Water Research 169 (2020) 115263

Contents lists available at ScienceDirect

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

Electrochemical tap water softening: A zero chemical input approach

Peter Clauwaert a, b, 1, Jolien De Paepe a, b, 1, Fu Jiang a, Bernabe
Alonso-Farin
~ as a, c,
a, 2 b, d a, b, *
Eleni Vaiopoulou , Arne Verliefde , Korneel Rabaey
a
Center for Microbial Ecology and Technology (CMET), Ghent University, Coupure Links 653, 9000, Gent, Belgium
b
CAPTURE, Ghent, Belgium3
c
Departamento de Ingeniería Química y Ambiental, Universidad de Sevilla, Higher Technical School of Engineering, Camino de los Descubrimientos s/n,
41092, Seville, Spain
d
Particle and Interfacial Technology group (PainT), Ghent University, Coupure Links 653, 9000, Ghent, Belgium

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

Article history: Electrochemical water softening was proposed as a sustainable alternative for ion exchange softening,
Received 4 July 2019 avoiding the input of salt to drinking water and the production of a concentrated brine. Here we
Received in revised form demonstrated two novel modes of operation combining an electrochemical cell with a fluidized bed
28 October 2019
crystallizer. The first approach relied on an electrochemical cell consisting of an anode and cathode
Accepted 31 October 2019
separated by a cation or anion exchange membrane. The feed water was first directed into a crystallizer
Available online 4 November 2019
where it was blended with alkaline cathode effluent. The effluent of the crystallizer, softened water, was
in part recirculated to the cathode to generate alkalinity, in part to the anode compartment, where the
Keywords:
Electrochemical cell
pH was again decreased. Average removal efficiencies for calcium and magnesium of 75e86% and 7e21%
Calcium respectively, could be sustainably reached, at a specific energy consumption of 7.0e10.1 kWh kg
1 CaCO3
Ion exchange membrane (0.86e1.39 kWh m
3 water). This configuration allowed reagent-free water softening, albeit with an
Water treatment effluent with a pH between 3.0 and 3.6. In a second mode of operation, the process influent to soften was
Chemical free also directed to the crystallizer and recirculated over the cathode, which was separated from the anode
Softening using an anion exchange membrane. In this mode of operation, the cathode effluent was sent through
the crystallizing unit, and the anode compartment was operated in closed-loop. Average calcium and
magnesium removal efficiencies of 73e78% and 40e44% were obtained at specific energy consumptions
of 5.8e7.5 kWh kg
1 CaCO3 (0.77e0.88 kWh m
3 water). Although the softened water had an elevated
pH (~9.4), the advantage of this configuration is concomitant removal of anions and the formation of
acids/disinfectant in the anode compartment. Both methods of operation thus showed reagent-free
water softening at a relatively low specific energy consumption. These novel methods of softening
could be used in remote locations where access to chemicals or discharge of ion exchange brines proves
to be difficult, or in case addition of chemicals for softening is unwanted. Further research is needed to
further decrease the specific energy consumption during long-term operation.
© 2019 Elsevier Ltd. All rights reserved.

1. Introduction
Water softening minimizes the risk of scaling in industrial and
domestic piping systems when using aqueous solutions in which,
depending on the conditions, the salts can exceed the solubility
limit and deposit on piping surfaces (Hasson et al., 2010). Scaling
* Corresponding author. Coupure Links 653, 9000, Gent, Belgium.
E-mail address: korneel.rabaey@ugent.be (K. Rabaey).
1
Equally contributed.
2
Concawe, Bld du Souverain 165, 1160 Brussels.
3 Ion exchange is commonly used to replace calcium and
www.capture-resources.be
with mostly calcium and magnesium salts is not only undesired
because of the effects and costs of eventual clogging of piping, but it
also reduces the energetic efficiency in heat exchangers and boilers
and decreases the efficiency of water membrane filtration appli-
cations (Gabrielli et al., 2006; Hasson et al., 2010; Sung et al., 2008;
Venkatesan and Wankat, 2011; Yu et al., 2019). In households, hard
water is known to cause scale deposition on domestic devices
(washing machines, coffee machines …), to reduce the efficiency of
water boilers, to increase the soap and detergent consumption, and
to decrease comfort (e.g., in showers) (Van der Bruggen et al.,
2009).

