Tailoring PES nanofiltration membranes through systematic investigations of prominent design, fabrication and operational parameters

Abstract

Design and fabrication of nanofiltration (NF) membranes with the desired characteristics and separation performance is of paramount importance. In this study various asymmetric nanofiltration membranes were designed and fabricated using poly(ethersulfone) (PES) via phase inversion technique. The effects of variation in polymer concentration, solvent type, additives in the dope solution and composition of coagulating agent were studied as the selected design parameters. The effects of variation in solvent evaporation time, coagulation bath temperature, casting speed (shear rate) and membrane thickness were also investigated as the selected fabrication parameters. The results obtained reveal that increasing the polymer concentration, promoting a delay in demixing through a change of solvent and composition of coagulating agent as well as decreasing the coagulation bath temperature, increasing the solvent evaporation time and membrane thickness all result in NF membranes with less overall porosity and mean pore size, lower water flux and higher salt rejections. Furthermore, addition of hydrophilic organic acids, (e.g., ascorbic and citric acids) in the dope solution and incrementally increasing the casting shear rate promote overall porosity of the membrane, water flux and salt rejection. Membranes derived from PES/N-methyl-2-pyrrolidone (30/70 wt%) could offer maximum salt rejections of 47.35% and 99.84% for sodium chloride and magnesium sulfate, respectively. The pure water flux in the membranes could be enhanced up to 54.88 l m⁻² h⁻¹ by addition of 1 wt% citric acid into the dope solution. In terms of operational parameters, increase in both feed pressure and pH could enhance the membrane flux and salt rejections. The findings in this study provide useful guidelines and methods for the design and fabrication of high performance asymmetric NF membranes with the desired microstructure, productivity and separation performance.

---

1. Introduction

Membrane technology has gained widespread acceptance for a variety of separation applications ranging from microfiltration to gas separation, pervaporation and fuel cells.^1–6 Many activities are in progress in various parts of the world on material synthesis, membrane fabrication, membrane modification, process design and development, process modeling and optimization, all aiming to overcome the remaining obstacles and to improve the competitiveness of the membrane technologies.^7–13 Nanofiltration (NF) is a relatively new type of pressure driven membrane technology that lies between ultrafiltration and reverse osmosis.¹⁴ NF membranes are employed effectively for the separation of charged and/or uncharged species in the size range of about 1–10 nm. The governing mechanisms in NF membranes are solution diffusion, Donnan effect, dielectric exclusion, electromigration or a combination of them.¹⁵ Numerous advantages such as energy efficiency, high flux as well as a higher rejection of multivalent than monovalent ions^15,16 have led to the widespread application of NF membranes in various industries, including, but not limited to, water softening,^17,18 seawater desalination,^19,20 brackish water treatment,^21–23 dye removal,^24–26 industrial wastewater treatment and reuse,^27–29 food and beverage processing³⁰ and the pharmaceutical industry.³¹

Development of high performance NF membranes requires a thorough understanding of proper material selection, formulation, design and fabrication procedures in addition to the adoption of appropriate knowledge and skills. Phase inversion can be described as a process of making membranes in which an initially homogeneous polymer solution is transformed into a solid structure through a complicated, yet controlled procedure. Non-solvent induced phase separation (NIPS), thermal induced phase separation, evaporation induced phase separation and vapor induced phase separation are among the well-established phase inversion processes. NIPS is commonly used for fabrication of asymmetric NF membranes. According to this procedure, a polymer solution is cast on a proper substrate, then immersed in a coagulation bath containing a non-solvent or a blend of non-solvents.³² The non-solvent diffuses into the polymer solution and the solvent escapes from the polymeric solution to the non-solvent bath, known as the demixing process, through which a system comprised of a polymer rich and polymer lean phases is formed. The former results in the formation of a solid membrane matrix, while the latter creates membrane pores.³³

Essentially, various kinetic and thermodynamic parameters including selection of polymer, solvent, non-solvent and additive to the polymer solution and its composition play important roles in the properties and characteristics of the ultimate membranes.³⁴

These parameters may generally be classified into two main groups of design and fabrication as shown in Fig. 1. Design parameters may include material selection, polymer concentration in the dope, solvent type, additives in dope solution, and composition of the coagulating agent. Various materials including but not limited to polysulfone,³⁵ cellulose acetate³⁶ and poly(vinylidene fluoride)³⁷ have been used for the preparation of asymmetric NF membranes. Among them, poly(ethersulfone) (PES) has been widely used for the preparation of a variety of membranes mainly because of its particular characteristics such as prominent oxidative stability, thermal hydrolysis stability (T_g ∼ 220 °C), chemical stability and mechanical stability.^38,39 It is also shown that the polymer concentration can play a major role in the structure of the membrane. A study by Balta et al.⁴⁰ revealed that the increase in the polymer concentration led to the formation of NF membranes with less porosity and less finger-like pores and consequently decreased water flux and increased salt rejection. Membranes with different morphologies and performance can be achieved but it depends on the type and nature of the solvents and non-solvents used. Various approaches such as bulk modification, blending, additive embedding, surface coating, interfacial polymerization, grafting (e.g., photon-induced, gamma ray, electron/ion beam induced, plasma induced, thermal induced), plasma treatment, immobilization and surface initiated atom transfer radical polymerisation have been applied for altering the hydrophobic nature of the PES derived membranes.⁴¹ The use of additives consisting of functional groups has also been considered as a possible technique for altering the hydrophilicity, surface roughness, surface charge and the pore size of the membranes.^42–44 For example, pore forming agents, such as poly(vinylpyrrolidone) and poly(ethylene glycol) have been added to the polymer solution for improving the permeability of the membranes.⁴⁵ Another effective parameter is composition of coagulating agent. While water has been the dominant coagulating agent, several types of additive such as ethanol,^46,47 isopropanol,^46–48 2-butanol⁴⁹ and N-methyl-2-pyrrolidone (NMP)⁵⁰ have been examined separately or as additives to water in order to tailor the characteristics of the membranes.

Fig. 1 Classification of prominent design and fabrication parameters involved in the preparation of asymmetric membranes.

Apart from the design parameters, fabrication parameters also play an important role in the properties of the resultant membranes. For example, the influence of solvent evaporation time was investigated in some previous research.^51,52 It was also shown that increasing the coagulation bath temperature (CBT) resulted in an accelerated phase inversion rate and consequently an improved water flux.^53–55 However, in some cases, a decrease of the water flux has been reported.⁴⁶ The casting speed also affects the membrane morphology, porosity and pore size which is attributed to the alternation of the molecular orientation caused by the induced shear rate.^35,56–58 It was also shown that the overall thicknesses of the membrane as well as the top layer increased by increasing the casting knife gap.^59–61

The main objective of the present research was to perform a systematic investigation on the effect of key parameters involved in the development of asymmetric NF membranes from the viewpoint of formation, structural characteristics, productivity and separation performance. For this purpose, the effect of various design parameters such as polymer concentration in the dope solution, solvent type, additives in the dope solution, composition of the coagulating agent as well as various fabrication parameters such as solvent evaporation time, coagulation bath temperature, casting speed (shear rate) and membrane thickness on the characteristics and performance of the membranes were investigated and determined. The effect of feed pressure and pH as operational parameters were also studied in detail. To the best of our knowledge, and based on the detailed review and analysis of the previous research, this is the first comprehensive report on the systematic investigation of NF membranes. Furthermore, the effect of the addition of selected hydrophilic additives (i.e., ascorbic and citric acids) into the dope solution was assessed for the first time for PES NF membranes.

2. Experimental

2.1. Materials

Polyethersulfone [PES, Ultrason E 6020 P, molecular weight (MW) = 58 000 g mol⁻¹] was supplied by BASF (Germany). N- Methyl-2-pyrrolidone (NMP) and N,N-dimethylacetamide (DMAc) both from Merck (Germany) were used as the solvent for the preparation of dope solutions mainly because of their good miscibility with PES. Ascorbic acid (C₆H₈O₆, pK_a = 4.10) and citric acid (C₆H₈O₇, pK_a = 3.15) both from Merck (Germany) were employed as additives for the dope solutions. Sodium chloride (NaCl) and magnesium sulfate (MgSO₄) salts were obtained from Merck. Ethanol and distilled water were used as coagulating agents. The chemical structure of the materials used in this study are illustrated in Fig. 2.

Fig. 2 Chemical structures of (a) PES (b) NMP (c) DMAc (d) ascorbic acid (e) citric acid.

2.2. Membrane preparation

Dope solutions were prepared by dissolving the specified amounts of PES flakes in the respective solvents. Ascorbic and citric acids were used in the specified amounts to the selected dope solutions as additives. The solutions were stirred for at least 24 h at 300 rpm and at room temperature to ensure complete dissolution. After formation of a homogeneous solution, the dope solutions were held at the ambient temperature for about 24 h and then degassed. Then the solutions were cast onto a glass plate using a semi-automatic film applicator with an adjustable gap at controlled temperature and relative air humidity. The nascent films on the glass plate were immersed immediately in the coagulation bath (except in one case which was used to investigate the effects of the solvent evaporation time) and remained there for 24 h to allow the residual solvents to leach out. Finally, the membranes were stored between two filter papers at room temperature and dried. The controlled conditions for the membrane formation are given in Table 1. The effect of various design parameters including polymer concentration in the dope solution, solvent type, additive type and amount in dope solution, composition of coagulating agent as well as fabrication parameters such as solvent evaporation time, coagulation bath temperature, casting speed (shear rate) and membrane thickness on the characteristics and performance of the membranes were studied systematically.

