Preparation and characterization of an amphiphilic polyamide nanofiltration membrane with improved antifouling properties by two-step surface modification method

Membrane fouling is an urgent problem needing to be solved for practical application of nanofiltration membranes. In this study, an amphiphilic nanofiltration membrane with hydrophilic domains as well as low surface energy domains was developed, to integrate a fouling-resistant defense mechanism and a fouling-release defense mechanism. A simple and effective two-step surface modification of a polyamide NF membrane was applied. Firstly, triethanolamine (TEOA) with abundant hydrophilic functional groups was grafted to the membrane surface via reacting with the residual acyl chloride group of the nanofiltration membrane, making the nanofiltration membranes more hydrophilic; secondly, the 1H,1H,2H,2H-perfluorodecyltrichlorosilane (PFTS), well-known as a low surface energy material, was covalently grafted on the hydroxyl functional groups through hydrogen bonding. Filtration experiments with model foulants (bovine serum albumin (BSA) protein solution, humic acid solution (HA) and sodium alginate solution (SA)) were performed to estimate the antifouling properties of the newly developed nanofiltration membranes. As a result of surface modification proposed in this study the antifouling properties of an amphiphilic modified F-PA/PSF membrane were enhanced more than 10% compared to the PA/PSF specimen in terms of flux recovery ratio.


