Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 32

ON THE SALT REJECTION PROPERTIES OF

NANOFILTRATION POLYAMIDE

MEMBRANES FORMED BY INTERFACIAL

POLYMERIZATION

Haochen Zhu1,2, Anthony Szymczyk1,2 *, Béatrice Balannec1,2 

1
 Sciences Chimiques de Rennes, UMR 6226 CNRS ­ Université de Rennes 1 ­ ENSCR, 263

Avenue du Général Leclerc, Bâtiment 10 A, CS 74205, 35042 Rennes Cedex, France

2
Université Européenne de Bretagne, France

* Corresponding author, anthony.szymczyk@univ­rennes1.fr

Abstract

Nanofiltration   polyamide   membranes   obtained   by   interfacial   polymerization   possess   skin

layers containing amine and carboxylic acid groups that are distributed in an inhomogeneous

fashion, leading  to a bipolar fixed charge  distribution  (i.e. the sign of the surface charge

density changes over the skin layer thickness). In this work we have modified the standard NF

1
model (based on steric/Donnan exclusion and the extended Nernst­Planck equation) so as to

account for the spatial variations of the fixed charge inside pores.  The retention performances

of   NF   membranes   having   bipolar   charge   distributions   that   capture   the   main   electrostatic

features of polyamide membranes have been investigated by considering 1­2, 2­1 and 1­1

electrolytes. In the range of volume fluxes usually obtained with NF polymer membranes,

calculations show that the theoretical retention sequence of polyamide membranes for the

above­mentioned   electrolytes   is   1­2   >   2­1   >   1­1,   in   agreement   with   experimental   data

available   in   the   literature.   This   retention   sequence   has   been   shown   to   be   specific   of

membranes with bipolar charge distributions since both homogeneously charged membranes

and membranes with unipolar charge distributions (i.e. the concentration of charged surface

groups can vary over the skin layer thickness but the sign of surface groups remains the same)

would  be more  permeable  to  asymmetric  electrolytes  having  divalent  counterions  than  to

symmetric mono­monovalent electrolytes. 

Conclusions drawn in this work are also likely to benefit the design of new NF membranes

with targeted distribution of ionizable surface groups for desalination and water treatment. 

Keywords: Nanofiltration; Modelling; Polyamide; Charge distribution; Rejection.

1. Introduction

The last two decades, considerable effort has been invested into the development of

membranes that combine the high retention of reverse osmosis (RO) materials with the lower

pressures used in ultrafiltration (UF) [1]. This has resulted in the development of

2
nanofiltration (NF) membranes. This promising technique has attracted increasing attention in

various industrial sectors and today, it appears as the pressure-driven separation process with

the highest expected growth in coming years.

Thin­film composite polyamide membranes are currently the main class of nanofiltration (NF)

membranes available on the market. The skin layer of these membranes is formed on top of a

mesoporous   support   by   interfacial   polymerization   [2,3].   The   technique   is   based   on   a

polycondensation   reaction   between   two   monomers,   that   is,   a   polyfunctional   amine   (e.g.

piperazine   or  m­phenylenediamine)   and   a   polyfunctional   acid   chloride   (e.g.   trimesoyl

chloride) [4]. Monomers are dissolved in immiscible solvents (water and an organic solvent)

and   the   aqueous   amine   solution   initially   impregnates   the   mesoporous   support.   Next,   the

impregnated support is brought into contact with the organic solution containing acid chloride

monomers and a polyamide film quickly forms at the interface. Due to the low solubility of

acid chloride in water and the better solubility of amine in the organic solvent, the polyamide

film grows on top of the mesoporous support [3­5]. Practically, the polymerization stops when

the interfacial polyamide film prevents any further diffusion of amine towards the organic

phase.

Several factors can affect the structure of the polyamide film and so the performances of the

resulting   membrane,   among   which,   the   nature   and   concentration   of   monomers   [6­9],   the

solvent nature [10,11], the use of cross­linking agents [12] or phase­transfer catalysts [10, 13].

An important feature of skin layers of polyamide membranes is the presence of ionizable

moieties due to unreacted amines and carboxylic acids (formed from hydrolysis of unreacted

acid chlorides) [4,5,14]. This issue has been recently addressed by Coronell et al. who showed

3
the   presence   of   both   positive   and   negative   surface   groups   by   Rutherford   backscaterring

spectrometry [15­17]. It has been demonstrated that these ionizable moieties are distributed in

a  highly  non­uniform  fashion  within  the  skin layer   of polyamide  membranes.   Indeed,  by

direct visualization of the internal structure of the skin layer of various thin­film composite

polyamide membranes by Transmission Electron Microscopy (TEM), Freger has shown that

the   skin   layers   of   polyamide   membranes   are   inhomogeneously   charged   and   built   of   a

negatively charged outer region sitting on top of an inner region possessing a positive charge

density [4]. A more recent TEM analysis performed by Pacheco et al. has led to conclusions

consistent with most Freger’s results except for the positively charged region [18]. Besides,

bipolar   charge   distributions   of  polyamide   membranes   have   been   predicted   by  Freger   and

Srebnik with their model for interfacial polymerization [14] and confirmed more recently by

coarse­grained molecular dynamics simulations [5,19].

Although it is now well established that the skin layers of polyamide membranes have bipolar

charge   distributions,   no   theoretical   investigation   of   the   influence   of   inhomogeneous

distribution of positive and negative surface groups on the separation performances has been

reported yet in the membrane community. Indeed, most NF models assume implicitly that the

fixed charge density is constant over the skin layer thickness [20­32]. Tsuru et al. [33] and de

Lint et al. [34] investigated the retention properties of bi­layered membranes considering two

layers   of   opposite   charge   but   keeping   the   assumption   of   a   homogeneous   distribution   of

ionizable surface groups within each layer. In this work, we have dropped this assumption in

order   to   investigate   the   retention   properties   of   NF   membranes   with   bipolar   charge

4
distributions,   considering   continuously   varying   charge   distributions   that   are   expected   to

capture the main electrostatic features of polyamide membranes.