Table 1 The effect of solvent type and polymer concentration in dope solution on the characteristics of the PES NF membranes^a

Solvent	Polymer concentration (wt%)	Membrane thickness (μm)	Mean pore size (nm)	Porosity (%)	Area fraction of macrovoids (%)	Contact angle (°)
a Conditions: casting temperature: 25 ± 3 °C; relative air humidity: 30 ± 5%; casting knife gap: 200 μm; casting shear rate: 23.54 s^−1; solvent evaporation time: 0 min; coagulating agent: pure water; coagulation bath temperature: 20 ± 2 °C.
NMP	24	125	4.93	55.5	54.96	63.4
	27	130	3.19	52.3	—	72.6
	30	133	1.20	46.2	40.58	75.5
DMAc	24	100	4.96	58.9	65.78	60.5
	27	107	4.09	55.7	—	67.3
	30	111	1.64	47.6	48.54	72.9

---

2.3. Membrane characterizations

2.3.1. Morphology analysis. A VEGA (TESCAN, Czech Republic) scanning electron microscope (SEM) was used for characterization of the cross-sectional morphology of the fabricated membranes. The membrane samples were fractured in liquid nitrogen and then sputtered with gold. ImageJ software (version 1.48) was used for further analysis of the types and population of macrovoids in the cross-section of membranes.

2.3.2. Porosity and pore size measurements. Overall porosity (ε) was calculated as a function of the membrane weight using the following equation:⁶²where w₁ and w₂ are the weights (kg) of the wet and dry membranes, A is the membrane effective area (m²), ρ is the density of water (998 kg m⁻³) and l is the membrane thickness (m). The Guerout–Elford–Ferry equation (eqn (2)) was used to determine the mean pore radius (r_m) of the membrane using pure water flux and porosity data:⁶³where μ is the viscosity of water (8.9 × 10⁻⁴ Pa s), Q is the volumetric flow rate of the permeated water (m³ s⁻¹) and ΔP is the operating pressure (Pa). The average pore size, r_m, was determined by filtration velocity method, in which the pure water flux of the wet membrane was measured by applying pressure (1 MPa) for a limited period.

2.3.3. Contact angle measurements. To evaluate membrane surface properties, contact angles were measured using the sessile drop method with a G10 goniometer (KRÜSS, Germany). Deionized water was used as the probe liquid in all the measurements. To minimize the experimental error, contact angle was measured at five random locations for each sample at room temperature, and the average value was reported.

2.4. Performance evaluation

The performance of the membranes was evaluated using a bench scale set-up as shown in Fig. 3. Rectangular shaped membrane sheets with an effective surface area of 15 cm² were cut and sandwiched between the test cells. The feed was pumped into the cell by using a high pressure pump, flowing tangentially to the membrane surface. The permeate stream was directed to the permeate tank while the retentate stream returned to the feed tank. In this way, the concentration of the feed remains almost unchanged because of the limited amount of permeate stream. The feed flow rate and stream pressures were adjusted using a back pressure regulator and by-pass valves. The following procedure was used in all experiments:

Fig. 3 Cross-flow filtration system: (1) feed tank (2) pump (3) valve (4) by-pass stream (5) pressure gauge (6) membrane module (7) back-pressure regulator (8) flowmeter (9) retentate stream (10) permeate stream (11) permeate tank (12) temperature control system.

(1) Before starting the experiments, each membrane was soaked in deionized water for 2 h to provide the required saturation and wetting in the membrane structure.

(2) Each membrane sheet was firstly subjected to pre-compaction at 1.5 MPa for 2 h using pure water. However, pure water permeate flux (J) was measured at 1 MPa and feed flow rate of 5 l min⁻¹ using eqn (3):

where V is the volume of permeated water (l), A is the membrane area (m²) and Δt is the permeation time (h).

(3) The single salt rejection of the membranes was measured using a synthetically designed salt solution (formula based on typical seawater properties) containing 10 500 ppm of NaCl and 1300 ppm of MgSO₄.

(4) The feed solution was circulated for about 1 h until the system reached steady state, and then the permeate flux was measured at 1 MPa and feed flow rate of 5 l min⁻¹. Samples from the feed and permeate were collected and analyzed separately using a standard MIC 99702 conductivity meter (Mic Meter, Taiwan).

(5) The single salt rejection (R) was determined using eqn (4):

where C_p and C_f are the concentrations of ion in the permeate and feed, respectively.

3. Results and discussion

3.1. Effect of the polymer concentration in dope solution on the structure and performance of NF membranes

The effect of polymer concentration in the dope solution on the water flux and salt rejection of the NF membranes prepared using NMP and DMAc as solvents is shown in Fig. 4. It can be seen that polymer concentration has a considerable effect on the membrane permeation and separation performance. According to Fig. 4(a), the pure water flux decreased from 38.35 l m⁻² h⁻¹ to 2.78 l m⁻² h⁻¹ for the membranes prepared using 24 wt% and 30 wt% PES in DMAc, respectively. Similarly, the pure water flux decreased from 27.65 l m⁻² h⁻¹ to 1.24 l m⁻² h⁻¹ for the membranes prepared using 24 wt% and 30 wt% PES in NMP, respectively. This can be attributed to the fact that an increase in the PES concentration results in an increment in the dope viscosity. This increase in viscosity consequently brings about a reduction of the solvent and non-solvent exchange rate, and thus the polymer concentration increases at the interphase of the dope solution and the coagulating agent. Thus, smaller amounts of coagulating agent can penetrate to the dope solution and the precipitation rate decreases through delayed demixing. It is well demonstrated that the instantaneous demixing leads to a finger-like structures, while delayed demixing leads to a sponge-like structures.⁶⁴ Fig. 5 presents SEM cross sectional images of membranes prepared using dopes having different polymer concentrations. It can clearly be seen that regardless of the solvent type, membranes formed from a low polymer concentration (i.e., 24 wt%) were comprised of finger-like macrovoids with interconnections extended to the bottom of the membranes. However, increasing the polymer concentration suppressed the formation of the finger-like macrovoids and created more spongy-like porous structures. Thus, it can be seen that an increase in polymer concentrations and delayed demixing resulted in membranes with lesser porosity, less finger-like pores, lower mean pore size, while there was a denser and thicker layer on the top. Data in Table 1 provide a good overview on the trend of changes in the morphological and structural characteristics of the membranes as a result of changes in polymer concentration. The data reveal that an increase in the polymer concentration led to the formation of up to about 11% thicker NF membranes. Similar results have been reported elsewhere.⁴⁶ This may arise from the fact that a higher polymer concentration provides less opportunity for the penetration of the coagulating agent and accordingly less contraction ratio is achieved because of the slowed demixing rate.

Fig. 4 The effect of PES concentration in the dope solution on (a) pure water flux (b) NaCl rejection (c) MgSO₄ rejection of NF membranes prepared using NMP and DMAc as solvents.

Fig. 5 Cross section SEM images of PES NF membranes: (a) polymer concentration: 24 wt%; solvent: NMP, (b) polymer concentration: 30 wt%; solvent: NMP, (c) polymer concentration: 24 wt%; solvent: DMAc, (d) polymer concentration: 30 wt%.; solvent: DMAc.

On the other hand, by increasing the polymer concentration, the mean pore size decreased from 4.93 nm to 1.20 nm and from 4.96 to 1.64 in NF membranes prepared using NMP and DMAc, respectively. Similarly, the porosity decreased by about 17% and 19% in NF membranes prepared using NMP and DMAc, respectively. The higher porosity can be attributed to a higher diffusional exchange between the solvent and coagulating agent at a lower polymer concentration.⁴⁶ According to the data in Table 1 and also shown in Fig. 5, an increase in the polymer concentration produced a reduction in the area fraction of macrovoids in the cross-section of membranes, a reduced number of macrovoids and turned the overall membrane morphology into a more spongy-like structure. In addition, the contact angle of the prepared membrane increased from 63.4° to 75.5° and from 60.5° to 72.9° in NF membranes prepared using dope solutions containing 24 wt% and 30 wt% of polymer in NMP and DMAc, respectively. Sotto et al.⁶² determined that increasing the polymer concentration decreases the solvent and coagulating agent diffusional exchange rate and thus, the surface porosity and the pore size decreases. It should be noted that the wettability of the membrane is influenced by the membrane material as well as the surface porosity and roughness. Thus, because the materials are the same, the greater contact angle of the membrane prepared using the dope with a higher concentration can be mainly attributed to the decreased surface porosity.⁶⁵ In other words, a higher polymer concentration leads to a lower surface porosity causing a decrease in the hydrophilicity of the membrane and the water flux.

The effect of the polymer concentration in the dope solutions on the performance of NF membranes was investigated in terms of monovalent and divalent salt rejections. Generally, an increase in polymer concentration led to improvements in the rejection of both NaCl and MgSO₄ by the membranes. According to Fig. 4(b), NaCl rejection increased from 23.36% to 47.35% and from 18.62% to 40.80% in NF membranes prepared using dope solutions containing 24 wt% and 30 wt% of polymer in NMP and DMAc, respectively. Membranes gave even higher rejections for MgSO₄. According to Fig. 4(c), MgSO₄ rejection increased from 86.35% to 99.84% and 82.78% to 99.10% in NF membranes prepared using dope solutions containing 24 wt% and 30 wt% polymer in NMP and DMAc, respectively. Because of the negative surface charge of the PES membranes, the separation mechanism of these membranes is not only affected by steric hindrance, but also by charge effect interaction. According to the Donnan effect, if a negatively charged ion is rejected because of electrostatic surface charge of the PES membrane, then the counter ion will also have to be rejected to neutralize the electric charge across the membrane. For salts, an increase in co-ion charge and a decrease in counter ion charge improve the rejection of salts.⁶⁶ Because of the higher charge of sulfate ions, the rejection of Mg²⁺ ions also increased because of the Donnan effect. On the other hand, as reported by Ali et al.,⁵⁸ the surface charge density of the membrane increases as the membrane pore size decreases. As shown in Table 1, membranes prepared using solutions with high polymer concentrations possessed smaller pores and high surface charge density. Consequently, a smaller pore size increases the rejection of salts through both physical size sieving and electrostatic repulsion mechanisms.

