ARTICLE IN PRESS

WAT E R R E S E A R C H 42 (2008) 2605 – 2617

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Treatment of brackish produced water using carbon

aerogel-based capacitive deionization technology

Pei Xu, Jörg E. Drewes, Dean Heil, Gary Wang

Environmental Science & Engineering Division, Colorado School of Mines, Advanced Water Technology Center (AQWATEC),
Golden, CO 80401-1887, USA

art i cle info ab st rac t

Article history: Capacitive deionization (CDI) with carbon-aerogel electrodes represents a novel process in
Received 27 September 2007 desalination of brackish water and has merit due to its low fouling/scaling potential,
Received in revised form ambient operational conditions, electrostatic regeneration, and low voltage requirements.
8 January 2008 The objective of this study was to investigate the viability of CDI in treating brackish
Accepted 10 January 2008 produced water and recovering iodide from the water. Laboratory- and pilot-scale
Available online 20 January 2008 experiments were conducted to identify ion selectivity, key operational parameters,
evaluate desalination performance, and assess the challenges for its practical applications.
Keywords:
The performance of the CDI technology (CDTs) system tested was consistent throughout
Carbon aerogel
the laboratory- and field-scale experiments. Deterioration of the carbon-aerogel electrodes
Capacitive deionization
was not observed during testing. The degree of ions adsorbed to the carbon aerogel (in mol/g
Desalination
aerogel) during treatment of brackish water was dependent upon initial ion concentrations
Brackish water
in the feed water with the following selectivity I4Br4Ca4alkalinity4Mg4Na4Cl. The
Produced water
preferential sorption of iodide revealed merit to efficiently recover iodide from brackish
Water reuse
water even in the presence of dominant co-ions. The research findings derived from this
study identified parameters that merit further improvements regarding design and
operation, including modification of pore-size distribution of aerogel, development of high
capacitance and low-cost electrode materials, reducing the dead volume after regeneration
and rinsing, minimizing energy consumption, and maximizing system recovery.
& 2008 Elsevier Ltd. All rights reserved.

1. Introduction (IOGCC/ALL Consulting, 2006; ALL Consulting, 2003; Veil et al.,

Due to population growth, reductions in fresh or traditional
water supplies, growing vulnerability due to droughts,
competition for conventional water sources, development of
unconventional water supplies (such as wastewater, brackish
water, and seawater) are increasingly promoted worldwide. In
the US, a large amount of brackish water, termed produced
water, is generated during natural gas operations (Benko and
Drewes, 2008). Produced water management is critical to a
sustainable natural gas operation because of environmental
concerns, high disposal costs, and limited disposal options
2004). Treating produced water for beneficial use is being
investigated as an attractive alternative to traditional disposal
methods, such as surface discharge or deep well injection
(IOGCC/ALL Consulting, 2006; Veil et al., 2004) offering an
opportunity to augment water supplies in areas that lack
fresh water supplies. Produced water is characterized by total
dissolved solids (TDS) concentrations exceeding 3000 mg/L
and desalination is often required prior to surface discharge
and beneficial use of produced water. In addition, some
produced waters are characterized by an elevated concentra-
tion of valuable elements, such as iodide. Iodine is an

Corresponding author. Tel.: +1 303 273 3401; fax: +1 303 273 3413.
E-mail addresses: pxu@mines.edu (P. Xu), jdrewes@mines.edu (J.E. Drewes), dheil@mines.edu (D. Heil), ljwang@mines.edu (G. Wang).
0043-1354/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2008.01.011
 ARTICLE IN PRESS
2606 WAT E R R E S E A R C H
essential and rare element with increasing demand in many
industrial applications. Recovery of iodide from produced
water provides an additional economic benefit to produced
water management by off-setting treatment costs and mini-
mizing brine volume for disposal.
Today, ion-exchange (IX) resins and membrane technolo-
gies are the most commonly used methods to deionize
produced water. Produced waters with a predominant sodium
biocarbonate make-up are frequently treated by the Higgins
Loop continuous IX process (IOGCC/ALL Consulting, 2006).
The Higgins Loop IX process is achieving high water
recoveries and is cost-efficient only in early field operations
with high water volumes (as water volumes decrease in a field
development and the water to the plant diminishes, the fixed
costs drive up the cost per volume) (Matthews, 2007). A
significant cost disadvantage of IX processes is the (a) use of
hydrochloric acid or sulfuric acid for resin regeneration that
needs to be delivered to remote sites, (b) the generation of a
sodium-heavy brine that must be hauled over difficult terrain/
dirt roads, and (c) the need for wastewater injection disposal
into class III disposal wells. Since the Higgins Loop IX process
is limited to treatment of sodium bicarbonate type of water
only, and produced water streams are dominated by sodium
chloride, these waters are not cost-efficiently treatable by this
process. Membrane technologies including reverse osmosis
and electrodialysis have been applied to treat produced
waters with different water compositions, but are typically
still in pilot testing (IOGCC/ALL Consulting, 2006; Funston
et al., 2002; Spitz, 2003; Xu et al., 2008). While membrane
treatment is also a highly efficient desalination technology, it
requires pre-treatment, such as conventional coagulation/
flocculation, followed by granular media filtration or micro-
filtration/ultrafiltration. The complexity of these treatment
processes, large volumes of waste generated during opera-
tion, sensitivity of membranes, and low tolerance to system
upsets (due to well water quality changes, temperature
variations, filter plugging) have hindered a widespread
application of membrane treatment in the field of produced
water treatment. New desalination technologies need to be
evaluated to overcome the limitations of existing desalination
processes, promoting beneficial use of produced water.
Capacitive deionization (CDI) with electrodes represents a
novel process in desalination of brackish water or production
of high-grade deionized water. CDI can be viewed as an
electrochemical process that takes place in a ‘‘flow through’’
double layer capacitor. During CDI treatment, water is flowing
between the porous surface electrode pairs usually having a
potential difference of 1.0–1.6 direct current (DC) voltage. The
controlling separation mechanism of CDI is the electrostatic
attractive force between the ionic species in the solute and
the charged electrodes. The negative electrodes attract
positively charged ions such as calcium, magnesium, and
sodium. Conversely, the positive electrodes attract negatively
charged ions such as chloride, nitrate, and sulfate, producing
deionized water. After the electrodes are saturated with salts
or impurities, the electrodes are regenerated by eliminating
the electric field. The adsorbed ions are desorbed from the
surface of the electrodes, and released into a relatively small,
concentrated regenerant stream. In addition to the electro-
static attractive force, Faradaic reactions between the dis-
42 (2008) 2605– 2617
solved impurities and the electrodes also contribute to the
separation mechanisms of the electrodes. For instance,
functional groups such as carbonyl and phenolic groups on
the carbon electrode can react with cations and form
chemical bonds (Yang et al., 2001). A decrease in the specific
adsorption between the electrodes and the ions is a promis-
ing way to eliminate the affinity of the ions held on the
electrodes during the regeneration step, increase the regen-
eration efficiency, and enhance CDI performance. Carbon
aerogel also exhibited high electrosorption capacity to certain
ions such as iodide (Ying et al., 2002), thus can be used
to selectively recover valuable constituents from water
solutions.
Previous research demonstrated that the efficiency of CDI
strongly depends upon the surface property of electrodes
such as surface area, microstructure and size distribution of
pores, chemical functional groups, and adsorption properties
(Pekala et al., 1998; Gabelich et al., 2002; Yang et al., 2001;
Hou et al., 2006). Carbon aerogel is the most common
electrode material in CDI treatment. It is a special class of
organic aerogels with unique porous structure consisting of
interconnected nanometer-sized particles (3–30 nm) and
small interstitial pores (o100 nm) (Pekala et al., 1998; Tran
et al., 2003). The monolithic structure, high electrical con-
ductivity (10–100 Siemens/cm), high specific surface area
(400–1000 m2/g), and controllable pore size distribution of
carbon aerogels make them ideal as electrode materials for
electrochemical double layer capacitors and electrosorptive
processes including CDI (Farmer et al., 1996; Pekala et al.,
1998; Ying et al., 2002). Besides carbon aerogel, other
materials are also being investigated as CDI electrodes, such
as nanoporous activated carbon cloth (Oh et al., 2006), titania-
incorporated activated carbon cloth (Ryoo et al., 2003), carbon
aerogel–silica gel composite electrodes (Yang et al., 2005),
graphitized-carbon monolithic column (Hou et al., 2006), and
multiwall carbon nanotubes (Zhang et al., 2006; Bordjiba
et al., 2007).
The CDI process exhibits several advantages such as a
simple, modular design, no need for a high-pressure pump or
heater, as well as operation at ambient conditions and low
voltages. The CDI process removes ions by charge separation
and the polarity of the electrodes can be reversed between
separation cycles. The scaling problems commonly asso-
ciated with membrane processes therefore may be avoided.
Unlike ion exchange, no additional chemicals are required for
regeneration of the electrosorbent in this system. The down
time of the system due to cleaning and maintenance is
expected to be low. Because of the low fouling and scaling
potential, the pretreatment for CDI is minimum (e.g.,
cartridge filtration to remove particles), and the chemicals
required for scaling control and cleaning are optional and site
dependent. Given the fact that CDI is only effective in the
removal of charged constituents from water, post-treatment
is often required to remove non-ionic organics and pathogens
to meet product water quality standards. As an emerging
desalination technology, CDI is still in the developing stage.
Most of the studies previously reported on CDI are theoretical
and experimental in nature. Research is needed to gain a
better understanding of CDI’s potential as a practical desali-
nation technology.
 ARTICLE IN PRESS
WAT E R R E S E A R C H 42 (2008) 2605 – 2617 2607