https://doi.org/10.1016/j.watres.2019.115263
0043-1354/© 2019 Elsevier Ltd. All rights reserved.
2 P. Clauwaert et al. / Water Research 169 (2020) 115263
magnesium with monovalent ions. The salinity of the water in-
creases since the bivalent cations are replaced by a double molar
concentration of monovalent ions (e.g. sodium). The ion exchange
process in addition consumes salts or acid for the regeneration of
the resins and a brine waste product, containing salts (>50 g L
1),
hardness ions and organic compounds is produced, typically dis-
charged into the sewer (Ariono et al., 2016). If brines cannot easily
be discharged, they often need expensive (e.g. thermal crystalli-
zation) or nuisance creating (e.g. evaporation ponds) treatment,
while discharge to the sewer system can result in toxicity to the
associated activated sludge systems downstream and/or corrosion
in the sewer itself (Ariono et al., 2016; Chen et al., 2016; Romero
Barranco et al., 2001). In most cases still, ion exchange brines are
discharged to the sewer system and contribute, together with the
discharge of used softened water with a higher salinity, to a sali-
nification of the watershed (ground and surface water to which the
treated wastewater is discharged) which might eventually lead to a
reduced availability of fresh water (Chen et al., 2016). This might
not be a global concern, but has already led to bans or restrictions in
specific critical areas, e.g. in many communities in California, on
water softeners that rely on salts or acids for the regeneration of the
ion exchange resins (Chen et al., 2016; Li et al., 2016).
Besides strategies to reuse ion exchange brines, e.g. for acid and
base production (Chen et al., 2016), electrochemical methods have
been proposed and implemented as a way to remove calcium (and
magnesium) from water without the need to dose reagents and
without brine production (Gabrielli et al., 2006; Hasson et al., 2008;
Nair and Misra, 1978; Sanjua
n et al., 2019a). The system relies on a
pH increase in the proximity of e.g. an oxygen or water reducing
cathode, which also changes the carbonate equilibrium and results
in the precipitation of CaCO3 and Mg(OH)2 (Hasson et al., 2008;
Zeppenfeld, 2011). A disadvantage is the high specific electrode
area requirement, since the scale mainly deposits on the cathodic
electrode surface (Luan et al., 2019; Yu et al., 2018a). Regeneration
of the cathode has been described to be challenging as the electrical
resistance increases as more deposits adhere to the cathode
(Gabrielli et al., 2006; Luan et al., 2019; Yu et al., 2018b).
Hasson et al. (2010) presented a novel configuration of an
electrochemical water softening system, in which the acidifying
anode was separated from the alkalinizing cathode by a cation
exchange membrane (CEM). This proposed system allowed pre-
cipitation in an external crystallizer downstream of the cathode. In
this way, the electrode area requirements could drastically be
reduced. As the electrochemical cell is continuously forcing cations
(e.g. sodium) from the anode to the cathode through the CEM by
electrical migration, the anolyte salt needs to be replaced upon
depletion. In this way, the proposed technology cannot be consid-
ered as a zero chemical input technology. Zaslavschi et al. (2013)
designed two configurations for electrochemical water softening
of synthetic brackish and seawater that would allow for operation
without the need to provide salts by combining bipolar membranes
(BPM) and CEMs in a four compartment configuration. In one
configuration used by these authors (‘Ia’, two CEMs, one central
BPM), a similar flow rate of feed water as the flow rate of hard water
that was treated, was used as an external input of sodium, calcium
and magnesium to enable electromigration through the CEM
resulting in a cathodic pH increase. This configuration resulted in a
higher CaCO3 hardness removal at lower charge densities, electrode
surface and energy consumption compared to their previous elec-
trochemical CEM configuration, where anode and cathode were
separated by only one CEM (Hasson et al., 2010), although chlorine
generation could not be prevented in configuration ‘Ia’. Another
configuration (‘II’; two bipolar membranes, one central CEM) was
proposed where the cations from the cathodically treated water
could theoretically be used for cation recycling, thus alleviating the
need to foresee an input of salts to maintain the operation of the
electrochemical water softening. Although this concept itself with
cation recycling was not tested, a similar concept (‘IId’) where a
synthetic brackish feed was used in the third, acidic compartment
instead of treated, softened brackish water, could successfully
achieve CaCO3-precipitation without chlorine formation, but
without internal cation recycling (Zaslavschi et al., 2013).
Applying electrochemical water softening on tap water is ex-
pected to be energetically less favorable than seawater, as the
electrical conductivity of tap water is much lower (~0.5 mS cm
1)
than seawater (~50 mS cm
1), resulting in an increase of the Ohmic
voltage drop for the same electrode-membrane configuration.
Therefore, minimizing the number of ion exchange membranes and
the distance between the electrodes is expected to minimize the
Ohmic voltage drop when treating tap water.
The aim of this study was to demonstrate two novel concepts for
electrochemical water softening of local tap water to achieve a
reagent-free approach towards effluent with a lower electrical
conductivity. Two new process schemes are proposed that combine
the advantages of an electrochemical softening system with a single
ion exchange membrane and a zero chemical input system. Both
configurations were evaluated in terms of bivalent cation removal,
final composition of the softened water and power consumption,
and the optimal charge density for tap water softening was
identified.
2. Materials and methods
An electrochemical cell (2 Perspex frames 20  5  2 cm3 in-
ternal dimensions) was equipped with a stainless steel mesh
(Solana, Belgium, 100 cm2) as cathode on one side of an ion ex-
change membrane and a titanium anode coated with Iridium oxide
(Magneto Special Anodes BV, The Netherlands, 100 cm2) on the
other side. A DC power supply (LABPS3005, Velleman, Belgium)
was used to apply a fixed current. The cell voltage was measured
and logged every minute with an Agilent 34972A LXI Data Acqui-
sition/switch unit (USA). Hourly averages were calculated for
further analysis. The Ohmic resistance was determined with a VSP-
300 (Bio-logic SAS, France) potentiostat with the current interrupt
method.
Local tap water (average composition in SI, Section B) was fed
from a 200L process influent vessel into the lower part of a fluidized
bed precipitation vessel (‘crystallizer’, volume 4L), where existing
precipitates served as seed crystals for further precipitation (Fig. 1).
The catholyte was recirculated over the precipitation vessel
(5 mL s
1) and the anolyte was recirculated at a rate of 0.47 mL s
1.
In configuration A, a CEM (Ultrex CMI-7000s, Membranes Inter-
national Inc., USA; “CEM M.I.”) was used during the short-term tests
where the effect of the charge density was evaluated and during
runs A1, A2 and A3. During run A4, an anion exchange membrane
(AEM) (Fujifilm, type 2 AEM, The Netherlands; “AEM Fujifilm”) was
used. In configuration B, a Fujifilm type 2 AEM (The Netherlands;
“AEM Fujifilm”) was used in run B1, whereas an AEM produced by
Membranes International Inc. (Ultrex AMI-7001, USA; “AEM M.I.”)
was used in run B2.
During the short-term tests (<162 h operation time) with
configuration A, whereby the effect of the charge density was
evaluated, every charge density was maintained for at least four
times the hydraulic retention time (HRT), so every charge density
was tested for at least 49 h. The flow rate was measured several
times per week and the average flow rate over a whole run was
used for further calculations. The average process influent flow rate
was between 0.32 and 0.34 L h
1 and one HRT between 12.9 and
13.7 h during the short-term tests. During runs A1, A2, A3 and A4,
the flow rate was between 0.52 and 0.56 L h
1 (HRT between 7.8
 P. Clauwaert et al. / Water Research 169 (2020) 115263 3

Fig. 1. Schematic overview of the configuration and operation of configuration A and B. Process influent was fed into the lower part of the crystallizer to promote precipitation
by contact with settled crystals. The content of the crystallizer was recirculated over the cathodic compartment of the electrochemical cell. The inlet of the recirculation loop was
located at the top of the crystallizer in order to minimize the amount of crystals entering the loop and cell, whereas the outlet of the recirculation loop was located at the bottom to
promote precipitation by contact with the settled crystals. In configuration A, softened water was pumped from the crystallizer into the anolyte recirculation vessel at the same flow
rate as the process influent flow rate (to keep a constant level in the crystallizer), where it was recirculated over the anodic compartment of the electrochemical cell. During run A4,
configuration A was used, but the CEM was replaced by an AEM. In configuration B, softened water was extracted from the crystallizer (effluent) and the anode loop was operated in
closed-loop. Cat.: cathode, An.: anode, P$S.: power supply. Figure is not drawn to scale.