3.2. The effect of solvent type on the structure and performance of NF membranes

Several studies have demonstrated the important role of solvent selection on the membrane properties and performance. This was also investigated in the present study and the effects of solvent type on pure water flux and rejection of salts are shown in Fig. 4. According to Fig. 4(a), membranes prepared using DMAc as the solvent exhibited higher water flux than those prepared using NMP. This trend was regardless of the polymer concentration in the dope solution. It should be noted that the difference between the pure water flux of the membranes prepared by the two solvents was minimal when the polymer concentration in the dope solution was 30 wt%. The pure water flux in the NF membranes prepared using dope solutions containing 24 wt% and 30 wt% of PES in DMAc was 38.35 l m⁻² h⁻¹ and 2.78 l m⁻² h⁻¹, respectively. However, the pure water flux in NF membranes prepared using dope solutions containing 24 wt% and 30 wt% PES in NMP was 27.65 l m⁻² h⁻¹ and 1.24 l m⁻² h⁻¹, respectively. Therefore, the membrane prepared using DMAc exhibited a higher pure water flux compared to those prepared using NMP. As shown by the data in Table 1, this result can be attributed to the higher porosity, lower thickness and hydrophilic nature of the membrane prepared using DMAc. As porosity and hydrophilicity increases and membrane thickness decreases, the water permeation through the membrane increases.

Analysis of the morphology of the membranes can provide useful insights about the effect of solvent type on the properties of the membranes. As shown by the cross-sectional SEM images in Fig. 5, switching the solvent from DMAc to NMP enabled suppression of the number and progress of macrovoids to a certain extent. Instead, relatively large macrovoids were formed.

The following points provide insights about the effect of solvent type on the morphology and performance of NF membranes:

(1) The solubility parameter and mutual diffusivity between the solvent and non-solvent in the coagulation bath are important factors in membrane preparation.⁶⁷ A smaller difference between the solubility parameters results in faster/instantaneous demixing causing formation of a porous top layer and finger-like pores in the support layer. However, a delay in demixing is often related to the formation of a dense layer. The solubility parameter data for various components are shown in Table 2. They indicate that the difference in solubility parameters between DMAc and water is the highest and this is attributed to the fact that the porosity, number of finger-like pores and mean pore size of membranes prepared using this solvent is increased. In addition, the diffusivity coefficient of the intended solvent and coagulating agent systems was calculated using the Wilke–Chang equation and is shown in Table 3. According to eqn (5), the diffusivity of solute a into solvent b (D_a–b(cm² s⁻¹)) can be described by:

where φ, μ, V_a, M_b and T are the association factor, viscosity (g cm⁻¹ s⁻¹), molar volume at boiling point (cm³ kmol⁻¹), molecular weight of the solvent (g kmol⁻¹), and temperature (K), respectively. The average mutual diffusivities of the DMAc–water system is higher than that of the NMP–water system. By increasing the diffusion rate between the solvent and the non-solvent, instantaneous demixing occurred which led to an increase in porosity in the top and the support layers.⁶⁷ A strong interaction between solvent and non-solvent is related to a small difference between solubility parameters of the solvent and non-solvent. A lower mutual diffusivity and lower porosity are the results of a strong interaction between the solvent and non-solvent.⁶⁸

Table 2 Solubility parameters of the materials and compounds used in this study

Material/compound	Solubility parameter (MPa^0.5)
	δd	δp	δh	δt
PES	19.6	10.8	9.2	24.19
NMP	18.0	12.3	7.2	22.9
DMAc	16.8	11.5	10.2	22.7
Water	15.5	16	42.3	47.8
Methanol	15.1	12.3	22.3	29.6
Water–methanol (75/25 vol%)	15.4	15.08	37.3	43.07
Water–methanol (50/50 vol%)	15.3	14.14	32.26	38.4

---

Table 3 Mutual diffusivity of various coagulating agent (c) and solvent (s) systems and the average values (D_m)

System	Diffusivity coefficient (cm^2 s^−1)
	Dc–s × 10^6	Ds–c × 10^6	Dm × 10^6
Water–NMP	11.56	7.92	9.74
Water–DMAc	20.89	8.10	14.49
Methanol–DMAc	12.01	6.12	9.06

---

(2) Another aspect is a consideration of the delay time (time between immersion of a polymer film into the coagulation bath to the inception of liquid–liquid demixing) and gelation time (time between demixing and solidification) and their role in the phase diagram (Fig. 6(a)).⁶⁹ A finger-like structure with very thin skin layer is generally expected with the instantaneous demixing process, whereas a sponge-like structure with a thick skin layer is usually observed with the delayed demixing process.⁶⁷ According to the three phase diagram for water–DMAc–PES and water–NMP–PES systems (Fig. 6(b)), the theoretical and experimental binodal curve of the water–DMAc–PES system is closer to the polymer–solvent axis than that of the water–NMP–PES system. Therefore less water is needed for the precipitation of PES (instability of polymer film) in the water–DMAc–PES system.⁷⁰ The time to first precipitation may be almost instantaneous for the water–DMAc–PES system. The final structure of the membrane depends on whether the delayed or instantaneous demixing is dominant. As shown by Fig. 6(b), instantaneous demixing occurred for the water–DMAc–PES system when compared to that of the water–NMP–PES system; resulting in the formation of a thin skin layer and finger-like pores in the support layer.⁶⁵

Fig. 6 (a) Schematic representation of a ternary phase diagram;⁶⁹ (b) theoretical binodal (—) and spinodal (– –) curves for water–DMAc–PES and water–NMP–PES systems. Experimental cloud point data (•) are used for verification of the theoretical calculations.⁷⁰

(3) As reported by Hasbullah et al.,⁷¹ the boiling point of the solvent contribute positively to the dope solution viscosity. Accordingly, the higher boiling point of NMP (202 °C) compared to that of DMAc (166 °C) results in dope solutions with increased viscosity. As mentioned previously, this increase in viscosity resulted in a reduction of the solvent and non-solvent exchange rate and delayed demixing occurred. Consequently, a membrane with lower porosity, less finger-like pores and lower mean pore size was formed.

The data in Table 1 provide a good overview on the trend of changes in the morphological and structural characteristics of membranes as a result of solvent selection. For example, the data revealed that membranes prepared using DMAc as the solvent experienced more contraction and possessed a lower thickness than their counterparts prepared using NMP. In addition, the mean pore size of the membranes prepared using DMAc was relatively high especially when higher dope concentrations were used. This is in good agreement with the pure water flux data. The exchange rate between solvent and coagulating agent affects the pore size. It should be noted that the diffusional exchange rate between the solvent and coagulating agent in the DMAc–water system is faster than that in the NMP–water system and because of delayed demixing, the solvent outflow is faster than the coagulating agent inflow. In this case, the pores tend to collapse and shrink.⁷² However, the porosity of the prepared membrane decreased from 55.5% to 46.2% and from 58.9% to 47.6% in NF membranes prepared using dope solutions containing 24 wt% and 30 wt% polymer in NMP and DMAc, respectively. In fact, when DMAc was used as solvent, instantaneous demixing occurred which led to the formation of a more porous structure as also reported previously by others.⁶⁷ As shown in Table 1, when NMP was used as a solvent, the solvent and coagulating agent diffusional exchange rate decreased and thus, the surface porosity and pore size decreased. The greater contact angle of NMP-based membranes in comparison with DMAc-based membranes could be attributed to the decreased surface porosity.⁶² Thus, DMAc-based membranes exhibited a higher pure water flux. This improvement in pure water flux could be attributed to an increase of membrane hydrophilicity.

Considering the effect of solvent type on the membrane performance, as shown in Fig. 4(b), it can be seen that NaCl rejection increased from 23.36% to 47.35% and from 18.62% to 40.80% in NF membranes prepared using dope solutions containing 24 wt% and 30 wt% polymer in NMP and DMAc, respectively. According to Fig. 4(c), MgSO₄ rejection increased from 86.35% to 99.84% and 82.78% to 99.10% in NF membranes prepared using dope solutions containing 24 wt% and 30 wt% polymer in NMP and DMAc, respectively. The rejection of both monovalent and divalent salts was higher in NMP-based membranes in comparison to DMAc-based membranes which is related to the smaller pore size in the NMP-based membrane which increases the rejection of salts through both physical size sieving and electrostatic repulsion mechanisms.

3.3. The effect of additive type and concentration in dope solution on the properties and performance of NF membranes

The effect of the addition of ascorbic acid as a polyhydroxy acid and citric acid as a polyacid and the effect of their concentration on the characteristics and performance of NF membranes were investigated. This was based on the idea that the incorporation of these acids may enhance the performance of the membranes by altering their surface and structural characteristics. The investigation was carried out on the NF membranes prepared using a dope solution containing 24 wt% PES dissolved in DMAc. As shown in Fig. 7(a), it can be seen that addition of both acids resulted in enhanced pure water flux in the membranes. The pure water flux of PES membranes increased from 38.35 l m⁻² h⁻¹ (bare) to 52.99 l m⁻² h⁻¹ and 54.88 l m⁻² h⁻¹ by addition of 1 wt% of ascorbic and citric acids, respectively. However, further addition of the acids beyond 1 wt% had negative effects. The pure water flux of the modified membrane was expected to be higher than that for the unmodified membrane. As shown in Table 4, improvement of water flux can be attributed to membrane thickness, porosity and hydrophilicity. Structural analysis and characterization of membranes can provide useful information about the performance of the membranes. The porosity of the membranes increased to 75.87% and 77.68% upon addition of ascorbic and citric acids up to 1.0 wt%, respectively. This could be attributed to the interactions between the additive with a component of the casting solution (polymer and solvent). By addition of both organic acids into the dope solution, hydrogen bonds were formed between the hydrogen atoms of the organic acids and the oxygen atoms of the PES, thus, weaker interactions occurred between polymer chains. In addition, the hydrophilic nature of the organic acids and the formation of hydrogen bonds between additive molecules and DMAc, decreases the ratio of the solvent outflow to the coagulating agent (water) inflow which causes instantaneous demixing.^73,74 Instantaneous demixing results in the formation of macrovoids, a porous top and support layers and a lower concentration of the polymer at the surface which leads to a thinner skin layer.⁶⁷ However, further addition of organic acids beyond 1 wt% resulted in an increase in binding force between the polymer chains and additive molecules and thus, the remaining free volume was reduced because of the tighter and stronger entanglements. Furthermore, according to the data in Table 4, and by comparing to the thickness of the reference membrane (100 μm), thinner membranes were obtained by using dope solutions containing either ascorbic or citric acids. The addition of additives decreased the ratio of the solvent outflow to the coagulating agent (water) inflow providing more opportunity for the penetration of coagulating agent and accordingly a bigger contraction ratio was obtained because of the faster demixing.⁷³ It was also found that the addition of both acids improved the hydrophilic properties of the membranes and this was largely attributed to the hydrophilic functional groups of the organic acids.⁷³ The effect of ascorbic acid on the contact angle of the membranes was greater than that of citric acid. The contact angle reduced from 60.5° to 38.6° and 45.3° upon addition of 1.5 wt% of ascorbic and citric acids, respectively.