This study investigated the viability and ion selectivity of

0.02470.006
CDI in treating brackish water generated during natural gas

2.170.5
Alkalinity 235720 mg/L as CaCO3

Sr
operation for beneficial water use and iodide recovery.
Laboratory- and pilot-scale tests were conducted to identify
key operational parameters, optimize system design and
operation, evaluate treatment performance, and critically

4.070.45 L/m mg
Specific UVA
0.09670.021
assess benefits and technical challenges of CDI in practical

0.3970.06
2.770.6

5078
applications.

Si

I
2. Materials and methods

97.8714.2
22507327
2.1. Water quality

Na
Water tested during this study was collected from a natural
gas operation site in Eastern Montana, USA. A comprehensive

0.001370.0005
water analysis was conducted to identify water quality

Total hardness 124723 mg/L

0.0770.03
constituents in produced water critical to CDI treatment and

Mn
beneficial use. Analytical methods and sampling strategies
followed protocols described in standard methods (APHA,

as CaCO3
AWWA, WEF, 2005) or published elsewhere (Xu and Drewes,
2006). Table 1 presents average concentrations of major

0.7070.41 mg/L
Oil and grease
0.45770.078
constituents (in mg/L and mmol/L) from 22 grab samples

0.6470.21
11.171.9

51717
collected over 4 years from the gas production well.

Mg

Br
The produced water was characterized as a brackish
groundwater of sodium chloride type with a pH of 8.457
0.22. The TDS concentration was quantified as 55207718 mg/L
Table 1 – Average water quality of produced water extracted from a natural gas operation site

with a specific conductance of 10,5107934 mS/cm. Besides

0.17670.028
6.971.1
sodium and chloride, major constituents with concentrations TDS 55207718 mg/L

K
of less than 100 mg/L were calcium, magnesium, bromide,
and iodide. The well water was elevated in iodide with a
concentration of 5078 mg/L. The water was classified as hard
(total hardness of 124723 mg/L as CaCO3) with an alkalinity of

10.074.3 m1
0.73870.133
29.575.3

UVA-254
235720 mg/L as CaCO3. Minor constituents with concentra-
Ca

tions of less than 10 mg/L were aluminum, boron, barium,

potassium, silicon, and strontium. No other inorganic con-
stituents were detected in the well water above the detection
limits of the analytical methods applied.
0.014670.004

93.3724.1
33067854

The total organic carbon (TOC) concentration in the

2.070.5
Ba

Cl

well water was 1.7570.20 mg/L, and organics were character-

10,5517934 mS/cm

ized by moderate to high aromaticity (specific UV absor-

Conductivity

bance (i.e., calculated as the ratio between UVA-254 and

DOC) equaled 4.070.45 L/mg m). Oil and grease were
1.7570.20 mg/L
0.35270.028

detected at low concentrations. Hydrocarbons and BTEX

3.870.3

compounds (i.e., benzene, toluene, ethylbenzene, m, p-xylene,

TOC
B

o-xylene), however, were not detected in the well water

samples.

2.2. Carbon aerogel and aquacell assembly

pH 8.4570.22

0.004170.0078
0.1170.21

The carbon aerogel was manufactured by the CDT Systems,

Al

Inc. (Dallas, TX), using a resorcinol/formaldehyde polymer-

ization and pyrolization process. Carbon veil was soaked with
a formaldehyde/resorcinol resin in a polypropylene mold. The
polymerization process was completed in a temperature-
Analytes conc.

controlled oven at 85 1C for 48 h. The sheets were removed

Conc. (mg/L)

Conc. (mg/L)

