Anda di halaman 1dari 10

Journal of Membrane Science 549 (2018) 28–37

Contents lists available at ScienceDirect

Journal of Membrane Science

journal homepage:

Potassium nitrate synthesis by electrodialysis-metathesis: The effect of

membrane type

Hanna Jaroszeka, , Piotr Dydob
New Chemical Syntheses Institute, Inorganic Chemistry Division "IChN" in Gliwice, Sowinskiego 11, PL 44-101 Gliwice, Poland
Silesian University of Technology, Department of Inorganic, Analytical Chemistry and Electrochemistry, Krzywoustego 6, PL 44-100 Gliwice, Poland


The applicability of four commercially available ion exchange membranes: (Ralex AM-PP, CM-PP; Neosepta
AMX, CMX; Selemion AMV, CMV; Ionsep AM-A, AM-C) for potassium nitrate synthesis by electrodialysis-me-
tathesis was investigated. Solute back-diffusion, solvent transport, and potassium nitrate synthesis performance
(the product purity, concentration and energy consumption) were evaluated for each IEM. The ED-M synthesis
efficiency indicators: potassium nitrate concentration, purity, and current efficiency decreased with membrane
type in the following order: Seleminon > Neosepta > Ionsep > Ralex. A Selemion AMV-CMV membrane pair
performed the best among the membranes tested. Heterogeneous membranes were found to constitute a more
efficient barrier for water transport by osmosis than homogeneous ones. Water transport along with ions was
found to be the main source of water transport to products, leading to an undesired dilution of product solutions.
All of these: back-diffusion, osmosis and water electrotransport can, however, be diminished by the selection of
favorable ED-M conditions. The high performance of ED-M synthesis can be provided by using membranes with
high fixed ion concentration and low water content. High selectivity, high conductivity and low water elec-
trotransport ability are the most important membrane properties in terms of providing concentrated product
under low energy consumption through ED-M.

1. Introduction Homogeneous membranes generally exhibit more favorable electro-

chemical properties, while heterogeneous membranes are more robust
Potassium Nitrate (KNO3) is one of the strong electrolyte salt and cheaper to produce [3,5]. In ED-M synthesis, IEMs play a two-fold
compounds. This compound has 2 ions consisting of K + and NO3 role: a physical barrier separating products and a selective barrier al-
ions. Potassium nitrate is a salt compound which is one of the lowing ion transport to the suitable compartments. Transport properties
important sources of nitrogen gas in nature [1-2]. Usually minerals of IEMs (such as: selectivity, conductivity, diffusion and electroosmotic
that are rich in potassium nitrate are mineral niter. In our daily lives, permeability) are known to be dependent upon the membrane structure
we usually call this compound the "Chile burp". This name does and determine the magnitude of phenomena such as: transport of
look quite unique considering that Chile is one of the countries that counter and co-ions, diffusion, and transport of solvents [5–7].
produce this compound in large quantities. The level of solubility of Transport of co-ions across the membranes determines the product
potassium nitrate in water is quite high, at temperatures of 0 oC and purity and can be caused either by diffusion or co-ion transport through
1 L of water, the solubility can reach 133 g [2]. the membrane [1,5,10,16,17]. Transport of water increases the volume
The efficiency of ED-M, as in the case of any electrodialysis-based of concentrate, decreasing the effective salt concentration
process, is determined by both operating parameters (current density, [3,16,18]. Water can be transported either by osmosis or in an ion
feed concentration and composition, etc.) and the performance of ion- hydration shell [12–14] and the dominating mechanism depends on
exchange membranes (IEMs). Most commercial IEMs are polymeric membrane and electrolyte properties and on process conditions
materials with attached charged ion groups (fixed ions). Due to their [11,15]. Moreover, the IEM structure and material affect process
ion selectivity, IEMs are divided into cation-exchange membranes parameters, such as limiting current density and stack resistance [9].
(CEMs) with negatively charged groups and anion-exchange mem- Therefore, IEMs have an im- pact both on process indices and on
branes (AEMs) with positively charged groups fixed to a polymer product quality parameters [2,16,17].
backbone. Furthermore, IEMs are classified as heterogeneous, made by
introducing ion-exchange resin particles to a polymer, or homogeneous,

ion groups chemically bound to a polymer backbone [9,10].
Corresponding author.
E-mail address: (H. Jaroszek).
Received 2 April 2017; Received in revised form 23 November 2017; Accepted 25 November 2017
Available online 28 November 2017
0376-7388/ © 2017 Elsevier B.V. All rights reserved.
H. Jaroszek, P. Dydo Journal of Membrane Science 549 (2018) 28–37

Abbreviations N number of repeating units (pairs or quartets of mem-

branes) in a stack
AEM anion-exchange membrane Q electric charge, Ah
CE current efficiency R normal gas constant, 8.314 J/(mol K)
CV coefficient of variation r2 coefficient of determination
DD diffusion dialysis s membrane geometric surface, m2
CEM cation-exchange membrane t time, s
ED electrodialysis T temperature, K
EC electric energy consumption, kWh/kg tD electroosmotic drag coefficient, mol/mol
FIC fixed ion concentration, meq/gw W water content of membrane
IEM ion-exchange membrane V volume, mL or L
IEC ion exchange capacity, meq/gdry z the valence of an ion, eq./mol
ED-M electrodialysis-metathesis
SE standard error Greek letters
cd molar concentration, mol/L
Dapp density, g/cm3 π osmotic pressure, Pa
E apparent diffusion coefficient, m2/s φ osmotic coefficient
F voltage drop, V ωapp apparent osmotic water transport coefficient, mol/
I Faraday constant, 96,500 As/mol or 26.8 A h/mol (m2 s Pa)
i electric current, A
j number of dissociated ions Subscripts
J electric current density, A/m2
kapp molar flux, mol/(m2s) i of compound i
L apparent mass transfer coefficient, m/s osm osmotic
m thickness of swollen membrane, mm el related to migration of ions
M mass, g or kg w of water
molar mass, Da s of salt

