Nano-ZnO impregnated inorganic–polymer hybrid thinfilm nanocomposite nanofiltration membranes: an investigation of variation in structure, morphology and transport properties

Inorganic–polymer hybrid, thinfilm nanocomposite nanofiltration (TFN-NF) membranes prepared by in situ interfacial polymerization of branched polyethyleneimine and trimesoyl chloride, with simultaneous impregnation of as-synthesized hexagonal wurtzite ZnO nanocrystals (nano-ZnO), either through aqueous or organic phase, have been extensively characterized. XPS analysis revealed that there was no inter-atomic charge transfer between nano-ZnO and host polyamide matrix, indicating that no formation of chemical bonding occurred between them in the skin layers of the membranes. The type of interaction present within the nanocomposite polyamide matrices of the membranes was through formation of noncovalent type secondary chemical interactions with peripheral hydroxyl groups of nano-ZnO and polyamide network as substantiated through FTIR analysis. SEM revealed the formation of distinct patterns and coils, through multiple-point interactions between the nano-ZnO and the polyamide network in the membranes' skin surfaces when introduced through aqueous amine phase. However, when introduced through the organic phase, nanomaterials remained distributed as discrete clusters within the membranes' skin layers because of lack of polar environment around the reaction zone, further emphasizing the role of the medium in which the nanomaterials are incorporated. AFM showed variation of surface roughness features with change in the precursor medium of introduced nano-ZnO. Nanofiltration performance towards different solutes, providing differential rejections in the order of MgCl2 > NaCl $ Na2SO4, revealed that the membranes were distinctly positively charged. Solvent fluxes of the membranes were significantly higher when nanomaterials were introduced through the aqueous phase as compared to the organic phase.