Introduction
Resource depletion and environmental pollution have become an exigent global issue that threatens daily life. Membrane technology becomes attractive in this context, as it is an energysaving and environmentally friendly technology. Nanoltration membranes (NF), which have a high permeation ux under low operational pressures, have been applied in water purication, pretreatment for seawater desalination, removal of heavy metals, textile dyeing and other elds. [1][2][3] Thin-lm composite (TFC) membranes are the most popular commercial NF membranes which usually consist of an ultra-thin selective layer upon the surface of a porous at sheet substrate membrane through the interfacial polymerization (IP) process, using diamine and acyl chlorides. 4,5 However, membrane fouling is a bottleneck for the practical application of NF in water and wastewater treatment. Fouling usually causes a signicant loss of productivity and increased operational cost. 6 Foulants in the water (like colloid substances, bacteria, and nature organic materials) can unsuitably absorb and deposit onto the membrane surface or into the pores and pore walls and thus cause membrane fouling, 7,8 which results in a ux decline during ltration operation and a shorter lifetime. Many researchers [9][10][11] have suggested that the dominant reason of membrane fouling is the intrinsic hydrophobicity of the membrane materials. Besides, the roughness and charge of the surface, feed solution chemistry and process conditions also have an important inuence on the membrane fouling. 8 The pores of NF membranes are very small ($1 nm), so that fouling is mainly manifested as surface fouling. 12 Currently, several surface modication methods like surface graing, 13 surface blending 14 and surface coating 15 have been used to improve the antifouling performance of NF membranes. Commonly, surface covalent graing of NF membrane is performed by using residual acyl chloride on the active layer to react with other functional groups capable of reacting with them. Zhu et al. 16 studied the structure and properties of TFC hollow ber NF membranes graed with poly (amidoamine) dendrimer (PAMAM). The hydrophilicity and water permeation ux increased with the PAMAM graing, while the salt rejection was hardly affected. The effect of the PAMAM graed TFC membrane on the removal of heavy metal ions like Pb 2+ , Cu 2+ , Ni 2+ , Cd 2+ , Zn 2+ and As 5+ was also signicant, the rejection was over 99%.
At present, antifouling membranes can be divided into several categories according to different criteria, on the basis of hydropathy property of membranes surfaces, the membranes can be classied into three categories: 17 hydrophilic antifouling surfaces, hydrophobic self-cleaning surfaces and amphiphilic surfaces. Hydrophilic modication of NF membranes is one of the main ways to enhance the antifouling properties of a membrane. It is based on the fact that the organic matters in the water, which are hydrophobic in nature, are more likely to attach on a membrane surface because of hydrophobic interactions. The advanced hydrophilicity of membrane surface is benecial to mitigate these molecular interactions. 18 Furthermore, the hydrophilic membrane surface promotes the formation of a hydration layer, which restricts the pollutant adsorption or deposition on the membrane surface. 19 Many hydrophilic materials have been used to prepare highly antifouling NF membranes, for example, poly(ethylene glycol) (PEG)-based materials, 20,21 zwitterionic materials; 22,23 among the most popular inorganic nanoparticles are titanium dioxide (TiO 2 ), 8,24 graphene oxide (GO), 25,26 and multi-wall carbon nanotubes (MWCNTs). 14,27,28 The strategy of hydrophilic modi-cation represents a fouling-resistance defense mechanism. [29][30][31] Unlike hydrophilic surfaces, hydrophobic membrane surfaces follow another mechanism: the fouling-release defense mechanism. [32][33][34] This works through the development of a low surface energy layer barrier on the membrane surface using siliconebased 35,36 and uorine-based materials, 37,38 with a surface energy of about 22 mJ m À2 and 10-18 mJ m À2 , respectively. Since the low surface energy characteristics of self-cleaning surfaces can make the intermolecular forces between pollutants and surfaces minimization, so that the pollutants are easily removed through low hydrodynamic shear force or simple mechanical cleaning, yielding an attractive "fouling-release" property. 39 Although many attempts have been made to enhance the antifouling abilities of membrane surfaces, there are still signicant problems to be addressed. With respect to hydrophilic modied membranes, the hydrophilic segments (like PEG) would be easily washed out of the matrix under the large drag force from the ltration pressure, because of their linear structures and low compatibility with the membrane matrix. 40,41 The surface of the membrane is still exposed to the risk of pollutant adsorption and deposition, thus reducing the ux. Therefore, the hydrophilic modication of NF membrane surfaces may is not a foolproof solution to solve the fouling effects. As for the hydrophobic modied lm, one of the biggest problems is that the introduction of low surface energy materials may cause the water ux decline. Thus, the amphiphilic surfaces containing both hydrophilic and low surface energy domains would be good candidates for membrane modication. The presence of hydrophilic domains guaranties high water uxes and low ux decline as result of hydraulic layer formation that mitigates the interaction between potential pollutants and the membrane surface. On the other hand, the hydrophobic anchoring groups are entangled with the hydrophobic membrane matrix because of the well compatibility, correlative hydrophilic segments can be stabilized in the membrane surface, thus weakens the interaction force between pollutants and membrane surface. 40,42 The amphiphilic modi-cation of NF membrane may thus meet the antifouling requirements of a much broader variety of pollutants.
For the past few years, many researchers have paid attention to develop the amphiphilic surfaces on ultraltration membranes with multi-defense fouling mechanisms. 43,44 The common method is to fabricate an amphiphilic polymer and then blend it into the casting solution to prepare antifouling membranes by different methods, such as surface segregation, 31,36 phase separation, 41,45 dip-coating 46 or phase inversion. 40 However, due to the large size of the synthesized copolymer molecules and much smaller pore size of the NF membrane, the abovementioned methods are not suitable for the preparation of the amphiphilic antifouling NF membrane taking into account the possible blockage of membrane pores. Thus, a new way is required to prepare amphiphilic NF membrane surfaces. Zhang et al. 17 fabricated an amphiphilic NF membrane by a two-step surface modication of a polyamide NF membrane, this amphiphilic NF membrane contains both hydrophilic domains and low surface energy domains, which can combine the fouling-resistance defense mechanism and the fouling-release defense mechanism. Triethylenetetramine with primary amine groups was rstly introduced to the membrane surface by reacting with unreacted acyl chloride on the surface of polyamide NF membrane. And then 2,2,3,4,4,4-hexa-uorobutylmethacrylate was covalently graed to the surface of the membrane by a Michael addition reaction with these primary amine groups. The antifouling experiments of BSA solution, HA solution and SA solution turned out that the ux recovery rate of the amphiphilic membrane was higher than of polyamide NF membrane, therefore possessed better antifouling property. However, this procedure is performed under harsh reaction conditions, which needs be carried out at a higher temperature, and the reaction time is relatively long. This will undoubtedly increase the energy consumption, and limit its industrial application. In this study, a simple and effective method is proposed avoiding the drawbacks mentioned above. The process involved in this approach can be performed at room temperature with the enhanced reaction kinetics.
An amphiphilic nanoltration membrane with hydrophilic domains and low surface energy domains was developed through a two-step surface modication. First, the polyamide NF membrane was fabricated by interfacial polymerization of piperazine (PIP) and trimesoyl chloride (TMC). Then TEOA having abundant hydrophilic functional groups, is cost-effective, environment-friendly and easily-obtained. It can be graed to the surface by reacting with the residual acyl chloride group of NF membrane, which makes the NF membranes more hydrophilic. Finally, the PFTS which is usually known as a low surface energy material was covalently graed on hydroxyl functional groups through hydrogen bonding. Morphology of the prepared membranes were characterized using atomic force microscopy (AFM), scanning electron microscopy (SEM), etc. Separation and rejection properties were investigated by ltration experiments of salts and PEGs. Filtration experiments with model foulants (BSA protein solution, HA solution and SA solution) were performed to estimate the antifouling property of the prepared NF membranes.