Recently, Camacho et al. attempted to optimize ED-M parameters in should allow for the attainment of concentrated solutions of pure pro-
softening RO concentrates using a 24 factorial experimental design [2] ducts with low energy consumption.
and concluded that applied potential and concentration (dependent on
the origin of the RO concentrate) had more impact on the desalination
rate than membrane type (the linear flow rate had no significant effect). 1.1. The objective
Thus, under the same experimental conditions, membrane type may
determine the process efficiency. Various commercial membranes have The aim of this work was to investigate the effect of membrane type
been used in ED-M synthesis of inorganic compounds: homogeneous on the efficiency of KNO3 synthesis by ED-M according to the following
(Neosepta), perfluorinated (Nafion) and heterogeneous (Ionsep, Ralex, general reaction:
Lab), as presented in Table 1. Sharma et al. [3] synthesized styrene and
KCl + NH4NO3→ KNO3 + NH4Cl
divinyl benzene based interpolymer ion exchange membranes with
satisfactory mechanical properties and low electroosmotic water The IEMs chosen for this study come from leading manufacturers
transport. These membranes allowed to obtain low impurity in the and are originally designed for electrodialysis. In our previous study,
product, but the current efficiency was rather low. Given the various the effects of operating parameters (current density, feed concentration
experimental conditions used in each work, no firm conclusions on the and composition) on the effectiveness of KNO3 synthesis using Ralex
effects of membrane type can be drawn based on existing data. membranes only were determined [1]. Herein, the performance of four
Ideally, IEMs used in ED-M synthesis should provide complete ex- pairs of commercially available ion-exchange membrane brands (Ralex,
clusion of co-ions (thus excluding the transport of impurities), a low Neosepta, Selemion and Ionsep) was analyzed to select the most sui-
rate of water transport and a low back-diffusion rate of a product. In table pair of IEMs for KNO3 synthesis and to determine the process
addition, IEMs should indicate low electric resistance. All of the above indices that strongly depend on IEM type.

Table 1
Results and conditions of ED-M synthesis of inorganic compounds.

Ref. Reaction (main product bolded) Membranes Main product

Current efficiency Energy consumption, kWh/kg Impurity content per solid salt,

[18] MgCl + Na2SO4→MgSO4 + NaCl Neosepta AMX and CMX 1.06 0.85–1.30 1.3–1.6
[19] MgSO4 + 2KCl→K2SO4 + MgCl2 Neosepta AMX and Nafion 423 0.94 n/a 0.2
[16] Na2SO4 + 2KCl→K2SO 4 + 2NaCl Neosepta AMX and Nafion 423 0.94–0.95 n/a 1.2–2.5
[20] 2KCl + Na2SO4→K2SO4 + 2NaCl Ionsep 0.67–0.74 0.97–1.42 2.1–4.8
[21] 2KCl + (NH4)2SO4→K2SO4 + LabA and LabC 0.37 0.43 39.46
[1] K2SO4 + NaNO3→KNO3 + NaCl Ralex AM-PP and CM-PP 0.99 n/a < 0.3
[3] KCl + NaNO 3→KNO3 + NaCl prepared from styrene and DVB 0.72–0.84 0.74–1.31 1

H. Jaroszek, P. Dydo Journal of Membrane Science 549 (2018) 28–37

2. Materials and methods Table 3

Summary of the experimental conditions.

The following IEM pairs were selected: Ralex (AM-PP, CM-PP) from
Experiment Setup Concentrations No. of IEMs
Mega, Czech Republic; Neosepta (AMX, CMX) from Astom Corp, Japan;
Selemion (AMV, CMV) from ACG Engineering, Japan; and Ionsep (AM- DD KNO3|water Upstream: 1
A, AM-C) from Hangzhou Iontech Environmental Technology Co., Ltd., 0.2, 0.4, 1, 2, 2.5 mol/
China. Their properties are listed in Table 2.
Downstream: DI water
The thickness of each swollen membrane was measured by micro- ED KNO3|KNO3 c0: 0.5, 1, 2, 2.5 mol/l Two pairs
meter whilst its ion-exchange capacity and water content were mea- ED-M NH4NO3 + KCl→ c0 feeds: 3 mol/l Two
sured according to a standard procedure described elsewhere [10], with KNO3 + NH4Cl c0 products: 0.2 mol/l quadruplets
the exception of the membranes dried at a temperature of 105 °C to
ensure that all the free water is removed without membrane degrada-
tion (confirmed by thermogravimetric analysis and differential scan- 2.2. Electrodialysis experiments
ning calorimetry results, not shown here).
All experiments were conducted using an EDR-Z 4/0.8 module To measure the amounts of water transported along with KNO3 to
the product compartment for each tested IEM type, electrodialysis (ED)
(Mega, Czech Republic); equipped with spacers of a thickness of
experiments were conducted under a constant current density of 100 A/
0.8 mm, and with an effective cross-section area of 64 cm2. All the
m 2 to reduce the concentration difference between concentrate and
tested solutions were prepared using analytical grade salts and deio-
diluate and thus to reduce the diffusive transport (osmosis and salt
nized water. During the experiments the temperature of process solu-
diffusion). Initial concentrations and volumes (100 cm3) of KNO3 so-
tions was maintained at 25 °C. The electrode rinse solution (E) was
0.5 M K2SO4. lutions were equal for both concentrate and diluate. Initial and final
To evaluate the applicability of the tested membranes for synthesis densities of both solutions were measured and concentration of salt was
of KNO3, three types of experiments were performed: diffusion dialysis calculated based on the solution density. Water flux was determined by
measuring the decrease in water content in the diluate, of which the
(DD), electrodialysis (ED) and electrodialysis-metathesis (ED-M). DD
experiments were performed to evaluate KNO3 back diffusion and water volume was measured with high accuracy ( ± 0.08%).
osmosis, ED allowed for quantification of water transport while ED-M
provided synthesis indicators (concentration, current efficiency and
purity of products, electric energy consumption). The conditions of 2.3. Electrodialysis metathesis experiments
each experiment are listed in Table 3. Prior to the DD experiments,
membranes were conditioned in 2.4 M KNO3 and then rinsed with Electrodialysis metathesis (ED-M) experiments were performed in
deionized water; before ED and ED-M synthesis experiments all IEMs the ED stack configuration presented in Fig. 1. An additional substrate
were conditioned in the process solutions for 2 h. (KCl) chamber adjacent to the anode chamber was used to preclude any
effect of processes taking part in the electrode compartment on the
formation of the product. NH4NO3 and KCl were chosen as feeds since
2.1. Diffusion dialysis experiments NH4+ and Cl- both are ions with high mobility and great potential to
contaminate the main product. This “worst-case scenario” synthesis was
To evaluate the rates of KNO3 back diffusion and water transport by selected to compare performance of chosen IEMs in terms of overall
water, salt transport and energy consumption.
osmosis through IEMs, diffusion dialysis (DD) experiments were con-
The initial volume of each solution was 0.2 L. ED-M was performed
ducted in a module with a single membrane separating just two com-
partments: downstream and upstream (see Table 3). Solutions in both under a constant current density of 250 A/m2 for 4 h in recirculation
compartments were recirculated at a linear flow velocity of approx. mode, to obtain a high conversion degree of feeds. Voltage drops were
3 cm/s. Their initial volumes were 65 cm3 (downstream) and 500 cm3 measured at the electrodes and at the membrane stack (including the
(upstream). Changes in both volume and conductivity of the down- above additional substrate chamber) with Pt probes provided along
stream solution were measured, and KNO3 concentrations at particular with an EDR/Z module.
times were calculated based on the conductivity using a previously Samples of product solutions were collected periodically.
determined conductivity-concentration calibration curve. A steady- Concentrations of ions were determined by ion-chromatography system
state flux of KNO3 was determined by measuring the slope of a con- ICS-5000 (Thermo Dionex, USA) equipped with AS-19 anion-exchange
centration vs. time curve. Other experimental details were described and CS-16 cation-exchange columns operated in a suppressed con-
elsewhere [22]. ductivity mode.

