Desalination 228 (2008) 10–29

Capacitive deionization (CDI) for desalination and water

treatment — past, present and future (a review)
Yoram Oren
Department for Desalination and Water Treatment–Zuckerberg Institute for Water Research and Unit for
Environmental Engineering, Ben-Gurion University of the Negev, PO Box 653, Beer-Sheva 84105, Israel
Tel. +972 (52) 579-3088; Fax: +972 (8) 647-2960; email: [email protected]

Received 16 February 2007; Accepted 1 August 2007

Abstract
Capacitive deionization (CDI) has attracted the interest of the community investigating water treatment
technologies since the mid-1960s. The technology is based on the recognition that high-surface-area electrodes, when
electrically charged, can quantitatively adsorb ionic components from water, thereby resulting in desalination. The
article reviews the theoretical and technological background of CDI, the history of its development, and past and
present attempts towards scaling up and commercialization. It also provides a critical review of the advantages and
limitations of the technology. In addition, further research approaches necessary to render CDI a mature and
competitive technology for the treatment of brackish and surface waters are outlined.

Keywords: Capacitive deionization (CDI); Capacitive deionization technology (CDT); Electronic water purifi-
cation (EWP); Electrosorption

1. Introduction
of salt adsorbed on a unit weight of electrode
Capacitive deionization (CDI) is an electro- material. This factor renders the electrodsorption
chemically controlled method for removing salt process quantitative and attractive for water
from aqueous solutions by taking advantage of treatment.
the excess ions adsorbed in the electrical double Early studies on CDI date to the mid-1960s
layer region at an electrode-solution interface and the early 1970s. The concept was first intro-
when the electrode is electrically charged by an duced by Caudle et al. [1], who used porous
external power supply. When the electrode has a carbon electrodes made of activated carbon
high specific surface area, this excess may powder in a flow-through mode for water desali-
become significant in terms of number of grams nation. Later, Johnson et al. [2] studied CDI as a

0011-9164/08/$– See front matter © 2008 Published by Elsevier B.V.

doi:10.1016/j.desal.2007.08.005
 Y. Oren / Desalination 228 (2008) 10–29 11

reversible process. Their work combined para- of the advantages and limitations of CDI; and
metric studies with investigations of the theo- (5) to outline the future R&D activities that are
retical basis of CDI and of various electrode necessary for making CDI a more mature
materials [2–5]. The intensive studies undertaken technology.
by the group of Johnson were eventually dis-
continued, mainly because of the instability of the
electrodes, particularly the anode. However, a 2. Background
preliminary cost evaluation performed by John-
2.1. Electrosorption of ions on electrodes
son’s group showed that efficient, low-cost
desalination based on this technology could At any interface between an electrolyte solu-
indeed be achieved, provided that sufficiently tion and another medium (a solid or an immi-
stable high-surface area electrodes could be scible liquid phase), there exist regions with
produced. Following this work, Johnson and excess charge while the whole interfacial system
Newman [4] published a comprehensive theore- is, of course, electrically neutral. The two sides of
tical analysis of ion adsorption on porous elec- the interface that carry an excess of opposite elec-
trodes. In the light of its importance, this work trical charge are described by the term “electrical
will be discussed at length below. double layer”.
The pioneering works of the groups of Caudle The interface properties and structure between
and Johnson triggered an enormous amount of an electronic conductor (electrode) and an aque-
work with a fundamental orientation aimed at ous electrolyte solution have been well docu-
reaching an understanding of the basic factors mented, both by the thermodynamic approach and
governing CDI. This research was accompanied by widely accepted models such as the Gouy-
by studies directed at developing novel electrode Chapman-Stern theory [6]. According to the latter
materials for efficient CDI and small and semi- theory, the electrical capacity of the interface, CT,
pilot-sized devices for desalination and water is expressed as the sum of two capacitors in
treatment. However, despite the vast amount of series, namely as:
CDI-related activity that spanned more than half

= +
a century and the intensive worldwide activity 1 1 1
devoted to developing technologies for creating (1)
new affordable water sources (e.g., seawater and CT CM − H CH − S
brackish water desalination) and for water saving
(e.g., municipal and industrial wastewater recla- where CM-H is the capacity of a “compact” double
mation), CDI has not yet emerged from the layer between the electrode surface, M, and the
laboratory or from the smallest pilot scale and plane of closest approach for the ions, H [this
demonstration units to become a valid and mature plane is sometimes known as the outer Helmholtz
commercial technology. plane (OHP)] and CH-S is the capacity of the
The objectives of this article are: (1) to pre- “diffuse” double layer spanning from plane H
sent the theoretical and technological basis for into the electrolyte solution.
CDI; (2) to describe broadly the experimental and By treating ions as point charges and assum-
technological achievements in developing tar- ing an ideally polarized electrode (no electro-
geted electrode materials, modes of operation and chemical reactions, namely, no electron transfer
process variations; (3) to describe past and through the interface) and the absence of ion
present attempts towards scaling up and commer- adsorption associated with specific interactions,
cialization of CDI; (4) to provide a critical review Gouy and Chapman calculated the dependence of
12 Y. Oren / Desalination 228 (2008) 10–29

the excess charge, q on the potential at plan (H),

NH, in z–z electrolyte solution of bulk concen-
tration Cs as

⎛ 2 RT εCs ⎞ ⎛ z ℑφH ⎞
q=⎜ ⎟ sinh ⎜ ⎟
1/ 2

⎝ π ⎠ ⎝ 2 RT ⎠
(2)

The electrical capacity of the diffuse double layer

is then obtained by differentiating Eq. (2) with
respect to NH:

⎛ εCs ⎞ ⎛ z ℑφH ⎞
CH − S = z ℑ ⎜ ⎟ cosh ⎜ ⎟
1/ 2

⎝ 2 RT π ⎠ ⎝ 2 RT ⎠
(3)
Fig. 1. Schematic representation of charge distribution in
an electrolyte solution adjacent to a negatively charged
electrode.
The excess charge distribution at the interface
between a charged electrode-electrolyte solution
according to the Gouy–Chapman–Stern model is
described schematically in Fig. 1.
When two oppositely charged electrodes are in
equilibrium with the same electrolyte solution, an
excess of charge is built up at the interface near
each of the electrodes. The spatial charge distri-
bution between the electrodes will then be that
given by the sum of the separate distributions, as
shown schematically in Fig. 2.
As can be seen from Eq. (3), the diffuse
double layer capacity varies as a function of CS1/2. Fig. 2. Excess charge distribution in a two-electrode
Another peculiar outcome of this equation is the system (not to scale).
minimum capacity at NH = 0 (the so-called
“potential of zero charge”). Some typical values Table 1
of experimental and calculated diffuse double Diffuse double layer capacity of Hg/NaF interface at the
layer capacities at the potential of zero charge as potential of zero charge
a function of solution concentration are given in
Table 1. It is evident that the electrical capacity Concen- Diffuse double layer specific capacity
increases with increasing the solution concen- tration (N) (µF cm!2)
tration. In addition, there is a large discrepancy Experimental Calculated from Eq. (3)
between the measured and the calculated capacity
values, but an in-depth discussion of this point is 0.001 6.0 7.2
0.01 13.1 22.8
beyond the scope of this article. Nevertheless,
0.1 20.7 72.2
Table 1 shows that the specific electrical capa- 1.0 25.7 228.0
cities can reach high values, as a result of the
extremely short distances involved in both the Adapted from Bockris and Reddy [7].
 Y. Oren / Desalination 228 (2008) 10–29 13

compact and the diffuse double layer structures.