Materials and chemicals
A at-sheet nonwoven-reinforced polysulfone (PSF) support membrane with average molecular weight cut-off (MWCO) of 80 000 g mol À1 was supplied by Hangzhou Water Treatment Technology Development Center (Hangzhou, China). Piperazine (PIP, 99%), n-hexane ($97%), trimesoyl chloride (TMC), sodium dodecyl sulfate (SDS, analytically pure), triethanolamine (TEOA, 98%), bovine serum albumin (BSA, analytically pure), sodium alginate (SA, analytically pure), humic acid (HA, analytically pure), 1H,1H,2H,2H-peruorodecyltrichlorosilane (PFTS, 95%) and polyethylene glycol (PEG) with different molecular weights (M w ¼ 200, 300, 400, 600, 800 and 1000 Da) were purchased from Macklin (Shanghai, China). Alcohol, sodium sulfate (Na 2 SO 4 ), and sodium chloride (NaCl) were purchased from Shanghai Lingfeng Chemical Reagent Co. Ltd, China. All reagents were used as received without further purication. Pure water used throughout in this work was derived from a Milli-Q system (Millipore, US). Fig. 1 shows the preparation procedure of the amphiphilic NF membrane, which contained a two-step surface modication of NF membrane. The nascent polyamide NF membrane was fabricated by interfacial polymerization of PIP and TMC on the PSF support membrane. The PSF support membrane which xed by polyuortetraethylene framework was rstly soaked into 0.5 wt% PIP aqueous solution containing 0.1 wt% SDS for 2 min, and then the aqueous solution was poured out. The membrane surface was rinsed with pure water to remove the remaining liquid droplets, followed by air dried at room temperature. Subsequently, the PIP impregnated membrane surface was brimming with 0.1 wt% TMC organic solution (n-hexane) for 1 min, and then the organic solution was poured out. The membrane surface was rinsed thoroughly with n-hexane, heat treatment of 60 C for 10 min was further carried out to make sure completion of the polymerization reaction. The resultant NF membrane was named PA/PSF membrane. On this basis, the amphiphilic membrane was prepared by a two-step modication.

Membrane preparation
Firstly, the surface of a newly prepared PA/PSF membrane was coated with TEOA ethanol solution for 2 min and washed thoroughly with ethanol to remove the remaining TEOA solution and then pure water to remove the residual ethanol. Membranes prepared with TEOA contents of 1, 2, 3, 4 and 5% (v TEOA /v solution ) are referred to as H 1 -PA/PSF, H 2 -PA/PSF, H 3 -PA/PSF, H 4 -PA/PSF and H 5 -PA/PSF, respectively. The resultant hydrophilic modied membranes were stored in pure water. Secondly, the surface of the hydrophilic modied membrane was immersed in PFTS alcohol solution with for 2 min. The membrane surface was ashed with alcohol to remove the remaining PFTS solution and then pure water to remove the remaining ethanol, followed by drying at 80 C for 10 min. Aer heat treatment, the fabricated membrane (designated as the F-PA/PSF membrane) was stored in pure water for the next experiment.