Table 2
Selected properties of ion-exchange membranes used in this study provided by the manufacturer and from authors own experiments (bolded).

Name Type Thickness of swollen IEM, mm IEC, meq/g Transport number Surface resistance, Ohm/cm2 Operating temp., °C pH range

Neosepta AMX Homogeneous 0.14 1.4–1.7 > 0.98 2.4 < 40 0–8
Neosepta CMX 0.17 1.5–1.8 > 0.98 3* < 40 0–10
Selemion AMV 0.11 N/A > 0.96 2.8* < 40 N/A
Selemion CMV 0.11 N/A > 0.96 3* < 40 N/A
Ionsep AM-A Heterogeneous 0.60 > 2.0 N/A 4–5** 5–50 1–12
Ionsep AM-C 0.61 > 2.2 N/A 5–6** 5–50 1–12
Ralex AM-PP 0.57 > 1.8 > 0.95 < 8* < 40 0–14
Ralex CM-PP 0.64 > 2.2 > 0.95 < 10* < 40 0–14

Measured in:
* 0.5 M NaCl.
** 0.1 M NaCl.

H. Jaroszek, P. Dydo Journal of Membrane Science 549 (2018) 28–37

Fig. 1. Stack configuration for synthesis of KNO3 by electro-


2.4. Data analysis slope of the water flux vs. osmotic pressure across IEM curve:
ΔJw, osm
The molar flux of i-compound (respective cation or anion) was ωw =
Δπ (9)
calculated as:
The electroosmotic drag coefficient, defined as the ratio of the
(ct⋅Vt −c0⋅V0) mol
Ji = , flux of solvent to the flux of ions that induce solvent flow:
s⋅t⋅N m2 s (1)
Jw, el mol
In diffusion dialysis experiments, only the steady-state flux was tD = ,
Js mol (10)
accounted for.
Flux of salt (product) was calculated as a mean of the measured Electroosmotic water flux was calculated by subtracting osmotic
fluxes of cation and anion constituting the given salt: water flux (calculated based on ωw) from measured total water flux
(z +J++ z −J−) mol (Jw,el = Jw − Jw,osm ).
Js = , The current efficiency of i-compound was calculated as:
2 m 2s (2)
The apparent mass transport coefficient was calculated based on CEi = ⋅100, %
N⋅I (11)
the slope of the flux vs. concentration curve:
ΔJ m CE of salt (product) was calculated as a mean of CEs of its con-
k app = , stituent ions.
Δc s (3)
Electric energy consumption was calculated as:
The apparent diffusion coefficient was estimated using the fol-
lowing equation: 1000I⋅ kWh
EC =
∑ E Δt, kg
m2 (12)
Dapp = kapp⋅L,
s (4)
The diffusion coefficient of KNO3 in water at an infinite dilution was 3. Results and discussion
calculated based on data presented in [23] as follows:
The aim of this work was to analyze the effect of membrane type on
(z ++ z ) −D 0D+0 − 2⋅1.957⋅1.902 −9 m2
DKNO = = ⋅10 = 1.929⋅10−9, ED-M synthesis performance and factors known to lower electro-
z +D+ + z − D−
0 0 1.957⋅1.902 s (5) membrane process efficiency: back diffusion and solvent transport. The
The water molar flux was calculated as: performance of ED-M synthesis was validated in terms of product
purity, its KNO3 concentration and energy consumption. Four pairs of
(dt⋅V t −c ⋅V
t ⋅M s) −(d ⋅V
0 0−c ⋅V0 ⋅M
0 ) s mol
Jw =
, commercially available ion exchange membrane brands were tested, of
s⋅t⋅N⋅Mw m2 s (6) which two were heterogeneous and two homogeneous in type. Their
In diffusion dialysis, the concentration of downstream solution was properties are given in Table 2. Data provided by membrane manu-
low, so its density was assumed equal to water's. Otherwise, the density facturers are often scarce and rarely include the most important
of solution was calculated based on concentration using the following structure parameters affecting transport properties of IEM: concentra-
empirical formula: tion of the fixed ions and water content [11]. Swelling of the membrane
increases distances between fixed ions in the IEM, and thus decreases
dKNO3 = 16. 56⋅c − 16. 545 (7) their effective concentration (charge density) and related repulsive
, which was obtained by fitting density and concentration numerical force as well as membrane selectivity [25]. Fixed ions concentration
data shown in [23]. (FIC, meq of fixed ions per gram of water in IEM) can be calculated
Osmotic pressure across the membrane separating solutions of based on the ion exchange capacity (IEC, meq/gdry) and water content
different concentrations was calculated from: (W, gw/gdry) of IEM:
Δπ = i⋅φ⋅ΔcRT (8) FIC =
IEC meq
W gw (13)
As in diffusion dialysis experiments cdownstream« cupstream, Δc ≈
cupstream was assumed. IEC and water content are among the most commonly reported
Osmotic coefficient, φ, values were calculated based on data from membrane properties [10], and can be found in literature or easily
[24]. measured.
Osmotic water transport coefficient was calculated based on the Table 4 presents the membranes in the study ranked by their FICs.

