Desalination 467 (2019) 257–262

Contents lists available at ScienceDirect

Desalination
journal homepage: www.elsevier.com/locate/desal

Desalination via chemical energy: An electrodialysis cell driven by T

spontaneous electrode reactions
⁎
S. Abu Khallaa, M.E. Sussa,b,
a
Grand Technion Energy Program (GTEP), Technion – Israel Institute of Technology, Israel
b
Faculty of Mechanical Engineering, Technion – Israel Institute of Technology, Israel

A B S T R A C T

The form of the energy input for widely investigated desalination technologies include thermal energy (distillation), mechanical work (reverse osmosis) or electrical
energy (electrodialysis and capacitive deionization). We here propose and characterize an electrodialysis-type desalination cell which is driven instead by sponta-
neous redox reactions occurring at electrodes. Thus, this system utilizes solely a chemical energy input to perform desalination, and requires no electricity input,
instead producing electricity while desalinating. With our custom-built prototype system based on high performance zinc-bromine chemistry, we demonstrate the
desalination of feedwater with an initial salinity of ~30 g/L while simultaneously generating up to 23.5 kWh of electricity per m3 of desalted water. Further, we show
our prototype cell can recover up to 85% of the input chemical energy as electricity during operation without needing a recovery device. The net energy usage,
defined as chemical energy input minus electricity output, was measured to be 3.9 kWh/m3 when desalting to near-zero concentration at 2 mA/cm2 with our first-
generation cell. Our proposed concept decouples reactant production from its usage in the cell, and we show that this decoupling can potentially lead to net negative
operating costs when low-cost reactants are employed.

1. Introduction cycling. Generally, desalination occurs during cell charging, and during
discharging brine is created and electrical energy can be recovered. Like
Currently, the most utilized technology for water desalination is energy storage systems, such cells are net users of electricity over a full
reverse osmosis (RO). Sea water RO plants deliver on the order of 10 charge/discharge cycle, although the electrical energy requirements for
million m3 of treated water per day, via pressurizing the feedwater such systems have shown significant promise for brackish water desa-
above its osmotic pressure of ~25 bar in order to drive water through a lination [12,16,25,26].
semi-permeable membrane that rejects salts [1,2]. As water scarcity An alternative strategy is to develop desalination systems based on
increasingly affects more diverse locales [3], the requirements of a electrochemical cells performing continuous energy conversion rather
technological solution to combat scarcity are becoming more diverse, than cyclic energy storage. The cell we propose and characterize here is
and technologies with novel functionalities should be explored. Many shown schematically in Fig. 1a, and desalinates water by leveraging a
alternatives to RO have been investigated. Electrodialysis (ED) is an chemical-to-electrical energy conversion process. Taken as a black box
electrochemical system employing alternating anion and cation ex- (Fig. 1a), our cell requires solely chemical energy as input and no
change membranes, and has been investigated for several decades as a electricity, and desalinates water while simultaneously producing
desalination technology. In an ED cell, electricity is applied to drive electricity. As with an ED cell, our cell consists of one anion and one
electrochemical reactions at electrodes, and the resulting steady ionic cation exchange membrane which sandwiches a desalination channel
current enables salt ion transport from a diluate channel into an ad- fed with feedwater. On the opposite side of the anion exchange mem-
jacent brine channel. While ED cells are highly scalable and can desa- brane is an anode and anolyte, while a cathode and catholyte are placed
linate at low (sub-osmotic) pressures, they are not as energy efficient as opposite to the CEM. During operation, the reductant present in the
RO when performing sea water desalination [4]. In recent years, there anolyte (here Zn) and oxidant present in the catholyte (here Br2) react
has been a surge of interest in emerging electrochemical systems to- spontaneously at the anode and cathode surfaces, respectively, pro-
wards water desalination, such as capacitive deionization [5,6], desa- viding an electric current between the anode and cathode which can be
lination batteries [7–13], and desalination redox flow batteries [14,15]. delivered to a load. The reductant and oxidant are carefully chosen to
Unlike ED, such systems are generally adapted from energy storage ensure they are electrostatically inhibited from entering the desalina-
cells, and operate in a cyclic fashion by undergoing charge/discharge tion channel by the AEM and CEM, respectively. The half-reactions also