Membrane characterization
The NF membrane was submerged in water for 24 hours prior to the determination of chemical composition analysis and morphologies. Attenuated Total Reectance-Fourier transform Infrared spectroscopy (ATR-FTIR, Nicolet 6700) and X-ray photoelectron spectroscopy (XPS, Kratos AXIS Ultra DLD, Japan) were used to characterize the chemical composition and structure of membrane surface before and aer modication. The samples were scanned from 400 to 4000 cm À1 by ATR-FTIR with a resolution of 1 cm À1 for each spectrum. XPS were operated at constant pass energy of 100 eV for wide scans and 20 eV for detailed scans setting the O1s peak at BE 528.8 eV. The morphological characteristics of the membranes were observed with a scanning electron microscope (SEM, HITACHI, S4700A). The membrane samples were dried under vacuum and sputtercoated with gold for producing electric conductivity before observation. Atomic force microscopy (AFM, Bruker Dimension station) was used to characterize the surface topology of the membranes with a scanning area of 10 mm Â 10 mm.
The water contact angle of pristine and modied membrane surfaces was measured using a contact angle goniometer (OCA-20, Data-physics, Germany). For each membrane sample, at least ve measurements were taken at different locations at room temperature and the average value was reported. The surface charge of the membranes were analyzed by an electrokinetic analyzer (SurPASS 3, Anton Paar, Austria) with 1 mM KCl as the electrolyte solution at a pH of 6.5-7.5 at room temperature. Zeta potential of the membrane was measured by an electrokinetic analyzer (SurPASS 3, Anton Paar, Austria) with 1 mM KCl as the electrolyte solution at a pH of 3.1-9.7 at room temperature.

Membrane ltration testing
The permeability and rejection tests were measured in a batch cross-ow NF membrane test unit at 25 C. All the membrane samples with effective area 19.6 cm 2 (A) were pre-compacted at 1.1 MPa with pure water for 90 min to obtain a stable ux before testing and then evaluated for 30 min at a pressure of 1.0 MPa. The penetrating uid is collected for the calculation of the water ux (J) and salt rejection (R) as follows: where t is the permeation time (h). C p and C f are the solute concentrations in permeate and feed, respectively. 1.0 g L À1 of Na 2 SO 4 or NaCl solution was used to estimate the rejection performance of membrane. Molecular weight cut-off of membrane was tested using 1 g L À1 PEGs aqueous solution with different molecular weights (200, 300, 400, 600, 800, and 1000 Da). A total organic carbon analyzer (TOC Analyzer, Shimadzu TOC-VCNP) were used to determine the concentrations of the PEGs. The salt concentrations were measured by a conductivity meter (DDSJ-308A, Leizi, China), since the electric conductivity is proportional to the total salt concentration in a certain range. The average value of the three sets of parallel experimental data was used to calculate water ux and salt rejection.

Antifouling performance of NF membranes
Experiments with model foulants (BSA protein solution, HA and SA solution) were performed to estimate the antifouling property of NF membranes, which were chosen to represent protein, natural organic matter (NOM) and polysaccharide, respectively. The concentration of foulants solution was 1 g L À1 . Before the antifouling experiment, the pure water ux (J W1 ) of NF membranes was measured at 1.0 MPa, and then the solution reservoir was lled with 1.0 g L À1 of foulant solution. Fouling tests were performed under the same pressure, the penetrating uid was collected every 30 min for 24 h, a steady ux (J p ) for foulant was obtained over a period of time. Aer that, the permeation returned to the feed tank to maintain a constant feed concentration. When the ltration of the foulants solution is nished, the membrane was cleaned with pure water for 30 minutes and the recovery ux aer washing was measured (J W2 ). The ux recovery ratio (FRR) and total ux decline ratio (DR t ) were used to signify the antifouling effect of NF membranes, dened as follows: 47