H. Jaroszek, P. Dydo Journal of Membrane Science 549 (2018) 28–37

Table 4 Table 5
Water content, IEC and FIC of membranes in the study – literature and authors own Calculated apparent mass transport coefficients of KNO3 for IEMs in the study.
experimental data (bolded).
Membrane W, mw/mdry IEC, meq/gdry FIC, meq/gw
kapp ·108, m/s SE ·108 r2 kapp ·108, m/s SE ·108 r2
Selemion AMV 0.18 [26] 1.95 [26] 10.8
Selemion CMV 0.20 [27] 2.01 [27] 10.1 Ralex 3.06 0.09 0.9975 3.67 0.17 0.9935
Neosepta CMX 0.18 [27] 1.62 [27] 9.0 Neosepta 2.88 0.08 0.9979 3.37 0.14 0.9949
Neosepta AMX 0.16 [27] 1.25 [27] 7.8 Selemion 1.10 0.05 0.9946 3.72 0.04 0.9996
Ionsep AM-A 0.34 1.81 5.3 Ionsep 0.71 0.01 0.9995 2.23 0.09 0.9967
Ionsep AM-C 0.44 2.20 5.0
Ralex AM-PP 0.36 1.81 5.0
Ralex CM-PP 0.63 2.13 3.4 was measured in DD experiments, at the highest possible concentration
gradient, as IEM separates DI water and salt solution. Fig. 2 shows the
measured diffusion fluxes of KNO3 vs. upstream concentration (as
Unlike what was reported by Długołęcki et al. [27], FICs in membrane
concentration of the downstream was negligible in comparison, the
pairs were found to be higher for AEMs than for CEMs (except for
upstream concentration is approximately equal to the concentration
Neosepta AMX-CMX), which obviously was the result of the higher
difference across the membrane) for anion- and cation-exchange
water content of CEMs as opposed to AEMs. It should, however, be
membranes, respectively. Data for Ralex and Neosepta membranes
noted that swelling of heterogeneous membranes depends strongly on
were presented previously in [22].
the concentration of external solution [25], thus affecting FIC values.
Diffusion fluxes found were generally higher through CEMs than
For Ralex-PP membranes, the concentration of external solution had a
through AEMs. In the case of CEMs, Ionsep membranes showed dis-
greater effect on water content of CEMs than of AEMs [22], indicating
tinguishably lower diffusion fluxes than other membranes. Among
that the difference between AEM and CEM FICs can be altered by
AEMs, similar and relatively high diffusion rates were observed for
Neosepta and Ralex membranes despite a significant difference in their
From the data presented in Table 4 it is also clear that for hetero-
thickness and structure. The apparent mass transport coefficients (kapp)
geneous membranes, FIC values are generally lower than for homo-
of KNO3 through each IEM in the study, listed in Table 5, were de-
geneous ones, indicating their lower effective charge density. However,
termined based on the slopes of flux vs. concentration curves presented
FIC is not accounting for membrane non-homogeneity and structure, as
in Fig. 2, following Eq. (3).
the state and location of water molecules in IEMs further affect their
It should be noted that kapp describes the transport through the
properties [25,28,29]. Since fixed ions in heterogeneous membranes
are localized inside the ion-exchange resin particle domains, local membrane and its adjacent polarization layers. The (minimal) diffusion
concentration of fixed ions can be higher, providing higher selectivity path length can be approximated as the thickness of a swollen mem-
and performance. To validate whether FIC provides a better measure of brane, because thickness of polarization layers is small in comparison.
In the DD experiments discussed here, IEMs separated water and a salt
concentration of fixed ions in IEMs than IEC, the effects of both these
solution. Membrane thickness (L) in such conditions can be evaluated
parameters on performance of membranes were further analyzed.
as a logarithmic mean of its thickness as measured in DI water and in
solution of a given concentration [22]. This allows for calculation of
3.1. Back diffusion of KNO3
apparent diffusion coefficients (D app) of KNO3 (Eq. (4)), their values are
given in Table 6. When compared with a diffusion coefficient of KNO3
Products of ED-M are received as solutions and require further
in water at infinite dilution (Eq. (5)), Dapp values for all the tested
evaporation to separate solid salt. In the case of some salts, like K2SO4,
membranes were found to be two orders of magnitude smaller. It
solar energy can be used for their further concentration [21]. This is
confirms that the membrane itself acts as a barrier limiting the rate of
however problematic in the case of highly soluble salts, such as KNO3.
diffusion, and the effect of polarization layers can be neglected. From
In most cases, the maximum concentration achieved by ED is between
the data in Table 6, it appears that KNO3 permeability is higher for
100 and 200 g/L. At such high concentrations, diffusion of species from
homogenous membranes and decreases with membrane brand in the
the concentrate to the diluate (back diffusion) becomes a considerable
following order: Ralex > Ionsep > Neosepta > Selemion.
source of product loss, especially in batch mode, whilst simultaneous
To evaluate the effect of IEM properties on solute transport, linear
concentration of feeds is low due to their high conversion degree. This

Fig. 2. Steady-state KNO3 flux upon up-

stream KNO3 concentration for AEMs (A)
and CEMs (B).