Fig. 7 The effect of type and concentration of organic acids as additives to the dope solutions on (a) pure water flux (b) NaCl rejection (c) MgSO₄ rejection in NF membranes.

Table 4 The effect of additives in dope solutions on the structural characteristics of PES NF membranes^a

Additive type	Additive concentration (wt%)	Solvent concentration (wt%)	Membrane thickness (μm)	Mean pore size (nm)	Porosity (%)	Contact angle (°)
a Conditions: dope solution: 24 wt% PES in DMAc, casting temperature: 25 ± 3 °C; relative air humidity: 30 ± 5%; casting knife gap: 200 μm; casting shear rate: 23.54 s^−1; solvent evaporation time: 0 min; coagulating agent: pure water; coagulation bath temperature: 20 ± 2 °C.
None	0	76.0	100	4.96	58.9	60.5
Ascorbic acid	0.5	75.5	93	4.63	67.8	53.1
	1.0	75.0	86	4.51	75.8	43.4
	1.5	74.5	91	4.46	73.8	38.6
Citric acid	0.5	75.5	96	4.73	66.9	55.7
	1.0	75.0	86	4.44	77.6	48.6
	1.5	74.5	89	4.21	74.7	45.3

---

Also, as shown in Fig. 7(b) and (c), it can be seen that the rejection of both monovalent and divalent salts increased by increasing the concentration of both organic acids in the dope solutions. Size exclusion (steric hindrance) and electrostatic charge repulsion are two important mechanisms that may affect the rejection of ionic solutes. It was shown that PES membranes possess a negative surface charge because of the breakdown of surface functional groups or adsorption of ions from the aqueous solution.⁷⁵ As shown in Table 4, mean pore size of the modified membrane was smaller than that of the unmodified membrane. An increase in salt rejection by addition of organic acids might be attributed to two factors: (i) reduction of pore size of the membrane which promoted the salt rejection by a size exclusion mechanism, and (ii) increase in surface negative charge and reduction of pore size (higher surface charge density) which promoted the salt rejection through the electrostatic charge repulsion.⁷³

3.4. The effect of type and concentration of coagulating agent on the properties and performance of NF membranes

The effects of adding methanol to the coagulating agent (distilled water) were investigated for the NF membranes prepared using a dope solution containing 24 wt% PES dissolved in DMAc. In order to study the effects of the composition of the coagulating agent on the structure and performance of the PES membrane, three different compositions were used (Table 5). The characteristics and performance of the membranes prepared at different compositions of coagulation agents are shown in Table 5. The pure water flux for different membranes are given in Table 5. The pure water flux of the membranes reduced from 38.35 l m⁻² h⁻¹ to 29.15 l m⁻² h⁻¹ when methanol was added to the coagulation bath from 0 to 50 vol%. This is because of the formation of membranes with lower porosity, smaller pore size and lesser hydrophilicity and increased thickness. Data are in good agreement with those for the membrane structural properties. By using a mixture of methanol and water as coagulation agent, the porosity of prepared membrane decreased from 58.9% to 52.74% and the mean pore size decreased from 4.96 nm to 4.68 nm in NF membranes prepared using a coagulating agent containing water and mixtures of water–methanol (50/50 vol%). Furthermore, according to the data in Table 5, a thicker membrane (107 μm) was obtained from using the coagulation bath containing a mixture of water–methanol (50/50 vol%) when compared to the reference membrane (100 μm).

Table 5 The effect of coagulating agent on the structural characteristics of PES NF membranes^a

Coagulating agent	Membrane thickness (μm)	Mean pore size (nm)	Porosity (%)	Contact angle (°)	Pure water flux (l m^−2 h^−1)	NaCl rejection (%)	MgSO4 rejection (%)
a Conditions: dope solution: 24 wt% PES in DMAc, casting temperature: 25 ± 3 °C; relative air humidity: 30 ± 5%; casting knife gap: 200 μm; casting shear rate: 23.54 s^−1; solvent evaporation time: 0 min; coagulation bath temperature: 20 ± 2 °C.
Water	100	4.96	58.9	60.5	38.35	18.62	82.78
Water–methanol (75/25 vol%)	103	4.93	57.2	62.1	35.13	19.17	89.35
Water–methanol (50/50 vol%)	107	4.68	52.7	65.9	29.15	25.20	91.22

---

The following points provide more insights about the effect of the composition of the coagulating agent on the structure and performance of NF membranes:

(1) If alcohols are used as the coagulating agent, the binodal curve in the ternary phase diagram shrinks significantly and shifts toward the coagulating agent–polymer axis.⁷⁶ This signifies that more amounts of coagulating agent are needed for the casting solution to become thermodynamically unstable and to reach a binodal curve. As found in previous work,⁷⁷ the delay time increases gradually when the coagulating agent was changed from water to alcohol. This results in delayed demixing which leads to the formation of a thicker membrane with a smaller pore size in the top layer and lower porosity in the support layer.⁴⁶

(2) As explained in Section 3.2, polymer–solvent and solvent–coagulating agent interactions as well as the mutual diffusivity between the solvent and the non-solvent have great effects on membrane performance and structure. If the difference between the solubility parameter of solvent and coagulating agent is low, the coagulating agent penetrates slowly into the cast film and delayed demixing occurs.⁶⁷ According to the solubility parameters shown in Table 2, by addition of methanol into the coagulation bath, the difference between the solubility parameters of the solvent–coagulating agent decreased and led to delayed demixing.⁷⁸ From the kinetic viewpoint, it is possible to correlate the diffusional exchange of solvent–coagulating agent to the demixing process. Generally, a higher diffusion of coagulating agent in the solvent results in a faster precipitation. The average mutual diffusivity for the intended solvent–coagulating agent systems are given in Table 3. The diffusivity coefficient between the solvent and coagulating agent was calculated using the Wilke–Chang equation (eqn (5)). The average mutual diffusivity of the DMAc–water system is higher than that of DMAc–alcohol system. Therefore, the lower exchange rate between solvent and methanol as coagulating agents leads to delayed demixing which is associated with a thicker membrane, smaller pore size in the top layer and lower porosity.

(3) The difference in solubility parameters between PES and the coagulating agent is in the order: (PES–water) > [PES–(25%methanol–75%water)] > [PES–(50%methanol–50%water)]. The decrease in the difference between the solubility parameter of the polymer and the coagulating agent usually implies a slower precipitation rate which results in delayed demixing and as a consequence, a thicker membrane with a lower porosity and smaller pore size is formed.⁷⁹

The contact angle of the prepared membrane increased as methanol was added to the coagulation bath. Because the materials and dope solution composition were the same, the greater contact angle of the membrane prepared in the coagulation bath containing methanol can be mainly attributed to the decreased surface porosity.⁶⁵ The addition of methanol into the coagulation bath decreased the solvent and coagulating agent diffusional exchange rate and thus, the average membrane surface porosity decreased and as a consequence, the wettability of the membrane decreased.⁶²

The effect of composition of the coagulating agent on the performance of the NF membranes was investigated in terms of rejection of monovalent and divalent salts. Generally, an increase in the methanol fraction in the coagulating agent led to improvements in rejection of both NaCl and MgSO₄. As shown in Table 5, the salt rejection increased from 18.62% to 25.2% for the NaCl solution and from 82.78% to 91.22% for the MgSO₄ solution when the mixture of water and methanol were used as the coagulating agent. The improvement in salt rejection can be explained by a reduction in the mean pore size of the membrane and enhanced size sieving effect.