and 8.5 h). Samples were taken for analysis once the pH in the
crystallizer reached a steady-state value and average values and
standard deviations were calculated based on the different samples
taken throughout the same run. The duration of runs A1, A2, A3 and
A4 was 324 h, 150 h, 199 h and 192 h respectively.
Treated effluent from previous operation was used as fresh
anolyte in configuration B. During run B1, 500 mL anolyte was
replaced approximately every 70 h, mainly because of difficulties
maintaining a constant anolyte volume (water displacement be-
tween cathode and anode through the AEM). During run B2, the
anolyte (750 mL) was replaced twice after approximately 100 h,
after which the anode was operated with 1500 mL of fresh anolyte
for another 200 h. Samples were taken for analysis once the pH in
the crystallizer reached a steady-state value and average values and
standard deviations were calculated based on the different samples
taken throughout the same run. During runs B1 and B2, the flow
rate was between 0.56 and 0.58 L h
1 (HRT between 7.6 and 7.9 h).
The duration of runs B1 and B2 was 214 and 504 h respectively.
The concentrations of anions and cations were determined on a
Metrohm 930 and 761 compact Ion Chromatograph, respectively,
equipped with a Metrosep A Supp 5e150/4.0 (anions, eluent:
1,0 mM NaHCO3 - 3,2 mM Na2CO3) or Metrosep C6 e 250/4 column
(cations, eluent: 1,7 mM HNO3, 1.7 mM dipicolinic acid) and a
conductivity detector (Metrohm, Switzerland). Samples were
filtered (0.20 mm ChromafilXtra filter, Machery-Nagel, USA) prior to
analysis. Total inorganic carbon (TIC) was measured with a TOC-
5000 analyzer (Shimadzu, Japan), equipped with a non-dispersive
infrared analyzer. The pH and electrical conductivity (EC) were
measured with a 744 pH meter (Metrohm, Switzerland) and a
Consort C6010 meter (Consort, Belgium) respectively.
The specific energy consumption either was calculated per cubic
meter of tap water (Em3; equation (1)) or per kilogram of calcium
carbonate removed (ECaCO3; equation (2)), to allow for easy com-
parison with literature.
 
DV*I*86400 ds *103 mL3 kWh
Em3 ¼ J
(1)
Qin *3; 6*106 kWh m3
4 P. Clauwaert et al. / Water Research 169 (2020) 115263

DV*I*86400 ds *103mmol g
*103 kg  
mol kWh
ECaCO3 ¼ J
(2)
3; 6*106 kWh*100:0869 gCaCO
mol
3
*ðCain
Caout Þ*Qin kgCaCO3

With DV the cell voltage [V], I the current [A], Qin the process dosed. In configuration A, the liquid of the crystallizer passed first
influent flow rate [L d
1], Cain and Caout the calcium concentration through the anode before being discharged to the effluent vessel,
in the process influent and effluent respectively [mM]. resulting in a pH of 3e4 in the effluent. A charge density of
175 C L
1 tap water was insufficient to establish a pH increase that
was sufficient to remove the calcium more than 80%. During the
3. Results
operation at a charge density of 175 C L
1 tap water, the voltage
over the cell needed to maintain this charge density increased from
3.1. Configuration A with a CEM: electrochemical water softening
~5.0 V in the beginning of the test to 8.7 V after 121 h of operation,
by internal sodium and potassium cycling
whereas for charge densities of 524 C L
1 tap water or higher, the
cell voltage did not tend to increase. At this low charge density of
3.1.1. Effect of charge density on the calcium removal
175 C L
1 tap water, the calcium concentration in the anode effluent
The effectiveness of configuration A was tested at different
was lower than in the cathodic effluent and crystallizer (Fig. 2b).
charge densities (175e3129 C L
1 tap water) with unmodified, local
The Ohmic resistance increased up to 1.38 ± 0.04 U m2 at the end of
tap water. In order to obtain a calcium concentration below 0.5 mM
the test, while it had remained relatively constant at
(~80% removal efficiency), a charge density of at least 524 C L
1 tap
0.38 ± 0.02 U m2 during the tests in the range of 524e3129 C L
1 tap
water was required (Fig. 2b,d), while the EC of the tap water
water.
decreased from 0.45 ± 0.02 mS cm
1 to 0.35 ± 0.04 mS cm
1.
At charge densities up to 1097 C L
1 tap water, no changes in the
From 524 C L
1 tap water on, a pH of 9.8 or higher could be
chloride concentration were observed, whereas at 2237 and
established in the crystallizer (precipitation vessel) by circulating
3129 C L
1 tap water, the chloride concentrations in the cathode
the crystallizer liquid over the cathodic compartment (Fig. 2a). The
effluent and the crystallizer were lower (0.1e0.2 mM) compared to
pH of the cathodic effluent was slightly higher than the pH in the
the process influent and effluent (0.7e0.8 mM) (Fig. 2c). This is an
crystallizer, where the process influent, which had a lower pH, was