⁎
Corresponding author.
E-mail address: [email protected] (M.E. Suss).

https://doi.org/10.1016/j.desal.2019.04.031
Received 3 February 2019; Received in revised form 21 April 2019; Accepted 27 April 2019
Available online 28 May 2019
0011-9164/ © 2019 Published by Elsevier B.V.
S. Abu Khalla and M.E. Suss Desalination 467 (2019) 257–262

Fig. 1. a) Schematic illustrating the concept and working principle of the electrodialysis-type cell tested here. Taken as a black box, the cell requires only chemical
energy as input, and outputs electricity while desalting water. Inside the black box, the cell undergoes spontaneous half-cell reactions at the zinc metal anode and
bromine cathode, resulting in generated electricity. The spontaneous reactions drive an ionic current through the desalination channel, transporting sodium ions
through a cation exchange membrane (CEM), and chloride ions through an anion exchange membrane (AEM). b) A picture of our custom-fabricated ED prototype
with 10 cm2 active area.

give rise to a spontaneous ionic current through the cell, which drives to form negatively charged tribromide (Br3−), which is the dominant
ion removal from the desalination channel. During operation, the feed oxidant species [17,18]. Thus, such a chemistry is compatible with the
is desalted as positively charged ions migrate from the desalination cell concept shown in Fig. 1a, as the zinc ion transport is inhibited by
channel to the catholyte and negatively charged ions to the anolyte. The the anion exchange membrane and Br− and Br3− ions by the cation
main differentiator from typical ED cells and the cell characterized here exchange membrane. While we here used zinc‑bromine to achieve high
is that typical ED cells operate in electrolytic mode and thus require desalination performance, numerous alternative chemistries are com-
input electricity, but our cell operates in galvanic mode and delivers patible, and in the future chemistries can be chosen with an emphasis
electricity. Compared to other emerging electrochemical desalination on optimizing either desalination energy efficiency, operational costs,
systems such as CDI and desalination batteries, which desalinate via or safety and water quality (see Fig. 3c for an analysis based on op-
two-stage charge/discharge processes and thus have limited salt storage erational cost). We made the anolyte by mixing zinc chloride salt into
capacity, our cell can instead be re-fueled, allowing for continuous deionized water to form 1 L of 2 M ZnCl2. For the catholyte, we mixed
desalination without interruption or capacity limitations. bromine and sodium bromide salt into deionized water to 1 L of 1 M
We provide an experimental characterization of the desalination Br2/2 M NaBr. Thus, during operation of our ED cell, zinc metal was
performance, electricity generation, and energy efficiency of our oxidized to zinc ions and tribromide ions reduced to bromide ions. The
custom-built prototype cell based on high-performance ZneBr2 redox relatively large volume of anolyte and catholyte was chosen in order to
chemistry. When operated at 2 mA/cm2 extracted current density, our maintain an approximately constant state of charge throughout a series
cell demonstrates the near complete desalination of water with an in- of desalination experiments. The feedwater was prepared by adding
itial salinity approximately that of sea water, while generating up to NaCl to 30 mL of deionized water to create a 500 mM (29.22 g/L) NaCl
23.5 kWh/m3 of electricity while desalting. We show that our prototype solution.
cell can recover up to 85% of the input chemical energy as electricity The cell consisted of a custom-milled planar graphite cathode,
during operation without needing a recovery device, and that the net 0.127 mm thick titanium sheet current collector at the cathode side and
energy usage, defined as chemical energy input minus electricity 0.62 mm thick zinc sheet for the anode (Alfa Aesar, United Kingdom),
output, was as low as 3.9 kWh/m3 when operated at 2 mA/cm2. Viton rubber gaskets, and Neosepta ion exchange membranes (Neosepta
Desalination at higher currents is also demonstrated, up to 16 mA/cm2, AMX and CMX, Tokuyama, Japan). The endplates used were acrylic for
which allowed for significantly higher salt removal rates but at the cost the zinc side and PVDF for the bromine side. The anolyte and catholyte
of increased input chemical energy per m3 desalted. For the ED concept flow channels were made by cutting 6.7 by 1.5 cm channels into 1 mm
developed here, we can decouple the production of the needed redox thick Viton rubber gaskets, and stacking two gaskets per flow channel
active chemicals from their use in the cell. We provide an analysis of the (flow channel thickness was 2 mm). Thus the active area of our cell
base operational costs and revenues for various potential redox che- used for current normalization was that of the anolyte and catholyte
mistries, showing that driving such cells with low-cost chemistries can flow channels, which was 10 cm2. The desalination channel was cut
potentially lead to net negative operating costs. into a single 1 mm thick Viton rubber gasket with the channel dimen-
sions of 12.5 by 1.5 cm. The desalination channel was longer than the
anolyte and catholyte channels in order to allow for convenient injec-
2. Experimental methods tion and removal of the feedwater from the cell. The cell was sealed
with fourteen M4, 48 mm long stainless steel bolts, which were plastic
The custom-built prototype cell tested here is shown in Fig. 1b, and wrapped to prevent internal short circuiting. The connection to an ex-
utilized zinc‑bromine chemistry. Such a chemistry is considered to be ternal load was made through tabs on the titanium and zinc metal
high-performance because it allows for kinetically fast electrochemical sheets.
reactions to minimize activation voltage losses at electrodes, and en- All three flows, anolyte, catholyte and feedwater were recirculated
ables use of highly conductive electrolytes to minimize Ohmic voltage through the cell using peristaltic pumps (Masterflex, Cole Parmer,
losses (around 100 mS/cm). Bromine (Br2) and bromide (Br−) complex