2. Theoretical Aspects

The structure of NF membranes was assumed to be a bundle of straight cylindrical capillaries

(pore radius rp and thickness Δz; see Fig. 1). The external solutions are assumed to be ideal

and perfectly stirred so that the concentration polarization is disregarded in the present study.

The system is considered isothermal with a temperature T of 298 K.

Figure 1. Schematic of a membrane pore used in the modeling: rp is the pore radius, Δz is the

pore length, 0- and Δz+ denote the axial coordinates just outside the pore, 0 + and Δz- denote the

5
axial coordinates just inside the pore, ci and Ci are the concentrations inside and outside the

pore, respectively.

Within the scope of the standard NF model, the transport equations of each ion i of charge zi

through the membrane are based on the extended Nernst-Planck (ENP) equation, accounting

for ionic diffusion under the action of the solute concentration gradient, electromigration

under the action of a spontaneously arising electric field, and convection due to solvent flow

in the membrane pores. In the case of NF membranes, the terms of convection and diffusion

are corrected through the hydrodynamics coefficients accounting for the effect of finite pore

size on the various components of the ionic transport [21-32, 35-37], the modified ENP for

ions i reads as follows

dci z i ci K i ,d Di , d J
ji   K i ,d Di ,  F  K ic ci v
dz RT dz Ak

(1)

The terms on the right hand side of Eq. (1) represent the transport due to diffusion,

electromigration and convection, respectively; ji is the molar flux of ion i through the pores,

Di,∞ is its diffusion coefficient in the external solution (value at infinite dilution), ci is the local

ion-concentration inside the nanopore, z is the axial coordinate in the flow direction, R is the

ideal gas constant, F is the Faraday constant, Ak is the membrane porosity, Jv is the solution

volume flux, ψ denotes the local axial electric potential inside the nanopore, Ki,c is a drag

factor accounting for the effects of the pores walls on the specie motion, and Ki,d represents the

effect of the pore to reduce the solute-solvent diffusion coefficient below its value in the free

bulk solution. The expressions for Ki,c and Ki,d can be found elsewhere [37].

6
It is worth mentioning that the issue of ion transport through inhomogeneous membranes with
continuously varying properties was investigated by Higuchi and Nakagawa [38] and by
Sorensen and Compan [39, 40] by means of a theoretical approach in which the external
solutions and the membrane were treated as two immiscible phases, the membrane being
described as a dense material without porosity. Within this framework, ions are transferred
from a phase #1 (the external solution) to a phase #2 (the membrane) and the standard
chemical potential of ions is different inside and outside the membrane. As a result, standard
chemical potential gradients should be included in the Nernst-Planck equations if
inhomogeneous membranes with continuously varying properties are considered [38-40]. In
the present work, however, we have used a different approach since our model takes into
account explicitly the presence of pores inside the membrane (as in most current NF models).
Within the scope of this “porous approach”, the membrane is modeled as a bundle of pores
which are filled with the same solvent (water) as the external compartments. Consequently,
when ions are transferred from the external solution to the pore inside, they are still dissolved
in the same solvent (the ions do not dissolve in the membrane phase in the “porous
approach”) and their standard chemical potential remains constant (if the solvent properties
are identical inside and outside the membrane pores, as assumed in the present work).
Therefore, in the present study there is no physical justification to include standard chemical
potential gradients in Eq. (1) even if the amount of fixed charges at the pore surface varies
continuously over the pore length.
The ionic molar fluxes ji are related to solution volume flux Jv (based on the membrane area)

as follows

ji 

J v C i z   (2)
Ak

Substitution of Eq. (2) into Eq. (1) allows the ionic concentration gradients within pores to be

expressed as:

z Fc ( z ) d ( z )
dci ( z )
dz

JV

K i ,d Di , Ak

K i ,c ci ( z )  C i (z  )  i i
RT dz

(3)

7
Eq. (3) has to be coupled with an explicit expression of the local electroneutrality inside

pores:

 z i ci  z   X ( z )  0                                                                                                                (4)
i

where  X(z)  denotes the local volume charge density of the membrane (i.e. the fixed charge

concentration at a point of coordinate  z  within a membrane pore) for cylindrical pores. The

fixed charge concentration is related to the surface charge density (σ) by

2 ( z )
X ( z) 
Fr p

(5)

Combining Eqs. (3) and (4), we can establish the expression of the axial electric field inside

pores as Eq. (6)

K
zi J V

K i ,c ci ( z )  C i ( z  ) 
dX ( z )
d ( z ) i i ,d D A
i , k dz
E(z)    
dz F F
 ci ( z ) z i2
RT i
 ci ( z ) z i2
RT i

(6)

the second terms in the right­hand side of Eq. 6 arises as a result of inhomogeneous charge

distribution.  

The distribution of ions at the membrane/solution interfaces is described by the following

modified Donnan equations, 

ci (0  )  Ci (0  )i exp
  zi D ( 0 0 )


 

(7)

c i ( z  )  C i ( z  ) i exp
  z i D ( z  z  )

                                                                             (8)
 

8
where 0­ | 0+ and Δz­ | Δz+ denote the membrane/solution interfaces at the feed side and the

permeate side, respectively.   i   is the steric partitioning coefficient of ion  i  defined at the

ratio between the available section (i.e. taking into account the finite size of the ion) and the

pore   cross   section  [35],     ΔΨD  being   the   dimensionless   Donnan   potential   arising   at   each

nanopore/external solution interface.