Fig. 2. Effect of the charge density on the pH (a), calcium concentration (b), chloride concentration (c), calcium removal efficiency and energy consumption (d) in configuration A
with a CEM. The values are average values from samples taken at different timepoints during the same run. The error bars represent the standard deviation.
 P. Clauwaert et al. / Water Research 169 (2020) 115263
indication for local cycling of chloride ions by electrical migration
from cathode to anode - even through the CEM which purpose is to
minimize anion migration, however without clear chloride removal
in the system.
The calcium removal efficiency could be increased up to ~94%
with a concomitant calcium effluent concentration below 0.2 mM
by increasing the charge density to 2237 C L
1 tap water or higher.
However, where the specific electrical energy consumption for
calcium removal at 524 C L
1 tap was 5.4 ± 0.2 kWh kg
1 CaCO3
removed (0.9 ± 0.0 kWh m
3 tap water treated), it increased up to
58.8 ± 2.5 kWh kg
1 CaCO3 removed (9.9 ± 0.1 kWh m
3 tap water
treated) at a charge density of 2237 C L
1 tap water (Fig. 2d). While
the specific electrical energy consumption was relatively low
(2.4 ± 0.9 kWh kg
1 CaCO3 removed & 0.3 ± 0.1 kWh m
3 tap water
treated) at 175 C L
1 tap water, the calcium removal efficiency was
only 56 ± 9% and the cell voltage could not be maintained constant
over time.
3.1.2. Long-term operation
The long-term stability of the process in configuration A was
evaluated during three separate runs, with a charge density of
646 C L
1 tap water (run A1), 480 C L
1 tap water (run A2) and
633 C L
1 tap water (run A3) respectively during at least 18 HRTs
(150e300 h, Fig. 3). The cathodically induced pH increase
(9.8e10.8) in the crystallizer could be maintained for all three runs
(Fig. 3), resulting in a calcium effluent concentration of
0.26 ± 0.12 mM for run A1 and 0.33 ± 0.04 mM for run A3, while a
calcium effluent quality of 0.41 ± 0.02 mM was obtained during run
A2, where a lower charge density (480 C L
1 tap water) was applied.
For runs A1 and A3, where the charge density was higher than
600 C L
1 tap water, average calcium removal efficiencies of more
than 82% were obtained, while the removal efficiency was 75 ± 1%
during run A2 (Table 1). The process influent magnesium concen-
tration ranged between 0.28 and 0.36 mM and the removal effi-
ciency in run A2 and A3 was only 21 ± 7% and 7 ± 7% respectively
(no data for run A1). After passage through the anode, the pH of the
effluent was in the range of 3.0e3.6 and the Ohmic resistance of the
cell could be maintained at 0.32 ± 0.04 U m2 during the three runs.
The EC of the process influent was 0.57 ± 0.13 mS cm
1 and
decreased on average with 14e18% after passage through the sys-
tem (Table 1). The sodium concentration in the crystallizer was
2.5e3.3 times higher than in the process influent, whereas the
sodium effluent concentration was similar as or slightly lower than
the process influent concentration (Table 1). An even higher con-
centration effect was observed for potassium in the crystallizer
(4.4e5.9) (Table 1). This concentrating of sodium and potassium in
the crystallizer and cathode indicates that monovalent cations are
Fig. 3. pH and calcium profile during long-term operation of configuration A at 646 C L
1 tap water (run A1, (a)), 480 C L
1 tap water (run A2, (b)) and 633 C L
1 tap water (run A3,
(c)).
5
recycled in the system by electromigration from anode back to
cathode through the CEM.
The chloride recovery efficiency in the effluent was 99 ± 7%,
97 ± 6% and 93 ± 4% for runs A1, A2, and A3 respectively (process
influent concentration: 0.92 ± 0.06 mM), which indicates that little
(max 7%) or no chloride oxidation occurred at the anode (Table 1).
Similar concentrations were observed in the crystallizer as in the
process influent and effluent. During run A1, nitrate and sulfate
were recovered in the effluent with an efficiency of 97 ± 6% and
97 ± 5% respectively, whereas the nitrate recovery was only
50 ± 16% and 45 ± 6% for run A2 and A3 respectively (nitrate pro-
cess influent concentration: 0.13e0.22 mM). The sulfate recovery
efficiency for run A2 and A3 was 85 ± 2% and 87 ± 2% respectively
(sulfate process influent concentration: 0.51e0.59 mM). The nitrite
concentration in the process influent was below 0.02 mM and no
increase nor decrease was observed in the effluent. At the end of
run A1, 98 ± 1% of the 2.4 ± 0.2 mM incoming inorganic carbon was
removed by the system. During a titration of the effluent, 0.9 mmol
of NaOH L
1 effluent was needed to increase the pH back to the
same level as the process influent.
3.2. Configuration A with an AEM: electrochemical water softening
by internal chloride cycling
Configuration A was also tested with an AEM (Fujifilm) for 24
HRTs at a charge density of 630 C L
1 tap water (run A4). The pH in
the crystallizer was 9.4 ± 0.3 and this resulted in an overall calcium
removal efficiency of 81 ± 3%. The chloride concentrations in the
process influent and effluent were 0.89 ± 0.01 mM and
0.87 ± 0.06 mM respectively, while in the crystallizer the chloride
concentration dropped to 0.42 ± 0.06 mM. The EC in the process
influent was 0.53 ± 0.01 mS cm
1, while the EC in the crystallizer
and the effluent was 0.25 ± 0.01 mS cm
1 and 0.40 ± 0.08 mS cm
1
respectively. The voltage required to maintain a fixed charge den-
sity started to increase after 12 HRTs from ~4.4 V to ~17 V at the end
of run A4, resulting in an increasing specific energy consumption of
8.6e20.0 kWh kg
1 CaCO3. The Ohmic resistance increased from
0.13 to 0.71 U m2. No increase in the nitrite concentration in the
effluent was observed.
3.3. Configuration B with an AEM: water softening combined with
anion removal and acid production
In configuration B, the effluent was directly extracted from the
crystallizer without passage through the anode, while the anode
loop was operated in batch mode. The performance of this
configuration was evaluated with two different AEM (AEM Fujifilm
6 P. Clauwaert et al. / Water Research 169 (2020) 115263

Table 1
Performance parameters of the three long-term operation tests of configuration A with a CEM at different charge densities.

Run A1 Run A2 Run A3

Duration [# HRTs] 40 18 25
Membrane CEM M.I. CEM M.I. CEM M.I.
Charge density [C L
1 tap water] 646 480 633
Current applied [mA] 100 70 100
Current density [A m
2] 10 7 10
Cell voltage applied [V] 7.2 ± 0.5 6.5 ± 0.3 7.9 ± 0.4
Calcium process influent concentration [mM] 2.06 ± 0.12 1.66 ± 0.04 1.68 ± 0.08
Calcium removal efficiency [%] 86 ± 5 75 ± 1 82 ± 3
Sodium concentration process influent [mM] 1.26 ± 0.06 1.23 ± 0.15 1.07 ± 0.08
Sodium concentration effluent [mM] 1.13 ± 0.13 1.11 ± 0.06 1.04 ± 0.08
Sodium concentration crystallizer [mM] 4.14 ± 0.26 3.12 ± 0.22 2.80 ± 0.11
Potassium concentration process influent [mM] 0.10 ± 0.04 0.07 ± 0.01 0.07 ± 0.01
Potassium concentration effluent [mM] 0.10 ± 0.06 0.14 ± 0.05 0.12 ± 0.03
Potassium concentration crystallizer [mM] 0.44 ± 0.10 0.41 ± 0.09 0.40 ± 0.05
Electrical conductivity ratio OUT/IN [
] 0.84 ± 0.16 0.82 ± 0.06 0.86 ± 0.14
Energy consumption [kWh kg
1 CaCO3] 7.4 ± 0.7 7.0 ± 0.4 10.1 ± 1.0
Energy consumption [kWh m
3 tap water] 1.28 ± 0.08 0.86 ± 0.04 1.39 ± 0.07

Table 2
Performance parameters of the long-term operation tests of configuration B with two different anion exchange membranes.