258
S. Abu Khalla and M.E. Suss Desalination 467 (2019) 257–262

Fig. 2. a) A polarization curve showing measured equilibrium cell voltage of our ED cell versus extracted current. b) Measured cell voltage during constant current
experiments at various currents from 2 to 16 mA/cm2. c) Measured middle channel effluent concentration and cumulative electricity production by the cell during
the experiment at 2 mA/cm2 shown in b). d) Measured middle channel effluent concentration and cumulative electricity production during the experiment at 16 mA/
cm2 shown in b).

USA), at a flow rate of 0.5 mL/min for the feedwater, 1.5 mL/min for current at about 30 mA/cm2. Such a limiting current density is roughly
the anolyte, and 2 mL/min for the catholyte. During polarization curve an order of magnitude lower than that predicted and observed for
measurements (Fig. 2a), a constant current was delivered from the cell galvanic electrochemical cells with planar electrodes which are limited
to a potentiostat in steps of 1 mA/cm2 (Biologic VSP, France), with a by bromine reduction (for similar bromine concentration) [18,19]. This
dwell time of at least 5 min per step to attain the equilibrium cell suggests that the limiting current observed may not be due to reactant
voltage. Also, for polarization curve measurements only, the feedwater exhaustion at the cathode, but rather that salt concentration in the
volume used was 1 L instead of 30 mL to achieve a steady-state (con- desalination channel approaches zero at the membrane surface. In
stant salt concentration profile) in the desalination channel. For the Fig. 2b–d, we show results of constant current experiments. Fig. 2b
constant current experiments (Fig. 2b–d), the ED cell delivered currents shows the measured voltage from our cell when extracting currents
ranging from 2 to 16 mA/cm2, while flow was maintained in all from 2 to 16 mA/cm2. As can be seen, setting a higher current resulted
channels. During these experiments, we measured both the resulting in a lower cell voltage, as is typical of fuel cell performance [20]. Each
cell voltage and the effluent conductivity leaving the desalination experiment was run until the cell voltage fell to 0.5 V, which was at-
channel (Tracedec, Innovative Sensor Technologies GmbH, Austria). tained via rapid voltage decay at the end of all curves shown in Fig. 2b,
Conductivity was converted to NaCl concentration via a prepared ca- a feature characteristic of a mass transport limitation. At the lower
libration curve. Constant current experiments continued until cell vol- currents tested (2 and 4 mA/cm2), a relatively stable cell voltage was
tage reached below 0.5 V, at which point the experiment ended. obtained, where cell voltage decreases slowly as the middle channel is
desalted. The relatively stable voltage lasts for several hours until the
3. Results and discussion cell voltage suddenly drops (at ~24 h for 2 mA/cm2 and ~10 h for
4 mA/cm2). The sudden drop in cell voltage coincides with the near
We here describe results of experiments performed with our pro- complete desalination of the middle channel, see Fig. 2c. When running
totype cell, shown schematically in Fig. 1a and pictured in Fig. 1b. In the cell at the highest current of 16 mA/cm2, the cell died roughly 1.5 h
Fig. 2a, we show the obtained polarization curve, by plotting measured into the experiment, without reaching a stable voltage.
equilibrium cell voltage versus extracted current density. From this In Fig. 2c, we show the measured concentration of the effluent
figure, we can observe that the cell's open circuit voltage (OCV) is leaving the desalination channel of the cell, and the cumulative elec-
measured to be about 1.74 V. A distinct activation region can be ob- tricity produced by the cell during an experiment at an extracted cur-
served at lower currents, until about 5 mA/cm2, after which a linear rent density of 2 mA/cm2. The conductivity data was taken simulta-
Ohmic region is observed until about 25 mA/cm2. At higher currents, neously to the voltage data of Fig. 2b. The initial concentration of the
significant mass transport limitations are observed, with a limiting salty water pumped through the middle channel was 500 mM NaCl,