These large specific capacities are important if
the pair of electrodes is to function in CDI, as
shown in the following simplistic calculation:
Considering an electrode–solution system with a
specific capacity of CT = 10 µF cm!2 and assum-
ing a voltage change of ΔV = 1 V, we would
obtain a charge excess at the double layer region
of ΔQ = ΔV×CT = 10 µcoulomb cm!2 or, in terms
of equivalent ions, 1.036×10!10 equivalents cm!2.
This figure may appear quantitatively not prac-
tical for desalination. However, for a porous elec-
trode (e.g. carbon) having a specific surface area
of 1000 m2 g!1 that is entirely accessible to ion
penetration (an issue discussed below), it can be
calculated that 1 g of this electrode material can
adsorb an excess of 1.036 meq ions, a number
that does appear promising for the desalination Fig. 3. Equivalent circuit representing a pair of charged
process. Thus, for two such electrodes, 50 g each electrodes.
would be required for the complete desalination
of 1 L of 3000 ppm brackish water. Although this
does not yet appear to be feasible for treating increasing the potential of an electrode towards
large water volumes because of the large quan- more positive values will result in an increase of
tities of electrode material required, there are anion adsorption and cation desorption. This is
several technological approaches to overcome shown in Fig. 4, which presents typical anion and
this hurdle. These will be discussed below. cation equilibrium adsorption curves on a single
porous electrode. An electroadsorption cell com-
prising of two electrodes is operated by transfer-
2.2. Analysis of desalting by charging a two- ring an amount of charge Δqe between the two
electrode system electrodes. While anions are adsorbed on the
Quantitative salt electrosorption on a pair of positively charged electrode, they are desorbed
identical porous electrodes for CDI has been from the negatively charged electrode. The same
analyzed by Oren and Soffer [8–10] and by John- holds for the cations on the negatively and
son and Newman [4] who discussed in detail positively charged electrodes, respectively. The
charge and energy efficiencies as well as total amount of salt ΔnS adsorbed under equili-
adsorption/desorption kinetics, respectively. brium by the unit cell can thus be less than the
The electrode pair separated by an electronic- charge equivalent Δqe/F exchanged between the
ally inert porous separator through which the salt electrodes. In some cases, the charge efficiency
solution flows has been described by an equi- for adsorption FΔnS/Δqe can reach a value of
valent electrical circuit comprised of a network of zero. Consequently, charge and energy efficiency
electrical double layer capacitors, CT, and elec- parameters of a two-electrode cell may be derived
trical resistors RM and RS, representing the resis- by using single electrode characteristics [9]. As a
tances of the electrode material and the solution, first step, it is convenient to define qe and qd as
respectively, as shown in Fig. 3. Charging CT by new charge coordinates for the whole cell by the
14 Y. Oren / Desalination 228 (2008) 10–29

Fig. 4. Typical anion (Δn!) and cation (Δn+) excess adsorbed on a single FC-12 carbon electrode as a function of electrode
charge. Solution: NaCl 0.12 M. (Reprinted with permission from the Journal of Applied Electrochemistry [9].)

properties of the individual electrodes 1 and 2 two electrodes the total cell charge (g1 + g2)qd
comprising the cell does not change. It can be changed only by
installing a third, auxiliary electrode within the
qd = μ1q1 + μ 2 q2 ⎫
cell through which electrode 1 or 2 can be
⎪
2qe = μ1q1 − μ 2 q2
charged independently. The average charge effi-
⎪
⎬
ciency for a charging step of the system com-

⎪
(4) prising two identical electrodes when charging

⎪
μ1 = g1 ( g1 + g 2 ) , μ 2 = g 2 ( g1 + g 2 ) ⎭
and from qe1 to qe2 is defined as:

⎛ ∂ns ⎞
∫q ⎜⎝ ∂qe ⎟⎠ d qe
Δns ℑ
ℑ =
qe 2

Δqe qe 2− qe1
(5)
In the above expressions, q1 and q2 are the e1 qd
charges per unit weight and g1 and g2, the weights
of electrodes 1 and 2 respectively. where (∂ns/∂qe) is the differential charge effi-
From these definitions, it can be shown that qd ciency at constant qd.
is the total cell charge, while Δqe is, as noted The net salt charge adsorbed per unit of elec-
above, the charge exchanged between the two trical charging work in the complete cell (the
electrodes. Both are calculated per unit weight of differential energy efficiency) is defined as
total electrode material. It should be noted that by F(Mns/Mw)qd, where w is the electrical work of
exchanging (g1 + g2)Δqe coulombs between the charging the electrical double layer, namely,
 Y. Oren / Desalination 228 (2008) 10–29 15

⎛q ⎞
d w = ΔE d qe = ⎜ e ⎟ d qe
⎝ CT ⎠
(6a)

Accordingly,

⎛ ∂n ⎞ ⎛ ∂ns ⎞
ℑ⎜ s ⎟ = ℑ T ⎜ ⎟
C
⎝ ∂w ⎠qd ⎝ ∂qe ⎠qd
(6b)
qe

The differential charge efficiency can be ex-

pressed by the individual charges as follows:

⎛ ∂n ⎞ ∂q − ∂q − ∂q + ∂q +
ℑ⎜ s ⎟ = 1 − 2 = 1 − 2
⎝ ∂qe ⎠ qd ∂q1 ∂q2 ∂q1 ∂q2
(7)
Fig. 5. Typical slopes Mq!i /Mqi vs. electrode charge q.
• FC-12 carbon electrodes; M Gouy–Chapman–Stern
At the extreme conditions, i.e., where electrodes model with g = 20;  KF solution on mercury electrode.
1 and 2 are highly positively and negatively (Reprinted with permission from the Journal of Applied
charged, respectively, the following approxi- Electrochemistry [9].)
mations hold:
The charge efficiency (Mns/Mqe) can be cal-
∂q +
∂q −
∂q +
∂q +
≈ 0;… ≈ 1;… ≈ 1;… ≈ 0 (8) provided that explicit expressions for ∂qi+ / ∂qi
culated from the single electrode properties

∂q1 ∂q1 ∂q2 ∂q2 and ∂qi− / ∂qi are known (such as that from the
1 1 2 2

data giving ∂qi+ / ∂qi and ∂qi− / ∂qi as a function

Gouy–Chapman model) or by using experimental
where q1!, q2!, q1+ and q2+, are the anion and
cation excess charges on electrodes 1 and 2, of the electrode charge or potential (Fig. 5).
respectively. By introducing the individual parts in Eq. (7)
When these limiting values are inserted into in terms of the Gouy–Chapman model and using
Eq. (7), the efficiency approaches unity. Thus, for the above charge coordinates, expression (9)
the two-electrode cell to approach its highest follows:
⎧
∂ns 1 ⎪ Qe + ( Qd / 2 )
desalination capacity and differential energy
ℑ = ⎨
∂qe 2 ⎪ ⎡( Q + Q / 2 )2 + μ 2 ⎤1/ 2
efficiency, the electrodes should be maintained at

⎩⎣ e 1⎦
the extreme opposite charge values.
The effect of the electrolyte concentration on d (9)

⎫
the charge efficiency must also be taken into

Qe − (Qd / 2) ⎪
+
account. As the bulk concentration increases, the
1/ 2 ⎬

⎣( d
⎡ Q / 2 − Qe ) + μ 22 ⎤ ⎪
concentration of ions bearing the same charge as

⎦ ⎭
that of the electrode at the double layer region 2

also increases. Therefore, ion depletion from that

region is less likely. Consequently, charge effi- where A = (2RTgg0CS)1/2 in which Qd = qd/2A,
ciency may become lower at higher concen- Qe = qe/2A where Cs is the bulk solution
trations and may never reach unity. concentration.
16 Y. Oren / Desalination 228 (2008) 10–29

Fig. 6. Differential charge efficiency as a function of Qe

for different Qd values.
Fig. 8. Deferential charge efficiency as a function of
solution concentration for different charge values, qe and
qd = 0.