Run B1 Run B2

Duration [# HRTs] 28 64
Membrane AEM Fujifilm AEM M.I.
Charge density [C L
1 tap water] 615 640
Current applied [mA] 100 100
Current density [A m
2] 10 10
Cell voltage applied [V] 5.1 ± 1.0 4.4 / 6.2
Process influent pH 7.9 ± 0.1 8.0 ± 0.1
Effluent pH 9.3 ± 0.4 9.5 ± 0.2
Calcium process influent concentration [mM] 1.81 ± 0.18 1.56 ± 0.17
Calcium removal efficiency [%] 73 ± 5 78 ± 2
Magnesium process influent concentration [mM] 0.37 ± 0.08 0.29 ± 0.05
Magnesium removal efficiency [%] 40 ± 19 44 ± 10
Chloride process influent concentration [mM] 0.99 ± 0.11 1.15 ± 0.06
Chloride removal efficiency [%] 51 ± 3 51 ± 10
Nitrate process influent concentration [mM] 0.20 ± 0.02 0.19 ± 0.02
Nitrate removal efficiency [%] 55 ± 20 93 ± 7
Sulfate process influent concentration [mM] 0.67 ± 0.02 0.69 ± 0.02
Sulfate removal efficiency [%] 58 ± 4 53 ± 6
Process influent electrical conductivity [mS cm
1] 0.57 ± 0.02 0.57 ± 0.03
Electrical conductivity ratio OUT/IN [
] 0.48 ± 0.05 0.48 ± 0.06
Energy consumption [kWh kg
1 CaCO3] 7.2 ± 1.0 5.8 / 10.5
Energy consumption [kWh m
3 tap water] 0.88 ± 0.18 0.77 / 1.11
Ohmic resistance [U.m2] 0.19 ± 0.05 0.08 / 0.14