from the molds and solvent washed to eliminate any

Organics
Cations

Anions

impurities and retained water. The air-dried polymerized

Conc.
mM

mM

sheets were then stacked in a furnace at 1000 1C for 72 h in a

continuous nitrogen flow environment. After pyrolization, the

2608
carbon aerogel sheets were cut to desired size, stacked, and
included as electrodes in a specially machined polypropylene
housing. Each carbon aerogel pair was separated by con-
ductive carbon spacers and a non-conductive glass fiber
screen to prevent short-circuiting. The distance between each
pair of carbon-aerogel sheets was 2.3 mm (about 90 mil).
A stack of carbon-aerogel sheets, spacers, and their asso-
ciated electrical bus connections were pressurized in a
polypropylene housing. No epoxy was used to assemble the
aquacell. Four pairs of graphite terminals provided connec-
tion to a power supply and measurement points to check the
voltage during charging and discharging.
The carbon aerogel had average density of 0.78 g/cm3,
bulk resistivity of 20 m ohm cm, and specific capacitance
of 2 Farad/cm2 (Welgemoed and Schutte, 2005). The Bru-
nauer–Emmet–Teller (BET) surface area and pore size dis-
tribution were measured using a five-point N2 gas adsorption
technique (ASAP 2020; Micromeritics, Norcross, GA). The
average pore size and pore size distribution were determined
from desorption of N2 according to a theory developed by
Barrett, Joyner, and Halenda (BJH) (Barrett et al., 1951).
Because the carbon-aerogel material was very hydrophilic
(with no measurable contact angle using sessile drop
method), the samples were pretreated at 250 1C in a vacuum
environment for 3 h to remove adsorbed water and carbon
dioxide gas prior to measurement. The carbon-aerogel sheet
samples were characterized to have an average BET surface
area of 113 m2/g, a BJH desorption pore volume of 0.177 cm3/g,
and a BJH desorption average pore size of 4.28 nm. The
incremental and cumulative pore area of the carbon aerogel
as a function of the pore size is shown in Fig. 1. About 96.2% of
the pores in carbon aerogel had sizes between 2.65 and
7.69 nm, and 41.8% of the pores had sizes between 3.0
and 3.4 nm.
180
160
140
Cumulative pore area (m2/g)
120
100
80
60
40
20
0 0
20 40
Fig. 1 – Pore size distribution of carbon aerogel.
ARTICLE IN PRESS
WAT E R R E S E A R C H 42 (2008) 2605– 2617
2.3. Laboratory bench-scale testing
Bench-scale CDI testing units used in this study consisted of
two carbon aerogel-based cells provided by CDT Systems, Inc.
The CDTs testing unit held 1.3 L of water and had a dry
weight of 4 kg. Each cell contained 24 sheets of carbon aerogel
(12 cathodes and 12 anodes) with dimensions of 159 
286  0.81 mm3 each, weighing about 650 g of total aerogel.
Synthetic brackish water or produced water was fed into
the cells by a Cole-Parmer Masterflex L/S pump (Barrington,
IL). Prior to testing, the produced water fed to the units was
processed through a 5 mm cartridge and a 0.45 mm filter bag
(Cole-Parmer, Vernon Hills, IL). The required power to charge
the cell was provided by a BK Precision 1746A DC power
supply (Yorba Linda, CA). Bench-scale CDTs cells were tested
with a focus on system stability, regeneration effectiveness,
ion selectivity, and desalting efficiency.
Experiments were initiated with the assessment of ion
removal properties of the cells with synthetic solutions, then
progressed to the treatment of a synthetic surrogate of the
produced water, and culminated in the treatment of produced
water from the natural gas production site.
Prior to each experiment, the cells were first equilibrated
with the tested feed water by flushing the cell until the
effluent electrical conductivity (EC) was equal to the EC of the
influent. Once equilibration was achieved, a voltage of 1.3 V
was applied to the cell with the pump off to pre-charge the
cell for 10 min. The treatment phase was initiated and the
effluent was collected in 5 min intervals for EC and other
measurements. Except as indicated specifically, regeneration
was accomplished by applying reverse polarity for 15–18 min
to discharge the cell with 2 L of fresh feed water flushed
through the cell for 1 h. After the regeneration step of each
cycle, the cell was rinsed for 10 min with feed water at the
100
Cumulative
90
Incremental
80
Incremental pore area (m2/g)
70
60
50
40
30
20
10
60 80 100 120 140
Pore size (A)
 ARTICLE IN PRESS
WAT E R R E S E A R C H
same flow rate, to bring the EC in the cell to near that of the
feed water. The cell was then pre-charged to initiate the next
treatment cycle.
2.4. Pilot-scale field testing
In order to verify findings derived from bench-scale experi-
ments, a field evaluation with two pilot-scale CDTs testing
cells was conducted at a gas operation site in Montana. The
consistency of bench- and pilot-scale testing regarding design
and operation would allow some assessments regarding up-
scaling of the CDTs process to full-scale treatment plants.
The pilot-scale CDTs test units were approximately 10
times larger than the bench-scale units, containing 6.2 kg of
aerogel each. Water was initially delivered from the wellhead
into a feed water reservoir consisting of three open drums of
2000 L. The three drums allowed dissipation of methane gas
prior to contact with the CDTs cell and served as a short-term
storage reservoir. The water was then pumped out of the
reservoir through a 5.0 mm filter followed by the CDTs units
(Fig. 2). After preliminary testing, the optimum treatment
cycle was determined and consisted of 80 min of treatment
(ion removal) followed by 20 min of regeneration, and 10 min
of purging using filtered raw water prior to the next treatment
stage. The cells were charged separately by two DC power
sources at a voltage of 1.3 V and a current varying from 60 to
70 A. The exact voltage and amperage were recorded to
determine power consumption. Electrical conductivity was
measured in the inlet and outlet from each cell using in-line
electrodes, and EC data were logged by a laptop computer
using a customized Labview program (National Instruments,
Inc.). In addition, influent and effluent samples were collected
at 6 min intervals and analyzed for iodide using an ion-
selective electrode and pH, conductivity, and temperature.
The samples were also shipped to the Colorado School of
Mines laboratory for further analysis including TOC, hard-
ness, alkalinity, UV absorbance, cations, and anions.
After the field testing in Montana, additional experiments
were conducted using synthetic water to investigate the
effect of flow rate and feed water concentration on sorption
Aquacell #101 Aquacell #102
DC power supplies DC power supplies
Fig. 2 – Flow schematic of CDTs pilot-scale test system employed during field trials.
42 (2008) 2605 – 2617 2609
performance using the CDTs pilot-scale testing unit in a
laboratory setting.
3. Results and discussion
3.1. Operational stability of CDTs cells
The reproducibility and operational stability of the CDTs cells
were assessed by treating a synthetic feed water containing
5000 mg/L sodium chloride and 50 mg/L iodide added as
potassium iodide. The bench-scale cell was operated
in a single pass continuous-flow mode with a flow rate of
250 mL/min for 50 min treatment. A total of five cycles were
completed in succession. Mass balance of sodium chloride
was based on measurements of EC at 5 min intervals in
samples taken during the treatment phase, during the final
