. Gao, L. Pan, H. Li, Y. Zhang, Z. Zhang, Y. Chen, Z. Sun, Thin Solid Films 517 [70] M.W. Ryoo, G. Seo, Water Res. 37 (2003) 1527.
(2009) 1616. [71] M.W. Ryoo, J.H. Kim, G. Seo, J. Colloid Interface Sci. 264 (2003) 414.
[48] R. Chandrasekaran, M. Koh, A. Yamauchi, M. Ishikawa, Electrochemistry 77 [72] B.E. Conway, E. Ayranci, H. Al-Maznai, Electrochim. Acta 47 (2001) 705.
(2009) 51. [73] A. Afkhami, B.E. Conway, J. Colloid Interface Sci. 251 (2002) 248.
[49] K.L. Yang, T.Y. Ying, S. Yiacoumi, C. Tsouris, E.S. Vittoratos, Langmuir 17 (2001) [74] E. Ayranci, B.E. Conway, J. Appl. Electrochem. 31 (2001) 257.
1961. [75] K.K. Park, J.B. Lee, P.Y. Park, S.W. Yoon, J.S. Moon, H.M. Eum, C.W. Lee, Desali-
[50] H. Chia-Hung, P. Taboada-Serrano, S. Yiacoumi, C. Tsouris, J. Chem. Phys. 128 nation 206 (2007) 86.
(2008) 044705. [76] X.Z. Wang, M.G. Li, Y.W. Chen, R.M. Cheng, S.M. Huang, L.K. Pan, Z. Sun, Elec-
[51] B. Kastening, M. Heins, Electrochim. Acta 50 (2005) 2487. trochem. Solid-State Lett. 9 (2006) E23–E26.
[52] A. Alfarra, E. Frackowiak, F. Beguin, Electrochim. Acta 47 (2002) 1545. [77] X.Z. Wang, M.G. Li, Y.W. Chen, R.M. Cheng, S.M. Huang, L.K. Pan, Z. Sun, Appl.
[53] A. Seron, H. Benaddi, F. Beguin, E. Frackowiak, J.L. Bretelle, M.C. Thiry, T.J. Ban- Phys. Lett. 89 (2006) 053127.
dosz, J. Jagiello, J.A. Schwarz, Carbon 34 (1996) 481. [78] Y. Gao, L.K. Pan, Y.P. Zhang, Y.W. Chen, Z. Sun, Surf. Rev. Lett. 14 (2007)
[54] B.E. Conway, Electrochim. Acta 40 (1995) 1501. 1033.
[55] C.J. Gabelich, T.D. Tran, I.H. Suffet, Environ. Sci. Technol. 36 (2002) 3010. [79] D. Zhang, L. Shi, J. Fang, K. Dai, J. Mater. Sci. 42 (2007) 2471.
[56] L. Zou, L. Li, H. Song, G. Morris, Water Res. 42 (2008) 2340. [80] Y. Gao, H.B. Li, Z.J. Cheng, M.C. Zhang, Y.P. Zhang, Z.J. Zhang, Y.W. Chen, L.K. Pan,
[57] F.G. Dwyer, P. Chu, in Mobile Oil Corporation N.Y., U.S. Patent 4,526,879 (1985). Z. Sun, Thin Solid Film 217 (2008) 1616.
[58] B.P. Pelrine, N.J. Trenton, in Mobile Oil Corporation N.Y., U.S. Patent 4,100,262 [81] E. Avraham, B. Yaniv, A. Soffer, D. Aurbach, J. Phys. Chem. C 112 (2008) 7385.
(1978). [82] L. Zou, G. Morris, D. Qi, Desalination 225 (2008) 329.
[59] K.C. Leonard, J.R. Genthe, J.L. Sanfilippo, W.A. Zeltner, M.A. Anderson, Elec- [83] A.G. Pandolfo, A.F. Hollenkamp, J. Power Sources 157 (2006) 11.
trochim. Acta 54 (2009) 5286. [84] B.J. Bladergroen, V.M. Linkov, Sep. Purif. Technol. 25 (2001) 347.
[60] H.H. Jung, S.W. Hwang, S.H. Hyun, L. Kang-Ho, G.T. Kim, Desalination 216 (2007) [86] E. Avraham, M. Noked, Y. Bouhadana, A. Soffer, D. Aurbach, J. Electrochem. Soc.
377. 156 (2009) 157.
[61] C.M. Yang, W.H. Choi, B.K. Na, B.W. Cho, W.I. Cho, Desalination 174 (2005) 125. [87] C.C. Nesbitt, J. Chem. Soc. Abstr. 224 (2002) U529.
[62] W.C. Li, H.Q. Hu, Y.D. Zhu, S.W. Zhu, New Carbon Mater. 20 (2005) 217. [88] M.I. Kerwick, S.M. Reddy, A.H.L. Chamberlain, D.M. Holt, Electrochim. Acta 50
[63] P. Ranamararia, N. Mohan, C. Rajagopal, B.S. Garg, J. Sci. Ind. Res. 63 (2004) 938. (2005) 5270.
[64] S.W. Hwang, S.H. Hyun, J. Non-Cryst. Solids 347 (2004) 238. [89] R.E. Perez-Roa, D.T. Tompkins, M. Paulose, C.A. Grimes, M.A. Anderson, D.R.
[65] T.Y. Ying, K.L. Yang, S. Yiacoumi, C. Tsouris, J. Colloid Interface Sci. 250 (2002) Noguera, Biofouling 22 (2006) 383.
18. [90] M.A. Andelman, U.S. Patent 5,779,891 (1998).
[66] L.W. Hrubesh, J. Non-Cryst. Solids 225 (1998) 335. [91] K. Christen, Environ. Sci. Technol. 40 (2006) 639.
[67] P. Xu, J.E. Drewes, D. Heil, G. Wang, Water Res. 42 (2008) 2605.
[68] H.J. Ahn, J.H. Lee, Y. Jeong, J.H. Lee, C.S. Chi, H.J. Oh, Mater. Sci. Eng. A 449–451
(2007) 841.