3.5. The effect of solvent evaporation time on the properties and performance of NF membranes

The time interval between the polymeric solution casting and immersion of the cast film into the coagulation bath is called the solvent evaporation time. This period significantly alters the structure and performance of the fabricated membrane. Structural analysis and characterization of the membranes can provide useful information about the performance of the membranes. According to the data shown in Table 6, and in comparison with the data for the reference membranes, the porosity of the membranes decreased from 58.90% to 53.76% and mean pore size decreased from 4.96 nm to 4.53 nm at 0 and 3 min solvent evaporation times, respectively. The changes in porosity and mean pore size could be attributed to the increase of polymer concentration in the top layer of the nascent film as well as the entanglement of macromolecules caused by the relaxation of the polymer chains at longer solvent evaporation times.⁸⁰ As the polymer concentration increases in the top layer, it acts as a more impenetrable barrier against the solvent outflow and coagulating agent inflow, which decreases the diffusional exchange rate. As a consequence, a membrane with a lower porosity and smaller pore size is formed.⁸¹ Furthermore, the thickness of the membrane increased from 100 μm to 107 μm over longer solvent evaporation times. This may arise from the fact that the increase in the polymer concentration at the top layer of the cast film provides less opportunity for the penetration of coagulating agent and therefore, less contraction ratio is provided because of the slowed demixing rate.⁴⁶

Table 6 The effect of solvent evaporation time and coagulation bath temperature on the structural characteristics of PES NF membranes^a

Solvent evaporation time (min)	Coagulation bath temperature (°C)	Membrane thickness (μm)	Mean pore size (nm)	Porosity (%)	Contact angle (°)	Pure water flux (l m^−2 h^−1)	NaCl rejection (%)	MgSO4 rejection (%)
a Conditions: dope solution: 24 wt% PES in DMAc, casting temperature: 25 ± 3 °C; relative air humidity: 30 ± 5%; casting knife gap: 200 μm; casting shear rate: 23.54 s^−1; coagulating agent: pure water.
0	20 ± 2	100	4.96	58.9	60.5	38.35	18.62	82.78
3	20 ± 2	107	4.53	53.7	63.4	26.73	30.65	91.58
0	0 ± 2	111	4.58	52.8	63.3	25.35	31.33	92.25
0	20 ± 2	100	4.96	58.9	60.5	38.35	18.62	82.78
0	50 ± 2	93	5.13	65.2	54.8	52.12	9.25	72.95

---

Membranes prepared over a longer solvent evaporation time exhibited a lower water flux. The pure water flux decreased from 38.35 l m⁻² h⁻¹ to 26.73 l m⁻² h⁻¹ as the solvent evaporation time increased from 0 min to 3 min. The pure water flux was directly proportional to the porosity and pore size and inversely proportional to the membrane thickness. At longer solvent evaporation times, the membrane porosity and pore size decreased and the membrane thickness increased causing a lower water flux. Increase in solvent evaporation time considerably affected the rejection of salts. As shown in Table 6, the rejection of salts increased from 18.62% to 30.65% for NaCl solution and from 82.78% to 91.58% for MgSO₄ solution at 0 and 3 min solvent evaporation time, respectively. This is associated with the diminution of pores that improve the rejection of salts through both size sieving and electrostatic charge repulsion (because of the increase in surface charge density).⁵⁸ About the contact angle, no clear changes can be inferred. A minor increase in contact angle by increasing the solvent evaporation time may be attributed to the lower surface porosity.⁶⁵

3.6. The effect of coagulation bath temperature (CBT) on the properties and performance of NF membranes

The effects of three different levels of CBT (0 °C, 20 °C, 50 °C) on membrane structure and performance are shown in Table 6. The pure water flux of the PES NF membrane prepared at a CBT of 0 °C was 25.35 l m⁻² h⁻¹. When the CBT increased to 50 °C, the pure water flux increased greatly to 52.12 l m⁻² h⁻¹. The increase of pure water flux may be attributed to three reasons. The first is the lower resistance caused by the thinner top layer; the second is the enlarged pores on the top layer; and the third is the higher porosity. In general, the porous structure with macrovoids occurred with an instantaneous demixing condition and demixing is faster at a higher CBT. On the other hand, at a lower CBT, delayed demixing occurs causing inhibition of the free growth of limited nuclei on the top layer, and instead many small nuclei are formed and distributed throughout the cast film. Accordingly, despite the instantaneous demixing, formation of macrovoids is suppressed and denser membranes are formed.⁶⁴

The porosity of the prepared membranes increased from 52.89% to 65.29% and the membrane thickness decreased from 111 μm to 90 μm in NF membranes prepared in coagulation bath temperatures of 0 °C and 50 °C, respectively. The diffusivity coefficient between the solvent and the coagulating agent is temperature dependent. According to eqn (5), an increase in temperature enhances the diffusional exchange rate between the solvent and the coagulating agent in the casting solution during the solidification process which leads to instantaneous demixing and thus, induces the formation of a membrane with a more porous structure. Decrease in the diffusivity coefficient by temperature slows down the penetration of the coagulation agent and thus, the contraction ratio is less because of the slowed demixing rate. Accordingly a thicker membrane was obtained.

The CBT values have a great influence on membrane performance in rejection of salts. By decreasing the CBT, the NaCl rejection increased from 9.25% to 31.33% and the MgSO₄ rejection increased from 72.95% to 92.25% in NF membranes prepared in coagulation bath at temperatures of 0 °C and 50 °C, respectively. Decrease of the CBT can also increase the dope viscosity, the chain rigidity, and the surface tension of the solution which will cause slow precipitation and a membrane with a smaller pore size is formed.⁸² As a result, the rejection of salts is increased.

The contact angle can be improved by increasing the CBT. Similar results have been reported previously.⁸² It should be noted that the wettability of the membrane as well as the surface porosity and roughness are influenced by the membrane material. Thus, because the materials and dope solution composition were the same, the greater contact angle of the membrane prepared at the higher CBT can mainly be attributed to the decreased surface porosity.⁶⁵

3.7. The effect of membrane casting speed on the properties and performance of NF membranes

Casting speed significantly affects the membrane performance and structure. The shear rate experienced during casting was calculated using the following relationship:

As shown in Table 7, the membrane thickness decreased by increase in the applied shear rate. Consequently, the pure water flux was affected by the membrane thickness and increased from 25.13 l m⁻² h⁻¹ to 41.32 l m⁻² h⁻¹ in membranes prepared at shear rates of 9.23 s⁻¹ and 42.76 s⁻¹, respectively. Based on the structural data, a thinner membrane with more porosity was formed at a higher shear rate resulting in higher pure water flux.⁵⁶ These findings are in agreement with other results reported in the literature.^56,58 Kusworo et al.⁵⁷ reported that at higher shear rates, a decrease in dope viscosity occurred because of the reduction of chain entanglement. Boussu et al.⁴⁶ claimed that at a lower dope viscosity, the diffusional exchange rate between the solvent and the coagulating agent increased causing the formation of a porous structure. Furthermore, by increasing the shear rate, the pore size of the membranes was decreased. The decrease in pore size resulted in better solute rejection. The NaCl rejection increased from 17.12% to 23.32% and the MgSO₄ rejection increased from 80.32% to 89.12% in NF membranes prepared at shear rates of 9.23 s⁻¹ and 42.76 s⁻¹, respectively. These results agree well with those reported by Ali et al.⁵⁸ The significant improvement in rejection of salts can be attributed to the molecular orientation induced by the shear rate during the casting of the polymer film.⁸³ As shown in Fig. 8, macromolecules exposed to a higher shear rate tend to have a more ordered alignment than those under a lower shear rate. Enhancement of the molecular orientation causes the polymer chains to pack closer leading to a decrease in free volume or pore size.^58,83,84 The formation of a smaller pore size and an increase in surface charge density of the membrane (because of the lower pore size) enhance the separation performance of PES NF membranes.⁵⁸ No change in contact angle of the membranes prepared at different shear rates was identified.

Table 7 The effect of casting shear rate and casting knife gap on the structural characteristics of PES NF membranes^a

Casting shear rate (s^−1)	Casting knife gap (μm)	Membrane thickness (μm)	Mean pore size (nm)	Porosity (%)	Contact angle (°)	Pure water flux (l m^−2 h^−1)	NaCl rejection (%)	MgSO4 rejection (%)
a Conditions: dope solution: 24 wt% PES in DMAc, casting temperature: 25 ± 3 °C; relative air humidity: 30 ± 5%; solvent evaporation time: 0 min; coagulating agent: pure water; coagulation bath temperature: 20 ± 2 °C.
9.23	200	110	5.06	48.3	59.1	25.13	17.12	80.32
23.54	200	100	4.96	58.9	60.5	38.35	18.62	82.78
42.76	200	92	4.54	64.4	60.1	41.32	23.32	89.12
23.54	100	73	5.09	63.6	57.6	56.41	13.54	80.98
23.54	200	100	4.96	58.9	60.5	38.35	18.62	82.78
23.54	300	162	4.59	52.9	65.8	17.87	28.32	90.83

---

Fig. 8 A hypothetical mechanism for the conformation of polymer chains induced by shear rate.⁸⁴

3.8. The effect of membrane thickness on the properties and performance of NF membranes

Membrane thickness plays an important role in determining flux and rejection properties of NF membranes. The effect of membrane thickness on pure water flux is shown in Table 7. The pure water flux decreased from 56.41 l m⁻² h⁻¹ to 17.87 l m⁻² h⁻¹ in membranes prepared with a thickness of 100 μm and 300 μm, respectively. The results demonstrate that the membrane with the lower thickness was more permeable because of the lower membrane resistance, higher porosity and larger pore size. At the beginning of the phase inversion process, because of the increase in the ratio of the solvent out-diffusion on the coagulating agent (water) inflow, the polymer concentration at the top layer increases. When the delay time is long, a higher polymer concentration is achieved at the interphase between the polymer film and the coagulating agent, which leads to the fast growth of the dense layer. As a result, a thicker and denser top layer with a smaller pore size was formed and as consequence, the pure water flux decreased.⁸⁵ The membrane porosity decreased significantly from 63.60% to 52.91% in the membrane prepared with a casting knife gap of 100 μm and 300 μm, respectively. Using a higher casting knife gap decreased the diffusional exchange rate between the solvent and the coagulating agent and thus, the membrane structure became thicker and less porous.⁵⁹ The mean pore sizes of the membranes are given in Table 7. As the initial casting thickness increases, the diffusion rate of the coagulating agents in the cast films decreases and thus, delayed demixing occurs. Therefore, the mean pore size was shifted to a lower value as the membrane thickness increased.⁵⁹

The effect of casting solution thickness on rejection of salts is shown in Table 7. By increasing the membrane thickness, the NaCl rejection increased from 13.54% to 28.32% and the MgSO₄ rejection increased from 80.98% to 90.83% in NF membranes prepared with thicknesses of 100 μm and 300 μm, respectively. The decrease of pore size enhanced the rejection of salts through both size sieving and electrostatic repulsion mechanisms. As the membrane thickness increased, the contact angle decreased. Because the materials and the dope solution are the same, the greater contact angle of the thicker membrane can mainly be attributed to the decreased surface porosity.⁶⁵