solution for the regeneration phase, and in the combined
solution from the rinse phase of each cycle. The concentra-
tion of both iodide and iodine were measured in 15 min
intervals for the effluent from the treatment phase, in the
final solution from the regeneration phase, and the combined
solution from the rinse phase.
The effluent conductivity and iodide concentration over five
consecutive sorption/desorption cycles are presented in Fig. 3.
The asymmetrical nature of the treatment and regeneration
curves is characteristic of the different kinetics associated
with ion sorption onto and desorption from the carbon-
aerogel surface. Iodide exhibited a different adsorption
kinetic as compared with chloride (in terms of conductivity).
Iodide concentrations were still at a minimum level, while
sodium chloride species were reaching feed water concentra-
tion levels in the cell effluent. The effluent conductivity
reached a minimum of 8.5 mS/cm after 15 min of each
treatment phase, after which the effluent conductivity slowly
approached feed water levels (10.2 mS/cm). Iodide concentra-
tion decreased from an initial concentration of 50 mg/L in the
influent to 17–25 mg/L in the effluent during the sorption
phase. The iodide retention was above 50%, as compared with
only 5–6% removal of sodium chloride. The average concen-
tration of dissolved iodine generated in the effluent varied
Data collection and analysis
Feed
Pump water
reservoir
5µm cartridge
filter
 ARTICLE IN PRESS
2610 WAT E R R E S E A R C H
14 140
One cycle Conductivity
12 120
10 100
EC (mS/cm)
Iodide (mg/L)
8 80
6 60
4 Iodide 40
2 20
0 0
0 100 200 300 400 500 600 700
Time (min)
Fig. 3 – Reduction in EC and iodide over five consecutive
cycles (laboratory-scale experiments: single pass
continuous-flow operation at 250 mL/min with feed water
composition of 5000 mg/L NaCl and 50 mg/L iodide).
between 1.1 and 5.7 mg/L. Given the half-cell potential for
I2/2I of 0.536 V, a portion of iodide ions was reduced to
iodine at the applied potential of 1.3 V during CDI treatment:
2HI-H2+I2. The depletion of H+ ions was consistent with the
increase of effluent pH from 8.3 to 9.4.
The final concentration of iodide in the regenerant water
varied between 71 and 89 mg/L, while the final conductivity in
the regenerant water was between 11.9 and 12.2 mS/cm. For
each cycle, 77–107% of adsorbed iodide was recovered from
the carbon aerogel during the regeneration phase. (Note: a
recovery greater than 100% was due to the desorption of
iodide from the previous run.) During the same time, the
regeneration cycle could only remove between 68% and 94%
of adsorbed sodium chloride from the carbon aerogel.
To illustrate the reproducibility between each treatment
and regeneration cycle, the relative standard deviation (RSD)
regarding minimum effluent conductivity was 1.3%. The RSD
regarding minimum effluent iodide was 14.7%. The higher
relative testing deviation of measured iodide data was likely
due to the longer sampling intervals (15 min versus 5 min for
EC). Error associated with measuring solution conductivity
and iodide is expected to be in the same range.
The potential for short-term decline in sorption capacity of
carbon aerogel was studied over five cycles. The minimum
observed sodium chloride concentration in the effluent
increased by 2.8%. The total sodium chloride adsorbed to
carbon aerogel decreased gradually by 25.6% between the first
and fifth cycle. However, the accumulated sodium chloride in
the cell over the five cycles was recovered nearly completely
after an overnight (16 h) regeneration cycle. The decline in
sorption capacity during consecutive runs is likely due to slow
kinetics of ion desorption from the carbon surface. No
appreciable decline in carbon-aerogel performance was ob-
served for iodide over the five consecutive cycles, which
might be due to the fact that electrosorption equilibrium
of iodide/iodine was not achieved over the five successive
cycles.
Throughout the laboratory- and pilot-scale testing, each
experiment was conducted by applying 2–5 consecutive
cycles under the same operational conditions. The pilot-scale
experiments treating produced water in the field exhibited
42 (2008) 2605– 2617
higher relative error than the laboratory-scale experiments,
likely due to a fluctuation of feed water quality and challenges
of operating larger-scale testing units at a remote site.
3.2. Maximum sorption capacity of bench-scale and pilot-
scale cells
To measure the maximum sorption capacity of the CDTs
cells, batch equilibrium experiments were conducted in a
continuous-recycling regime. For bench-scale cells, 2 L of
synthetic sodium chloride solution (2000 mg/L) was pumped
through the cells at a flow rate of 250 mL/min and an applied
power of 1.3 V. The conductivity of the solution in the feed
reservoir was monitored continuously until steady-state
conductivity was achieved. The total capacity of the cells
was determined by mass balance calculations. The maximum
sorption capacity of the cells was determined to be 7.0 mg
NaCl/g carbon aerogel. For the pilot-scale cells, 38.75 L of
produced water (5200 mg/L TDS, other water quality para-
meter listed in Table 1) was pumped through the cells at a
flow rate of 250 mL/min and an applied power of 1.3 V. The
maximum sorption capacity of the cells was 7.1 mg TDS/g
carbon aerogel. Thus, the bench- and pilot-scale CDTs cells
exhibited almost the same sorption capacities.
3.3. Effect of flow rate and initial feed concentration on
electrosorption performance
The effects of flow rate and initial feed concentration on
electrosorption performance were investigated using syn-
thetic sodium chloride solutions at a single-pass continuous
flow regime (Fig. 4). At the feed flow rate of 250 mL/min using
the bench-scale testing unit, the increase of dissolved salts in
feed stream resulted in an initially linear increase of
operational sorption capacity (Fig. 4(a)). After the feed
concentration was greater than 2000 mg/L, the operational
sorption capacity of the electrodes leveled off reaching 6 mg
TDS per gram of carbon aerogel. The average sorption rate
followed the same trend: increasing initially and stabilized at
0.1 mg TDS per gram of carbon aerogel per minute of
treatment time (Fig. 4(a)).
The effect of flow rate on electrosorption of CDI was
examined using the pilot-scale testing unit treating synthetic
sodium chloride solutions. As compared with the feed
concentration, the increase in feed flow rate exhibited a
limited effect on the electrosorption performance of the
electrodes (Fig. 4(b)). At the feed concentration of 500 mg/L,
the operational sorption capacity of the electrodes decreased
from 3.2 to 2.5 mg TDS per gram of carbon aerogel when the
flow rate increased from 700 to 3000 mL/min. The average
sorption rate, however, increased from 0.013 to 0.052 mg TDS
per gram carbon aerogel per minute with the increase of flow
rate. This implies that high flow rate should be employed for
industrial CDI process design and operation since fewer CDI
modules will be required to produce the same amount and
quality of desalinated water than at low flow rate.
To better understand the effect of feed flow rate and
concentration on electrosorption performance of the electro-
des, the experimental results were plotted as a function of
TDS loading—mass of TDS loaded per gram of aerogel per
 ARTICLE IN PRESS
WAT E R R E S E A R C H 42 (2008) 2605 – 2617 2611