Introduction
Pressure-driven nanoltration (NF) membrane processes have become signicantly useful in recent years in aqueous stream separations. They offer exciting options for various industrial applications due to their ability to achieve high selectivity between mono and multivalent ionic species from mixed electrolyte streams, with additional advantages such as low transmembrane pressure, low operational cost and high permeate ux. 1,2 Thin lm composite (TFC) type membranes having an ultrathin selective barrier layer, generally produced by in situ interfacial polycondensation of reactive monomers on a suitable substrate material, constitute the majority of commercially available polymeric nanoltration membranes. The barrier skin layer, made of synthetic polymeric materials, additionally and preferentially containing either negative or positive chargebearing functionalities, inuences the separation behaviors of the membrane by preferentially retaining certain components and allowing the passage of certain others, depending on interaction between surface charge of the membrane and charges of the solutes under consideration. 3,4 The major areas of application for these membranes include bulk wet-stream processing such as water purication for production of drinking water, effluent treatment for removal of toxic elements, 5 dairy industries, 6 food and pharmaceutical industries, 7,8 textile industries, 9 and biotechnological industries. 10 For almost a decade, when it seemed that membrane performance in terms of solute rejection, and especially the solvent ux, had reached its limits, the advent of nanotechnology provided a new dimension to scientists by changing the scenario. 11 Membranologists explored promising ways to overcome some of the limitations associated with synthetic polymeric membranes by developing new types of organicinorganic hybrid nanocomposite materials. [12][13][14] Extensive research on various metals and metal oxides as well as ceramic and zeolite nanoparticles has been conducted to obtain improved physicochemical properties in the membranes. Production of membranes by incorporating nanomaterials resulted in better chemical, mechanical and thermal resistance and also prevented adhesion as well as adsorption of the undesired fouling materials on their surfaces, thereby maintaining membrane permeability and selectivity as well as minimizing the load on pretreatment and chemical cleaning. It has been observed that microstructural features of the membranes are affected by site-specic interaction between the metal oxide nanoparticle and the host polymer matrix, altering the skin layer morphology, pore size, pore size distribution, surface roughness, and so on, which eventually inuence their performances. Various metal oxide nanoparticles such as silica (SiO 2 ), 15,16 alumina (Al 2 O 3 ), 17,18 titania (TiO 2 ), [19][20][21] ferric oxide (Fe 2 O 3 ), 22 zinc oxide (ZnO), 23,24 and zeolites, 25,26 have been employed as nanollers in nanocomposite membranes' polymeric host matrix leading to promising improvements like better chemical and thermomechanical properties.
ZnO nanoparticles have recently attracted membranologists' attention as a suitable nontoxic, bio-safe and bio-compatible material with strong inhibitory and antibacterial effects over a broad spectrum of both G + and G À bacteria. [27][28][29] Adams et al., 30 Brayner et al. 31 and Jeng and Swanson 32 have reported that ZnO is more efficient in controlling the growth of micro-organisms than other metal oxides such as SiO 2 , MgO, TiO 2 and CaO, by signicantly inhibiting the growth of a wide range of pathogenic bacteria under visible light conditions. The commonly proposed antimicrobial mechanisms for nano-ZnO are (1) generation of reactive oxygen species like hydrogen peroxide (H 2 O 2 ), hydroxyl radical (cOH) and singlet oxygen ( 1 O 2 ) from the surface of zinc oxide, penetrating the cell membrane and inhibiting cell growth, and (2) rupture of the cell membrane and leakage of cytoplasmic content because of binding between the bacterial cell and ZnO nanoparticles due to electrostatic forces. [33][34][35] The present study emphasizes investigations of structural and physicochemical features of inorganic-polymer hybrid, thinlm nanocomposite nanoltration (TFN-NF) membranes, prepared by impregnation of as-synthesized nano-ZnO into the polymer skin layer during the in situ interfacial polycondensation of branched polyethyleneimine (PEI) and trimesoyl chloride (TMC). The synthesized nano-ZnO has been characterized by X-ray powder diffraction (XRD) analysis, Fourier transform infra-red (FTIR) spectroscopy, Raman spectroscopy and X-ray photoelectron spectroscopy (XPS). For membranes, XPS as well as FTIR analysis of skin layers showed the inuence of nano-ZnO, with variation in the ways of incorporation, on the electronic-and molecular-level interactions occurring within the nanocomposite polyamide matrices of the TFN-NF membranes. Changes in surface morphologies of the skin layer of nanocomposite membranes were investigated via scanning electron microscopy (SEM). Quantitative surface roughness analysis via atomic force microscopy (AFM) was applied to probe the role of nanomaterials in inuencing membrane topographies. Energy dispersive X-ray (EDX) analysis has been employed to determine the elemental composition of TFN-NF membrane skin layers. Cross-ow permeation experiments were performed to correlate the variation in physicochemical features of the membranes with their transport parameters such as solute rejection and solvent ux behaviors.

Synthesis of ZnO nanocrystals
ZnO nanocrystals (nano-ZnO) were synthesized by thermal decomposition of anhydrous Zn(acetylacetonate) 2 in n-octylamine at temperature as low as 110 C. The precursor anhydrous Zn(acetylacetonate) 2 was prepared following the Rudolph and Henry method. 36 In a typical preparation, 0.5 g (1.89 mmol) of anhydrous Zn(acetylacetonate) 2 was combined in a 50 mL three-neck ask with 10 mL ($60.5 mmol) of n-octylamine. The reaction ask was evacuated to a vacuum level of 2 mbar for 30 min at 40 C. The reaction system was then heated ($10 C min À1 ) to 110 C under argon ow and maintained at this temperature for 2 h. The obtained white suspension was cooled to 60 C and 5 mL of methanol was added to the reaction ask. The nanocrystals of ZnO were recovered by centrifugation, dispersed in toluene, and precipitated by adding methanol. The redispersion and precipitation steps were repeated twice to remove any unreacted precursors and excess solvent.

Preparation of Psf base membrane
The at-sheet microporous Psf base membrane was prepared following the nonsolvent induced phase inversion method, where a homogeneous casting solution, comprised of 30 (w/v)% Psf in NMP, along with the pore-forming additive, PVP (40 w/w% of Psf), was cast on a nonwoven polyester substrate to a thickness of 100 mm using an automated casting machine. The membrane was immediately dipped in a water bath and maintained at room temperature for immersion precipitation. The water of the bath was changed several times to ensure complete removal of the solvent (NMP) and additive (PVP) from the membrane matrix via leaching. The entire casting machine was kept in an environmentally controlled atmosphere where temperature and relative humidity were maintained at 25 C and 35-40%, respectively. This Psf membrane was subsequently used as a base material for preparation of the thinlm nanocomposite nanoltration (TFN-NF) membranes.