For the sake of clarity, dielectric effects [20,22,23,37,41] have been disregard in the present

work. Finally, the calculation of salt rejection rate can be obtained for a given solution volume

flux (Jv) according to its definition:

C i (z  )
Rsalt  1                                                                                                                   (9)
C i (0 - )

3. Results and Discussion

The pore radius (rp) and the thickness to porosity ratio (z/Ak) of the skin layer were set at

0.66 nm and 4.8 m, respectively, for all membranes. These values correspond to the average

structural features determined by Bowen and Mohammad from twenty­nine commercial NF

membranes [42].

Since bipolar membranes are considered in the present work, the sign of the membrane charge

changes over the skin layer thickness. Throughout this manuscript, the terms “coions” and

“counterions” are defined conventionally with respect to the sign of the fixed charge density

at the pore entrance (i.e. the outer surface of the skin layer).

9
Fig. 2 shows the theoretical salt rejection rate of a millimolar Na 2SO4 solution versus

permeate volume flux by two NF membranes exhibiting bipolar fixed charge distributions

(the variation of the local volume charge density X as a function of the axial position inside

pores of the skin layer is shown in the inset of Fig.2 for both membranes). Each charge

1
distribution is fully symmetric so that the average volume charge density ( X avg  0
X ( z ) dz

where z  z / z is the dimensionless axial coordinate inside pores) is zero for both

membranes. For comparison, the rejection properties of an uncharged membrane (i.e. X = 0

mmol/L everywhere inside pores) have been plotted in Fig. 2 too. The uncharged membrane

can reject only ~20% of Na2SO4 due to moderate steric hindrance whereas salt rejections are

much higher for both inhomogeneously charged membranes (although their average charge

density is zero too). These first results illustrate the significant role of spatial inhomogeneities

in ion transport through NF membranes.

The two bipolar distributions considered in Fig. 2 have been chosen to be symmetric with

respect to each other and so, the Na2SO4 solution enters the membrane either by the negatively

charged end or by the positively charged end depending on the membrane under

consideration. Interestingly, the membrane with the positively charged outer surface (i.e. the

pore entrance) is able to reject efficiently an asymmetric electrolyte with divalent counterions

like Na2SO4.

Examples of the electric field and the ion concentrations inside pores of the inhomogeneously

charged membranes considered in Fig. 2 are shown in Figs. 3 and 4, respectively. The pioneer

works devoted to the influence of inhomogeneous fixed charge distribution on ion transport

focused on diffusive transport through synthetic membranes. By using the usual Henderson

assumption that consists in considering constant ion concentration gradients inside the

membrane, Takagi and Nakagaki investigated the membrane potential phenomenon through

asymmetrically charged membranes [43].

10
Figure   2.  Theoretical   rejection   rate   of   10­3  mol/L   Na2SO4  solution   versus   the   permeate

volume flux by (i) an uncharged membrane (open symbols) and (ii) two bipolar membranes

(closed symbols). The fixed charge distributions are shown in the inset (the arrow indicates

the volume flux direction) and the corresponding functions are given in Appendix.

As pointed out by Higuchi and Nakagawa [38] and by Sorensen and Compan [39, 40], the

treatment of the Nernst-Planck equations proposed by Takagi and Nakagaki violates basic

laws of thermodynamics since it produces physically unrealistic results like e.g. a non-zero

steady-state salt flux in a situation with two identical solutions on the two sides of the

membrane. Sorensen and Compan [39] suggested that standard chemical potential gradients

11
had to be included in the Nernst­Planck equations so as to avoid  these anti­thermodynamic

results but it was shown by Manzanares et al. [44] that basic laws of thermodynamics are not

violated   when   the   non­linear  Nernst­Planck   equations  are   solved   numerically  without   the

Henderson approximation (see Fig. 4 in ref [44]) even if standard chemical potential gradients

are not included in the Nernst-Planck equations. Actually, applying the Henderson assumption

for ion concentrations leads to the further assumption that the electric field arising inside

pores is constant. Although such an assumption may be reasonable for homogeneously

charged membranes filled with a binary electrolyte, it does not hold anymore for

inhomogeneously charged membranes [45]. This is clearly illustrated in Fig. 3 which shows

that the variations of the electrostatic potential inside pores are strongly non-linear. Here, we

have solved the full non-linear NP equations by considering Eqs. (3) and (6) and obtained

results in full agreement with the first law of thermodynamics, i.e. a zero steady-state salt flux

is obtained when no volume flux is applied through the membrane and two identical solutions

are considered on the two sides of the membrane: more precisely, a non­zero diffusive ion­

flux is induced by the fixed charge concentration gradient but it is exactly balanced by the

electromigrative   flux     induced   by   the   electric   field   that   spontaneously   arises   through   the

inhomogeneously charged nanopores (even in the absence of any volume flow), and then

there   is   no   net   molar   flux.   Besides,   for   equal   salt   concentrations   on   the   two   sides   of

inhomogeneous  membranes  the computed electromotive force (emf) is zero as a result of

canceling values of the sum of Donnan potentials (at membrane / solution interfaces) and the

electrical   potential   difference   arising   through   the   membrane   pores.   Identical   results   were

obtained by Sorensen and Compan [39] and Manzanares et al. clearly showed that the use of

the Henderson approximation could lead to a different conclusion, i.e. a non­zero emf for

12
equal salt concentrations on the two sides of the membrane (see Fig. 3 in ref [44]). Finally, it

can be noted that “standard” NPEs equations were numerically solved in several other studies

dealing with nanopores with continuously inhomogeneous charge distributions without

leading to any anti-thermodynamic results [46-48]. 

Figure 3. Electric field inside inhomogeneously charged membranes considered in

Fig. 2 (square and triangle symbols correspond to the charge distributions shown in the inset

of Fig.2); Jv = 2x10­5 m/s.

13
Figure 4. Ion concentrations inside inhomogeneously charged membranes considered in

Fig. 2 (square and triangle symbols correspond to the charge distributions shown in the inset

of Fig.2); small symbols: Na+; big symbols: SO42­;  Jv = 2x10­5 m/s.