7.0 0.20 3.5 0.10

Sorption (mg TDS/g aerogel)

Sorption (mg TDS/g aerogel)

6.0 3.0

Sorption rate (mg TDS/g

Sorption rate (mg TDS/g
0.08
5.0 0.15 2.5

aerogel/min)
aerogel/min)
2.0 0.06
4.0
0.10
3.0 1.5 0.04
2.0 0.05 1.0
0.02
1.0 Sorption Sorption rate 0.5 Sorption Sorption rate

0.0 0.00 0.0 0.00

0 2000 4000 6000 0 1000 2000 3000 4000
Feed TDS concentration (mg/L) Flow rate (mL/min)

Fig. 4 – Effect of flow rate and feed concentration on operational sorption capacity and average sorption rate of the carbon
aerogels using synthetic water. The error bars represent the standard deviation of 3–5 runs of each testing condition. (a) Feed
flow rate 250 mL/min using laboratory bench-scale unit; (b) feed TDS 500 mg/L using pilot-scale testing unit.
Sorption rate (mg TDS/g aerogel/min)

0.14 micropores, resulting in lower electrosorption capacity (Yang

0.12
0.10
0.08
0.06
0.04
Laboratory testing with NaCl solution
0.02 Field testing with produced water
0.00
0 0.5
TDS loading (mg TDS/g aerogel/min)
Fig. 5 – Effect of TDS loading on sorption rate of carbon
aerogel. The error bars represent the standard deviation of
3–5 runs of each testing condition.
minute during the sorption stage (i.e., feed flow rate  feed
concentration/mass of the electrodes) (Fig. 5). At lower feed
concentrations (i.e., lower TDS loading), the sorption
rate increased linearly. When TDS loading exceeded
0.55 g/(g aerogel min), the sorption rate leveled off at approxi-
mately 0.09 mg TDS/(g aerogel min).
Previous studies have demonstrated that the adsorption
capacity of the carbon aerogels is mainly affected by electrical
double-layer capacity due to the electrostatic attractive force
between the ions and the electrode (Yang et al., 2001; Gabelich
et al., 2002; Ying et al., 2002). In porous electrodes such as
carbon aerogel, the electrical double layers are formed inside
the pores instead of adjacent to the electrode surface. The
pores can greatly increase the effective surface area of the
porous electrodes as well as the electrical capacity. However,
when the pore size is of similar magnitude as the electrical
double layer thickness, the electrical double layer inside the
pores overlaps, resulting in a loss of electrical capacity (Yang
et al., 2001). This overlapping effect exists in microporous
(o2 nm) and part of mesoporous (2–50 nm) regions. Given the
microporous structure of the carbon aerogel (see Fig. 1), this
overlapping effect likely prevents ions from entering the
et al., 2001; Gabelich et al., 2002). The electrical double layer
thickness is primarily affected by ion solution concentration
and applied voltage. Farmer et al. (1996) reported that the
electrical double layer thicknesses were 1 and 20 nm for 0.1
and 104 M electrolytes solutions, respectively. Our study
confirmed that increasing electrolytes loading to the elec-
trode surfaces through increasing feed concentration could
reduce electrical double layer thickness, and consequently,
improve electrical capacity of the carbon aerogel. However,
the mass transfer limitations may still be in effect due to
1 1.5 2 micropores overlapping and monolayer sorption behavior of
carbon-aerogel electrodes. These effects may explain the
observed trend that the sorption rate increased with initially
increasing electrolytes loading but leveled off after the
loading rate reached 0.55 g TDS/(g aerogel min) as illustrated
in Fig. 5.
No significant correlation between iodide sorption and flow
rate was observed during produced water treatment. Within
the tested flow rate ranging from 600 to 1500 mL/min, the
retained iodide in the carbon-aerogel electrodes varied
between 0.13 and 0.15 mg I/g aerogel and the sorption rate
varied in the range of 0.004–0.006 mg I/(g aerogel min).
3.4. Ion selectivity and effect of temperature on
electrosorption performance
Ion selectivity of the carbon aerogel was examined using pre-
filtered produced water at ambient room temperature of 23 1C
and temperature of 12 1C, considering in-situ well water
conditions. Laboratory-scale batch treatment was conducted
at each temperature and four treatment cycles were com-
pleted. During the treatment phase, 6.25 L of produced water
was re-circulated at a flow rate of 250 mL/min for a period of
50 min. For the experiments conducted at 12.5 1C, the
temperature of the water discharged from the CDTs bench-
scale cell was checked periodically and ranged between 11.6
and 13.2 1C.
During CDI treatment, reduction in EC at 23 1C was very
similar to that achieved at 12.5 1C and resulted in reduction
between 6.7% and 7.0% (Fig. 6). Iodide reductions in both the
 ARTICLE IN PRESS
2612 WAT E R R E S E A R C H 42 (2008) 2605– 2617

treated water and regenerant water were also similar at the
two temperatures examined and varied between 66% and
69%, with iodide concentrations close to 15 mg/L in the
treated water and close to 90–100 mg/L in the regenerant
solution. Divalent cations, such as Ca, Mg, and Sr, exhibited
between 21.7% and 22.5% removal on average at 23 1C and
between 16.3% and 18.0% at 12.5 1C, respectively. The removal
of sodium, however, was higher at 12.5 1C than at room
temperature (13.3% versus 8.2% on average). The removal of
potassium was similar to the divalent cations, 21.3% and
15.7% at 23 and 12 1C, respectively. The carbon aerogel did not
remove any silica and boron from the produced water likely
due to the undissociated state of these constituents within
the operating pH range (8.3–8.5). Increasing the pH to above
9.5, boron will dissociate and could be removed by the CDTs
process potentially meeting water quality standards for
beneficial use such as irrigation.
The amount of adsorbed ions by carbon aerogel (in mol/g
aerogel) in treating produced water increased in the order of
Na (7.9  105)bCa (1.6  106)4Mg (9.7  107)4K (4.6  107)
for cations, and Cl (7.7  105)bBr (3.6  106)4I (2.6  106)
for anions (Table 2). This sorption trend in treating produced
80
70
23 degree C
60 12.5 degree C
Removal (%)
water is consistent with results reported by Gabelich et al.
(2002) in treating Colorado River Water.
Previous studies suggested that, for ion species having
similar initial solution concentration (in terms of molarity),
the hydrated radius might control their sorption capacity of
carbon-aerogel electrodes (Gabelich et al., 2002; Ying et al.,
2002). Monovalent ions such as sodium with smaller hydrated
radii were preferentially removed from solution over multi-
valent ions (such as calcium) on a percent or molar basis.
There was also evidence that the counter ion present could
play an important role in an individual ion’s sorption capacity
(Gabelich et al., 2002; Ying et al., 2002). In this study, it was
observed that, in a competitive multi-ionic solution, feed
concentration seemed to play a more important role in ion
uptake than the ionic hydrated radius. For instance, potas-
sium has an effective ionic radius of 3.31 Å in aqueous
solution, much smaller than sodium (3.58 Å), calcium
(4.12 Å), and magnesium (4.28 Å). Therefore, it is expected
that the sorption capacity of potassium should be higher than
sodium, calcium, and magnesium based on hydrated radius.
The experimental results, however, showed that the mass of
adsorbed ions was dependent upon the feedwater concentra-
tion (in terms of molarity) rather than the hydrated radius,
and followed the order of NabCa4Mg4K. A similar conclu-
sion can be drawn for anions. Ying et al. (2002) reported that
the sorption capacity of anions followed the trend of Br4Cl
considering similar feed concentrations (in terms of molar-
ity). This study observed the opposite trend of sorption
capacity, which corresponded to the initial feed concentration

50 (in terms of molarity) of Cl4Br (Table 2).

The removal of iodide in the produced water reached 69%,
40
much higher than other ions (Table 2). Ying et al. (2002)
30 attributed high sorption capacity of iodide to its higher partial
20 charge-transfer coefficient than chloride and bromide. The
partial charge-transfer coefficient is an indicator of how many
10
electrons can be released from the adsorbate to an adsorbent.
0 In this study, a concentration between 1.1 and 5.7 mg/L of
EC I Ca Mg Sr Na K B
iodine was detected in the CDI effluent during produced water
Fig. 6 – Effect of temperature on removal of different treatment. Besides electrostatic adsorption, iodide and iodine
constituents from produced water (batch recycling
operation using laboratory bench-scale unit).

Table 2 – Ion sorption and hydrated area of the carbon aerogel electrodes in treating produced water (batch recycling
operation using laboratory-scale unit at 23 1C)

Ions Hydrated Feed concentration Sorption capacity Removal Hydrated area

radius* (Å) (mM) (  105 mol/g aerogel) (%) (m2/g aerogel)

Na 3.58 97.83 7.9 8.1 19.15

K 3.31 0.18 0.046 21.3 0.10
Ca 4.12 0.74 0.16 22 0.51
Mg 4.28 0.46 0.097 21.7 0.34
Cl 3.32 93.26 7.7 7.6 16.05
Br 3.3 0.63 0.36 50 0.74
I 3.31 0.39 0.26 69.7 0.54

Total cations 8.20 20.09

Total anions 8.32 17.33

Note: *Reference: Nightingale (1959).