From the behavior of the two functions, the

practical limits for qd and the range qe1 to qe2 for
attaining maximum efficiencies can be deduced:
The total cell charge qd should be as low as
possible (practically, zero). The lower limit qe1
should be somewhat above zero, while the high-
est limit qe2 should be as large as possible, with
the obvious limit being set by parasitic Faradaic

from Eq. (9) it can be shown that ℑ ( ∂ns / ∂qe )

processes such as water electrolysis. Furthermore,

∝ CS−1/ 2 , namely, the charge efficiency decreases

Fig. 7. Differential energy efficiency as a function of Qe
for different Qd values.
as the bulk solution concentration increases, in
accordance with the above discussion. This rela-
The differential charge efficiency and the tionship is depicted in Fig. 8, which shows also
differential energy efficiency as derived from that the differential charge efficiency decreases as
Eqs. (6) and (9) are shown as a function of Qe for the amount of charge Δqe exchanged between the
different Qd values in Figs. 6 and 7, respectively. electrodes decreases. It should be pointed out that
It is apparent that unlike the adsorption charge since the data in Fig. 8 were generated based on
efficiency, the adsorption energy efficiency drops the Gouy–Chapman model equations and on the
for high Qe values and acquires maxima for assumption that the electrodes are non-porous, the
intermediate values. This is due to the square absolute numbers for the differential charge effi-
power dependence of the capacitor energy on ciency may be far from reality. However, the
charge, while ion adsorption is, at most, pro- relative change of the differential charge effi-
portional to the charge. For both efficiencies, the ciency with concentration may reflect the real
optimal values are obtained for Qd = 0. situation quite well.
 Y. Oren / Desalination 228 (2008) 10–29 17

Johnson and Newman [4] analyzed the hours, respectively. For a process based on the
electrosorption kinetics of ions on porous elec- immediate availability of the surface electrical
trodes in a two-electrode system. Taking into capacitance, 2 is extremely — and impractically
account that the adsorption rate is generally not — long. This issue is discussed further below.
rate determining, they found that while the In the above models, the porous structure of
capacity for electrosorption depends on the the high-surface-area electrode and its effect on
double layer electrical capacity, on the available the true capacity for ion adsorption was not
surface area of the electrode and on the applied considered explicitly. Lin et al. [11] and Yang et
potential, the adsorption rate depends on the al. [12] pointed out that the surface area measured
transport rate of charges to these surfaces and, by traditional methods such as gas adsorption
hence, on the ohmic potential drop in the solution (BET using either N2 or CO2 as probes) and
and the electrode matrix. Two characteristic times mercury intrusion porosimetry provide values
are associated with the process: For short that are much larger than those available for ion
charging times, the capacity of the electrodes is adsorption. A model was developed by Yang et
utilized only near the electrode separator and near al. [12] using the Gouy–Chapman–Stern theory
the current collector. This time, 1 is expressed by for the double layer structure within a single

R1 = + S
cylindrical pore. Results for pore size distribution
2L L
σ+κ κ
(10) for a carbon aerogel (see below) were incor-
porated into the model equations, and experi-
mental results on salt adsorption capacity as a
where function of bulk concentration and applied

⎡ R ( κ + σ ) κσ ⎤
voltage were compared to the theory. It was

⎢⎣ 2 α ( κ + σ ) ⎥⎦
2

τ1 = ⎢ 1 ⎥
concluded that only pores with diameter larger
2 2
than a specific cutoff value really contribute to
electrosorption of ions. The cut-off pore diameter
reflects a maximum occupancy of the pore
where κ and are the specific conductivity of the volume by the electrical double layer. The degree
solution and the electrode, respectively, α is the of occupancy is a function of the solution concen-
salt diffusivity, and L and LS are half the thick- tration and the applied voltage and full occupancy
ness of the electrodes and the solution layer prevents ions from entering the pores, thereby
between them, respectively. lowering electrosorption capacity.
If the applied potential is maintained for a
sufficient length of time, the entire electrode
capacity will eventually be used up. The charac- 3. Technology development
teristic time, 2, for this is Four main topics concerned with developing
CDI into a commercially feasible technology for
τ2 =
L2
α
(11) water treatment can be identified in the literature:
1. Development of an electrode material that
is suitable for the process;
Taking some ordinary values of the conducti- 2. Arrangement of the various components
vities and diffusivity and reasonable values for within the basic cell unit;
the electrode and solution layer thicknesses (mm 3. Turning the process into a continuous or
to fraction of mm), it was found that 1 and 2 semi-continuous process; and
have the characteristic order of seconds and 4. Energy recovery.
18 Y. Oren / Desalination 228 (2008) 10–29

3.1. Electrode material

Following the discussion above, it is clear that
the correct choice of electrode material is the
most critical issue for the success of electro-
chemical separation processes, including CDI. A
suitable CDI electrode material should have the
following properties:
1. Specific surface area (i.e., the surface area
per unit weight) available for electrosorption as
large as possible;
2. Electronic conductivity as high as possible;
3. Fast response of the entire surface area to
electrosorption–electrodesorption changes;
4. Chemical and electrochemical stability over
a wide range of pH values and in the presence of
oxidants (such as dissolved chlorine) and ability
to tolerate frequent voltage changes;
zation. ×500,000. Reprinted with permission from
Fig. 9. Carbon aerogel microstructure: TEM visuali-
5. Easily shaped according to design
requirements; Galelich et al. [30] and Environmental Science and
6. Low propensity for scaling, biofouling and Technology.
organic fouling.
In the light of these requirements, particularly aerogels as the most promising available mater-
the first one, it is not surprising that carbon—with ials for this purpose. These unique porous
its wide variety of available forms and porosity materials consist of interconnected uniform
—has been found to be the electrode material of particles, 3 to 30 nm in diameter, with interstitial
choice for CDI and other electroadsorption pores that are less than 50 nm in diameter
processes. The numerous studies dedicated to the (Fig. 9). This type of structure gives high density
fundamental electrochemical properties of carbon and specific surface area in the range 400 to
and graphite have shed light on the characteristics 1100 m2 g!1, low hydraulic resistance, and an
of the electrical double layer structure and its exceptionally high electrical conductivity of ~25–
ability to quantitatively adsorb ions and on the 100 S cm!1.
effect of porous structure of the carbon electrode Carbon aerogels are prepared by pyrolysing
on the adsorption rate (e.g., Oren et al. [13]; Oren organic aerogels made from resorcinol-formal-
and Soffer [14,15]; Kinoshita [16]). dehyde, phenol-furfural, phenol-resorcinol-for-
Among the various forms of carbon inves- maldehyde, melamine-formaldehyde, polyure-
tigated for separation of electroactive materials thanes, polyureas and polyvinyl chloride by the
by electroadsorption and CDI are activated car- sol-gel process [32]. The organic aerogel porous
bon [17], activated carbon cloth [18], activated precursor is prepared by polymerization of the
carbon cloth modified with titania [19,20], carbon relevant monomers into cross-linked polymer
felt [21,22], carbon black [8–10], sintered acti- clusters forming wet gels. These gels are then
vated carbon [23], carbon nanotubes [24–26], and dried before pyrolization under conditions that
carbon aerogels [12,27–31]. The largest number prevent the collapse of the porous structure, such
of investigations has been devoted to carbon as CO2 supercritical drying [33] and solvent
 Y. Oren / Desalination 228 (2008) 10–29 19