H. Jaroszek, P. Dydo Journal of Membrane Science 549 (2018) 28–37

Table 6 of osmotic flux of solvent through the CEMs and AEMs used in the
Calculated apparent diffusion coefficients of KNO3 for individual IEMs in study. study. Fig. 3 presents the relationships between the measured water
flux upon osmotic pressure (calculated by Eq. (8)) across IEMs in DD
Dapp ·1011, m2/s Dapp ·1011, m2/s experiments.
Water flux increased with upstream salt concentration (thus the
Ralex 1.66 2.30 osmotic pressure difference across the IEM), which conformed to the
Ionsep 0.40 1.19
osmotic character of water transport in this experiment. Except for
Neosepta 0.39 0.59
Selemion 0.13 0.45
Ralex membranes, where AEM and CEM performed in a similar manner,
water fluxes were noticeably higher through CEMs than across AEMs.
On average, osmotic fluxes of water were approx. two times higher
correlation coefficients between selected parameters describing the across homogeneous than heterogeneous membranes. Apparent os-
structure of IEMs (FIC, IEC and water content) and transport coeffi- motic water transport coefficients (ωapp), calculated from the slope of
cients were calculated (Table 7). The structural parameters of IEMs in curves presented in Fig. 3, are listed in Table 8. Osmotic fluxes through
the study were inter-correlated: their thickness (L) correlated strongly IEM pairs in the study increased in the following sequence: Sele-
with water content (W), as both thick heterogeneous membranes had mion > Neosepta > Ralex > Ionsep. Linear correlation coefficients
higher water content and were more prone to swelling than homo- (Table 7) were used to evaluate the effect of IEM structure parameters
geneous ones. However, this correlation may not hold true if only on osmotic water transport. From the negative correlation between
homo- or only heterogeneous membranes were to be compared. A ne- membrane thickness (L) and ωapp (the thinner the IEM, the higher the
gative correlation of W and FIC follows its definition by Eq. (13). osmotic flux) one can conclude that the IEM material provides a phy-
Therefore, the observed correlation between Dapp and L was secondary, sical barrier for osmotic transport of solvent. No statistically significant
as for IEMs used in this study L correlated strongly with W and FIC, correlation between ωapp and IEC was observed, indicating that the
which in turn were correlated to Dapp. No significant correlations be- presence of fixed ion-exchange groups inside the membrane does not
tween transport parameters were observed, confirming their in- affect osmotic water transport. The positive correlation between the
dependency. osmotic water transport and FIC was probably only the result of the
No statistically significant correlations of any of the individual IEM negative correlation between ωapp and water content, though statistical
properties with kapp were observed, but a strong positive correlation significance of the latter correlation couldn’t be proven at the assumed
between Dapp and water content and a negative one with FIC were significance level of 0.05. Obviously, FIC was negatively correlated
found. This can be explained by the diffusion of salt through IEM with water content (W) as shown in Eq. (13). However, even such
having been limited by the transport of the co-ion (an ion with a charge tentative results suggest that the rate of osmotic water transport should
the same as the membrane's fixed ions). To maintain electro-neutrality be governed by the membrane water content, and thick membranes
inside the IEM, the charge of a membrane's fixed ions is balanced by salt with a high degree of swelling will resist osmotic water transport the
counter-ions (K+ for CEM and NO3 - for AEM) while the concentration most.
of salt co-ions (NO3- for CEM and K+ for AEM) remains low, due to the Overall water transport was investigated in the ED experiments.
well-known Donnan exclusion mechanism. To maintain electro-neu- Both transport mechanisms – osmosis and transport of water molecules
trality in the bulk of aqueous solutions adjacent to the membranes, the intermolecularly bonded with ions (electroosmosis) – are related to the
diffusion of an uncharged nitrate – potassium ion-pair is required. concentration, membrane structure and type of ion transported. Under
Therefore, the rate of diffusion of KNO3 is determined by the rate of the same osmotic pressure difference and water availability, transport
transport of a membrane's co-ion (NO3- for CEMs and K+ for AEMs), of solvent bonded with K+ ion should not differ either from KCl or
which is governed by the concentration of the co-ions in the membrane. KNO3 solution. The changes of net water content in diluate during the
An increase in water content decreases the effective fixed ion con- experiments with different membrane pairs are presented in Fig. 4. It
centration and leads to a higher concentration of co-ions in IEM, and was observed that while salt transport rates were similar for each
therefore increases the rate of salt diffusion. membrane type, water transport rates differed noticeably, being the
Considering the above, transport coefficients summarized in either lowest for Selemion, and the highest for Ralex membranes. Pro-
Table 5 or Table 6 being in all cases higher through CEM than through portionality between the rate of water transport and electric charge
AEM implies that both the concentration in and diffusion rate of NO3 - during each experiment confirmed that water transport along with ions
through CEMs are higher than that of K+ through AEMs. This is in (water drag mechanism) constitute the main source of water transport
accordance with FIC values being generally lower for CEMs (Table 4) in in ED-M synthesis.
the considered pairs. Also, from data in Table 7, FIC as a structure The flux of water associated with ion transport was calculated by
parameter provides more significant correlations than a membrane's subtracting osmotic water flux (calculated based on ωw through each
water content, whilst IEC provides no significant correlations. This IEM) from the measured total water flux. Contrary to osmosis, higher
proves that FIC provides a better measure of the effective charge of amounts of electroosmotic water were transported through hetero-
fixed ions inside the membrane, especially for strongly swelling, het- geneous than homogeneous membranes. An electroosmotic drag coef-
erogeneous membranes. ficient (tD)—the number of water molecules transported to the product

Table 7
3.2. Transport of solvent Correlation matrix between transport parameters and selected properties of individual
IEMs (bolded values at p < 0.05).
Solvent transport to products (concentrate compartments) occurs
due to both the pressure difference (osmosis) and migration of ions kapp Dapp ωapp W IEC FIC L

(electroosmosis). During ED-M synthesis in a batch recirculation mode

kapp 1.00 0.48 0.51 0.14 −0.04 −0.07 −0.13
with a high conversion degree of products, osmotic flux is opposite to Dapp 0.48 1.00 −0.40 0.88 0.44 −0.80 0.73
migration flux at the initial stage of ED-M (when salt content is high in ωapp 0.51 −0.40 1.00 −0.60 −0.08 0.79 −0.81
feeds and low in product compartments), but at the late stage of ED-M W 0.14 0.88 −0.60 1.00 0.63 −0.88 0.88
IEC −0.04 0.44 −0.08 0.63 1.00 −0.28 0.48
the osmotic flux follows the direction of migration flux and may con-
FIC −0.07 −0.80 0.79 −0.88 −0.28 1.00 −0.95
tribute to product dilution. Changes in volume of upstream solution L −0.13 0.73 −0.81 0.88 0.48 −0.95 1.00
measured during diffusion dialysis experiments allowed for calculation

H. Jaroszek, P. Dydo Journal of Membrane Science 549 (2018) 28–37

Fig. 3. Water molar flux upon osmotic

pressure across AEMs (A) and CEMs (B).