 ARTICLE IN PRESS
WAT E R R E S E A R C H 42 (2008) 2605 – 2617 2613

can have complex intermolecular interactions with the
carbon-aerogel material. Iodide and iodine can react with
various carbon groups of the aerogel, such as carbonyl and
phenolic functional groups, resulting in higher sorption
capacity.
To further investigate the effective surface area of the
carbon aerogel in produced water treatment, the amount of
aerogel surface area covered by each ion specie (m2/g aerogel)
was calculated by multiplying the molar sorption capacity of
the ions (mol/g aerogel) by Avogadro’s number (6.022  1023)
and the two-dimensional area based on the hydrated
radius (m2):
Hydrated area ¼ Sorption capacity  Avogadro’s number
 p  hydrated radius2 .
The carbon aerogel exhibited similar sorption capacities of
total cations and anions, 8.07  105 versus 8.32  105
equivalents per gram aerogel (Table 2). Based on the hydrated
radius, cations and anions covered similar surface areas of
the carbon aerogel, about 20 and 17 m2/g aerogel, respectively.
The utilized surface area was low, about 33% of the total
surface area as compared with the measured BET surface area
of 113 m2/g aerogel. Given the average pore size of the aerogel
in the range of 4 nm, the solutes could not effectively diffuse
into the inner micropores and resulted in lower effective
surface area for sorption. Gabelich et al. (2002) reported that
only 10% of the carbon aerogel was available for sorption.
3.5. Field performance of CDTs cells
The sorption and desorption pattern of various constituents
present in the produced water during the 1-stage CDI
treatment (single pass with two CDTs cells in series) is
presented in Fig. 7. The sorption and desorption trends of
each studied constituents during field trials were consistent
with bench-scale observations. The maximum removal of the
constituents followed the order of aromatic organic carbon (in
terms of UVabs at 254 nm, 83.3%)4iodide (77%)4Br (62.5%)4
Ca (40.7%)4alkalinity (as CaCO3, 40.0%)4Mg (34.3%)4Na
(18.4%)4Cl (16.0%). The adsorbed ion mass to carbon aerogel
(in mol/g aerogel) in treating the produced water followed the
order of NabCa4Mg4K for cations, and ClbBr4I for anions.
(1) The operational sorption capacities (mol/g aerogel) were
much lower during the field tests than during laboratory
experiments, likely due to low Reynolds numbers during the
operation of the pilot-scale CDTs cells. Reynolds numbers
varied between 55 and 150 for field testing, indicating laminar
flow conditions between carbon-aerogel sheets. The laminar
flow may significantly increase the electrical double layer
thickness, resulting in low operational electrosorption
capacities.
Similar to the observations made during the bench-scale
study, TOC concentrations were actually lower in the
regenerant solution than in the treated effluent (Fig. 7(a)).
Adsorption of TOC to the aerogel material during regenera-
tion when the cell is uncharged could result in potential

16 120 5000
1.6
Br I Cl
TOC (mg/L), UVA 254 (/cm)

EC UVA254 TOC
14 1.4 100
4000

Cl (mg/L)
I and Br (mg/L)

12 1.2
80
EC (mS/cm)

10 1.0 3000
8 0.8 60
6 2000
0.6 40
4 0.4
20 1000
2 0.2
0 0.0 0 0
0:00 1:12 2:24 3:36 4:48 0:00 1:12 2:24 3:36 4:48
Operating time Operating time

60 3500 10 300
Ca Mg Na
50 3000 8 250
pH and B (mg/L)
Ca and Mg (mg/L)

2500
Alkalinity (mg/L)

40 200
Na (mg/L)

6
2000
30 150
1500 4
20 100
1000
10 2
500 B pH Alkalinity 50
0 0 0 0
0:00 1:12 2:24 3:36 4:48 0:00 1:12 2:24 3:36 4:48
Operating time Operating time