exchange and controlled evaporation [34]. Pyro- experiments showed that after a month of con-
lysis is conducted at elevated temperatures (800– tinuous operation the electrodes lost 6–8% of
1100EC) at nitrogene atmosphere. The chemical their adsorption capacity, particularly when
composition, microstructure, and physical proper- operating at lower cell voltages. No explanation
ties of the aerogels precursors can be controlled at was given for this phenomenon, but this loss
the nanometer scale, giving rise to the unique could be recovered almost entirely upon periodic-
electrical and electrochemical properties of the ally reversing the electrode polarity [45].
pyrolized material. Carbon aerogels can be pre- Yang et al. [31] used carbon aerogel-silica
pared in the form of monoliths, powders, micro- composites for CDI. Silica was added in various
spheres and thin film composites [27]. Pekala et proportions to the carbon matrix to improve the
al. [32], who were the first to prepare these mechanical properties of the electrode, its ease of
materials, characterized their structure, properties production, and its wettability. They showed that
and performance as electrodes for electrochemi- adding silica increased the electrode capacity by
cal separations in terms of the preparation con- up to 28% resulting, as speculated by the authors,
ditions. They found that aerogel electrical capa- from increased wettability of the electrode mater-
citance is significantly influenced by the pyro- ial and, therefore, facilitating better accessibility
lysis temperature and activation time, namely, in to ions. An additional advantage was that prepa-
both cases it reaches a maximum with respect to ration time of the electrode material was reduced
these parameters. This implies that at sufficiently by 50%.
long activation times and sufficiently high tem- Perhaps the most comprehensive study on the
peratures the porous structure becomes partially use of carbon aerogels in CDI under real condi-
closed, resulting in a decrease of the surface area. tions is that of Gabelich et al. [30]. They con-
As mentioned above, carbon aerogels are ducted a thorough study to elucidate the effect of
considered as promising materials for use in ionic charge, radius and mass on the electro-
supercapacitors and in CDI and other separation sorption of ions on carbon aerogels. A laboratory-
processes based on electroadsorption [29,35–38]. scale CDI stack containing carbon aerogels that
The materials, their uses for different applica- had been prepared under different conditions was
tions, and the methods of their preparation are used to separate salts from artificial single and
covered by a series of patents held by the Uni- multi-salt solutions. In addition, more complex
versity of California and the USA DOE (e.g., streams of an unfiltered blend of 75% Colorado
[28,39–41]) and other bodies (e.g., [42,43]). River water and 25% water from the California
Carbon aerogel electrodes in the form of a State Water Project, with and without small
CDI stack were tested by Farmer et al. [44,45] for amounts of organic material, were studied so as to
the removal of Na2SO4, Na3PO4 and Na2CO3, determine the effect of natural organic matter
from aqueous solutions and for removal of NaCl (NOM) on CDI performance. It was found that
and NaNO3 from single salt solutions. They also monovalent ions with smaller hydrated radii were
investigated the use of the same electrode stack preferentially removed from the solutions over
for the removal of ammonium perchlorate from multivalent ions. Two findings of practical impor-
aqueous solutions originating from dissolving tance emerged from this study: (1) Only a few
solid rocket propellants during decommissioning percents (14–42 m2 g!1) of the BET electrode sur-
[45]. They found that the adsorption capacity for face area (400–590 m2 g!1) were available for
larger ions such as NO3! and ClO4! is lower than electrosorption; and (2) the presence of organic
that for the smaller Cl! ions. In addition, aging material in the treated water significantly reduced
20 Y. Oren / Desalination 228 (2008) 10–29

the electrosorption capacity for ions, as a result of including pores that are not penetrable by ions
fouling of the aerogel surface and perhaps also due to size differences. As a result, the BET
the porous system. In water treatment applica- surface area is much larger than that accessible by
tions, the two effects determine the overall the hydrated ions, as clearly stated by Gabelich et
adsorption capacity per unit volume of a CDI al. [30]. As noted above, this reduces the specific
cell, as discussed further below. electrosorption capacity, which is, in turn, reflec-
Another type of electrode material was sug- ted in the size of the CDI device.
gested by Gadkaree et al. [46]. They developed a 2. The fact that hindered diffusion of ions into
family of high-surface-area carbon electrodes in extremely small pores is involved, leads to
honeycomb form. The technology is based on decreasing rates of the electrosorption/desorption
synthetic carbon precursors combined with a processes. This phenomenon becomes crucial
ceramic backbone that does not contain any when a cyclic, adsorption-regeneration procedure
binders. These structures contain up to 95% of is considered.
carbon, which can then be activated to give sur- 3. Organic fouling, biofouling and scaling by
face areas in excess of 1000 m2 g!1. The elec- sparingly soluble salts of the electrodes are dis-
trodes were optimized for ion electrosorption cussed only superficially in the literature. As
with respect to surface area and pore structure. noted above, this is a critical issue when the
The point of fouling and scaling of CDI treatment of “real” waters is considered. In the
electrode was also addressed in the US patent of absence of reasonable solutions, these phenomena
Andelman [23]. This patent makes claims for may render CDI prohibitive because of electrode
various CDI reactor configurations, including blocking, leading to loss of electrosorption capa-
“washer-type”, spiral-wound, “multiple mani- city and increase of energy consumption due to
fold”, flat type, “hexagonal road”, that may reduction of cell conductivity.
potentially reduce the fouling propensity of the
electrodes due to a reduced concentration polari-
3.2. Design and modes of operation of CDI units
zation of the foulants as a result of an intensive
mixing on the electrode surface. Electroadsorp- The first attempt to operate CDI in a pseudo-
tion in each configuration should be accompanied continuous mode was that of Oren and Soffer
by polarity reversal of the electrodes in every [8,10] in a process that was termed “electro-
charge cycle. However, although the patent chemical parametric pumping” due to its simi-
claims for a foul-resistant flow through capacitor, larity to the previously developed cyclic “thermal
it provides just a single example with a saturated parametric pumping” for separation of liquid
CaSO4 solution that does exhibit electrode scaling mixtures [47]. The electrochemical parametric
through precipitation. No discussion is provided pumping column for the separation of salt from
on the possible adverse implication of organic water consisted of two identical high-surface-area
and bio-fouling of the electrodes. carbon black electrodes separated by an elec-
In summary, the following points should be trically inert porous barrier and filled with the
emphasized: appropriate solution. Each electrochemical para-
1. Referring to the BET surface area of a metric pumping cycle comprised four consecutive
porous electrode as the “true” surface area avail- operations, as shown in Fig. 10:
able for electrosorption of ions is misleading. In C (A) Adsorption, which in a generalized sense
the BET analysis, a gas, usually nitrogen is used is a step of transfer of material to be separated
as a probe to measure pore size distribution from the solution to the electrode phase,
 Y. Oren / Desalination 228 (2008) 10–29 21

Fig. 10. Four basic stages of the electrochemical parametric pump. Reprinted with permission from the Journal of the
Electrochemical Society [8].

which is induced by changing the potential

difference across the electrodes;
C (B) Forward axial pumping of the solution,
during which the salt bound to the electrodes
remains stationary;
C (C) Desorption, which is also an electrically
induced step that results in back transfer to the
solution of exactly the same amount of salt
that had been adsorbed onto the electrodes in
the first step and;
C (D) Backward axial pumping of the solution,
now containing an excess of salt.