Results shown in Fig. 2 cannot be obtained with homogeneously charged membranes since

the Donnan theory predicts that such membranes strongly reject electrolytes with divalent

coions but are much more permeable to electrolytes with divalent counterions (strictly

speaking, this is true only if dielectric exclusion is negligible since dielectric effects are

controlled by the square of the ion charge [23, 41]). Moreover, for moderate volume fluxes,

both membranes exhibit similar rejection performances although their outer surfaces have

opposite surface charge densities. The reason is that pressure-driven transport of ions through

porous membranes is controlled by the most repulsive region inside pores (which may well

not be located at the pore entrance) when Peclet numbers are low enough [49] (this may be

the case in NF because skin layers of thin-film composite NF membranes are very thin [50,

51] but further studies focusing on the physical properties of nanoconfined fluids are needed

14
to clarify this issue since ion diffusion inside pores is expected to be significantly hindered

due to confinement effect). Interested readers can find a more detailed explanation in our

previous works [45, 49]. We believe that it is also the reason why NF/RO membranes coated

with a (neutral) polyvinyl alcohol layer in a post-treatment step are still able to exhibit high

salt rejections [52]. In that case, the pore entrance is uncharged and the most repulsive region

of the membrane (located within the skin layer of the original uncoated membrane) is just

shifted away from the feed / membrane interface with almost no consequences on the ion

rejection performances of the membrane (provided Peclet numbers are low enough as

mentioned previously).

In the case of the membrane with the positive charge at the pore entrance, Fig. 2 shows that

there is an  optimal   driving   force   beyond   which   the   intrinsic   rejection   rate   of   Na 2SO4

decreases.   As   shown   in   our   previous   works,   for   very   high   driving   forces   (i.e.   for   Peclet

numbers much higher than unity), the fixed charge distribution has no more influence on the

membrane separation properties and the rejection rate is ruled only by the charge density at

the pore entrance [45, 49]. It implies that a maximum rejection rate is predicted if the most

repulsive region within the skin layer does not correspond to the pore entrance (it should be

stressed, however, that in most cases, this maximum occurs at volume fluxes higher than

those achieved experimentally in NF). 

Conclusions drawn from Fig. 2 have important consequences since they suggest that NF

membranes with bipolar charge distributions should be able to reject efficiently both 1-2 and

2-1 asymmetric electrolytes. This is illustrated in Fig. 5 which shows the theoretical

separation performances of a bipolar membrane with respect to various electrolytes. The

charge distribution of the bipolar membrane is shown in the inset of Fig. 5. The lack of

information about the molecular structure of the skin layer of polyamide membranes prevents

15
us from asserting that the charge distribution we chose is quantitatively relevant.

Nevertheless, it captures the main electrostatic features of most thin-film composite

polyamide membranes formed by interfacial polymerization, namely,

 Charged surface groups are distributed in a highly non uniform fashion within the skin

layer [4].

 The top surface of the skin layer is negatively charged due to the presence of

unreacted acid chloride groups [4, 5, 14, 19].

 The inner part of the skin layer is positively charged due to the presence of unreacted

amines and it is thicker than the negatively charged one [4].

 The charge density in the central part of the skin layer is expected to be quite low [4,

5, 14, 19].

 The average charge density is negative as observed experimentally from membrane

potential measurements performed with most commercial polyamide membranes [53,

54] (the average charge density of the membrane considered in Fig. 5 is

-4.15 mmol/L).

It should be noted however, that for the sake of simplification we have considered straight

cylindrical pores whereas more realistic geometries should consider narrower pores in the

central part of the skin layer. Indeed, this zone is expected to be the densest due to a high

cross-linking degree, whereas toward the outer surfaces (face and back) the polymer grows

looser [19, 55, 56].

As shown in Fig. 5, the retention sequence predicted for the bipolar membrane follows the

order Na2SO4 > CaCl2 > NaCl, i.e. both 1-2 and 2-1 asymmetric electrolytes are more

efficiently rejected than 1-1 electrolytes. It must be stressed that the same retention sequence

is observed experimentally with thin-film composite polyamide membranes. An example

taken from the literature is shown in Fig. 6 (with MgCl 2 instead of CaCl2). Other examples

16
obtained with different polyamide membranes but leading to similar qualitative conclusions

can be found in the literature (see e.g. Fig. 3 in ref [57] and Fig. 4 in ref [58] showing similar

rejection sequences with NF45 and UTC 20 polyamide membranes, respectively). As

mentioned above, for the usual range of volume fluxes obtained in NF, salt rejection is

controlled by the part of the skin layer that is the most repulsive with respect to the electrolyte

under consideration, i.e. by the (negative) outer part of the skin layer for polyamide

membranes when these membranes face asymmetric electrolytes with multivalent anions (e.g.

Na2SO4) and by the (positive) inner part of their skin layer for asymmetric electrolytes with

multivalent cations (MgCl2, CaCl2…).

Figure 5. Theoretical rejection rate of Na2SO4, CaCl2 and NaCl solutions (CNa SO = CCaCl = 10-
2 4 2

3
mmol/L, CNaCl = 2×10-3 mmol/L, so that concentrations expressed in equivalent of charge per

liter are identical for all electrolytes) versus the permeate volume flux by a “model”

polyamide membrane. The fixed charge distribution of the “model” polyamide membrane is

shown in the inset (the arrow indicates the volume flux direction) and the corresponding

17
function is given in Appendix.

Figure  6.  Experimental  salt  rejection  rate of symmetric  and asymmetric  electrolytes  as  a

function of concentration  by a NF40 polyamide  membrane  (reprinted  from ref. [58] with

permission).