259
S. Abu Khalla and M.E. Suss Desalination 467 (2019) 257–262

which is approximately the salinity level of seawater. At the start of the RO), as electricity is naturally outputted by the ED cell during opera-
experiment, a quick reduction to ~470 mM is observed within the first tion. Recovery efficiency decreases with increasing current density, as is
10 min, followed by slower and fairly steady salt removal as the salt expected as higher currents represent further departures from equili-
water is recirculated through the desalination channel of the cell. Salt brium (from OCV), and drops to 60% at 16 mA/cm2.
removal continued until the salt concentration reached about 10 mM at To gain additional insight, we can compare the energy used to drive
24 h, at which point the cell voltage fell quickly to 0.5 V and the ex- desalination in our ED cell, (kWh/m3)chem - (kWh/m3)elect, to the cal-
periment ended (see Fig. 2b). Thus, the total salt concentration reduc- culated thermodynamic minimum energy requirement for desalting a
tion was nearly two orders of magnitude during the 2 mA/cm2 ex- 500 mM NaCl feedwater, ΔGsep. The energy used to drive desalination is
periment. Electricity was produced by the cell during desalination, measured to be 3.9 kWh/m3 at 2 mA/cm2, and increases with applied
which can be calculated by the integral in time of cell voltage shown in current to 5.2 kWh/m3 at 16 mA/cm2. The minimum energy, ΔGsep, is
Fig. 2b multiplied by the extracted current. In Fig. 2c, we show the calculated assuming a thermodynamically reversible separation pro-
cumulative electricity produced by the cell, which is roughly linear in cess, and thus is unachievable in practice and independent of the
time due to the near constant voltage of the cell during most of the technology used to perform the desalination. The calculation was per-
experiment (Fig. 2b). The total electricity delivered by the cell during formed using previously developed expressions for minimum energy
desalination was ~23.5 kWh/m3 of desalted water. In Fig. 2d, we show [23], and using 1.5% water recovery, as this was the ratio of feedwater
the salt concentration and electricity produced for an extracted current volume to the combined anolyte and catholyte volume in our prototype
of 16 mA/cm2. Upon beginning the experiment, a distinct sharp drop in system. The calculated minimum thermodynamic energy ranges from
salt concentration to 350 mM is observed within about 10 min, similar ~0.33 to 0.63 kWh/m3, because of the range of desalted water con-
to the initial transient seen at 2 mA/cm2, but significantly stronger. centration achieved at the end of our desalination experiments (see
After this initial transient, salt removal proceeds at a much faster rate Fig. 2c and d). We note that the latter calculation serves as an estimate
relative to Fig. 2c, until the cell voltage suddenly drops to 0.5 V at about of the minimum thermodynamic energy for our cell, as the process we
1.4 h (Fig. 2b). As can be seen in Fig. 2d, the cell died before the middle performed was not solely a separation, but also we generated ionic
channel was completely desalted, as at the end of the experiment the species in the brine streams via spontaneous electrochemical reactions.
salt concentration was just above 100 mM. The latter points to a po- The ratio of energy used to drive desalination to the calculated ther-
tentially important role of mass transport boundary layers in the de- modynamic minimum energy is 6.3 at 2 mA/cm2 (see Fig. 3b). Im-
salination channel on the cell desalination performance. We hypothe- proved desalination efficiencies can undoubtedly be demonstrated in
size that while the bulk salt concentration was ~100 mM, the salt the future by undertaking optimization of the cell geometry and elec-
concentration adjacent to the membranes in the desalination channel trolyte chemistry to reduce major sources of resistance. The results
approached zero [21], causing the observed sharp drop in cell voltage shown here are in stark contrast to other electrochemical systems for
(Fig. 2b). Total electricity production during desalination at 16 mA/ water desalination, such as capacitive deionization, which are largely
cm2 was noticeably lower than for 2 mA/cm2, and was ~7.8 kWh/m3. inefficient for seawater desalination and restricted to brackish water
Overall, the results of Figs. 2c and d demonstrate that the extracted feeds [5].
current density must be carefully chosen if complete desalination is In order to develop insight into the cost proposition of ED cells