In each cycle, the stream moving forward is

poorer while the stream moving backwards is
richer in the separated salt. In a multi-cycle
operation, salt gradient is established along the
electrode column while diluate and concentrate
Fig. 11. Variation of the ratio between bottom (CB) and
streams could be withdrawn from its ends. top (CT) concentrations as a function of number of cycles
In the studies of Oren and Soffer [10], the in an electrochemical parametric pumping column. Salt
electrochemical parametric pump was operated concentration 0.05 M, M- ΔV/Vo = 0.5, •- ΔV/Vo = 0.25.
fully automatically in two basic modes: total ΔV and Vo are the volume step and column volume,
respectively. Reprinted with permission from the Journal
reflux and product withdrawal. Some typical
of Applied Electrochemistry [10].
results are presented in Fig. 11 for the two cases.
In addition, two models for the build-up of the phase mass transport equation with the proper
concentration profile within the column were boundary and initial conditions, while the second
elaborated. The first involved a solution of a two- was based on the mixed-cells approximation. In
22 Y. Oren / Desalination 228 (2008) 10–29

ration stage, the flow is discontinued, while the

salt is desorbed from the electrodes. In this
manner, the waste (concentrate) volume is mini-
mized. An example of an adsorption-regeneration
cycle is shown in Fig. 13. It is suggested that two
CDI units be operated simultaneously, one unit to
work in the adsorption mode while the other is
being regenerated. In this way, pseudo-con-
tinuous operation is achieved.
Farmer [47] has listed some possible important
applications of the separation process, including:
Removal of various ions, including radionuclides
from low-level radioactive wastes or from waste-
water without the generation of acid, base, or
other similar secondary wastes; treatment of
boiler water in nuclear and fossil power plants;
production of high-purity water for semi-
conductor processing; electrically driven water
softening for homes (the CDI system softens
drinking water without the need for sodium
Fig. 12. Schematic representation of an electrochemical chloride and does not require salt additions for
adsorption-regeneration cell. Reprinted with permission regeneration); removal of salt from water for
from Farmer et al. [48] and the American Chemical
agricultural irrigation; and desalination of
Society.
seawater.
An interesting version of a pseudo-continuous
both cases, electrochemical equilibrium between CDI setup was suggested by Shiue et al. [51] in
the solution and the electrode surface was which two or more CDI units are installed in a
assumed. For both models, a computer algorithm two-compartment carousel (Fig. 14). One set of
was used to simulate the multi-cycle operation of the CDI units is designated for deionization while
the electrochemical parametric pump. the other set, for regeneration. As the carousel
Following the above work, Farmer, Tran and rotates, the CDI units switch positions and func-
their coworkers designed an electroseparation set- tions. An additional option is design for energy
up based on carbon aerogel multiple electrode recovery, namely, while the deionizing CDI cells
cells [28,48–50]. The electrochemical stack, are charged by an external power supply, the
shown schematically in Fig. 12, is composed of regenerating cells release their residual electrical
separated electrode pairs held together by a titan- energy to an external accumulator for further use.
ium frame, through which the treated solution Shiue et al. [52] also filed a patent proposing
flows in a serial manner. All the even-numbered a CDI unit (or a “flow through capacitor” as
electrodes are electrically connected, as are the termed by them) in which the two electrodes are
odd-numbered electrodes. The two sets of elec- wound together with porous inert dividers in a
trodes are connected to an external power supply. concentric (spiral wound) manner. The electrode
The solution is desalted by polarizing the material comprises black magnetite (Fe3O4) nano-
electrode pairs, and regeneration is accomplished particles, as the active material, attached to
by short-circuiting the electrodes. At the regene- current collectors made of aluminum, copper,
 Y. Oren / Desalination 228 (2008) 10–29 23

Fig. 13. Three superimposed (schematic) timing charts taken from the system shown in Fig. 12. Solution: NaCl, 100 µS
cm!1. A, electrolyte conductivity; B, cell current; C, cell voltage; T, deionization-regeneration cycle. Reprinted with
permission from Farmer [47] and the US Patents Office.

3.3. Commercialization and case histories

The first attempts towards commercialization
of CDI began between 1999 and 2000, when two
US companies, Biosource, Inc. and Sabrex of
Texas, Inc., were awarded a contract from the
Defense Advanced Research Products Agency
(DARPA) to develop a prototype portable water
treatment unit based on CDI for the US Army. At
the same time, the US Far West Group, Inc. (now
Capacitive Deionization Technology Systems
Inc.) received a development contract from
Arizona Public Service (APS) to supply a CDI
prototype [termed capacitive deionization tech-
nology (CDT)] for a water purification system for
study by APS.
Fig. 14. Two CDI units in a carousel arrangement for Based on the above activity, Sabrex launched
deionization and energy recovery. Reprinted with per- the electronic water purifier (EWP), declaring it
mission from Shiue et al. [51] and the US Patents Office. to be capable of removing dissolved ions, includ-
ing iron, arsenic, chrome, perchlorate and nitrate,
and of treating brackish, sea- and wastewaters.
titanium, stainless steel or nickel foils. It is According to the developer, the unit is also suit-
claimed that this arrangement is suitable for treat- able for use as home water purifier in conjunction
ing waste streams, water purification and desali- with UV. However, no record could be found of
nation. However, no reference is given to the the fate of the portable CDI unit for military use.
question of the stability of these electrodes and Recently, two other companies have declared
the possible contaminating potential. achievements in CDI technology. ENPAR
24 Y. Oren / Desalination 228 (2008) 10–29