It must be stressed that the retention sequence shown in Fig. 5 (Na 2SO4 > CaCl2 > NaCl) can

be   obtained   only   with   membranes   having   bipolar   charge   distributions.   Indeed,   neither

homogeneously charged membranes nor membranes with unipolar charge distributions (i.e.

the fixed charge concentration is spatially changing but the sign of the surface charge remains

the same) can exhibit similar separation performances. It is illustrated in Figs. 7 and 8 which

show that membranes  with either  homogeneous  or unipolar charge distributions  lead to a

different retention sequence, namely Na2SO4  > NaCl > CaCl2  (we chose arbitrarily a linear

18
distribution of the fixed charge in Fig. 8: additional calculations performed with various kinds

of unipolar distributions led to the same retention sequence). The weak negative retention of

CaCl2  in   Figs.   7   and   8   occurs   because   the   membrane   is   only   slightly   charged   and

Dcoions>Dcounterions [59]. This behavior has been also observed experimentally [60].

Figure 7. Theoretical rejection rate of Na2SO4, CaCl2 and NaCl solutions (CNa SO  = CCaCl  = 10­


2 4 2

3
 mmol/L, CNaCl = 2×10­3 mmol/L, so that concentrations expressed in equivalent of charge per

liter   are   identical   for   all   electrolytes)   versus   permeate   volume   flux   for   a   homogeneously

charged membrane (with X = ­4.15 mmol/L corresponding to the average charge density of

the bipolar membrane considered in Fig. 5).

19
Again this conclusion is supported by experimental data. An example (taken from the

literature) obtained with a sulfonated polyethersulfone NTR 7450 membrane is shown in Fig.

9 (with MgCl2 instead of CaCl2). The skin layer of this membrane has only negatively charged

groups (SO3-) and the experimental sequence of salt rejection is Na2SO4 > NaCl > MgCl2. The

same retention sequence was obtained by Urairi et al. with the NTR 7450 membrane [61].

Other examples obtained with different unipolar membranes but leading to similar qualitative

conclusions can be found in the literature (see e.g. Fig. 2 in ref [57] which shows that a

MPF21 alumina membrane gives the retention sequence CaCl2 > NaCl > Na2SO4, as predicted

for a unipolar positive membrane).

Figure 8. Theoretical rejection rate of Na2SO4, CaCl2 and NaCl solutions (CNa SO  = CCaCl  = 10­


2 4 2

3
 mmol/L, CNaCl = 2×10­3 mmol/L, so that concentrations expressed in equivalent of charge per

liter  are   identical   for  all   electrolytes)   versus  permeate   volume   flux  for  a  membrane   with

20
unipolar   charge   distribution   (shown   in   the   inset;   the   corresponding   function   is   given   in

Appendix). The arrow in the inset indicates the volume flux direction through the membrane.

The  average  volume   charge  density  is   ­4.15 mmol/L,  i.e.  identical   to  the  average  charge

density of the bipolar membrane considered in Fig. 5.

Figure  9.  Experimental  salt  rejection  rate of symmetric  and asymmetric  electrolytes  as  a

function of concentration by a NTR 7450 (sulfonated polyethersulfone) membrane (reprinted

from ref. [58] with permission).

To our best knowledge, this is the first study that demonstrates that the ion rejection properties

of polyamide membranes can be explained by the inhomogeneous distribution of amine and

carboxylic acid groups within the skin layer.

21
Until now, it was argued that the behavior of these membranes could not be ascribed to

electrostatic interactions between ions and the proper charge of the membrane (i.e. the

membrane charge due to ionization of surface groups).

Adsorption of multivalent cations on the membrane material, leading to charge reversal from

negative to positive values, was postulated by some authors [62, 63]. Divalent cations like

Ca2+ and Mg2+ are known to adsorb on many materials but this adsorption does not necessarily

lead to charge reversal (especially at low salt concentrations). For example, Szymczyk and

Fievet performed tangential streaming potential measurements with the AFC 30 polyamide

membrane in MgCl2 solutions of various concentrations and did not observe any charge

reversal even at electrolyte concentrations up to 0.025 mol/L [23]. Similar conclusions were

obtained with the AFC 40 polyamide membrane and millimolar solutions of various 2-1

electrolytes (CaCl2,   CuCl2,   NiCl2  and   ZnCl2)  [64].   In   a   recent   work   based   on   a

phenomenological analysis of transport of mono­ and divalent ions in NF membranes, Bason

et al. suggested, however, that the reduced dielectric constant inside pores may lead to a

localized   binding   of   counterions   with   fixed   charges   accompanied   by   charge   reversal   for

divalent ions [65]. There would be no contradiction with tangential streaming potential results

since the dielectric properties  of the solution outside and inside pores are expected to be

different.  

The higher rejection obtained with 2-1 electrolytes with respect to 1-1 electrolytes was also

ascribed to steric effects [57, 63]. The Stokes radii of both Ca2+ and Mg2+ are indeed

significantly larger than that of Na+ but the difference in steric hindrance is not large enough

to explain the huge difference in rejection rate observed in Fig. 6.

More recently, dielectric effects were considered in the NF theory [22, 23, 41]. Dielectric

phenomena may explain the rejection sequence obtained in Figs 5 and 6 since excess

22
solvation energy due to dielectric exclusion varies with the square of the ion charge [23, 41].

Electrolytes with divalent cations and / or anions are then expected to be more strongly

repelled from membrane pores than 1-1 electrolytes. As mentioned previously, we did not

include dielectric effects in the present study for the sake of clarity since dielectric effects and

inhomogeneous distribution of the membrane fixed charge would produce similar effects

(qualitatively speaking) on the retention sequence (it can be noted that the NTR 7450

membrane is a relatively loose membrane [58], which could explain the experimental

sequence of salt rejection observed in Fig. 9, i.e. Na 2SO4 > NaCl > MgCl2 , since dielectric

exclusion is a strong function of pore size.