3.9. Effect of operating parameters on the properties and performance of NF membranes

3.9.1. Feed pressure. The effect of feed pressure on water flux and rejection of salts of the reference membrane (PES–DMAc 76%/24%) is shown in Fig. 9. Because of the porous nature of the membrane, the Hagen–Poiseuille equation can be considered under the hypothesis of constant pressure gradient along the membrane pore.⁸⁶ According to the Hagen–Poiseuille flow as expressed in eqn (7), increase in water flux is proportional to the increase in pressure difference across the membrane:where J_V (m s⁻¹), r_p (m), η (Pa s⁻¹), δ (m), ΔP (Pa) and L_p (m s⁻¹ Pa⁻¹) are volume flux, average pore radius, water dynamic viscosity inside the pore, effective membrane thickness, pressure difference across the membrane and hydraulic membrane permeability, respectively. According to Fig. 9(a), water flux data versus feed pressure is approximately linear and calculated water permeability is 16.47 l m² h⁻¹ bar⁻¹. Also, the salts rejection increased with the increase in feed pressure (Fig. 9(b)). This is because the water permeate flux through the membrane is linearly related to the trans-membrane pressure, whereas the salt flux is relevant to both the concentration gradient across the membrane and the water flux. When the feed pressure increases, the water permeate flux increases relatively more compared to the salt permeate. This causes a decreasing salt concentration in the permeate stream and an increasing rejection of salts.⁸⁷

Fig. 9 Effect of feed pressure on (a) pure water flux (b) NaCl and MgSO₄ rejection of PES NF membranes. (Conditions: dope solution: 24 wt% PES in DMAc, casting temperature: 25 ± 3 ° C; relative air humidity: 30 ± 5%; casting knife gap: 200 μm; casting shear rate: 23.54 s⁻¹; solvent evaporation time: 0 min; coagulating agent: pure water; coagulation bath temperature: 20 ± 2 ° C; feed composition: 10 500 ppm NaCl, 1300 ppm MgSO₄, pH = 7).

3.9.2. Feed pH. The rejection of salts and the water flux as a function of feed pH were investigated for the reference membrane (PES–DMAc 76%/24%) and the results are shown in Table 8. The zeta potential of the bare membrane was also measured to determine the membrane surface charge as a function of pH, as shown in Fig. 10. The higher zeta potential of the PES membrane at a higher solution pH value indicates the membrane surface is more negative. It is believed that the negative charge of the PES membrane originates from the functional groups of PES (O=S=O).⁷⁵ Because of the electrostatic charge repulsion between SO₄²⁻ and Cl⁻ molecules and the negative charges on the membrane surface, SO₄²⁻ and Cl⁻ molecules are rejected. An increase in solution pH results in an increase in the negative surface charge of membrane, which leads to a stronger electrostatic charge repulsion between the anions and the membrane surface and increases the rejection of anions. Subsequently, the rejection of cations (Mg²⁺, Na⁺) increases because of the Donnan effect. In other words, if anions (SO₄²⁻ and Cl⁻) are rejected because of electrostatic charge repulsion with the negative surface charge of the PES membrane, then the cations will also have to be rejected to retain the electric charge neutrality of the solution. An increase in co-ion charge and a decrease in counter ion charge improves the rejection of salts.⁶⁶ Because of the higher charge of the sulfate ions, the rejection of Mg²⁺ ions also increases caused by the Donnan effect. The electrostatic charge repulsion between the anion molecules and the membrane surface increases by increasing the solution pH, which hinders adsorption of anion molecules on the membrane surface. Hence, the barriers for the passage of solutions across the membrane decreases and therefore the water flux increases.⁷³

Table 8 The effect of feed pH on water flux and salt rejection of PES NF membranes^a

Feed pH	Water flux (l m^−2 h^−1)	NaCl rejection (%)	MgSO4 rejection (%)
a Conditions: dope solution: 24 wt% PES in DMAc, casting temperature: 25 ± 3 ° C; relative air humidity: 30 ± 5%; casting knife gap: 200 μm; casting shear rate: 23.54 s^−1; solvent evaporation time: 0 min; coagulating agent: pure water; coagulation bath temperature: 20 ± 2° C; feed composition: 10 500 ppm NaCl, 1300 ppm MgSO4; testing pressure = 10 bar.
5	33.12	12.43	75.47
7	35.21	17.96	81.97
9	37.19	19.65	86.09

---

Fig. 10 The zeta potential of the bare PES membrane as a function of pH.⁷⁵

3.10. Comparison with literature

Table 9 provides the salt rejection and water flux performance of several commercial and non-commercial NF membranes reported in literature. It can be seen that the water flux for the membranes developed in this study were almost higher than all the others in the table, except that of NF90. This can be attributed to high porosity, low thickness and formation of more free volume in the structure of the developed membranes. On the other hand, the salt rejection performance of the fabricated membranes seems to be satisfactory considering the salt concentration used for testing in comparison with other research. The membrane separation performance for divalent ions was better than that for monovalent ions. The authors are working for further improvements in this respect.

Table 9 Performance comparison for several commercial and non-commercial NF membranes

Membrane type	Water flux (l m^−2 h^−1)	NaCl rejection (%)	MgSO4 rejection (%)	Operating conditions	Ref.
Polyethersulfone with citric acid as additive	54.88	26.83	89.36	P = 10 bar, CNaCl = 10 500 ppm, CMgSO4 = 1300 ppm	Present study
Oxidized multiwalled carbon nanotube/polyethersulfone nanocomposite	8	20	58	P = 4 bar, C = 200 ppm	90
Amine-functionalized multiwalled carbon nanotubes	5.23	20	50	P = 4 bar, C = 200 ppm	75
NF90 (commercial)	80	22	99	P = 10 bar, C = 800 ppm	88
DS-5 DL (commercial)	18	1.3	93	P = 14 bar	89
NE2540-70 (commercial)	NA	7.2	3.56	P = 14 bar	89

---

4. Conclusions

The properties and performance of asymmetric PES NF membranes were investigated through a systematic analysis on the effects of design, fabrication and operational parameters, with the main aim of obtaining membranes with high rejection of monovalent and divalent salts while possessing a reasonable water flux. The results obtained revealed that among the studied parameters, polymer concentration greatly affected the membrane morphology and performance. It was also found that the membrane structure and performance were considerably dependent on the solvent–coagulating agent and solvent–polymer interactions. Overall, membranes prepared using DMAc exhibited a higher water flux than those prepared using NMP as solvent. Modified membranes containing organic acids exhibited structures with a lower thickness, higher porosity and smaller pore size compared to the unmodified membrane. This was attributed to the hydrophilic functional groups of additives as proven by the water contact angle data indicating that the wettability of membranes was increased by both citric and ascorbic acids. The water flux of the modified membrane was improved by the addition of 1 wt% of organic acids, although the rejection of salts improved for all concentrations of organic acids. Structural analysis of the membranes indicated that the membranes prepared with coagulating agent containing methanol had a lower porosity, smaller pore size, less hydrophilic surface and higher thickness because of slower diffusional exchange between the solvent and the coagulating agent. Despite the decline in pure water flux, rejection of both NaCl and MgSO₄ was improved in membranes prepared using a water–methanol mixture as a coagulating agent. Furthermore, membranes cast at a lower temperature, with a higher thickness and higher solvent evaporation time were found to be less porous with a lower mean pore size and were less permeable to water and more capable of rejecting salts. In all the membranes, the rejection of divalent ions was greater than monovalent ones, owing to the negative surface charge of the PES membranes. As a consequence, addition of organic acids and increase in casting shear rate simultaneously improved both water flux and salt rejection. However, there was a trade-off between rejection and permeability for other parameters. These findings provide valuable information and guidelines for the development of high performance nanofiltration membranes for specialized separation applications.

Acknowledgements

The authors would like to thank the Iran Nanotechnology Initiative Council (INIC) for partial financial support to this project.