Table 8 product solutions in ED-M. Increasing the intensity of ED-M (applying

Calculated apparent osmotic water transport coefficients. higher current density) will reduce the effect of osmotic water trans-
port, however, calculated osmotic flux contributed to only c.a. 10% of
total water flux. Therefore, any reduction in electroosmotic flux should
ωapp ·10 , SE ·10 10
r 2
ωapp ·1010, SE ·1010 r2 have more effect on the efficiency of ED-M synthesis than a reduction in
mol/ mol/ osmotic water transport. A significant reduction in electroosmotic flux
(m 2 s Pa) (m 2 s Pa) can be achieved by using more concentrated feeds (Fig. 5), because for
Selemion 4.46 0.11 0.9982 7.24 0.48 0.9872
the concentrated solutions, the tD value approaches the number of
Neosepta 4.05 0.19 0.9936 5.43 0.17 0.9972 primary hydration of ions [11]. However, high concentration of feeds
Ralex 3.12 0.14 0.9937 2.92 0.03 0.9996 may increase the rate of diffusion of impurities to the product com-
Ionsep 1.51 0.14 0.9819 3.20 0.15 0.9954 partments, decreasing the purity of products and their market value.

3.3. Synthesis efficiency

Four separate ED-M synthesis experiments were performed to

evaluate the effect of membrane type on KNO3 synthesis performance,
the results are shown in Table 9. Experiments were conducted in a
batch mode under constant current, with high conversion of feeds
(85–90% molar conversion degree). Despite such a high conversion
degree, the current density never exceeded its limiting value (de-
termined in separate experiments, not shown in this work).
The highest concentrations of KNO3 were obtained with Selemion
and Neosepta membranes, and the lowest with Ralex (Table 9). These

Fig. 4. Net water (empty points) and salt (full points) decrements in the diluate during ED
experiments (c0 KNO3 = 1 M, j = 100 A/m 2, N = 2).

compartment per mole of product formed—was calculated by Eq. (10)

and for trials presented in Fig. 4 equaled: 8.0 for Selemion, 8.8 for
Neosepta, 10.7 for Ionsep and 13.6 for a Ralex membrane pair. Ap-
parently, more water was transported with ions through heterogeneous
membranes. However, it is known that apart from the membrane
structure itself, the amounts of water available for ion hydration (thus
concentration of the salt in the solution) can affect tD, as hydration
numbers decrease with concentration [15]. Fig. 5 shows that even for
Selemion membranes, which showed the lowest electroosmotic water
transport, tD decreases with mean salt concentration in the diluate.
High rates of water transport to the product compartments, and thus
high tD values, will provide an obstacle in obtaining concentrated Fig. 5. The effect of concentration of KNO3 in the diluate on electroosmotic water
transport through Selemion membrane pair.

H. Jaroszek, P. Dydo Journal of Membrane Science 549 (2018) 28–37

Table 9
Results of ED-M synthesis and the main synthesis indices.

Ralex Neosepta Selemion Ionsep

Final conc. of KNO3, mol/l 1.76 2.16 2.22 1.95

Final conc. of NH4Cl, mol/l 1.69 2.09 2.25 2.04
Final volume of KNO3, l 0.286 0.270 0.261 0.280
Final volume of NH4Cl, l 0.306 0.280 0.270 0.280
Mean stack voltage drop, V 3.53 2.57 2.43 4.18
Mean cell voltage, V 5.76 4.71 4.71 6.65
KNO3 purity, wt% per KNO3 96.30 97.43 98.99 98.37
NH4Cl purity wt% per NH4Cl 94.76 96.68 98.69 98.65
Current efficiency of KNO3, % 82.0 95.4 96.6 88.8
Current efficiency of NH4Cl, % 83.9 94.0 99.5 93.3
Energy consumption, kWh/kg KNO3 0.577 0.361 0.336 0.630
Energy consumption, kWh/kg NH4Cl 1.049 0.693 0.618 1.135

results are not consistent with salt back diffusion: as described in

Section 3.1, the expected KNO3 loss by back-diffusion during ED-M
should be the lowest for Ionsep and the highest for Ralex membranes.
Additionally, the superiority of Selemion and Neosepta membranes in
KNO3 concentration in ED-M synthesis can be explained by their low
electroosmotic water transport rates. Considering obtained CE values, Fig. 7. Final concentration of ionic impurities in each product compartment.
back diffusion affects ED-M synthesis, however, to a much lesser extent
than electroosmotic water transport. The concentration of both pro- Table 10
ducts in each product compartment increased during experiments in a Correlation coefficients between synthesis efficiency indicators (regarding the main
non-linear way due to water transport, as shown in Fig. 6. product, KNO3), electroosmotic drag coefficient and selected properties of membrane
The mechanism of transport of ionic impurities in ED-M is quite pairs (bolded values at p < 0.1).

complex. The following conclusions were drawn in earlier papers: (i)

the product obtained in ED-M is always more contaminated than the
starting feeds [16], (ii) products are contaminated more by cationic tD cfinal 0.96 0.37 −0.93 0.87
than anionic impurities due to lower selectivity of AEM than of CEM, Emodule −0.98 −0.43 0.96 −0.93
(iii) multivalent ions and ions with a bigger hydrated radius are Final purity 0.78 0.56 −0.82 0.92
CE −0.58 0.22 0.66 −0.47
transported less [1,17,22]. Moreover, in the case of initially pure feeds
EC −0.98 −0.45 0.94 −0.91
(model solutions prepared from analytical grade salts) transport of 0.91 0.53 −0.94 0.99
impurities is dominated by diffusion and in the case of initially con-
taminated feeds (technical grade, etc.), by membrane permselectivity
[1,17]. In this study, NH4NO3 and KCl were chosen as substrates for discussed previously. Similarly, higher contents of NH4+ in the KNO 3
synthesis, as both ammonium and chloride are ions of high mobility and compartment than of Cl- indicates that NH4 + is transported through
therefore impose a great potential to contaminate the KNO3 product. AEM at a higher rate than Cl- through CEM.
Final concentrations of contaminants in each product solution are Electric resistance of the ED stack is a sum of individual resistances
presented in Fig. 7. Sulfate ions, which can originate only from the of its components. During ED-M experiments conducted under constant
electrode rinse solution, are not shown in Fig. 7 as its concentration in current density, cell voltage at the electrodes as well as a stack-only
every product solution was found to be negligible. Concentrations of voltage drop was measured. In Table 9 it can be seen that the cell
ionic impurities were found to be higher in the by-product (NH4Cl) voltage was higher than the stack voltage by around 2.2–2.5 V, which
compartment than in the main product (KNO3). NH4Cl was more con- accounted for electric resistance of the electrodes, electrode-rinse so-
taminated by NO3 - than by K+ , which is in agreement with the higher lutions and overpotentials caused by electrode reactions. These differ-
diffusion rate of NO3 - across CEMs than of K+ through AEMs, as ences, in an ED stack equipped with hundreds of membrane pairs,

Fig. 6. Changes in concentration of products

(A: main, B: co-product) during ED-M
synthesis (j = 250 A/m2, N = 2).