Fig. 7 – Treatment of various constituents in produced water using a 1-stage unit with two cells in series at a flow rate of
560 mL/min and regeneration rate of 1900 mL/min (T ¼ 18.6–27.6 1C).
 ARTICLE IN PRESS
2614 WAT E R R E S E A R C H
electrode fouling by organic matter by clogging the pores of
carbon-aerogel material. UV absorbance exhibited similar
sorption and desorption trends as anions during treatment
(Fig. 7(a)). This indicates that aromatic organic acids behave
similarly to anions during the CDI process. The carbon
aerogel exhibited a preferential sorption of iodide from
produced water regarding percent removal. Moreover, the
iodide adsorbed onto the electrodes exhibited incomplete
desorption during regeneration in comparison to chloride and
bromide (Fig. 7(b)). The initial specific adsorption capacity of
iodide to carbon aerogel (without an applied voltage to the
electrodes) was determined to be 1.3 mg iodide/g carbon
aerogel (10.4 mmol/g carbon).
Calcium, magnesium, and sodium had the same sorption
and desorption trend during produced water treatment
(Fig. 7(c)). The carbon aerogel did not remove any silica and
retained very little boron from produced water due to the
undissociated state of these constituents within the operating
pH range (Fig. 7(d)). The maximum removal of boron was
about 16.9% during the first stage of treatment and the boron
concentration in the effluent of the third stage was still
exceeding 3 mg/L. The pH varied between 7.5 and 9.1 during
sorption and regeneration cycles (Fig. 7(d)). The carbon
aerogel also exhibited a high retention of alkalinity about
40% during produced water treatment (Fig. 7(d)).
At a higher flow rate during regeneration (1700 mL/min
using prefiltered produced water), the regeneration time was
shortened to one-third of the production time. More than 80%
of adsorbed ions were desorbed from the electrodes during
regeneration. However, a rather high volume of brine was
produced, for instance, at least the same amount of product
water for the 1-stage treatment.
To simulate the treatment efficiency of multiple CDTs
stages, the product water (effluent) from the previous stage
was collected and used as influent of the subsequent stage.
The sorption and regeneration cycles by 2-stage and 3-stage
CDTs cells were operated at a flow rate of 550 mL/min. The
cells were regenerated with filtered well water at a flow rate of
1700 mL/min. As compared with 1-stage operation, the
treatment efficiencies of salts were constant over the multiple
stages with 2.8–2.9 mg TDS/g aerogel (Table 3). Iodide recovery
during the regeneration step was effective, with iodide
concentrations increasing from 45 mg/L in the influent to
values in excess of 120 mg/L in the brine during the
regeneration of the first stage, and in excess of 80 mg/L
during the regeneration of the second stage. This was similar
Table 3 – Comparison of TDS and iodide removal by 1-, 2-, and 3-stage treatment
Stage Conductivity (mS/cm)
Infl. Effl. Sorption
min. (mg/g aerogel)
1 11.68 9.15 2.9
2 9.86 7.83 2.8
3 6.70 5.33 2.8
42 (2008) 2605– 2617
and even better than the results obtained with the bench-
scale unit. Using original well water for regeneration, the
iodide concentration in the regenerant of a third stage could
be lower than the influent as a result of dilution.
3.6. CDTs operational performance and sorption efficiency
during multi-stage treatment of brackish water (5000 mg/L)
Because of the high TDS level in the produced water, the field
experiments employing a 3-stage treatment still could not
meet the water quality standards for beneficial use (Table 3).
Successive batch experiments were conducted using the
CDTs bench-scale unit to simulate a multi-stage desalination
treatment that would reduce the TDS of produced water from
5000 mg/L to a level suitable for irrigation (i.e., 1000 mg/L) for
potential beneficial use. Approximately 6.25 L (1.65 gallons) of
10 mS/cm synthetic water (5000 mg/L NaCl and 50 mg/L iodide
as KI) was fed into the cell at a flow rate of 250 mL/min at
room temperature. Treated product water was collected and
subsequently used as feed water for the next experiment
simulating a successive multi-stage treatment. Initially,
10 mS/cm water was used as regenerant water for the first
stage and then recycled for the following experiments to
reduce regenerant volume and increase system recovery.
After regeneration, the cell was rinsed using 10 mS/cm water
for 10 min. The recovery is defined as the percentage of
product water based on the total amount of water used for
treatment, regeneration, and rinsing.
During the 10 cycles of treatment, the conductivity of the
water reduced from an initial value of 10 to 5.4 mS/cm. The
incremental reduction in EC (difference between feed water
EC and product water EC) decreased from 1.04 mS/cm for the
first cycle to 0.22 mS/cm by the ninth cycle, apparently
associated with deficiency of regeneration and rinsing.
Improvements in the increment of EC reduction were
observed after overnight regeneration. Iodide concentration
in the regenerant water was 74.4 mg/L after the first cycle, and
then gradually decreased during the first day to 66.4 mg/L in
the recycled regenerant at the end of the fourth cycle. Iodide
concentration in the regenerant then increased to 89.9 mg/L
after the overnight regeneration, and then decreased to
50 mg/L at the end of the tenth cycle due to adsorption of
iodide onto the carbon aerogel. It is evident that recycling of
regenerant water is detrimental to iodide recovery from
produced water due to the high affinity of iodide ions to
carbon aerogel.
Iodide (mg/L)
Infl. Effl. Regenerant Sorption
min. (max.) (mg/g aerogel)
45.0 10 123 0.10
13.5 3.7 85 0.03
6.0 3.5 42–50 0.005
 ARTICLE IN PRESS
WAT E R R E S E A R C H
After the tenth cycle, the EC of the feed water was
decreased in steps to approximately 4.0, 3.0, and 2.0 mS/cm
in order to accelerate completion of the experiments.
This was deemed acceptable because the iodide concentra-
tion in the treated water had stabilized at approximately
10–13 mg/L, and iodide concentration in the regenerant
water was near 50 mg/L. Treatment of the 2.0 mS/cm water
yielded a very low value for the increment of EC reduction of
only 0.08 mS/cm. Consequently, with the given experimental
set-up at least 10 cycles would be required to decrease EC
from 2.0 to 1.0 mS/cm. Because the regenerant water was
reused, the EC in the regenerant stage reached values as high
as 13.33 mS/cm. However, after the treated water EC was
below 5.0 mS/cm, EC in the regenerant water tended to
remain near a value of 10.0 mS/cm. This is probably due to
carryover of relatively low EC water remaining in the cell after
the treatment step.
In order to estimate the effect of carryover from regenera-
tion on the treatment of water in the following cycle, the
reduction in operational adsorption capacity was calculated.
It is assumed that the maximum adsorption capacity avail-
able was 5.54 mg/g aerogel based on the results from the first
cycle. The residual concentration of sodium chloride in the
cell after the rinse stage was slightly above 5000 mg/L
(EC ¼ 10.0 mg/L) throughout the entire range of initial EC
values because produced water was used for rinsing. After
draining the cell following the end of the rinse step, at least
800 mL of this water remained in the cell. Therefore, the
adsorption capacity needed to reduce the EC of the residual
water in the cell from 10.0 mS/cm to match the EC of the feed
water at a given stage of treatment was calculated as
0:80 Lð5000  Ci Þ
Mlost ¼ , The efficiency and production capacity of the system,
650 g
where Mlost is the mg of adsorption capacity per g aerogel lost
due to carryover, and Ci is the initial concentration of NaCl in
mg/L at a given stage of the treatment process.
Consequently, the percentage of the adsorption capacity
available for treating water at a given initial sodium chloride
concentration is
5:54  Mlost
Mavail ð%Þ ¼  100.
5:54
These calculations suggest that the percentage of available
adsorption capacity decreases from 100% at initial EC of
10.0 mS/cm to only 11% for an initial EC of 2 mS/cm. The
predicted 10-fold decrease in available adsorption capacity is
consistent with the experimentally measured reduction in EC
increment from approximately 1 mS/cm at an initial EC of 10
to 0.08 mS/cm at an initial EC of 2 mS/cm. The very low
increments of EC reduction at low initial EC levels can be
accounted for based on consideration of the carryover of ions
from regeneration.
To minimize the carryover problem and improve sorption
capacity, 1 mS/cm water was used in the rinse step. As
compared to rinsing with 10 mS/cm water, the operational
sorption capacity of the CDTs cell increased significantly
using 1 mS/cm water for rinsing (Table 4). The EC of the rinse
water was increased from 1.0 to 3.5 mS/cm, indicating that
the rinse water could be reused when treating water in the
42 (2008) 2605 – 2617 2615
Table 4 – Treatment of 10.0 mS/cm water (5000 mg/L NaCl
plus 50 mg/L I) using 1.0 mS/cm water to rinse cell
between treatment cycles
Cycle EC Increment Increment
(mS/cm) (mS/cm) (mS/cm)
(1 mS/cm rinse) (10 mS/cm rinse)
1 10.0648.70 1.36 1.04
2 7.1046.16 0.94 0.57
3 5.1244.46 0.66 0.42
4 3.0042.63 0.37 0.15
early stages of the process. Iodide recovery was poor due to
rinsing the cell with low-iodide (I ¼ 10 mg/L)-treated water.
3.7. Benefits and technical challenges
The performance of the CDTs system was maintained at a
consistent level throughout the duration of the field trials.
Scaling or fouling of the aerogel electrodes did not occur
during treatment of this water source. Although the system
as configured did not reduce TDS to the extent needed to
meet the targeted water quality standards, the scale at which
the test was conducted would allow an extrapolation to
determine the number of cells and configuration needed to