Finally, a recent work by Ba and Economy [66] gives additional credit to our theoretical

analysis based on bipolar charge distributions. These authors first synthesized a positively

charged NF membrane by chemical modification of a polyimide (PI) membrane using

polyethylenimine (PEI). Rejection measurements carried out with this membrane led to the

sequence CaCl2 > NaCl > Na2SO4 (see Fig. 7 in ref [66]) in agreement with Donnan exclusion

for a positively charged membrane. The PI-PEI membrane was further modified by coating its

top surface by a layer of negatively charged sulfonated poly (ether ether ketone) (SPEEK),

resulting in a new membrane with a bipolar charge distribution. The separation performances

of the SPEEK-coated PI-PEI membrane were then investigated, leading to the rejection

sequence CaCl2 ≈ Na2SO4 > NaCl, in agreement with our approach based on inhomogeneous

fixed charge distributions. A similar modification of the salt-retention sequence was obtained

by Urairi et al. after surface modification of sulfonated polyethersulfone NF membranes by

adsorption of cationic polyelectrolytes [61].

4. Conclusion

23
Salt rejection properties of NF membranes with bipolar fixed charge distributions have been

investigated by means of a standard NF model (i.e. based on steric/Donnan exclusion and the

extended Nernst­Planck equation) modified to account for the spatial variations of the fixed

charge inside pores.

Special   attention   has   been   paid   to   bipolar   charge   distributions   that   capture   the   main

electrostatic features of thin­film composite polyamide membranes which represent the main

class of NF membranes available on the market. The usual electrolyte retention sequence

observed experimentally with polyamide membranes is 1­2 > 2­1 > 1­1.

This retention sequence cannot be predicted on the basis of electrostatic effects if charged

surface sites are assumed to be homogeneously smeared on the pore walls. Indeed, in this case

the sequence 1­2 > 1­1 > 2­1 or 2­1 > 1­1 > 1­2 (depending on the sign of the membrane

charge) is expected. For volume fluxes observed in NF, it has been shown that the retention

sequence 1­2 > 2­1 > 1­1  can be explained by the presence on the pore walls of functional

groups with opposite charge (amine and carboxylic acid groups in the case of polyamide

membranes) that are distributed in a non uniform fashion. It is proposed that this feature of

polyamide membranes obtained by interfacial polymerization results in the specific retention

sequence obtained with these membranes. These findings are well supported by experimental

data (available in the literature) obtained with NF membranes made of different polymers.

Finally, this work also benefits the design of new NF membranes with targeted distribution of

ionizable surface groups for advanced membrane separations.

Appendix

24
Expressions of the bipolar fixed charge distributions shown in the inset of Fig. 2: 

 1 1 
X ( z )  20   1                            (closed square symbols)
 exp10 z   1 exp  10  z  1   1 

 1 1 
X ( z )  20   1                             (closed triangle symbols)
 exp10 z   1 exp  10  z  1   1 

Expression of the bipolar fixed charge distribution shown in the inset of Fig. 5:

 1.1   2.4 
X ( z )  20  0.5   40  0.5 
 exp  30  z  1   0.8   exp 9 z   1 

Expression of the unipolar fixed charge distribution shown in the inset of Fig. 8:

X ( z )  8.2978 z  8.2978

Acknowledgment

The authors are grateful to the EU commission (FEDER), the “Conseil Régional de Bretagne”

and the “Université Européenne de Bretagne” for their financial support through the research

program EPT – MOMEN.

References

25
[1] W. B. S. de Lint, Transport of electrolytes through ceramic nanofiltration membranes,

Ph. D Thesis, University of Twente, 2003.

[2] P. W. Morgan, Condensation polymers by interfacial and solution methods, Interscience:

New York, 1965.

[3] J. Petersen, Composite reverse osmosis and nanofiltration membranes, J. Membr. Sci. 83

(1993) 81-150.

[4] V. Freger, Nanoscale heterogeneity of polyamide membranes formed by interfacial

polymerization, Langmuir 19 (2003) 4791-4797.

[5] R. Nadler, S. Srebnik, Molecular simulation of polyamide synthesis by interfacial

polymerization, J. Membr. Sci. 315 (2008) 100-105.

[6] C.K. Kim, J.H. Kim, I.J. Roh, J.J. Kim, The changes of membrane performance with

polyamide molecular structure in the reverse osmosis process, J. Membr. Sci. 165 (2000) 189-

199.

[7] Y. Zhou, S. Yu, M. Liu, C. Gao, Polyamide thin film composite membrane prepared from

m-phenylenediamine and m-phenylenediamine-5-sulfonic acid, J. Membr. Sci. 270 (2006)

165-168.

[8] N.K. Saha, S.V. Joshi, Performance evaluation of thin film composite polyamide

nanofiltration membrane with variation in monomer type, J. Membr. Sci. 342 (2009) 60-69.

[9] S. Verissimo, K.-V. Peinemann, J. Bordado, Influence of the diamine structure on the

nanofiltration performance, surface morphology and surface charge of the composite

polyamide membranes, J. Membr. Sci. 279 (2006) 266-275.

[10] J. Jegal, S.G. Min, K.-H. Lee, Factors affecting the interfacial polymerization of

polyamide active layers for the formation of polyamide composite membranes,

J.Appl.Polym.Sci. 86 (2002) 2781-2787.

26
[11] S.H. Kim, S.Y. Kwak, T. Suwuki, positron annihilation spectroscopic evidence to

demonstrate the flux-enhancement mechanism in morphology-controlled thin-film-composite

(TFC) membrane, Environ. Sci. Technol. 39 (2005) 1764-1770.

[12] Y. Zhou, S. Yu, M. Liu, H. Chen, C. Gao, Effect of mixed crosslinking agents on

performance of thin-film-composite membranes, Desalination 192 (2006) 182-189.

[13] M. Duan, Z. Wang, J. Xu, J. Wang, S. Wang, Influence of hexamethyl phosphoramide on

polyamide composite reverse osmosis membrane performance, Sep. Purif. Technol. 75 (2010)

145-155.