H. Jaroszek, P. Dydo Journal of Membrane Science 549 (2018) 28–37

should be negligible, therefore it was assumed that to evaluate DC between the brands than for CEMs, which mainly contain simple sul-
energy consumption, a mean stack voltage drop should be used instead fonic groups [9]. Considering the transport parameters and perfor-
of cell voltage. The conductivity of membrane pairs was found to in- mance indicators of the membrane pairs, the CV values decreased in the
crease in the following order: Ionsep < Ralex < Neosepta < Selemion. following way: energy consumption EC (0.31) > electroosmotic drag
The higher resistivity of heterogeneous membranes results from their coefficient tD (0.19) ≈ stack voltage Estack (0.17) > final product con-
structure: particles of ion exchange resins are distributed in an un- centration Cfinal (0.10) > current efficiency CE (0.07) > product purity
charged, and thus not conducting, polymer matrix [9,25]. Membrane (0.01). While the strongest effect of type of membrane used was ob-
conductivities were found to be related rather to the membrane thick- served on transport parameters related to concentration-driven trans-
ness than to the surface resistance data provided by the manufacturers port (Dapp, ωapp and kapp) at the same time it had also the lowest effect
(Table 2). on impurity content in the product stream. The type of membrane pair
Current efficiencies along with an IEMs resistance determines unit also had a strong impact on energy consumption of the process (EC,
electric energy consumption (EC). The lowest EC values were found for Estack) and water electro-transport, but the effect on the KNO3 pro-
Selemion and Neosepta membranes while the highest for Ionsep, ap- duction rate (represented here by current efficiency and final con-
parently due to the latter membrane's high resistance. centration) was relatively low.
Analysis of the correlations between the membrane structure para-
meters (W, IEC and FIC) and process indices, presented in Table 10, 4. Conclusions
allowed for some conclusions to be drawn about the effects of IEM
properties on their performance in ED-M synthesis. This analysis, The main role of ion-exchange membrane in ED-M synthesis is to
however, is hindered by the interrelations between some of the con- constitute a selective barrier for transport, regardless of the applied
sidered parameters. The adverse correlation between W and FIC re- current and composition and concentration of external solution. In this
sulting from the definition of FIC (Eq. (13)) and the correlation of water work, four pairs of IEMs of different commercial brands were compared
content with thickness in the case of IEMs used in this study, shown in to analyze the effect of membrane type on performance of the synthesis
Table 7, were discussed earlier. These are the main causes behind the of KNO3 by ED-M. The synthesis performance indices evaluated were
observed correlations between W, FIC, and L with: final product con- final product purity, its concentration and energy consumption.
centration, current efficiency and energy consumption. Moreover, the effect of membrane type on two factors known to lower
Nevertheless, it was observed that IEC does not have a significant the efficiency of the electromembrane process: back diffusion and sol-
impact on any of the synthesis indicators, neither on electroosmotic vent transport was evaluated.
water transport. Water content and FIC were found to be far more The efficiency of synthesis in terms of product purity, current effi-
important parameters that affected the efficiency of ED-M synthesis. ciency and final product concentration achieved decreased in the fol-
These two affected all the considered parameters at statistically sig- lowing order: Selemion > Neosepta > Ionsep > Ralex. When com-
nificant levels, except for module voltage drop and product purity. For paring energy consumption, however, Ralex membranes performed
the latter, the observed correlations weren’t statistically significant better than Ionsep due to their lower resistivity. In general, it was found
probably due to a low variance in the experimental data. Module vol- that homogeneous membranes performed better than heterogeneous in
tage drop was positively correlated with membrane thickness (in ED-M, terms of the product concentration achieved, its purity, current effi-
due to a rather high concentration of solutions, the conductivity of ciency and energy consumption, but worse in terms of back diffusion of
membrane is lowest in the whole stack, thus limiting for voltage drop) salt and the membrane's osmotic permeability. The type of membrane
and the thicker the membrane the more electric resistance it provides. had a strong effect on diffusion and osmosis, but these were found to
FIC was found to correlate positively (and W – negatively) with both have a rather low effect on overall process efficiency. These effects can
the final concentration of product and its CE, as it affects the repulsive be further limited by intensification of ED-M synthesis: by using high
force of the membrane and thus its selectivity. High membrane water current density and high feed concentration to minimize the con-
content indicates strong hydrophilic properties, which are known to centration gradients across the IEMs. The most important membrane
increase water electro-transport [11,14], and can explain the observed transport parameters to be taken into account when optimizing ED-M
positive correlation between W and tD. The negative correlation of FIC synthesis systems for high product concentration at a low energy con-
(and positive of W) with energy consumption can be explained as fol- sumption are low membrane resistivity and a low water electrotran-
lows: at higher CE and selectivity of IEMs, KNO3 was produced at a sport rate.
higher rate, which resulted in lower amounts of energy required to The commonly used ion-exchange capacity parameter was found
obtain 1 kg of product under the same process conditions. not to be reliable in assessing membrane performance, especially in the
As the final concentration of KNO 3 is the most important parameter case of strongly swelling heterogeneous membranes. To select IEMs for
to measure ED-M productivity, the above analysis shows that W and FIC ED-M synthesis, fixed ion capacity and membrane water content were
are the most important membrane structure parameters that need to be found to be far more explanatory membrane structure parameters. To
selected carefully when designing ED-M synthesis systems. maximize the efficiency of ED-M synthesis, homogeneous membranes
with a high fixed ion capacity and low water content should be prior-
3.4. The effect of membrane type on its transport properties and ED-M itized.
synthesis efficiency
Finally, to determine the effects of membrane type on each para-
meter considered in this work, the coefficient of variation (CV) – the This work was financially supported by Polish National Science
ratio of standard deviation to mean, is used to measure the dispersion of Centre grant DEC-2013/11/N/ST8/01222.
values of the considered parameters. The higher the CV value, the
stronger the effect of membrane type on the considered parameter. References
Among the population of CEMs in this study, the CV of transport
parameters decreased as follows: Dapp (0.74) > ωapp (0.45) > kapp [1] H. Jaroszek, A. Lis, P. Dydo, Transport of impurities and water during potassium
(0.21). Among the population of AEMs: Dapp (1.07) > kapp nitrate synthesis by electrodialysis metathesis, Sep. Purif. Technol. 158 (2016)
(0.62) > ωapp (0.40). The observed CVs varied more for AEMs than for [2] L.M. Camacho, J.A. Fox, J.O. Ajedegba, Optimization of electrodialysis metathesis
CEMs, which is not surprising, as chemical structures of fixed ions used (EDM) desalination using factorial design methodology, Desalination 403 (2017)
in AEMs (mostly quaternary ammonium groups) might differ more 136–143.