& AEM M.I.) in runs B1 and B2 respectively. In both cases, an in-
crease of the pH in the crystallizer could be established to achieve
effluent concentrations of 0.38 ± 0.09 mM calcium and
0.21 ± 0.05 mM magnesium (Table 2). Besides precipitation of cal-
cium and magnesium, concomitant removal of anions from the tap
water was established through migration through the AEM where
anions accumulated in the anolyte or were removed. As a result,
also the EC of the tap water was approximately halved (Table 2).
The removal efficiency of chloride and sulfate from the process
influent was typically between 50 and 60% for both types of
membrane, whereas the removal of nitrate was higher in the case of
the AEM M.I. (Table 2). The concentration of nitrite in the effluent
was 0.05 ± 0.02 mM during run B1 and B2, while the process
influent nitrite concentration was only 0.01 ± 0.00 mM.
The cell voltage fluctuated more during run B1 compared to run
B2 and minimizing water displacement between cathode and
anode was more challenging with the AEM used in run B1. During
the first 23 HRTs (~180 h) of run B2, the average cell voltage was
4.5 ± 0.1 V (Fig. 4b) and the specific energy consumption was
~6 kWh kg
1 CaCO3, but then the cell voltage increased resulting in
an increase in the specific energy consumption to ~7.5 kWh kg
1
CaCO3. Although the energetic efficiency decreased over time, the
pH gradient and calcium removal could be maintained throughout
run B2 (Fig. 4a). After 311 h of operation in run B2, the generated
acidic anolyte was used once to rinse the cathodic compartment for
6 h, after which the operation was resumed and the cell voltage and
the Ohmic resistance levelled off (Fig. 4b). Nevertheless, a decrease
in the calcium concentration in the process influent resulted in a
further increase in the specific energy consumption (~10 kWh kg
1
CaCO3) (Fig. 4b).
During run B2, the anolyte was replaced by 750 mL fresh anolyte
after 124 h and 198 h, while after 317 h the anolyte was replaced by
1500 mL fresh anolyte. The initial pH value in the anolyte rapidly
decreased to ~1.5 during operation. The concentration of chloride,
nitrate and sulfate and as a consequence also the EC increased over
time in the anolyte (Fig. 5), although the concentration levelled off
and decreased, especially for chloride, while no fluctuation was
observed in the concentration of the anions in the treated tap
water. The recovery in the anolyte of the chloride, nitrate and sul-
fate removed from the tap water over the 505 h of operation was
23%, 30% and 76% respectively.
 P. Clauwaert et al. / Water Research 169 (2020) 115263
Fig. 4. pH, calcium and magnesium profile during long-term operation of configuration B2 at 640 C L
1 tap water (a) and the evolution of cell voltage, energy consumption and
Ohmic resistance (b).
Fig. 5. The evolution of the chloride, nitrate and sulfate concentration and the elec-
trical conductivity in the anolyte over time for run B2. The anolyte was replaced by
750 mL fresh anolyte after 124 h and 198 h, while after 317 h the anolyte was replaced
by 1500 mL fresh anolyte.
4. Discussion
4.1. Water softening with internal anion or cation cycling or with
anion removal
Two different configurations for tap water softening were
compared in this study. In configuration A, the effluent from the
crystallizer passed through the anodic compartment before being
discharged, while in configuration B, the effluent was directly
extracted from the crystallizer and the anode loop was operated in
batch mode. Overall, both configurations had a similar calcium
removal efficiency and energy consumption at a charge density of
~600 C L
1 tap water, whereas configuration B achieved an addi-
tional decrease in the salinity due to electromigration from cathode
to anode. Ions migrate through the membrane to restore the elec-
troneutrality in the electrode compartments (to compensate the
proton and hydroxyl production at anode/cathode). In configura-
tion A, this did not result in a net removal of anions or cations, as
the tap water was sent through both compartments, resulting in an
internal cycling of anions or cations. With a CEM, cations (sodium,
potassium, calcium, magnesium and protons) migrated from the
anode back to the cathode, whereas with an AEM, anions (chloride,
nitrate, sulfate, carbonate and hydroxyl ions) migrated from the
cathode to the anode. Anion migration to the anodic compartment
7
through the AEM in configuration B resulted in a net removal of
anions, decreasing the EC with ~50% (compared to ~15% in
configuration A). The anions gradually accumulated in the anolyte,
creating an acidic salt solution which can be used to periodically
rinse the electrochemical cell in order to remove precipitates from
the electrode and membrane, or it can be discharged. Chen et al.
(2016) used the acidic stream to regenerate ion exchange media.
In Sanju
an et al. (2019b), electrochemical water softening was
combined with electrochemical production of a weak acid to
benefit from both the cathodic and anodic process. An ED
concentrate (produced from groundwater) was treated in the
cathodic compartment to decrease the hardness, while weak acids
(e.g. lactic acid) were obtained by protonation of their conjugate
bases (e.g. lactate) in the anodic compartment.
In configuration B, the calcium and magnesium removal and
decrease in EC was similar with both AEM (AEM M.I. and AEM
Fujifilm), but the cell voltage fluctuated more during the run with
AEM Fujifilm compared to the run with AEM M.I. and minimizing
water displacement between cathode and anode was more chal-
lenging with the AEM Fujifilm. Overall, water displacement is a
factor that needs to be managed which can be done by modulating
anode fluid pressure.
Another important difference between the two configurations is
the change in pH of the tap water (process influent pH ~7e8).
Configuration A creates softened tap water with an acidic pH (3e4)
due to the recirculation through the anodic compartment where
protons are produced through water oxidation which imbalance
the OH
generation as the softening leads to some removal. The tap
water treated in configuration B has an alkaline high pH (9.5) as the
acid is wasted. The pH of tap water is typically kept between 6.5
and 9.5. Further research is needed to evaluate whether the tap
water treated with these electrochemical water softening systems
is acceptable for potable use or is rather suited for softened process
water. It is important to note that due to the precipitation of car-
bonates and the lack of reagent addition, the alkalinity of the water
is very low, enabling neutralization without major acid addition.
Generating softened water at a high or low pH could be an
advantage to prolong the storage time of drinking water as the
extreme pH limits the risk for contamination. Further research
would be needed to verify this as well as verify whether extreme
pH have no other adverse impacts during storage. If a low or a high
pH in the treated tap water is undesired, a combined system can be
engineered in which a part of the water is treated with configura-
tion A and the other part with configuration B. Further research is
needed to verify that mixing the treated water of these two
8 P. Clauwaert et al. / Water Research 169 (2020) 115263
configurations (high pH and low pH) in a certain ratio could allow
to obtain the desired effluent pH.
4.2. Optimal charge density for tap water softening
The effect of the charge density on the calcium removal and
specific energy consumption was investigated in configuration A
with a CEM separating the two electrode compartments. Generally,
the higher the charge density, the more hydroxyl ions are produced
at the cathode, resulting in a higher pH in the crystallizer and,
hence, more calcium removal (higher supersaturation). A charge
density of 175 C L
1 tap water only increased the pH from 8.0 ± 0.2
(process influent pH) to 8.4 ± 0.1, removing less than 60% of the
calcium. Hence, a large fraction of the calcium remained in solution
(>40%), which could have caused the increase in Ohmic resistance
and cell voltage. Part of the calcium migrated from anode to cath-
ode through the CEM and probably caused scaling in or on the
membrane. Another possible explanation for the increase in Ohmic
resistance at a low charge density is scaling on the cathode. Due to
the relatively low pH increase in the crystallizer (pH of 8.4 at
175 C L
1 compared to pH > 9.8 at charge densities  524 C L
1),
there was a larger pH difference between the pH in the crystallizer
and the local pH at the surface of the cathode, which could have led
to an increased precipitation on the cathode surface rather than in
the crystallizer. A charge density to 524 C L
1 tap water was suffi-
cient to obtain >80% of calcium removal without an increasing
Ohmic resistance and cell voltage. Further stepwise increasing the
charge density to 3129 C L
1 tap water, only led to a slight increase
in calcium removal efficiency (from ~83% to ~93%), while the en-
ergy consumption (kWh m
3 tap water) and specific electrical
energy consumption (kWh kg
1 CaCO3 removed) increased with a
factor of >10. Despite the higher hydroxyl production at the cath-
ode (Table 3), the pH in the crystallizer only marginally increased at
higher charge densities (from 524 till 3129 C L
1 tap water). This
can be attributed to the logarithmic nature of pH values and the
limited availability of cations besides protons for electromigration.
To restore the electroneutrality in the electrode compartments,
cations migrated from the anodic compartment back to the
cathodic compartment. The tap water contained about 1.25 mM
sodium and 0.12 mM potassium, and less than 0.52 mM calcium
after precipitation in the cathode, while >5 mM hydroxyl is pro-
duced at charge densities above 524 C L
1 tap water. Hence, the
internal recycling of sodium and potassium from the anode to the
cathode was not sufficient to compensate the hydroxyl and proton
production at the electrodes. As a result, more protons migrated to
the cathode at higher charge densities, compensating the hydroxyl
production at the cathode. At 2237 and 3129 C L
1 tap water, even
chloride ions migrated from the cathode to the anode through the
CEM to balance the charge. Hence, the optimal charge density for
tap water softening to remove most of the calcium with a minimal
energy investment was around ~500e600 C L
1 tap water for the
system and the water used here.
Table 3
Theoretical hydroxyl and hydrogen production at the cathode and oxygen production at the anode in function of the charge density (assuming a faradaic efficiency of 100% and
no leakage of hydroxyl ions through the membrane).