desired, and that there is a trade-off between electricity delivered driven by spontaneous electrochemical reactions, we here provide an
during desalination and desalination rate (current density). Further- analysis of the basic operational costs and revenues of such cells. As
more, Fig. 2d shows that our first-generation cell can attain order of shown schematically in Fig. 3a, the aspects we consider here are the
magnitude higher desalination current densities relative to desalination cost of the reactants used to drive the cell, ($/m3)chem, and the income
batteries, which are thus far restricted to roughly 1 mA/cm2 from electricity produced by the cell, ($/m3)elect. The estimated cost of
[9,10,12,22]. a reactant per m3 of seawater desalinated can be calculated as QMp/nF,
Fig. 2 showed the ability of our cell to deliver desalted water and where Q is the ionic charge in units of Coulombs per m3 of feedwater, M
electricity without requiring an electricity input. To analyze the po- is the reactant molar mass, F is Faraday's constant, n is the moles of
tential of ED cells driven by spontaneous electrode reactions, we will electrons per mole of reactant, and p is the reactant price in units of
develop a methodology to understand both its energy efficiency and $/kg. In Fig. 3c, we show the calculated cost of reactants in units $/m3,
cost proposition towards seawater desalination. In Fig. 3a, we show a for a feedwater of 0.5 M NaCl, and for various chemistries which are
schematic representation of the combined operation of our cell with a compatible with the concept (either as shown in Fig. 1a, or requiring
chemical production plant supplying the needed redox active chemi- three or four membranes). The total cost of reactants requires summing
cals. The main energy flows are the chemical energy input to the cell the cost per m3 of both the reductant and oxidant used in the proposed
from the plant, (kWh/m3)chem, and the electrical energy output from cell. As can be seen in Fig. 3a, chemistries such as vanadium‑iodine
the cell during desalination, (kWh/m3)elect, both per m3 of desalted (VeI) and zinc‑bromine (ZneBr2) are likely cost-prohibitive for use in
water. The chemical energy input to our cell can be calculated as the large-scale water desalination. This is because their reactant cost alone
electricity it would have produced if the cell voltage had remained at is on the order of $102-103/m3, whereas for seawater desalination by
OCV for the entire desalination process, and this process has the same RO, the total cost of the desalted water is typically between $0.5/m3 -
duration as that shown in Fig. 2b. This input energy can also be in- $1/m3 (of which ~35% is the cost of electricity required during op-
terpreted as the electricity produced during a hypothetical lossless eration) [2,24]. Other chemistries, namely those utilizing low-cost
conversion of chemical-to-electrical energy (when there are no finite sulfur, hydrogen, or iron, and those that are air-breathing, demonstrate
resistances in the ED cell). In our analysis, we neglect chemical energy promising cost propositions, with reactant cost achieving on the order
losses due to reactant crossover through the membranes, as this loss of $1/m3. To gain further insight, the revenue resulting from the gen-
mechanism is expected to be small relative to energy loss via resistive erated electricity during cell operation should be subtracted from the
dissipation. We can then define an energy recovery efficiency as the reactant costs. As shown in Fig. 2c, when desalting seawater with our
ratio of the electricity output of the cell during desalination to the input prototype cell at 2 mA/cm2, we generate 23.5 kWh/m3. Thus, assuming
chemical energy, (kWh/m3)elect/(kWh/m3)chem. In Fig. 3b, we plot the the sale of electricity at $0.15/kWh, our ED cell provides up to ~$3.5/
measured energy recovery efficiency, showing that the prototype cell m3 of revenue (dashed line in Fig. 3c). This value of revenue is specific
can recover up to 85% of the input chemical energy as electricity when to our first-generation prototype ZneBr2 cell, while future cells opti-
operated at 2 mA/cm2, and thus at this current density 15% of the input mized for minimal resistive losses should be able to achieve similar
energy was used to drive the desalination process. This recovery occurs revenues at greater than 2 mA/cm2 current densities, enabling higher
without any dedicated energy recovery device (such as is needed in desalted water throughput. Thus, the cost analysis proposed here