References

1. Y. Xiao, B. T. Low, S. S. Hosseini, T. S. Chung and D. R. Paul, The strategies of molecular architecture and modification of polyimide-based membranes for CO₂ removal from natural gas—A review, Prog. Polym. Sci., 2009, 34, 561–580.
2. S. S. Hosseini and T. S. Chung, Carbon membranes from blends of PBI and polyimides for N₂/CH₄ and CO₂/CH₄ separation and hydrogen purification, J. Membr. Sci., 2009, 328, 174–185.
3. S. Najari, S. S. Hosseini, M. R. Omidkhah and N. R. Tan, Phenomenological modeling and analysis of gas transport in polyimide membranes for propylene/propane separation, RSC Adv., 2015 10.1039/c5ra05556c.
4. S. S. Hosseini, M. M. Teoh and T. S. Chung, Hydrogen separation and purification in membranes of miscible polymer blends with interpenetration networks, Polymer, 2008, 49, 1594–1603.
5. S. Ü. Çelik, A. Bozkurt and S. S. Hosseini, Alternatives toward proton conductive anhydrous membranes for fuel cells: Heterocyclic protogenic solvents comprising polymer electrolytes, Prog. Polym. Sci., 2012, 37, 1265–1291.
6. M. Tamaddondar, H. Pahlavanzadeh, S. Saeid Hosseini, G. Ruan and N. R. Tan, Self-assembled polyelectrolyte surfactant nanocomposite membranes for pervaporation separation of MeOH/MTBE, J. Membr. Sci., 2014, 472, 91–101.
7. S. S. Hosseini, Y. Li, T.-S. Chung and Y. Liu, Enhanced gas separation performance of nanocomposite membranes using MgO nanoparticles, J. Membr. Sci., 2007, 302, 207–217.
8. S. S. Hosseini, M. R. Omidkhah, A. Zarringhalam Moghaddam, V. Pirouzfar, W. B. Krantz and N. R. Tan, Enhancing the properties and gas separation performance of PBI–polyimides blend carbon molecular sieve membranes via optimization of the pyrolysis process, Sep. Purif. Technol., 2014, 122, 278–289.
9. S. S. Hosseini, N. Peng and T. S. Chung, Gas separation membranes developed through integration of polymer blending and dual-layer hollow fiber spinning process for hydrogen and natural gas enrichments, J. Membr. Sci., 2010, 349, 156–166.
10. V. Pirouzfar, S. S. Hosseini, M. R. Omidkhah and A. Z. Moghaddam, Modeling and optimization of gas transport characteristics of carbon molecular sieve membranes through statistical analysis, Polym. Eng. Sci., 2014, 54, 147–157.
11. V. Pirouzfar, A. Z. Moghaddam, M. R. Omidkhah and S. S. Hosseini, Investigating the effect of dianhydride type and pyrolysis condition on the gas separation performance of membranes derived from blended polyimides through statistical analysis, J. Ind. Eng. Chem., 2014, 20, 1061–1070.
12. W. B. Krantz, A. R. Greenberg, E. Kujundzic, A. Yeo and S. S. Hosseini, Evapoporometry: A novel technique for determining the pore-size distribution of membranes, J. Membr. Sci., 2013, 438, 153–166.
13. S. S. Hosseini, S. M. Roodashti, P. K. Kundu and N. R. Tan, Transport Properties of Asymmetric Hollow Fiber Membrane Permeators for Practical Applications: Mathematical Modeling For Binary Gas Mixtures, Can. J. Chem. Eng. DOI:10.1002/cjce.22215.
14. I. Koyuncu and M. Cakmakci, Nanofiltration for Water and Wastewater Treatment, Nanotechnol. Water Treat. Appl., 2010, 103.
15. B. A. M. Al-Rashdi, D. J. Johnson and N. Hilal, Removal of heavy metal ions by nanofiltration, Desalination, 2013, 315, 2–17.
16. A. Rahimpour, M. Jahanshahi, N. Mortazavian, S. S. Madaeni and Y. Mansourpanah, Preparation and characterization of asymmetric polyethersulfone and thin-film composite polyamide nanofiltration membranes for water softening, Appl. Surf. Sci., 2010, 256, 1657–1663.
17. A. R. Anim-Mensah, W. B. Krantz and R. Govind, Studies on polymeric nanofiltration-based water softening and the effect of anion properties on the softening process, Eur. Polym. J., 2008, 44, 2244–2252.
18. B. Van der Bruggen and C. Vandecasteele, Removal of pollutants from surface water and groundwater by nanofiltration: overview of possible applications in the drinking water industry, Environ. Pollut., 2003, 122, 435–445.
19. N. Hilal, H. Al-Zoubi, N. A. Darwish, A. W. Mohamma and M. Abu Arabi, A comprehensive review of nanofiltration membranes: Treatment, pretreatment, modelling, and atomic force microscopy, Desalination, 2004, 170, 281–308.
20. M. Telzhensky, L. Birnhack, O. Lehmann, E. Windler and O. Lahav, Selective separation of seawater Mg²⁺ ions for use in downstream water treatment processes, Chem. Eng. J., 2011, 175, 136–143.
21. A. Lhassani, M. Rumeau, D. Benjelloun and M. Pontie, Selective demineralization of water by nanofiltration application to the defluorination of brackish water, Water Res., 2001, 35, 3260–3264.
22. K. Walha, R. B. Amar, L. Firdaous, F. Quéméneur and P. Jaouen, Brackish groundwater treatment by nanofiltration, reverse osmosis and electrodialysis in Tunisia: performance and cost comparison, Desalination, 2007, 207, 95–106.
23. S. Zhao, L. Zou and D. Mulcahy, Brackish water desalination by a hybrid forward osmosis–nanofiltration system using divalent draw solute, Desalination, 2012, 284, 175–181.
24. S. Chakraborty, M. K. Purkait, S. DasGupta, S. De and J. K. Basu, Nanofiltration of textile plant effluent for color removal and reduction in COD, Sep. Purif. Technol., 2003, 31, 141–151.
25. S. K. Nataraj, K. M. Hosamani and T. M. Aminabhavi, Nanofiltration and reverse osmosis thin film composite membrane module for the removal of dye and salts from the simulated mixtures, Desalination, 2009, 249, 12–17.
26. B. M. Watson and C. D. Hornburg, Low-energy membrane nanofiltration for removal of color, organics and hardness from drinking water supplies, Desalination, 1989, 72, 11–22.
27. M. Barakat, New trends in removing heavy metals from industrial wastewater, Arabian J. Chem., 2011, 4, 361–377.
28. M. Muthukrishnan and B. Guha, Effect of pH on rejection of hexavalent chromium by nanofiltration, Desalination, 2008, 219, 171–178.
29. Y. Sato, M. Kang, T. Kamei and Y. Magara, Performance of nanofiltration for arsenic removal, Water Res., 2002, 36, 3371–3377.
30. A. P. Echavarria, C. Torras, J. Pagán and A. Ibarz, Fruit juice processing and membrane technology application, Food Eng. Rev., 2011, 3, 136–158.
31. L. D. Nghiem, A. I. Schäfer and M. Elimelech, Pharmaceutical retention mechanisms by nanofiltration membranes, Environ. Sci. Technol., 2005, 39, 7698–7705.
32. B. S. Lalia, V. Kochkodan, R. Hashaikeh and N. Hilal, A review on membrane fabrication: Structure, properties and performance relationship, Desalination, 2013, 326, 77–95.
33. A. Brunetti, S. Simone, F. Scura, G. Barbieri, A. Figoli and E. Drioli, Hydrogen mixture separation with PEEK-WC asymmetric membranes, Sep. Purif. Technol., 2009, 69, 195–204.
34. A. F. Ismail and A. R. Hassan, Formation and characterization of asymmetric nanofiltration membrane: effect of shear rate and polymer concentration, J. Membr. Sci., 2006, 270, 57–72.
35. A. Ismail and A. Hassan, Formation and characterization of asymmetric nanofiltration membrane: effect of shear rate and polymer concentration, J. Membr. Sci., 2006, 270, 57–72.
36. S. Yu, M. Liu, M. Ma, M. Qi, Z. Lü and C. Gao, Impacts of membrane properties on reactive dye removal from dye/salt mixtures by asymmetric cellulose acetate and composite polyamide nanofiltration membranes, J. Membr. Sci., 2010, 350, 83–91.
37. M. Buonomenna, L. Lopez, P. Favia, R. d'Agostino, A. Gordano and E. Drioli, New PVDF membranes: the effect of plasma surface modification on retention in nanofiltration of aqueous solution containing organic compounds, Water Res., 2007, 41, 4309–4316.
38. A. G. Fane, C. Y. Tang, and R. Wang, 4.11-Membrane Technology for Water: Microfiltration, Ultrafiltration, Nanofiltration, and Reverse Osmosis, in Treatise on Water Science, ed. W. Peter, Elsevier, Oxford, 2011, pp. 301–335.
39. V. Vatanpour, S. S. Madaeni, L. Rajabi, S. Zinadini and A. A. Derakhshan, Boehmite nanoparticles as a new nanofiller for preparation of antifouling mixed matrix membranes, J. Membr. Sci., 2012, 401, 132–143.
40. S. Balta, A. Sotto, P. Luis, L. Benea, B. Van der Bruggen and J. Kim, A new outlook on membrane enhancement with nanoparticles: the alternative of ZnO, J. Membr. Sci., 2012, 389, 155–161.
41. C. Zhao, J. Xue, F. Ran and S. Sun, Modification of polyethersulfone membranes–A review of methods, Prog. Mater. Sci., 2013, 58, 76–150.
42. A. L. Ahmad, A. A. Abdulkarim, B. S. Ooi and S. Ismail, Recent development in additives modifications of polyethersulfone membrane for flux enhancement, Chem. Eng. J., 2013, 223, 246–267.
43. P. Daraei, S. S. Madaeni, N. Ghaemi, H. A. Monfared and M. A. Khadivi, Fabrication of PES nanofiltration membrane by simultaneous use of multi-walled carbon nanotube and surface graft polymerization method: Comparison of MWCNT and PAA modified MWCNT, Sep. Purif. Technol., 2013, 104, 32–44.
44. J. Zhu, Y. Zhang, M. Tian and J. Liu, Fabrication of a Mixed Matrix Membrane with in Situ Synthesized Quaternized Polyethylenimine Nanoparticles for Dye Purification and Reuse, ACS Sustainable Chem. Eng., 2015, 3, 690–701.
45. M. Sadrzadeh and S. Bhattacharjee, Rational design of phase inversion membranes by tailoring thermodynamics and kinetics of casting solution using polymer additives, J. Membr. Sci., 2013, 441, 31–44.
46. K. Boussu, C. Vandecasteele and B. Van der Bruggen, Study of the characteristics and the performance of self-made nanoporous polyethersulfone membranes, Polymer, 2006, 47, 3464–3476.