Charge density Hydroxyl production
[C L
1 tap water] [mmol OH
L
1 tap water]
175 1,81
524 5,43
1097 11,37
2237 23,18
3129 32,43
4.3. Safety issues/concerns
In both configurations, chloride was present in the anodic
compartment, which can result in the production of chlorine (Cl2),
due to chloride oxidation at the anode (Zaslavschi et al., 2013). The
chlorine can be released to the atmosphere or further react to HOCl,
which can oxidize materials such as the membrane, and to HCl
(Zaslavschi et al., 2013). In configuration A, the chloride concen-
tration in the anodic compartment was low (<1 mM) and the
chloride recovery efficiencies were high, indicating that there was
no or very limited chloride oxidation at the anode. In configuration
B, however, chloride was concentrated in the anodic compartment
which was operated in batch mode. The concentration in the
anolyte increased up to 10e15 mM but levelled off and decreased
over time, suggesting that chloride was removed by oxidation at
the anode. Chlorine production can be prevented by applying more
selective electrodes (e.g., a boron doped diamond electrode)
instead of a mixed metal oxide anode. By including a BPM, chloride
migration to the anodic compartment can be limited (Zaslavschi
et al., 2013), but this will increase the Ohmic resistance and
negate the pH effects needed. Likely the best strategy is to
discharge regularly a small measure of anode fluid to minimize
chloride accumulation. Given the low chloride concentrations in
tap water, the risk for formation of chlorinated byproducts is low,
however, this requires further evaluation.
Up to ~93% removal of nitrate from the tap water was observed
in configuration B2, which could be an interesting feature for
generating drinking water with a low nitrate level because of the
correlation between high nitrate intake and high methaemoglobin
levels (Super et al., 1981). Since only 30% of the nitrate could be
recovered in the anolyte, it is likely that nitrate was electrochem-
ically removed in the system. Future research is needed to avoid the
formation of nitrite in the electrochemically treated tap water, as it
was found up to ~0.05 mM in configuration B.
A second safety concern is the production of oxygen and
hydrogen gas at the anode and cathode respectively (Table 3). A
current density of 600 C L
1 tap water results in 3.1 mmol H2 L
1
tap water and 1.6 mmol O2 L
1 tap water. Without proper separator
(e.g., gas-impermeable ion exchange membrane), a flammable,
potentially explosive mixture of H2 and O2 could be formed due to
convection or diffusion of the gasses across the membrane (Xiang
et al., 2016), so a proper risk management of the system is
required for applying this technology on a large scale. Likely the
introduction of an open air cathode with gas diffusion layer can
alleviate this concern by eliminating hydrogen production besides
minimizing energy requirements.
4.4. How much softening is actually needed?
The optimal charge density in this study was around
500e600 C L
1 tap water, which resulted in a calcium removal from
~1.6 to 2.2 mM Ca (16e22 F, classified as moderately hard to hard
water) to ~0.3e0.4 mM Ca (3e4 F, classified as very soft water). Too
H2 production O2 production
[mmol H2 L
1 tap water] mmol O2 L
1 tap water
0,9 0,5
2,7 1,4
5,7 2,8
11,6 5,8
16,2 8,1
 P. Clauwaert et al. / Water Research 169 (2020) 115263
soft water has corrosive properties, which can lead to leaching of
heavy metals (lead, copper…) from piping and metal surfaces,
resulting in the presence of toxic heavy metals in the drinking water.
In Flanders, Belgium, a minimal water hardness of 15 F is imposed
for the water distribution sector. As it is not recommended to soften
water below 15 F, a bypass could be installed. By mixing the soft-
ened water (≪15 F) with non-softened water (>15 F), the desired
water hardness could be obtained and this would also neutralize the
pH changes of the water (i.e., the acidification obtained with
configuration A or alkalinisation obtained with configuration B).
4.5. An electrochemical softening unit for tap water softening in
households
Electrochemical water softening is considered as alternative for
ion exchange softening. The cell will require an ion exchange
membrane to allow for the creation of a pH gradient. Typically
these are used in highly conductive streams, like seawater and/or
brackish water (Zaslavschi et al., 2013). Tap water has a much lower
conductivity (~0.5 mS cm
1 compared to >50 mS cm
1 for
seawater) and, therefore, relative to current density, the electrolyte
and membrane resistances entail a higher energy consumption
which implies that the current density needs to be limited.
The electrochemical methods of softening could be used in
remote locations where access to chemicals or discharge of ion
exchange brines proves to be difficult, or in case the addition of
chemicals for softening is unwanted. The relatively low specific
energy consumption (0.8e1.4 kWh m
3 tap water) also opens op-
portunities for household-based tap water softening. Current in-
stallations for softening at home level typically involve the use of an
ion exchange unit, which requires frequent regeneration with salt.
For a household of four people, consuming 120 L water IE
1 d
1,
less than 100 cm2 electrode surface area would suffice if the cell is
operated at a current density of 100 A m
2. Several aspects still
require further development. First, it appears advisable to avoid
chlorine formation at the anode which can be achieved with a
boron doped diamond electrode. Interestingly, such an electrode is
highly suited for the removal of trace contaminants, which delivers
an additional benefit of further water purification. Second,
hydrogen formation at the cathode needs to be avoided which can
be achieved via an open air cathode with a gas diffusion layer.
Calcium precipitate (mostly calcium carbonate) will accumulate in
the crystallizer, and needs to be harvested from the crystallizer.
About 12e60 kg calcium carbonate year
1 is produced, depending
on the initial water hardness and the softening depth. Further
research is necessary to determine the composition and value of
this calcium precipitate, which at household level could be dis-
charged via the household waste.
Besides the environmental benefits, electrochemical water
softening also seems a viable alternative to ion exchange in terms of
costs (SI, Section B). In contrast to ion exchange softeners which
consume power (~5 W), salt (0.024 kg m
3  F
1) and water (1.6L m
3
 F
1) for regeneration (Van der Bruggen et al., 2009), electro-
chemical water softening only requires energy to drive the oxidation
and reduction reactions at the electrodes. Depending on the initial
water hardness and softening depth, about 60 (20 F/15 F) to
160e300 (40 F/15 F) kWh household
1 year
1 is consumed. The
major investment cost are the electrodes, but by the use of the
external crystallizer, the required electrode surface is minimal.
5. Conclusions
 Two configurations achieving softening of tap water at energy
investments comparable to existing softeners (5.8e7.5 kWh
kg
1 CaCO3 removed) were demonstrated.
9
 Water reduction at the cathode increased the pH of the tap
water, removing ~73e86% of calcium by precipitation
 Depending on the application, the softened tap water created in
the cathodic compartment-crystallizer i) can be sent through
the anodic compartment (configuration A), or ii) can be
extracted directly from the crystallizer (configuration B).
Configuration A produces tap water with a low pH, whereas
configuration B produces tap water with a high pH and lower
salinity. Mixing the treated water of these two configurations
could allow to obtain tap water with the desired pH.
 Contrarily to conventional approaches, no chemical inputs are
needed and the system only generates a calcite precipitate to be
discharged, for example with the household waste. The elec-
trical conductivity of the softened water is lower than the
unsoftened water.
 Further developments can include the use of the anode to
achieve trace contaminant removal, thereby using the electricity
to the maximum and enabling a fine-tuned system requiring
limited maintenance.
Declaration of competing interest
The authors declare that they have no known competing
financial interests or personal relationships that could have
appeared to influence the work reported in this paper.
Acknowledgements
The authors would like to acknowledge the MELiSSA foundation
to support JDP via the POMP1 (Pool Of MELiSSA PhD) program. K.R.
was supported by the FDU project under ESA contract 4000113452/
15/NL/AT. We kindly acknowledge the financial support of the
Belgian Science Policy (BELSPO). B. Alonso-Farin ~ as acknowledges
the support of the mobility grant PP2015-4634/VPPI, granted by the
University of Seville. The authors thank Tim Lacoere for providing
graphical assistance and Greet Van de Velde for technical assistance
with the ion analyses.
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.watres.2019.115263.
References
Ariono, D., Purwasasmita, M., Wenten, I.G., 2016. Brine effluents: characteristics,
environmental impacts, and their handling. J. Eng. Technol. Sci. 48 (4), 367e387.
Chen, Y.Y., Davis, J.R., Nguyen, C.H., Baygents, J.C., Farrell, J., 2016. Electrochemical
ion-exchange regeneration and fluidized bed crystallization for zero-liquid-
discharge water softening. Environ. Sci. Technol. 50 (11), 5900e5907.
Gabrielli, C., Maurin, G., Francy-Chausson, H., Thery, P., Tran, T.T.M., Tlili, M., 2006.
Electrochemical water softening: principle and application. Desalination 201
(1e3), 150e163.
Hasson, D., Lumelsky, V., Greenberg, G., Pinhas, Y., Semiat, R., 2008. Development of
the electrochemical scale removal technique for desalination applications.
Desalination 230 (1e3), 329e342.
Hasson, D., Sidorenko, G., Semiat, R., 2010. Calcium carbonate hardness removal by
a novel electrochemical seeds system. Desalination 263 (1e3), 285e289.
Li, J., Koner, S., German, M., SenGupta, A.K., 2016. Aluminum-cycle ion exchange
process for hardness removal: a new approach for sustainable softening. En-
viron. Sci. Technol. 50 (21), 11943e11950.
Luan, J., Wang, L., Sun, W., Li, X., Zhu, T., Zhou, Y., Deng, H., Chen, S., He, S., Liu, G.,
2019. Multi-meshes coupled cathodes enhanced performance of electro-
chemical water softening system. Separ. Purif. Technol. 217, 128e136.
Nair, M.K.V., Misra, B.M., 1978. Electrolytic scale formation in sea-water distillation
systems. Desalination 25 (3), 263e268.
Romero Barranco, C., Brenes Balbuena, M., Garcıa Garcıa, P., Garrido Ferna
ndez, A.,
2001. Management of spent brines or osmotic solutions. J. Food Eng. 49 (2),
237e246.
Sanju