Technologies Inc., located in Ontario, Canada, compounds to below the target level of 10 mg L!1
announced the shipping of a small-scale DesEL total nitrogen from initial concentrations of 500
unit—a CDI-based technology—to the University and 270 mg L!1, respectively. Water recoveries
of Montreal for testing by Anjou Recherche, the up to 95% were achieved using a multistage
Research and Development Division of Veolia system.
Environment, and CDT Systems, Inc., a Texas- 3. Intake process water at Hamilton Harbor.
based company, announced the launch of a The objective of the tests was to reduce the con-
mobile CDT System. Their technology is also centration of Ca to 40 mg L!1 and that of Cl! to
presented in a document on new advanced 60 mg L!1 in the water stream. The concentration
technologies published by the Missile Defense of TDS in the treated water was decreased by
Agency, Advanced Systems Technology Appli- more than 80%. The concentration of Ca was
cation Program (MDA/AS) (2004) as a tech- reduced by 90% from 55 mg L!1 to 5.4 mg L!1,
nology suitable for water treatment in disaster while that of Cl! was reduced by more than 80%
areas. However, the commercialization status of from 121 mg/L to 21 mg L!1.
the mobile unit was not made clear in the presen- 4. City of Greater Sudbury, Lake Ramsey,
tation. It should be mentioned that all the Ontario. The tested water sample represented
technologies mentioned above make use of a typical surface water in the region, which may be
carbon aerogel as the electrode material. used as a source of drinking water. Recovery of
Case histories are important since they show treated water was in excess of 90%, and the
whether a technology is reliable under real ope- concentration of TDS was reduced by 92%. The
rating conditions. The literature on case histories waste stream represented less than 10% of vol-
regarding CDI is somewhat limited and is ume of the input sample.
restricted to the activity of ENPAR (DesEL) and 5. Mine wastewater northern Quebec,
a research group at Sandia NL [53]. Ontario, Canada. A four-stage system was devel-
ENPAR’s experience covers the following oped to recover up to 90% of the waste stream as
water treatment cases: clean water. The conductivity of the treated
1. Nitrate removal from drinking water in a water was reduced from 30 mS cm!1 to less than
Middle Eastern country. The main objective of 3.2 mS cm!1.
the work was to adapt the DesEL technology for The Sandia CDI experience is limited to the
the selective removal of nitrate from a drinking clean up of wastewater from coal-bed methane
water source with an average NO3! concentration (CBM) production. This technology was used
of 155 mg L!1 while preserving the original taste because it is considered to have potential for
of the water. The concentration of nitrate was cleaning up the large volumes of brackish water
reduced to less than 50 mg L!1 (which yet vio- extracted along with the natural gas. Sandia’s
lates the US primary drinking water standards) laboratory tests used CBM water, with TDS
and the TDS of the treated water was reduced levels of 2000–5000 ppm, which was generated
from 787 mg L!1 to 512 mg L!1, a value that in New Mexico’s San Juan Basin, where much of
approaches the USEPA secondary standard the CBM development has occurred to date.
(aesthetic standard). Removal rates ranged from 75 to 90%, with the
2. High ammonia/nitrate contaminated highest recoveries associated with the lowest
groundwater at an industrial site in southwestern TDS concentrations.
Ontario. The DesEL system was found to be It should be stressed that although the cases
effective in removing ammonia-N and nitrate-N described above appear promising, no infor-
 Y. Oren / Desalination 228 (2008) 10–29 25

mation was disclosed on the length of the tests or which have become the leading electrode material
how the electrodes behaved at long operation for CDI. As a result of the promotion of bacterial
times. growth by adsorbed organic material, the CDI
electrodes may become covered with a biofilm of
bacterial extra-cellular polymers. This biofouling
4. Discussion and conclusions will impede electric charging of the electrodes,
followed by loss of ion electrosorption efficiency
CDI could be potentially attractive for water and increase in energy consumption, as occurs in
treatment for the following reasons: membrane-based water treatment. To prevent this
1. Unlike membrane-based treatment technol- phenomenon, at least partially, the CDI influent
ogies such as reverse osmosis and nanofiltration, has to be thoroughly pretreated to reduce the con-
CDI does not require high pressures; thus equip- centrations of total organic carbon and scaling
ment such as housing, pipes, and pumps is less factors. This is in contrary to what has been stated
costly. elsewhere [55] that “Pretreatment is not a major
2. CDI requires low voltages that normally do issue…”.
not exceed several volts. Therefore, safety issues 2. As discussed above, electrode material for
are negligible. CDI has been developed to great extent in the last
3. CDI is suitable for operation in remote 15 years. However, ways has yet to be found for
areas since it can be operated on solar energy. In increasing the practically available surface area
this respect, the technology may be considered for ion electrosorption. This should certainly be
environmentally clean. done by a careful design of the pore structure of
4. As discussed above, energy can be recov- these materials, perhaps by utilizing well defined
ered rather simply, namely, by utilizing the elec- and highly oriented building elements such as
trical energy created by the regenerating unit. carbon nanotubes [57,58]. Achieving this will not
This advantage should be viewed in comparison only result in more compact CDI units due to
with modern RO energy recovery devices, which much larger specific adsorption capacity of the
are complicated and expensive. electrodes but will also bring about to faster
However, there are several clear drawbacks, response to changes imposed by the adsorption-
listed below, that yet limit the process from been regeneration cycles.
used for large-scale installations: 3. It should be also emphasized that unlike
1. Scaling by precipitation of sparingly soluble statements in some publications (e.g., [54,55])
salts, organic fouling and biofouling are ex- claiming that this technology is suitable for sea-
tremely important factors in any water treatment water treatment, desalination with CDI is limited
technology. They result in a reduction of water to water of low to medium salinity for two
quality, a decline in production rate, and a sharp reasons: (a) It may be seen from Fig. 8 and the
increase in energy consumption. As already related discussion above that for the salt concen-
mentioned above, none of the studies on CDI, tration of seawater (0.6–0.7 M as NaCl) the theo-
even those on field case studies, has taken these retical differential charge efficiency will decrease
factors into account to any significant extent. It is by a factor that is more than one order of magni-
well known that high-surface-area carbons, such tude larger than that for brackish water; and
as active charcoal and carbon aerogels, strongly (b) Fig. 15 shows a comparison of the specific
adsorb organic material, which may constitute energy consumption for the removal of salt from
nutrients supporting bacterial growth [56]. This brackish and seawater by CDI in comparison with
problem is also relevant to carbon aerogels, the currently available values for RO desali-
26 Y. Oren / Desalination 228 (2008) 10–29

such as scaling and fouling, which constitute

major drawbacks for all types of water treatment
technologies, have almost not been addressed for
CDI. As mentioned above, energy recovery,
although a less complex issue for CDI than for
other water treatment technologies, should also be
studied in depth. Thus, it is apparent that studies
of these issues along with improving the electrode
material should constitute the future milestones
on the road towards a more mature CDI
technology.

6. Symbols
A — Factor defined in Eq. (8)
CH-S — Capacity of diffused double layer,
Fig. 15. Specific energy consumption for CDI as a
function of NaCl concentration and for different voltages, µF cm!2
as compared with brackish and seawater RO desalination. CM-H — Capacity of compact double layer,
µF cm!2
Cs — Bulk concentration, M
nation. It is clear that at operative voltages the CT — Total double layer capacity, µF cm!2
energy consumption for CDI is much larger than E — Electrical potential, Volt
that for seawater RO desalination. However, the F — Faraday number, coulomb equiv!1
values for brackish water desalination by the two g1,g2 — Weight of electrodes 1 and 2, g
technologies are nearly similar. It should be L — Half the thickness of the electrodes
emphasized that energy recovery was not taken [Eqs. (9), (10)], cm
into account in Fig. 15. With current technolo- LS — Half the thickness of solution layer
gies, this factor is as high as 30–50% for seawater between electrodes [Eqs. (9), (10)],
RO. The experience with energy recovery in CDI cm
is rather limited, and the values provided in the ns — Number moles in the solution, mole
literature are extremely small (6–18%) [52].This Qd — Normalized qd [Eq. (8)]
point has to be addressed if CDI is to be com- qd — Two-electrode-cell charge coordi-
petitive with other technologies. nate, coulomb g!1
In summary, CDI has a long research history, Qe — Normalized qe [Eq. (8)]
both the pure scientific and the technological qe — Two-electrode-cell charge coordi-
aspects, and efforts have also been made towards nate, coulomb g!1
commercialization. It may be concluded that this q1,q2 — Total charge of electrodes 1 and 2,
technology may be potentially attractive for coulomb
desalination of low salinity sources such as q1!, q1+,
brackish and surface waters but not for seawater q2!, q2+ — Anion and cation charge access of
and for removing trace contaminants from pollu- electrodes 1 and 2, respectively,
ted streams, as also stated elsewhere [59]. How- coulomb
ever, it is somewhat disappointing that issues R — Gas constant, L.atm.mole!1 K!1
 Y. Oren / Desalination 228 (2008) 10–29 27