47. S. S. Madaeni and A. H. Taheri, Preparation of PES ultrafiltration membrane for treatment of emulsified oily wastewater: Effect of solvent and non-solvent on morphology and performance, J. Polym. Eng., 2009, 29, 183–198.
48. S. S. Madaeni and A. Rahimpour, Effect of type of solvent and non-solvents on morphology and performance of polysulfone and polyethersulfone ultrafiltration membranes for milk concentration, Polym. Adv. Technol., 2005, 16, 717–724.
49. S. Madaeni and A. Rahimpour, Effect of type of solvent and non-solvents on morphology and performance of polysulfone and polyethersulfone ultrafiltration membranes for milk concentration, Polym. Adv. Technol., 2005, 16, 717–724.
50. K. J. Baik, J. Y. Kim, J. S. Lee, S. C. Kim and H. K. Lee, Morphology of membranes formed from polysulfone/polyethersulfone/N-methyl-2-pyrrolidone/water system by immersion precipitation, Korea Polym. J., 2001, 9, 285–291.
51. M. Khayet, C. Cojocaru and M. García-Payo, Experimental design and optimization of asymmetric flat-sheet membranes prepared for direct contact membrane distillation, J. Membr. Sci., 2010, 351, 234–245.
52. D. Mosqueda-Jimenez, R. Narbaitz, T. Matsuura, G. Chowdhury, G. Pleizier and J. Santerre, Influence of processing conditions on the properties of ultrafiltration membranes, J. Membr. Sci., 2004, 231, 209–224.
53. M. Amirilargani, E. Saljoughi, T. Mohammadi and M. R. Moghbeli, Effects of coagulation bath temperature and polyvinylpyrrolidone content on flat sheet asymmetric polyethersulfone membranes, Polym. Eng. Sci., 2010, 50, 885–893.
54. B. K. Chaturvedi, A. K. Ghoshb, V. Ramachandhranb, M. K. Trivedi, M. S. Hantab and B. M. Misrab, Preparation, characterization and performance of polyethersulfone ultrafiltration membranes, Desalination, 2001, 133, 31–40.
55. C. B. Spricigo, J. C. C. Petrus, R. A. F. Machado, L. A. V. Sarmento and A. Bolzan, Preparation and characterization of polyethersulfone membranes for use in supercritical medium, J. Membr. Sci., 2002, 205, 273–278.
56. A. F. Ismail, A. R. Hassan and B. C. Ng, Effect of shear rate on the performance of nanofiltration membrane for water desalination, Songklanakarin J. Sci. Technol., 2002, 24, 879–889.
57. T. D. Kusworo, A. F. Ismail, A. Mustafa and T. Matsuura, Dependence of membrane morphology and performance on preparation conditions: The shear rate effect in membrane casting, Sep. Purif. Technol., 2008, 61, 249–257.
58. N. A. Ali, N. Halim, A. Jusoh and A. Endut, The formation and characterisation of an asymmetric nanofiltration membrane for ammonia–nitrogen removal: Effect of shear rate, Bioresour. Technol., 2010, 101, 1459–1465.
59. A. Ahmad, N. Ideris, B. Ooi, S. Low and A. Ismail, Synthesis of polyvinylidene fluoride (PVDF) membranes for protein binding: Effect of casting thickness, J. Appl. Polym. Sci., 2013, 128, 3438–3445.
60. A. Conesa, T. Gumí and C. Palet, Membrane thickness and preparation temperature as key parameters for controlling the macrovoid structure of chiral activated membranes (CAM), J. Membr. Sci., 2007, 287, 29–40.
61. N. Vogrin, Č. Stropnik, V. Musil and M. Brumen, The wet phase separation: the effect of cast solution thickness on the appearance of macrovoids in the membrane forming ternary cellulose acetate/acetone/water system, J. Membr. Sci., 2002, 207, 139–141.
62. A. Sotto, A. Boromand, S. Balta, J. Kim and B. Van der Bruggen, Doping of polyethersulfone nanofiltration membranes: antifouling effect observed at ultralow concentrations of TiO₂ nanoparticles, J. Mater. Chem., 2011, 21, 10311–10320.
63. V. Vatanpour, S. S. Madaeni, R. Moradian, S. Zinadini and B. Astinchap, Novel antibifouling nanofiltration polyethersulfone membrane fabricated from embedding TiO₂ coated multiwalled carbon nanotubes, Sep. Purif. Technol., 2012, 90, 69–82.
64. M. Amirilargani, E. Saljoughi, T. Mohammadi and M. Moghbeli, Effects of coagulation bath temperature and polyvinylpyrrolidone content on flat sheet asymmetric polyethersulfone membranes, Polym. Eng. Sci., 2010, 50, 885–893.
65. M. Omidvar, S. Mahmoud Mousavi, M. Soltanieh and A. A. Safekordi, Preparation and characterization of poly (ethersulfone) nanofiltration membranes for amoxicillin removal from contaminated water, Journal of Environmental Health Science and Engineering, 2014, 12, 1–10.
66. T. Dey, Nanotechnology for water purification, Universal-Publishers, 2012.
67. S. Madaeni and A. Taheri, Preparation of PES ultrafiltration membrane for treatment of emulsified oily wastewater: Effect of solvent and non-solvent on morphology and performance, J. Polym. Eng., 2009, 29, 183–198.
68. K.-Y. Chun, S.-H. Jang, H.-S. Kim, Y.-W. Kim, H.-S. Han and Y.-I. Joe, Effects of solvent on the pore formation in asymmetric 6FDA–4, 4′ ODA polyimide membrane: terms of thermodynamics, precipitation kinetics, and physical factors, J. Membr. Sci., 2000, 169, 197–214.
69. L. Wang, Z. Li, J. Ren, S.-G. Li and C. Jiang, Preliminary studies on the gelation time of poly (ether sulfones) membrane-forming system with an elongation method, J. Membr. Sci., 2006, 275, 46–51.
70. J. Barzin and B. Sadatnia, Correlation between macrovoid formation and the ternary phase diagram for polyethersulfone membranes prepared from two nearly similar solvents, J. Membr. Sci., 2008, 325, 92–97.
71. H. Hasbullah, A. F. Ismail, B. C. Ng and M. S. Abdullah, Effect of solvent types on structure and performance of polyethersulfone nanofiltration membrane in monovalent and divalent ions separation, in 2nd International Conference on Chemical and Bioprocess Engineering in Conjunction with 19th Symposium of Malaysian Chemical Engineers (SOMChE 2005), Kota Kinabalu, Sabah, Malaysia, 8–10 December 2005.
72. M. Ghosh, Polyimides: fundamentals and applications, CRC Press, 1996.
73. N. Ghaemi, S. S. Madaeni, A. Alizadeh, P. Daraei, M. M. S. Badieh and M. Falsafi, et al., Fabrication and modification of polysulfone nanofiltration membrane using organic acids: morphology, characterization and performance in removal of xenobiotics, Sep. Purif. Technol., 2012, 96, 214–228.
74. I. C. Kim and K. H. Lee, Effect of various additives on pore size of polysulfone membrane by phase-inversion process, J. Appl. Polym. Sci., 2003, 89, 2562–2566.
75. V. Vatanpour, M. Esmaeili and M. H. D. A. Farahani, Fouling reduction and retention increment of polyethersulfone nanofiltration membranes embedded by amine-functionalized multi-walled carbon nanotubes, J. Membr. Sci., 2014, 466, 70–81.
76. B. T. Swinyard and J. A. Barrie, Phase separation in non-solvent/dimethylformamide/polyethersulphone and non-solvent/dimethylformamide/polysulphone systems, Br. Polym. J., 1988, 20, 317–321.
77. D.-y. Zuo, B.-k. Zhu, J.-h. Cao and Y.-y. Xu, Influence of alcohol-based nonsolvents on the formation and morphology of PVDF membranes in phase inversion process, Chin. J. Polym. Sci., 2006, 24, 281–289.
78. A. Rahimpour, S. S. Madaeni, M. Jahanshahi, Y. Mansourpanah and N. Mortazavian, Development of high performance nano-porous polyethersulfone ultrafiltration membranes with hydrophilic surface and superior antifouling properties, Appl. Surf. Sci., 2009, 255, 9166–9173.
79. P. Sukitpaneenit and T.-S. Chung, Molecular elucidation of morphology and mechanical properties of PVDF hollow fiber membranes from aspects of phase inversion, crystallization and rheology, J. Membr. Sci., 2009, 340, 192–205.
80. M. Khayet and T. Matsuura, Membrane distillation: principles and applications, Elsevier, 2011.
81. C. Feng, B. Shi, G. Li and Y. Wu, Preparation and properties of microporous membrane from poly (vinylidene fluoride-co-tetrafluoroethylene)(F2. 4) for membrane distillation, J. Membr. Sci., 2004, 237, 15–24.
82. Y. Wang, B. Shi and X. Li, Effects of coagulation bath temperature on performances of polyethersulfone membranes modified by nanosilver particles in situ reduction, Polym. Eng. Sci., 2013, 53, 1614–1622.
83. T.-S. Chung, S. K. Teoh, W. W. Y. Lau and M. P. Srinivasan, Effect of shear stress within the spinneret on hollow fiber membrane morphology and separation performance, Ind. Eng. Chem. Res., 1998, 37, 3930–3938.
84. C. Cao, T.-S. Chung, S. B. Chen and Z. Dong, The study of elongation and shear rates in spinning process and its effect on gas separation performance of poly (ether sulfone)(PES) hollow fiber membranes, Chem. Eng. Sci., 2004, 59, 1053–1062.
85. S. Zhang, K. Y. Wang, T.-S. Chung, Y. C. Jean and H. Chen, Molecular design of the cellulose ester-based forward osmosis membranes for desalination, Chem. Eng. Sci., 2011, 66, 2008–2018.
86. S. Bandini and L. Bruni, Transport phenomena in nanofiltration membranes, Comprehensive Membrane Science and Engineering, Elsevier, Oxford, 2010, pp. 67–89.
87. K. Y. Wang, Y. Xiao and T.-S. Chung, Chemically modified polybenzimidazole nanofiltration membrane for the separation of electrolytes and cephalexin, Chem. Eng. Sci., 2006, 61, 5807–5817.
88. A. Somrani, A. Hamzaoui and M. Pontie, Study on lithium separation from salt lake brines by nanofiltration (NF) and low pressure reverse osmosis (LPRO), Desalination, 2013, 317, 184–192.
89. M. Telzhensky, L. Birnhack, O. Lehmann, E. Windler and O. Lahav, Selective separation of seawater Mg²⁺ ions for use in downstream water treatment processes, Chem. Eng. J., 2011, 175, 136–143.
90. V. Vatanpour, S. S. Madaeni, R. Moradian, S. Zinadini and B. Astinchap, Fabrication and characterization of novel antifouling nanofiltration membrane prepared from oxidized multiwalled carbon nanotube/polyethersulfone nanocomposite, J. Membr. Sci., 2011, 375, 284–294.

---