meet these standards. CDI technology (CDT) was revealed to
be promising for iodide recovery from the produced water
examined as well as other sources that contain iodide. CDI
needs only simple pre-treatment such as cartridge filtration.
however, need to be improved before CDT will become
economically feasible to treat water of this level of salinity.
The results of the field trials allowed identifying critical
equipment parameter and operational conditions that merit
further improvement. Development of high capacitance and
low cost of electrode materials is a key factor for a successful
and economic commercial application of CDI. By controlling
the pore size distribution of carbon aerogels, the electrosorp-
tion efficiency of the electrode material can be improved. The
slow kinetics of transport of ions into and out of the highly
porous electrodes should also be addressed. Several efforts
have been made to improve the performance of the capacitor.
Andelman and Walker (2004) used charge-blocking layers
(essentially IX membranes) on the surface of the electrodes to
limit migration of ions into and out of the electrodes and
decrease the transition times between purifying and purging
cycles.
The recovery of the CDT process is low for produced water
treatment (about 25–33%) due to the substantial amount of
feed water used for electrode regeneration and rinsing. The
disposal of high volume of brine generated during CDI may be
a potential barrier for its practical application. The enhance-
ment of recovery can be achieved by developing faster
charging and discharging cycles for the electrodes, reducing
the volume and time for regeneration and rinsing, as well as
reducing dead volume (residual water volume) after regen-
eration.
 ARTICLE IN PRESS
2616 WAT E R R E S E A R C H
The minimum energy consumption was 0.21 Wh/g salts
removed at applied current of 15 A for both sorption and
regeneration. The power consumption of CDI in treating
produced water could be as high as 0.84 and 0.95 kWh/m3
(3.18 and 3.58 kWh/kgal) to meet water quality standards of
1000 and 500 mg TDS/L, respectively. Although applying high
current could shorten the time for electrode charging and
discharging, the specific power consumption per gram of
removed salt is becoming excessively high. Approximately
half of the energy was used for discharging the electrodes
during the regeneration phase. It implies that the current CDT
process is not energy efficient in treating high-salinity water.
Energy consumption can be reduced if energy is retrieved
during the process of electrode discharge. For example, Shiue
et al. (2005) reported a significantly improved efficiency of CDI
by using spiral wound electrodes (activated carbon coated on
titanium foil) cartridge in combination with on-line electro-
lytic ozonation. Ozone was co-produced by low-voltage
electrolysis of water, and the electrolytic ozone could be
placed either before or after the CDI. The electricity retrieved
at the discharge of CDI operation could be reused for the
production of ozone. Water recovery of flow-through capa-
citor increased to more than 90%, and energy recovery rate
was exceeding 30%. Non-charged species including organics
and pathogens were reduced through ozonation.
4. Conclusions
 The performance of the CDTs system was consistent
throughout the bench-scale experiments and duration of
the field trials. Deterioration of the carbon-aerogel electro-
des or electrode fouling was not observed throughout the
course of the study, but might occur due to adsorption of
organics during regeneration during times when the
electrodes are uncharged.
 The quantity of ions adsorbed to the carbon aerogel (in
mol/g aerogel) in treating the brackish water was depen-
dent upon feed water concentrations, increasing in the
order of NabCa4Mg4K for cations, and ClbBr4IClb
Br4I for anions.
 The maximum percentage of removal, however, followed a
different trend in the order of organic acids (in terms of
UVabs at 254 nm) 4I4Br4Ca4alkalinity4Mg4Na4Cl.
 Due to the microporous nature of the aerogel structure
(average pore size of 4.28 nm), the actual available surface
area for ion sorption was about 33% of the measured BET
area. Therefore, modifying pore size distribution is critical
to improve sorption performance of the carbon-aerogel
electrodes.
 CDI could be an alternative for brackish water desalina-
tion, but the sorption capacity of the electrodes and
operational performance merit further improvement.
Acknowledgment
The authors thank the US Bureau of Reclamation (BOR) for its
financial, technical, and administrative assistance in funding
and managing the project through which this information
42 (2008) 2605– 2617
was derived. The authors also thank Paul Mendell with
Mendell Energy Inc., for his financial and technical support,
and CDT Systems Inc., for providing CDTs testing units and
technical support.
R E F E R E N C E S
ALL Consulting, 2003. Handbook on Coal Bed Methane Produced
Water: Management and Beneficial Use Alternatives. Tulsa,
Oklahoma /http://www.all-llc.com/CBM/pdf/CBMBU/
CBM%20BU%20Screen.pdfS.
Andelman, M.D., Walker, G.S., 2004. US Patent 6,709,560.
American Public Health Association (APHA), American Water
Works Association (AWWA), and Water Environment Federa-
tion (WEF), 2005. Standards Methods for the Examination of
Water and Wastewater, 21st ed.
Barrett, E.P., Joyner, L.G., Halenda, P.P., 1951. The determination of
pore volume and area distributions in porous substances. I.
Computations from Nitrogen isotherms. J. Am. Chem. Soc. 73,
373.
Benko, K., Drewes, J.E., 2008. Produced water in the Western
United States: geographical distribution, occurrence, and
composition. Environ. Eng. Sci. in press.
Bordjiba, T., Mohamedi, M., Dao, L.H., 2007. Synthesis and
electrochemical capacitance of binderless nanocomposite
electrodes formed by dispersion of carbon nanotubes and
carbon aerogels. J. Power Sources 172 (2), 991–998.
Farmer, J.C., Fix, D.V., Mack, G.V., Pekala, R.W., Poco, J.F.,
1996. Capacitive deionization of NH4ClO4 solutions with
carbon aerogel electrodes. J. Appl. Electrochem. 26,
1007–1018.
Funston, R., Ganesh, R., Leong, L.Y.C., 2002. Evaluation of
technical and economic feasibility of treating oilfield pro-
duced water to create a ‘‘New’’ water resource. In: Proceedings
of the 2002 GWPC Produced Water Conference, Colorado
Spring, CO, October 16–17, 2002.
Gabelich, C.J., Tran, T.D., Suffet, I.H.M., 2002. Electrosorption of
inorganic salts from aqueous solution using carbon aerogels.
Environ. Sci. Technol. 36, 3010–3019.
Hou, C.-H., Liang, C., Yiacoumi, S., Dai, S., Tsouris, C., 2006.
Electrosorption capacitance of nanostructured carbon-based
materials. J. Colloid Interface Sci. 302, 54–61.
IOGCC/ALL Consulting, 2006. A guide to practical management of
produced water from onshore oil and gas operations in the
United States. A final report prepared for US Department of
Energy, Tulsa, Oklahoma /http://www.all-llc.com/iogcc/
prodwtr/ProjInfo.htmS.
Matthews, M., 2007. Coalbed methane producers have some new
options. McGraw-Hill Construction/ENR 258, 19.
Nightingale Jr., E.R., 1959. Phenomenological theory of ion
salvation: effective radii of hydrated ions. J. Phys. Chem. 63,
1381–1387.
Oh, H.-J., Lee, J.-H., Ahn, H.-J., Joeng, Y., Kim, Y.-J., Chi, C.-S.,
2006. Nanoporous activated carbon cloth for capacitive
deionization of aqueous solution. Thin Solid Films 515 (1),
220–225.
Pekala, R.W., Farmer, J.C., Alviso, C.V., Tran, T.D., Mayer, S.T.,
Miller, J.M., Dunn, B., 1998. Carbon aerogels for electrochemi-
cal applications. J. Non-Cryst Solids 225, 74–80.
Ryoo, M.-W., Kim, J.-H., Seo, G., 2003. Role of titania incorporated
on activated carbon cloth for capacitive deionization of NaCl
solution. J. Colloid Interface Sci. 264, 414–419.
Shiue, L.-B., Hung, W.-T., Chou, G.-N., Wang, S.-Y., Chung, H.-C.,
Chiu, I.-C., 2005. Energy-effective desalination with online
sterilization using capacitors. In: Proceedings of the IDA World
Congress, Singapore, September 2005.
 ARTICLE IN PRESS
WAT E R R E S E A R C H
Spitz, L.R., 2003. Advanced mobile water treatment systems for
treating coal bed methane produced water. Paper presented at
the 2003 Ground Water Protection Council Meeting.
Tran, T.D., Farmer, J.C., Pekala, R.W., 2003. Carbon aerogels and
their applications in supercapacitors and electrosorption
processes /http://www.vacets.org/vtic97/tdtran.htmS.
Veil, J., Puder, M., Elcock, D., 2004. A White paper describing
produced water from production of crude oil, natural gas, and
coal bed methane. A final report prepared for US Department of
Energy, Argonne National Laboratory, Chicago, Illinois /http://
www.netl.doe.gov/publications/oil_pubs/prodwaterpaper.pdfS.
Welgemoed, T.J., Schutte, C.F., 2005. Capacitive deionization
technology: an alternative desalination solution. Desalina-
tion 183 (1–3), 327–340.
Xu, P., Drewes, J.E., 2006. Viability of nanofiltration and ultra low
pressure reverse osmosis membranes for multi-beneficial use
of methane produced water. Sep. Purif. Technol. 52, 67–76.
42 (2008) 2605 – 2617 2617
Xu, P., Drewes, J.E., Heil, D., 2008. Beneficial use of co-produced
water through membrane treatment: technical–economic
assessment. Desalination, in press.
Yang, K.-L., Ying, T.-Y., Yiacoumi, S., Tsouris, C., Vittoratos, E.S.,
2001. Electrosorption of ions from aqueous solutions by
carbon aerogel: an electrical double layer model. Langmuir 17,
1961–1969.
Yang, C.-Y., Choi, W.-H., Na, B.-K., Cho, B.W., Cho, W.I.,
2005. Capacitive deionizaton of NaCl solution with carbon
aerogel-silica gel composite electrodes. Desalination 174,
125–133.
Ying, T.-Y., Yang, K.-L., Yiacoumi, S., Tsouris, C., 2002. Electro-
sorption of ions from aqueous solutions by nanostructured
carbon aerogel. J. Colloids Interface Sci. 250, 18–27.
Zhang, D., Shi, L., Fang, J., Dai, K., Li, X., 2006. Preparation and
desalination performance of multiwall carbon nanotunes.
Mater. Chem. Phys. 97, 415–419.