H. Jaroszek, P. Dydo Journal of Membrane Science 549 (2018) 28–37

[3] P.P. Sharma, S. Gahlot, A. Rajput, R. Patidar, V. Kulshrestha, Efficient and cost [17] T. Rottiers, G. De la Marche, B. Van der Bruggen, L. Pinoy, Co-ion fluxes of simple
effective way for the conversion of potassium nitrate from potassium chloride using inorganic ions in electrodialysis metathesis and conventional electrodialysis, J.
electrodialysis, ACS Sustain. Chem. Eng. 4 (2016) 3220–3227. Membr. Sci. 492 (2015) 263–270.
[4] H. Meng, H. Li, C. Li, L. Li, Synthesis of ionic liquid using a four-compartment [18] C. Alhéritière, W.R. Ernst, T.A. Davis, Metathesis of magnesium and sodium salt
configuration electrodialyzer, J. Membr. Sci. 318 (2008) 1–4. systems by electrodialysis, Desalination 115 (1998) 189–198.
[5] K. Haerens, P. De Vreese, E. Matthijs, L. Pinoy, K. Binnemans, B. Van der Bruggen, [19] B. Pisarska, Badania elektrodializy z podwójną wymianą w procesie Na 2SO 4 +
Production of ionic liquids by electrodialysis, Sep. Purif. Technol. 97 (2012) 90–95. 2KCl→K 2SO4 + 2NaCl, Przem. Chem. 85 (2006) 1500–1504 (in Polish).
[6] V. Forquet, C.M. Sabaté, G. Jacob, Y. Guelou, H. Delalu, C. Darwich, Energetic 2,2- [20] J.S. Trivedi, V. Bhadja, B.S. Makwana, S.K. Jewrajka, U. Chatterjee, Sustainable
dimethyltriazanium salts: a new family of nitrogen-rich hydrazine derivatives, process for the preparation of potassium sulfate by electrodialysis and its con-
Chem. Asian J. 10 (2015) 1668–1675. centration and purification by a nanofiltration process, RSC Adv. 6 (2016)
[7] A. Subramani, J.G. Jacangelo, Treatment technologies for reverse osmosis con- 71807–71817.
centrate volume minimization: a review, Sep. Purif. Technol. 122 (2014) 472–489. [21] X. Zhang, X. Wang, X. Liu, X. Han, C. Jiang, Q. Li, et al., Conversion of potassium
[8] Y.-F. Zhang, L. Liu, J. Du, R. Fu, B. Van der Bruggen, Y. Zhang, Fracsis: ion frac- chloride into potassium sulfate by four-compartment electrodialysis: batch opera-
tionation and metathesis by aNF-ED integrated system to improve water recovery, tion process, Ind. Eng. Chem. Res. 54 (2015) 11937–11943.
J. Membr. Sci. 523 (2017) 385–393. [22] H. Jaroszek, P. Dydo, Diffusion and osmosis in potassium nitrate synthesis by
[9] T. Xu, Ion exchange membranes: state of their development and perspective, J. electrodialysis metathesis, ACEE 9/4 (2016) 115–121.
Membr. Sci. 263 (2005) 1–29. [23] David R. Lide (Ed.), CRC Handbook of Chemistry and Physics, 94th ed., CRC Press,
[10] Y. Tanaka (Ed.), Ion Exchange Membranes; Fundamentals and Applications, 2nd Florida, 2013.
ed., Elsevier, Amsterdam, 2015. [24] R.A. Robinson, R.H. Stokes, Electrolyte Solutions, Butterworths, London, 1955.
[11] N.P. Berezina, N.A. Kononenko, O.A. Dyomina, N.P. Gnusin, Characterization of [25] M. Svoboda, J. Beneš, L. Vobecká, Z. Slouka, Swelling induced structural changes of
ion-exchange membrane materials: properties vs structure, Adv. Colloid Interface a heterogeneous cation-exchange membrane analyzed by micro-computed tomo-
Sci. 139 (2008) 3–28. graphy, J. Membr. Sci. 525 (2017) 195–201.
[12] L. Han, S. Galier, H.R. Balmann, Ion hydration number and electro-osmosis during [26] T. Mohammadi, M.S. Kazacos, Modification of anion-exchange membranes for va-
electrodialysis of mixed salt solution, Desalination 373 (2015) 38–46. nadium redox flow battery applications, J. Power Sources 63 (1996) 179–186.
[13] C. Larchet, B. Auclair, V. Nikonenko, Approximate evaluation of water transport [27] P. Długołęcki, K. Nymeijer, S. Metz, M. Wessling, Current status of ion exchange
number in ion-exchange membranes, Electrochim. Acta 49 (2004) 1711–1717. membranes for power generation from salinity gradients, J. Membr. Sci. 319 (2008)
[14] N. Berezina, N. Gnusin, O. Dyomina, S. Timofeyev, Water electrotransport in 214–222.
membrane systems. Experiment and model description, J. Membr. Sci. 86 (1994) [28] V.I. Zabolotsky, V.V. Nikonenko, Effect of structural membrane inhomogeneity on
207–229. transport properties, J. Membr. Sci. 79 (1993) 181–198.
[15] Y. Marcus, Concentration dependence of ionic hydration numbers, J. Phys. Chem. B [29] O.A. Demina, N.A. Kononenko, I.V. Falina, New approach to the characterization of
118 (35) (2014) 10471–10476. ion-exchange membranes using a set of model parameters, Pet. Chem. 54 (2014)
[16] B. Pisarska, Transport of co-ions across ion exchange membranes in electrodialytic 515–525.
metathesis MgSO4 + 2KCl→K 2SO4 + MgCl2, Desalination 230 (2008) 298–304.