RM — Electrode resistance (Fig. 3) and R.H. Horowitz, Electrosorb process for desalting
RS — Solution resistance (Fig. 3) water, Office of Saline Water Research and Develop-
T — Absolute temperature, K ment, Progress Report No 516, US Department of the
Vo — Volume of an electrochemical para- Interior, Publication 200 056, 1970.
metric pumping column (Fig. 11), [3] A.M. Johnson, A.W. Venolia, R.G. Wilbourne and
J. Newman, The Electrosorb Process for Desalting
cm3
Water, Marquardt, Van Nuys, CA, 1970.
ΔV — Volume step of an electrochemical [4] A.M. Johnson and J. Newman, Desalting by means
parametric pump (Fig. 11), cm3 of porous carbon electrodes. J. Electrochem. Soc.,
w — Electrical charging work [Eq. (5a)], 118(3) (1971) 510–517.
joule [5] A.M. Johnson, Electric demineralization apparatus,
z — Charge number US Patent, 3,755,135, 1973.
[6] P. Delahay, Double Layer and Electrode Kinetics,
Interscience, Wiley, New York, 1966.
Greek [7] J.O'M. Bockris and A.K.N. Reddy, Modern
α — Salt diffusivity [Eqs. (9), (10)], cm 2 Electrochemistry, Vol. 2, Plenum Press, New York,
s!1 1970.
[8] Y. Oren and A. Soffer, Electrochemical parametric
κ — Specific conductivity of the solution
pumping, J. Electrochem. Soc., 125(6) (1978) 869.
[Eqs. (9), (10)], S cm!1
[9] Y. Oren and A. Soffer, Water desalting by means of
g,g0 — Dielectric constant and vacuum per- electrochemical parametric pumping. I The equili-
mittivity brium properties of a batch unit cell. J. Appl.
— Specific conductivity of the elec- Electrochem., 13 (1983) 473–487.
trode [Eqs. (9), (10), S cm!1 [10] Y. Oren and A. Soffer, Water desalination by means
μ1,μ2 — Weight fraction of electrodes 1 and 2 of electrochemical parametric pumping II:
1, 2 — Typical time scales [Eqs. (9), (10)], Separation properties of a multistage column. J.
s Appl. Electrochem., 13 (1983) 484.
φH — Potential at plane H at the electrode- [11] C. Lin, J.A. Ritter and B.N. Popov, Correlation of
solution interface, Volt double layer capacitance with the pore structure of
Δ — Difference sol-gel derived carbon xerogels. J. Electrochem.
Soc., 146(10) (1999) 3639–3643.
[12] K.-L. Yang, T.-Y. Ying, S. Yiacoumi, C. Tsouris and
Acknowledgments
E.S. Vittoratos, Electrosorption of ions from aqueous
The author is indebted to Prof. Ora Kedem for solutions by carbon aerogel: An electrical double
helful discussions and to Mrs. Vered Shapiro for layer model. Langmuir, 17 (2001) 1961–1969.
technical assistance. [13] Y. Oren, H. Tobias and A. Soffer, The electrical
double layer of carbon and graphite electrodes, part
I: Dependence on electrolyte type and concentration.
J. Electroanal. Chem., 162 (1984) 87.
References [14] Y. Oren and A. Soffer, The electrical double layer of
[1] D.D. Caudle, J.H. Tucker, J.L. Cooper, B.B. Arnold carbon and graphite electrodes, part II: Fast and slow
and A. Papastamataki, Electrochemical deminerali- charging processes. Electroanal. Chem., 186 (1985)
zation of water with carbon electrodes, Research 63.
Report, Oklahoma University Research Institute, [15] Y. Oren and A. Soffer, The electrical double layer of
1966. carbon and graphite electrodes, part III: Charge and
[2] A.M. Johnson, A.W. Venolia, J. Newman, R.G. dimensional changes at wide potential range. J.
Wilbourne, C.M. Wong, W.S. Gillam, S. Johnson Electroanal. Chem., 206 (1986) 101.
28 Y. Oren / Desalination 228 (2008) 10–29

[16] K. Kinoshita, Carbon-Electrochemical and Physico- deionization, electrochemical purification, and