Characterization of NF membrane
ATR-FTIR and XPS were used to understand the chemical structure of the skin layer of the NF membranes. Fig. 2 presents the ATR-FTIR spectra of the surfaces of PSF support membrane (a), nascent NF membrane PA/PSF (b), TEOA-graed NF membrane H 4 -PA/PSF (c) and PFTS-graed NF membrane F-PA/PSF (d).
Compared with the spectrum of the PSF support membrane, an apparent absorption peak at 1627 cm À1 corresponding to C]O stretching vibration of amide I appears in spectra (b), (c) and (d), conrming the formation of polyamide active layer aer IP process. The new peak at 1729 cm À1 of membrane H 4 -PA/PSF (spectra (c)) is ascribed to n C]O of ester compound, which is formed by the esterication reaction between the hydroxyl groups of TEOA and the residual acyl chloride groups on the polyamide active layer. However, there is no difference between spectrum (c) and spectrum (d), this may be because the concentration of PFTS in membrane F-PA/PSF is too low to be detected by ATR-FTIR.
To make sure that the PFTS was successfully graed onto the TEOA-modied membrane, an XPS characterization was further accomplished. Fig. 3  The surface morphologies of the PSF and PA NF membranes were observed by SEM and AFM. As shown in Fig. 4, the PA/PSF membrane had a smooth surface (b) compared with the support PSF membrane (a). Aer the rst-step of the modication, the H 4 -PA/PSF membrane surface appeared some small nodular structures (c), may be due to the aggregation of TEOA. Aer the second-step of the modication (d), the surface of the F-PA/PSF membrane was much rougher. AFM analysis was further carried out for quantitative investigation, and the results are shown in Fig. 4. It was found that the surface roughness of membranes slightly changed: the R a values of membrane PA/PSF, H 4 -PA/PSF and F-PA/PSF were 7.7 AE 0.3, 10.4 AE 1.4, and 12.2 AE 3.1, respectively. A smooth surface restricts the adsorption and adhesion of pollutants, which is favorable to antifouling property. The results were in agreement with the SEM morphology observations.
As shown in Fig. 5, the hydrophilicity of the membrane surface was measured by the surface water contact angle (WCA).
The value of WCA decreased aer the TEOA graing, which means that the hydrophilicity of the membrane surface was increased. The enhancement of membrane hydrophilicity is caused by the increase of hydroxyl groups density in TEOA. This effect is benecial for the water permeation as well as to mitigate the hydrophobic pollutant adsorption onto the membrane surface. This is conducive to the improvement of permeation ux and the reduction of fouling by hydrophobic substances. The membrane surface charge was evaluated through the zeta potential measurements at neutral pH, depicted in Fig. 6. Clearly that all the surface were negatively charged. The zeta potential of TEOA-modied membrane was less negatively charged in comparison to PA/PSF membrane. As increasing the TEOA content, the zeta potential negative charge decreases. This effect can be understood taking into account that some acyl chloride groups reacted with the hydroxyl functional groups of TEOA, thus the amount of the acyl chloride groups for hydrolysis decreased as result of increased amount of TEOA. 48 The zeta potential of the membranes decreased aer a two-step modication, which could be attributed to the introduction of HFBM. The substantial changes in the zeta potential of the membranes indicated the success of the chemical modication.

Separation performance of membranes
The preparation conditions during two-step surface modication (such as concentration of TEOA and PFTS) were examined to deduce the optimal separation performance of the modied NF membrane. The effect of the TEOA concentration on the permeation ux as well as on the NaCl and Na 2 SO 4 rejection of the PA/PSF membranes during the rst-step modication is displayed in Fig. 7. The water ux increased visibly when the TEOA concentration increased from 0 to 4%, and then decreased when the TEOA concentration increased to 5%. On the other hand, the NaCl and Na 2 SO 4 rejection changes only slightly. The improved surface hydrophilicity of the TEOA-modied membranes is the main reason to the improvement of water ux. In addition, ethanol as a solvent facilitates the improvement of ux. This is because the polyamide layer and the PSF support membrane have different swelling degree in the ethanol solvent (swelling degree of the polyamide layer is larger than the support membrane), so the gap between the two layers increases in the ethanol solvent, which is favorable to the passage of more water molecules. 17,50 The slight uctuation in salt rejection is due to the lower electrostatic repulsive effect caused by the reduced surface negative charge. 51 The effect of the PFTS concentration on the permeation ux as well as on the NaCl and Na 2 SO 4 rejection of the PA/PSF membranes during second-step modication is presented in Fig. 8. The ux clearly declined aer PFTS was graed due to the hydrophobicity of PFTS. When the content of PFTS increases from 0.05% to 1.0%, the ux decreases from 46.3 AE 1.0 L (m 2 h) À1 to 39.6 AE 0.9 L (m 2 h) À1 , while the rejection of Na 2 SO 4 and NaCl remains constants about 80% and 30%, respectively.
Molecular weight cut-off (MWCO) values of the membranes were also determined using PEGs as model solutes.  membranes. The rejection performance of modied membrane is hardly affected by the chemical modication proposed in this study. 48,52 This is consistent with the observed salt rejections.