260
S. Abu Khalla and M.E. Suss Desalination 467 (2019) 257–262

Fig. 3. a) Schematic of the energy and monetary inputs and outputs for an ED cell driven by spontaneous electrochemical reactions, with units of kWh or $ per m3 of
desalted seawater. The proposed cell utilizes reactants produced at minimized cost in a separate facility, thus decoupling reactant production from its usage in the
cell. b) Measured energy recovery efficiency (left axis, black circles) and desalination efficiency (right axis, red triangles) versus extracted current density, obtained
with our prototype. c) Evaluation of the cost of potential reactants for the ED cell, in units of $ per m3 of desalted seawater. The dashed line shows the calculated
revenue from the electricity production of our prototype cell.

suggests that by decoupling reactant production and its usage in the kWh/m3 of desalted water while desalinating a feed of ~30 g/L NaCl to
cell, as shown in Fig. 3a, it may be possible to desalt water with a net near-zero concentration at 2 mA/cm2 extracted current. We further
negative operational cost per m3. This may occur if the active chemicals provided an analysis of base operational costs, which demonstrates that
come at zero cost (e.g. O2 from air), or are produced inexpensively as a by decoupling reactant production and use as well as using low-cost
by-product from common chemical processes (e.g. S), so that the rev- reactants, such ED cells can potentially yield net-negative operational
enue from generated electricity surpasses the cost of the reactants costs. Future studies should focus on understanding and improving such
needed. Note that if we instead generated the reactants in the cell cell performance with low-cost reactants, and propose and test scale-up
electrochemically, obtaining net negative operating costs would not be strategies.
possible, as such production would necessarily require higher electrical
input to the cell than the electricity gained on discharge [14,15]. There
remain many future steps towards conclusively demonstrating ex- Acknowledgements
ceptionally low, or even negative, operational costs for such a cell,
including incorporating the cost of fouling protection, upstream filtra- This study is supported by The Nancy and Stephen Grand Technion
tion, anode replacement for metal-based chemistries, and disposal of Energy Program (GTEP), and the Ministry of National Infrastructures,
the spent anolyte and catholyte streams. Nevertheless, the base cost Energy and Water Resources via the undergraduate to graduate student
analysis provided here demonstrates that developing ED cells based on scholarships in the energy fields. The authors would like to acknowl-
low-cost reactants has the potential to lead to ultra-low cost desalted edge funding from the Israel Science Foundation in the framework of
water. the Israel National Research Center for Electrochemical Propulsion
(INREP) project.
4. Conclusion