chemical Properties, Wiley, New York, 1988. regeneration of electrodes, US 5,425,858, June 20,
[17] A. Ban, A. Schafer and H. Wendt, Fundamental of Assignee: The Regents of the University of
electrosorption on activated carbon for wastewater California (Oakland, CA), 1995.
treatment of industrial effluents. J. Appl. Electro- [29] J.C. Farmer, D.V. Fix, G.C. Mack, R.W. Pekala and
chem., 28 (1998) 227–236. J.F. Poco, Capacitive deionization of NaCl and
[18] E. Ayranci and B.E. Conway, Adsorption and NaNO3 solutions with carbon aerogel electrodes. J.
electrosorption of ethyl xanthate and thiocyanate Electrochem. Soc., 143 (1996) 159–169.
anions at high area carbon cloth electrodes studied by [30] C.J. Gabelich, T.D. Tran and I.H. Suffet, Electro-
in situ UV spectroscopy: Development of procedures sorption of inorganic salts from aqueous solution
for wastewater purification. Anal. Chem., 73 (2001) using carbon aerogel. Environ. Sci. Technol., 36
1181–1189. (2002) 3010–3019.
[19] M.-W. Ryoo and G. Seo, Improvement in capacitive [31] C.-M. Yang, W.-H. Choi, B.-K. Na, B.-W. Cho and
deionization function of activated cloth by titania W.I. Cho, Capacitive deionization of NaCl solution
modification. Water Res., 37 (2003) 1527–1534. with carbon aerogel – silica silica gel composite
[20] M.-W. Ryoo, J.-H. Kim and G. Seo, Role of titania electrodes. Desalination, 174 (2005) 125–133.
incorporated on activated carbon cloth for capacitive [32] R.W. Pekala, J.C. Farmer, C.T. Alviso, T.D. Tram,
deionization of NaCl solution. J. Coll. Interf. Sci., S.T. Mayer, J.M. Miller and B. Dunn, Carbon aero-
264 (2003) 414–419. gels for electrochemical applications. J. Non-
[21] E. Ayranci and B.E. Conway, Adsorption and Crystalline Solids, 225 (1998) 74–80.
electrosorption at high area carbon felt electrodes for [33] R.W. Pekala, Organic aerogels from the polycon-
wastewater purification: System evaluation with densation of resorcinol with formaldehyde. J. Mat.
inorganic S-containing anions. J. Appl. Electrochem., Sci., 24(9) (1999) 3221.
31 (2001) 257–266. [34] S.-W. Hwang and S.-H. Hyun, Capacitance control
[22] E. Ayranci and B.E. Conway, Removal of phenol, control of carbon aerogel electrodes. J. Non-Crystal-
phenoxide and chlorophenols from wastewater by line Solids, 347 (2004) 238–245.
adsorption and electrosorption at high area carbon [35] S.T. Mayer, R.W. Pekala, J.F. Poco and J.L.
felt electrodes. J. Electroanal. Chem., 513 (2001) Kaschmitter, The aerocapacitor: An electrochemical
100–110. double-layer energy storage device. J. Electrochem.
[23] M.D. Andelman, Non-fouling flow-through capa- Soc., 140 (1993) 446–451.
citor system. US Patent 5,779,891, 1998. [36] R.W. Pekala, S.T. Mayer, J.F. Poco and J.L.
[24] K. Dai, L. Shi, K/ Fang, D. Zhang and B. Yu, NaCl Kaschmitter, Structure and performance of carbon
adsorption in multi-walled carbon nanotubes. Mat. aerogel electrodes, in novel forms of carbons II:
Lett., 59 (2005) 1989–1992. MRS Symposium Proc., C.L. Renschler, D.M. Cox,
[25] K. Dai, L.-Y. Shi, J.-H. Fang, D.-S. Zhang and B.-K. J.L. Pouch and Y. Achiba, eds., Vol. 349, Material
Yu, Desalination techniques of carbon nanotube Research Society, Pittsburg, 1994.
electrodes by electric adsorption. Yingyong Kexue [37] J.C. Farmer, D.V. Fix, G.C. Mack, R.W. Pekala and
Xuebao, 23(5) (2005) 539–544 (in Chinese). J.F. Poco, Capacitive deionization of NH4ClO4
[26] D.-S. Zhang, L.-Y. Shi, J.-H. Fang and K. Dai, Study solutions with carbon aerogel electrodes. J. Appl.
on carbon nanotubes of the electroadsorption desali- Electrochem., 26 (1996) 1007–1018.
nator with flow-through capacitor. Gongneng Cailiao [38] J.C. Farmer, T.D. Tran, J.H. Richardson and D.V.
Yu Qijian Xuebao, 11(3) (2005) 287–290, 294 (in Fix, The application of carbon aerogel electrodes to
Chinese). desalination and waste treatment, Proc. Separation
[27] R.W. Pekala and C.T. Alviso, Carbon aerogels and Science and Technologies Conference, AIChE, Los
xerogels. Proc. Materials Research Society Symp., Angeles, 1997.
270 (Novel Forms of Carbon) (1992) 3–14. [39] R.W. Pekala, Melamine-formaldehyde aerogels. US
[28] J.C. Farmer, Method and apparatus for capacitive Patent 5,081,163, January 14, 1992, Assignee: The
 Y. Oren / Desalination 228 (2008) 10–29 29

United States of America as represented by the chemical purification and regeneration of electrodes,
Department of Energy, Washington, DC, 1992. US patent 6,309,532 B1, 2001.
[40] S.T. Mayer, J.L. Kaschmitter and W. Pekala, [50] T.D. Tran and D.J. Lenz, Alternative–polarity
Aquagel electrode separator for use in batteries and operation for complete regeneration of electro-
supercapacitors U.S. 5,260,855, March 28, Assignee: chemical deionization system, US patent 6,346,187
Regents of the University of California (Oakland, B1, 2002.
CA), 1995. [51] L.R. Shiue, A. Sun, C.C. Shiue, F.C. Hsieh, Y.H.
[41] S.T. Mayer; J.L. Kaschmitter and R.W. Pekala, Hsieh and J.J. Jou, Deionizers with energy recovery,
Carbon aerogel electrodes for direct energy con- US patent 6,661,643, 2003.
version, US 5,601,938, February 11, Assignee: [52] L.R. Shiue, A. Sun, C.C. Shiue, S. Wang, F.C. Yen
Regents of the University of California (Oakland, Hsieh, W.T. Lo and Y.H. Hsieh, Replacable flow-
CA), 1997. through capacitor for removing charged species from
[42] L. Ling, R. Zhang and C. Lu, Process for prepar- liquids. US patent 6,462,935,B1, 2002.
ing narrow size distribution charcoal aerogel, [53] K. Christen, Desalination technology could clean up
CN1401424, 2003. wastewater from coal-bed methane production.
[43] Z. Zhu, W. Wu and Z. Liu, Process for preparing Environ. Sci., February 1, 2006, p. 639.
carbon aerogel, CN1395988, 2003. [54] T.J. Welgemoed and C.F. Schutte, Capacitive deioni-
[44] J.C. Farmer, D.V. Fix, G.V. Mack, R.W. Pekala and zation technology: An alternative desalination solu-
J.F. Poco, Capacitive deionization with carbon tion. Desalination, 183 (2005) 327–340.
aerogel electrodes: Carbonate, sulfate and phosphate [55] R. Atlas, Purification of brackish or seawater using
removal, Proc. 27th International SAMPE Technical electronic water purification, Desalination Water
Conf., 1995, pp. 294–304. Reuse, 10(4) (2001) 10–17.
[45] K.P. Gadkaree, T. Tao and W.A. Cutler, High sur- [56] M. Kawai, J. Yamagishi, N. Yamaguchi, K. Tani and
face area carbon substrates for environmental M. Nasu, Bacterial population dynamics and com-
applications, in W.M. Kriven, ed., Proc. 64th Con- munity structure in a pharmaceutical manufacturing
ference on Glass Problems: Ceramic Engineering and water supply system determined by real-time PCR
Science, ACS, 25(3) (2004) 523–528. and PCR-denaturing gradient gel electrophoresis. J.
[46] R.H. Wilhelm, A.W. Rice and A.R. Bendelius, Appl. Microbiol., 97(6) (2004) 1123–1131.
Parametric pumping, a dynamic principle for separat- [57] B.E. Conway, J. Niu, and W.G. Pell, Electro-
ing fluid mixtures. Ind. Eng. Chem. Fund., 5(1) chemistry at high specific-area carbon electrodes:
(1966) 141–144. Applications to adsorptive purification of waters and
[47] J.C. Farmer, Method and apparatus for capacitive to charge-storage by supercapacitors. Kemija u
deionization and electrochemical purification and Industriji, 54(4) (2005) 187–198.
regeneration of electrodes, US patent 5,954,937, [58] C.-M. Yang, W.-H. Choi, B.W Cho, W.Il. Yun,
1999. K. Suk and H.-S. Han, Desalination effects of
[48] J.C. Farmer, S.M. Bahowick, J.E. Harrar, D.V. Fix, capacitive deionization process with porous carbon-
R.M. Martinelli, A.K. Vu and K.L. Carroll, Electro- nano materials. Kongop Hwahak, 15(3) (2004) 294–
sorption of chromium ions on carbon aerogel 299 (in Korean).
electrodes as a mean of remediating ground water. [59] J.C. Farmer, J.H. Richardson and D.V. Fix, Desali-
Energy Fuels, 11 (1997) 337–347. nation with carbon aerogel electrodes, UCRL-ID-
[49] T.D. Tran, J.C. Farmer and L. Murguia, Method and 125298, 1996.
apparatus for capacitive deionization and electro-