Preparation of Control-NF and inorganic-polymer hybrid TFN-NF membranes
A control, thinlm composite nanoltration membrane (Control-NF), devoid of any nanomaterial, was prepared by in situ interfacial polycondensation reaction (Scheme 1) between PEI (in aqueous medium) and TMC (in organic, n-hexane medium). A piece of microporous Psf base membrane was treated with 4 (w/v)% aqueous solution of PEI for 4 min. Aer removing the amine soaked base membrane from the aqueous medium and eliminating the excess solution by gently squeezing it between two rubber rollers, the membrane was dipped in a 0.1 (w/v)% TMC solution for 30 s. The nascent TFC membrane, comprised of a freshly deposited, ultrathin polyamide lm over the surface of the Psf base membrane, was further heat cured at 90 C in a hot-air circulated oven for 30 min.
For development of the inorganic-polymer hybrid TFN-NF membranes, the nano-ZnO was introduced in either aqueous or organic phase. The nano-ZnO was dispersed in aqueous or n-hexane medium by applying ultrasonication for 15 min prior to the start of the membrane-making process. When prepared by incorporation of the nano-ZnO through the aqueous phase, 4 (w/v)% aqueous solutions of PEI containing well-dispersed nano-ZnO of varying concentrations: 0.05, 0.1 and 0.2 (w/v)% were used. The membranes were prepared using the same procedure as for the Control-NF by performing an in situ interfacial polycondensation reaction with 0.1 (w/v)% TMC solution (schematics in Fig. 1a), followed by curing at 90 C for 30 min. These membranes were named Aq-ZnO:TFN-NF-1, Aq-ZnO:TFN-NF-2 and Aq-ZnO:TFN-NF-3, respectively.
Similarly, in another set of experiments, nano-ZnO of various concentrations of 0.05, 0.1 and 0.2 (w/v)% were ultrasonically dispersed in n-hexane, containing 0.1 (w/v)% TMC monomer. The in situ interfacial polycondensation was conducted adopting the same approach as for the Control-NF, with 4 (w/v)% aqueous PEI solution (schematics in Fig. 1b), followed by curing at 90 C for 30 min. These membranes were named Org-Scheme 1 Synthesis of polyamide layer of Control-NF from PEI and TMC following in situ interfacial polycondensation process. ZnO:TFN-NF-1, Org-ZnO:TFN-NF-2 and Org-ZnO:TFN-NF-3, respectively. They were washed with de-ionized water to remove the leachable chemicals before conducting performance evaluation tests.

Textural and structural characterizations of assynthesized nano-ZnO
Phase purity and structure of the ZnO nanocrystals were determined by X-ray powder diffraction (XRD) data, which were collected on a Philips X'Pert pro X-ray diffractometer using Cu-Ka radiation (l ¼ 1.5418Å) at 40 kV and 30 mA.
The FTIR spectrum of the ZnO nanocrystals was recorded in attenuated total reectance (ATR) mode using a Bruker Vertex 70 FTIR spectrometer, in the range of 500-1500 cm À1 .
Micro-Raman spectra of the nanocrystals were obtained on a LABRAM-I, ISA-make spectrometer using an Ar + ion laser (488 nm) equipped with a Peltier-cooled CCD detector.
For XPS characterization of the ZnO nanocrystals, a DESA-150 electron analyzer (Staib Instruments, Germany) equipped with a Mg-Ka X-ray source (1253.6 eV) was employed. The spectrometer's binding energy scale was calibrated with the Au-4f 7/2 photopeak at a binding energy (BE) of 83.95 eV. Curve ttings of intensity versus binding energy (BE) plots comprising the multiplex photo peaks were performed employing a leastsquares peak analysis soware (XPSPEAK 4.1). Using the Gaussian/Lorentzian sum function, peak areas as well as corresponding full width at half-maximum (FWHM) were also determined for the high-resolution Zn-2p and O-1s spectra. The tting error in both peak position and FWHM are $0.1 eV. Spectral characterization of skin layers of TFN-NF membranes by FTIR spectroscopy. A Bruker Vertex 70 FTIR spectrometer, equipped with an ATR attachment (ZnSe crystal, 45 angle of incidence and refractive index 2.4), was employed for ATR-FTIR analysis of membranes. The membrane's skin surface was kept face down on the ATR crystal element and light pressure was applied using a MIRacle high-pressure clamp with torque-limited press. The radiation penetration depth was 2 mm. All infrared spectra were recorded in transmittance mode in the range of 600-4000 cm À1 at 25 C. Evaluation of all spectra was carried out by acquiring 100 scans at a spectral resolution of 2 cm À1 .