Declaration of competing interest

We here characterized experimentally an electrodialysis cell which
takes solely chemical energy as input in the form of Zn and Br2 re-
There are no conflicts to declare.
actants, and outputs desalted water and electricity. With our custom-
fabricated prototype, we demonstrated electricity production of 23.5

261
S. Abu Khalla and M.E. Suss Desalination 467 (2019) 257–262

References A.P. Jose, A.R. Völkel, J.B. Rivest, ACS Energy Lett. 3 (2018) 375–379.
[15] S.H. Xianhua Hou, Qian Liang, Hu Xiaoqiao, Yu Zhou, Ru Qiang, Fuming Chen,
Nanoscale 10 (2018) 12308–12314.
[1] M. Elimelech, W.A. Phillip, Science 333 (2011) 712–717. [16] R. Zhao, S. Porada, P.M. Biesheuvel, A. Van der Wal, Desalination 330 (2013)
[2] N. Ghaffour, T.M. Missimer, G.L. Amy, Desalination 309 (2015) 197–207. 35–41.
[3] A.Y. Hoekstra, M.M. Mekonnen, A.K. Chapagain, R.E. Mathews, B.D. Richter, PLoS [17] M.E. Suss, K.M. Conforti, L. Gilson, C.R. Buie, M.Z. Bazant, RSC Adv. 6 (2016)
One 7 (2012). 100209–100213.
[4] H. Strathmann, Desalination 264 (2010) 268–288. [18] R. Ronen, I. Atlas, M. Suss, J. Electrochem. Soc. 165 (2018) A3820–A3827.
[5] M. Suss, S. Porada, X. Sun, M. Biesheuvel, J. Yoon, V. Presser, Energy Environ. Sci. 8 [19] W.A. Braff, M.Z. Bazant, C.R. Buie, Nat. Commun. 4 (2014) 2346.
(2015) 2296–2319. [20] R. O'hayre, S.-W. Cha, F.B. Prinz, W. Colella, Fuel Cell Fundamentals, John Wiley &
[6] S. Porada, R. Zhao, A. Van Der Wal, V. Presser, P.M. Biesheuvel, Prog. Mater. Sci. 58 Sons, 2016.
(2013) 1388–1442. [21] M. Tedesco, H.V.M. Hamelers, P.M. Biesheuvel, J. Memb. Sci. 510 (2016) 370–381.
[7] M.E. Suss, V. Presser, Joule 2 (2018) 10–15. [22] J. Lee, S. Kim, J. Yoon, ACS Omega 2 (2017) 1653–1659.
[8] M. Pasta, C.D. Wessells, Y. Cui, F. La Mantia, Nano Lett. 12 (2012) 23–27. [23] N.Y. Yip, M. Elimelech, Environ. Sci. Technol. 46 (2012) 5230–5239.
[9] S. Porada, A. Shrivastava, P. Bukowska, P.M. Biesheuvel, K.C. Smith, Electrochim. [24] M.K. Wittholz, B.K.O. Neill, C.B. Colby, D. Lewis, Desalination 229 (2008) 10–20.
Acta 255 (2017) 369–378. [25] Ashwin Ramachandran, et al., High water recovery and improved thermodynamic
[10] D. Nam, K. Choi, J. Am. Chem. Soc. 139 (2017) 11055–11063. efficiency for capacitive deionization using variable flowrate operation, Water re-
[11] P. Srimuk, F. Kaasik, B. Krüner, A. Tolosa, S. Fleischmann, N. Jäckel, M.C. Tekeli, search 155 (2019) 76–85.
M. Aslan, M. Suss, V. Presser, J. Mater. Chem. A 4 (2016) 18265–18271. [26] L. Wang, J. Dykstra, S. Lin, Energy Efficiency of Capacitive Deionization, Environ.
[12] T. Kim, C.A. Gorski, B.E. Logan, Environ. Sci. Technol. Lett. 4 (2017) 444–449. Sci. Technol. (2019).
[13] J. Lee, P. Srimuk, S. Carpier, J. Choi, ChemSusChem 11 (2018) 3460–3472.
[14] D. Desai, E.S. Beh, S. Sahu, V. Vedharathinam, Q. Van Overmeere, C.F. De Lannoy,

262