sito, E., Montiel, V., 2019a. Electrochemical water
an, I., Benavente, D., Expo
softening: influence of water composition on the precipitation behaviour. Separ.
10 P. Clauwaert et al. / Water Research 169 (2020) 115263
Purif. Technol. 211, 857e865.
Sanju

sito, E., Montiel, V., 2019b. Paired electrolysis for
an, I., García-García, V., Expo
simultaneous electrochemical water softening and production of weak acid
solutions. Electrochem. Commun. 101, 88e92.
Sung, S.-K., Suh, S.-H., Kim, D.-W., 2008. Characteristics of cooling water fouling in a
heat exchange system. J. Mech. Sci. Technol. 22 (8), 1568e1575.
Super, M., Heese, H.D., Mackenzie, D., Dempster, W.S., Duplessis, J., Ferreira, J.J., 1981.
An epidemiological-study of well-water nitrates in a group of south-west-
african namibian infants. Water Res. 15 (11), 1265e1270.
Van der Bruggen, B., Goossens, H., Everard, P.A., Stemge
e, K., Rogge, W., 2009. Cost-
benefit analysis of central softening for production of drinking water. J. Environ.
Manag. 91 (2), 541e549.
Venkatesan, A., Wankat, P.C., 2011. Simulation of ion exchange water softening
pretreatment for reverse osmosis desalination of brackish water. Desalination
271 (1), 122e131.
Xiang, C., Papadantonakis, K.M., Lewis, N.S., 2016. Principles and implementations
of electrolysis systems for water splitting. Mater. Horizon. 3 (3), 169e173.
Yu, Y., Jin, H., Jin, X., Yan, R., Zhang, L., Chen, X., 2018a. Current pulsated electro-
chemical precipitation for water softening. Ind. Eng. Chem. Res. 57 (18),
6585e6593.
Yu, Y., Jin, H., Meng, P., Guan, Y., Shao, S., Chen, X., 2018b. Electrochemical water
softening using air-scoured washing for scale detachment. Separ. Purif. Technol.
191, 216e224.
Yu, Y., Jin, H., Quan, X., Hong, B., Chen, X., 2019. Continuous multistage electro-
chemical precipitation reactor for water softening. Ind. Eng. Chem. Res. 58 (1),
461e468.
Zaslavschi, I., Shemer, H., Hasson, D., Semiat, R., 2013. Electrochemical CaCO3 scale
removal with a bipolar membrane system. J. Membr. Sci. 445, 88e95.
Zeppenfeld, K., 2011. Electrochemical removal of calcium and magnesium ions from
aqueous solutions. Desalination 277 (1), 99e105.