[14] V. Freger, S. Srebnik, Mathematical model of charge and density distributions in

interfacial polymerization of thin films, J. Appl. Polym. Sci. 88 (2003) 1162-1169.

[15] O. Coronell, B. Marinas, X. Zhang, D.G. Cahill, Quantification of functional groups and

modeling of their ionization behavior in the active layer of FT30 reverse osmosis membrane,

Environ. Sci. Technol. 42 (2008) 5260-5266.

[16] O. Coronell, B. Marinas, D.G. Cahill, Accessibility and ion exchange stoichiometry of

ionized carboxylic groups in the active layer of FT30 reverse osmosis membrane, Environ.

Sci. Technol. 43 (2009) 5042-5048.

[17] O. Coronell, M.I. Gonzalez, B. Marinas, D.G. Cahill, Ionization behavior, Stoichiometry

of association, and accessibility of functional groups in the active layers of reverse osmosis

and nanofiltration membranes, Environ. Sci. Technol. 44 (2010) 6808-6814.

[18] F.A. Pacheco, I. Pinnau, M.Reinhard, J.O. Leckie, Characterization of isolated polyamide

thin films of RO and NF membranes using novel TEM techniques, J. Membr. Sci. 358 (2010)

51-59.

[19] R. Oiwerovich-Honig, V. Raim, S. Srebnik, Simulation of thin film membranes formed

by interfacial polymerization, Langmuir 26 (2010) 299-306.

27
[20] A. E. Yaroshchuk, Non-steric mechanisms of nanofiltration: superposition of Donnan and

dielectric exclusion, Sep. Purif. Technol. 22-23 (2001) 143-158.

[21] J. Palmeri, P. Blanc, A. Larbot, P. David, Theory of pressure-driven transport of neutral

solutes and ions in porous ceramic nanofiltration membranes, J. Membr. Sci. 160 (1999) 141-

170.

[22] S. Bandini, D. Vezzani, Nanofiltration modeling: the role of dielectric exclusion in

membrane characterization, Chem. Eng. Sci. 58 (2003) 3303-3326.

[23] A. Szymczyk, P. Fievet, Investigating transport properties of nanofiltration membranes

by means of a steric, electric and dielectric exclusion model, J. Membr. Sci. 252 (2005) 77-88.

[24] X. -L. Wang, T. Tsuru, S. –I. Nakao, S. Kimura, The electrostatic and steric-hindrance

model for the transport of charged solutes through nanofiltration membranes, J. Membr. Sci.

135 (1997) 19-32.

[25] W. R. Bowen, A. W. Mohammad, N. Hilal, Characterisation of nanofiltration membranes

for predictive purposes-use of salts, uncharged solutes and atomic force microscopy, J.

Membr. Sci. 126 (1997) 91-105.

[26] C. Labbez, P. Fievet, A. Szymczyk, A. Vidonne, A. Foissy, J. Pagetti, Analysis of the salt

retention of a titania membrane using the "DSPM" model: effect of pH, salt concentration and

nature, J. Membr. Sci. 208 (2002) 315-329.

[27] X. Lefebvre, J. Palmeri, D. David, Nanofiltration Theory: An Analytic Approach for

Single Salts, J. Phys. Chem. B 108 (2004) 16811-16824.

[28] D. L. Oatley, B. Cassey, P. Jones, W. Richard Bowen, Modelling the performance of

membrane nanofiltration--recovery of a high-value product from a process waste stream,

Chem. Eng. Sci. 60 (2005) 1953-1964.

[29] J. Garcia-Aleman, J. Dickson, Mathematical modeling of nanofiltration membranes with

mixed electrolyte solutions, J. Membr. Sci. 235 (2004) 1-13.

28
[30] X. Lefebvre, J. Palmeri, Nanofiltration theory: good co-ion exclusion approximation for

single salts, J. Phys. Chem. B 109 (2005) 5525-5540.

[31] G. Hagmeyer, R. Gimbel, Modelling the salt rejection of nanofiltration membranes for

ternary ion mixtures and for single salts at different pH values, Desalination 117 (1998) 247-

256.

[32] W. R. Bowen, J. S. Welfoot, Modelling the performance of membrane nanofiltration-

critical assessment and model development, Chem. Eng. Sci. 57 (2002) 1121-1137.

[33] T. Tsuru, S.I. Nakao, S. Kimura, Ion separation by bipolar membranes in reverse

osmosis, J. Membr. Sci. 108 (1995) 269-278.

[34] W.B.S. de Lint, T. Zivkovic, N.E. Benes, H.J.M. Bouwmeester, D.H.A. Blank,

Electrolyte retention of supported bi-layered nanofiltration membranes, J. Membr. Sci. 277

(2006) 18-27.

[35] W. M. Deen, Hindered transport of large molecules in liquid-filled pores, AIChE J. 33

(1987) 1409-1425.

[36] J. Garcia-Aleman, J. Dickson, A. Mika, Experimental analysis, modeling, and theoretical

design of McMaster pore-filled nanofiltration membranes, J. Membr. Sci. 240 (2004) 247-

255.

[37] A. Szymczyk, M. Sbaï, P. Fievet, A. Vidonne, Transport Properties and Electrokinetic

Characterization of an Amphoteric Nanofilter, Langmuir 22 (2006) 3910-3919.

[38] A. Higuchi, T. Nakagawa, Membrane potential and ion transport in inhomogeneous ion-

exchange membranes, J. Chem. Soc., Faraday Trans. 1, 85 (1989) 3609-3621.

[39] T.S. Sorensen, V. Compan, Nernst-Planck model simulating the electromotive force

measured over asymmetric membranes with special reference to the initial time method for

investigation of surface layers, J. Phys. Chem. 100 (1996) 7623-7631.

29
[40] T.S. Sorensen, V. Compan, Salt flux and electromotive force in concentration cells with

asymmetric ion exchange membranes and ideal 2:1 electrolytes, J. Phys. Chem. 100 (1996)

15261-15273.