Antifouling property of membranes
The time-dependent normalized ux (J p /J w1 ) proles of the PA/ PSF, H 4 -PA/PSF and F-PA/PSF membranes during ltration of aqueous solutions containing different model foulants are shown in Fig. 10. It is apparent that the uxes of all membranes exhibited an obvious decline when a foulant solution was ltered. Aer some time, stable uxes were obtained for all membranes. There are some interactions between membrane surface and foulants. On the one hand, the foulants deposit onto the membrane surface under the driving pressure and cause serious membrane fouling and ux decline. On the other hand, the foulants deposited on the membrane surface are removed from the membrane surface under the near surface    shear stress. The ux tends to be stable when the adsorption and desorption of foulants reach to an equilibration under the joint effect of two kind of pressure. Aer this, a simple hydraulic cleaning was applied. The values (FRR, DR t ) calculated from uxes before and aer fouling and cleaning were used to evaluate the antifouling properties of the membranes. FRR and DR t values of PA/PSF, H 4 -PA/PSF and F-PA/PSF membranes for ltration of BSA, HA and SA aqueous solutions are given in Table 1. The FRR of the amphiphilic modied F-PA/PSF membrane is higher than that of the blank control PA/PSF membrane and the hydrophilic modied H 4 -PA/PSF membrane, and the DR t is lower than that of the other two membranes during ltration of three different model foulant solutions. This manifests that the F-PA/PSF membrane has better antifouling properties. The good antifouling ability can be mainly ascribed to the amphiphilic nature of the F-PA/PSF membrane. On the one hand, the hydrophilic domains are benecial to bind water molecules and form a compact hydration layer to alleviate the interaction between the membrane surface and potential foulants (fouling-resistant defense mechanism). On the other hand, the foulants can be readily detached under the hydrodynamic shear forces in the presence of hydrophobic segments with low surface energy barriers (fouling-release defense mechanism).   To make sure the PFTS was not washed out during the highpressure ltration testing or the hydraulic cleaning in the antifouling test. XPS characterizations of the ltrated and the fouled-cleaned F-PA/PSF membranes was further carried out to verify the existence of PFTS. The result was depicted in Fig. 11. Compared with the Fig. 3(a), the presence of F1s signal meant the stable existence of PFTS. So the amphiphilic modication of NF membrane was fairly stable under foulants ltration experiment and hydraulic cleaning.

Conclusions
In this study, a simple and effective method was employed to fabricate an amphiphilic NF membrane with hydrophilic domains and low surface energy domains through a two-step surface modication. The surface modication resulted in an unimportant decrease of the MWCO from 305 Da to 290 Da, while the Na 2 SO 4 rejection increased slightly from 80.0% AE 0.8 (PA/PSF) and 82.3% AE 1.1 (H 4 -PA/PSF) to 82.9% AE 1.3, and maintaining a constant NaCl rejection of around 30%. Compared with the control PA/PSF membrane and the hydrophilic modied H 4 -PA/PSF membrane, the antifouling properties of the amphiphilic modied F-PA/PSF membrane were clearly enhanced. FRR values were 96.9%, 97.5%, and 96.8% for a BSA solution, HA solution and SA solution, respectively. As expected, the two-step surface modication did not alter the pore size of membranes. The membrane aer amphiphilic modication was fairly stable, and PFTS on the surface were not washed out throughout the experiment. In addition, it should be noted that the introduction of hydrophobic PFTS in F-PA/PSF membrane promotes the water ux decline. Hence, this investigation can provide a method for fabrication of antifouling NF membranes under mild conditions.

Conflicts of interest
There are no conicts to declare.