[41] A. E. Yaroshchuk, Dielectric exclusion of ions from membranes, Adv. Colloid Interface.

Sci. 85 (2000) 193-230.

[42] W. R. Bowen, A. W. Mohammad, A theoretical basis for specifying nanofiltration

membranes - Dye/salt/water streams, Desalination 117 (1998) 257-264.

[43] R. Takagi, M. Nakagaki, Facilitated and reverse transport of electrolytes through an

asymmetric membrane, J. Membr. Sci. 27 (1986) 285-299.

[44] J.A. Manzanares, S. Mafé, J. Pellicier, Transport phenomena and asymmetry effects in

membranes with asymmetric fixed charge distributions, J. Phys. Chem. 95 (1991) 5620-5624.

[45] A. Szymczyk, H. Zhu, B. Balannec, Pressure-driven ionic transport through

nanochannels with inhomogenous charge distributions, Langmuir 26 (2010) 1214-1220.

[46] J. Cervera, B. Schiedt, R. Neumann, S. Mafé, P. Ramirez, Ionic conduction, rectification,

and selectivity in single conical nanopores, J. Chem. Phys. 124 (2006) 104706.

[47] P. Ramirez, V. Gomez, J. Cervera, B. Schiedt, S. Mafé, Ion transport and selectivity in

nanopores with spatially inhomogeneous fixed charge distributions, J. Chem. Phys. 126

(2007) 194703.

[48] D. Constantin, Z.S. Siwy, Poisson-Nernst-Planck model of ion current rectification

through a nanofluidic diode, Phys. Rev. E 76 (2007) 041202.

[49] A. Szymczyk, H. Zhu, B. Balannec, Ion Rejection Properties of Nanopores with Bipolar

Fixed Charge Distributions, J. Phys. Chem. B 114 (2010) 10143-10150.

[50] S. Bason, Y. Kaufman, V. Freger, Analysis of ion transport in nanofiltration using

phenomenological coefficients and structural characteristics, J. Phys. Chem. B 114 (2010)

3510-3517.

30
[51] A. Yaroshchuk, X. Martinez-Llado, L. Llenas, M. Rovira, J. D. Pablo, J. Flores, P. Rubio,

Mechanisms of transfer of ionic solutes through composite polymer nano-filtration

membranes in view of their high sulfate/chloride selectivities, Desalination and Water

Treatment 6 (2009) 48-53.

[52] C. Y. Tang, Y.-N. Kwon, J. O. Leckie, Effect of membrane chemistry and coating layer on

physiochemical properties of thin film composite polyamide RO and NF membranes II.

Membrane physiochemical properties and their dependence on polyamide and coating layers,

Desalination 242 (2009) 168-182.

[53] J. Schaep, C. Vandecasteele, Evaluating the charge of nanofiltration membranes, J.

Membr. Sci. 188 (2001) 129-136.

[54] Y. Lanteri, P. Fievet, A. Szymczyk, Evaluation of the steric, electric, and dielectric

exclusion model on the basis of salt rejection rate and membrane potential measurements, J.

Colloid Interface Sci. 331 (2009) 148-155.

[55] V. Freger, Kinetics of film formation by interfacial polycondensation, Langmuir 21

(2005) 1884-1894.

[56] V. Freger, Swelling and morphology of the skin layer of polyamide composite

membranes: an atomic force microscopy study, Environ. Sci. Technol. 38 (2004) 3168-3175.

[57] J. M. M. Peeters, J. P. Boom, M. H. V. Mulder, H. Strathmann, Retention measurements

of nanofiltration membranes with electrolyte solutions, J. Membr. Sci. 145 (1998) 199-209.

[58] J. Schaep, B. Van der Bruggen, C. Vandecasteele, D. Wilms, Influence of ion size and

charge in nanofiltration, Sep. Purif. Technol. 14 (1998) 155-162.

[59] P. Fievet, C. Labbez, A. Szymczyk, A. Vidonne, A. Foissy, J. Pagetti, Electrolyte

transport throuth amphoteric nanofiltration membranes, Chem. Eng. Sci. 57 (2002) 2921-

2931.

31
[60] C. Labbez, P. Fievet, A. Szymczyk, A. Vidonne, A. Foissy, J. Pagetti, Analysis of the salt

retention of a titania membrane using the “DSPM” model: effect of pH, salt concentration and

nature, J. Membr. Sci. 208 (2002) 315-329.

[61] M. Urairi, T. Tsuru, S.I. Nakao, S. Kimura, Bipolar reverse osmosis membrane for

separating mono- and divalent ions, J. Membr. Sci. 70 (1992) 153-162.

[62] J. Schaep, C. Vandecasteele, J.W. Mohammad, W.R. Bowen, Analysis of the salt

retention of nanofiltration membranes using the donnan-steric partitioning pore modes, Separ.

Sci. Technol. 34 (1999) 3009-3030.

[63] J. Schaep, C. Vandecasteele, A. W. Mohammad, W. R. Bowen, Modelling the retention of

ionic components for different nanofiltration membranes, Sep. Purif. Technol. 22-23 (2001)

169-179.

[64] A. Szymczyk, N. Fatin-Rouge, P. Fievet, C. Ramseyer, A. Vidonne, Identification of

dielectric effects in nanofiltration of metallic salts, J. Membr. Sci. 287 (2007) 102-110.

[65] S. Bason, V. Freger, Phenomenological analysis of transport of mono- and divalent ions

in nanofiltration, J. Membr. Sci. 360 (2010) 389-396.

[66] C. Ba, J. Economy, Preparation and characterization of a neutrally charged antifouling

nanofiltration membrane by coating a layer of sulfonated poly(ether ether ketone) on a

positively charged nanofiltration membrane, J. Membr. Sci. 362 (2010) 192-201.

32

You might also like