Characterization of TFN-NF membranes
Morphological investigation of TFN-NF membranes by SEM. The morphological analyses through SEM of the TNF-NF membranes' ultrathin skin surfaces were carried out using CamScan-CS3200LV (UK). In order to make the membranes electrically conducting, 2 mm 2 strips of the dry membrane samples were sputter coated on the skin surfaces with goldpalladium alloy for 60 s employing a 30 mA current. The skin layer images were recorded at an acceleration voltage of 15 kV and 4000Â magnication when operated in secondary electron mode. For acquiring the cross-sectional images of the TFN-NF membranes, the membrane samples (without support polyester fabric) were fractured in liquid nitrogen into smaller sized strips and sputter coated on the cross-sectional layer with goldpalladium alloy for 100 seconds at a current of 15 mA. The image acquisition was done in secondary electron mode at an acceleration voltage of 20 kV and 1000Â magnication.
Topographical characterization of skin surfaces of TFN-NF membranes by AFM. Topographical characterization of membrane skin surfaces was done by extracting quantitative data from the three-dimensional (3D) topographies acquired in the semi-contact mode by using an AFM instrument (NT-MDT SOLVER next; Russia). Small squares of approximately 1 cm 2 of the membrane were cut and glued onto a metal substrate. The rectangular cantilever NSG 10 (NT-MDT, Russia) was made from Si 3 N 4 with a spring constant of 11.8 N m À1 , having a typical resonance frequency of 240 kHz and a nominal tip-apex radius of 10 nm with a high aspect ratio. The scanning was done on a 10 mm Â 10 mm area of the membrane in air at room temperature, with a scanning frequency of 0.1 Hz. The scanned regions were attened using a second-order polynomial to remove curvatures and slopes from the image and then the resulting best t was subtracted from it. NOVA-P9 soware was used for the purpose of image acquisition and evaluation of surface roughness parameters of the membranes. Surface roughness parameters of the membranes were calculated from the height proles of the image in terms of average roughness (R a ), root mean square roughness (R q ) and ten point height (R 10Z ). 37 Elemental analysis and mapping of skin surfaces of TFN-NF membranes. Quantitative surface elemental analysis and mapping were performed by an energy-dispersive X-ray spectrometer (EDX) coupled to the SEM and a micro-analysis system (INCA Oxford Instruments, UK) employing an accelerating voltage of 20 kV and a magnication of 4000Â. It was equipped with an ultrathin beryllium window and 20 mm 2 Si detector. The spectra were acquired for 2 h to map the elements of interest such as N, Zn and O in each membrane. The maximum depth of penetration of the beam (beam diameter ¼ 7.5 nm) inside the sample until energy of $10 kV (BE of C-1s: 284.2 eV, BE of O-1s: 543.1 eV, BE of N-1s: 409.9 eV, BE of S-1s: 2472 eV and BE of Zn-1s: 9659 eV) was found (calculated using Casino V3, Monte Carlo simulation of electron trajectory in solids) to be about 3 microns with width of about 3 microns. Thus, the volume of penetration of the e-beam was very small with respect to the large sample volume. The elemental analysis of N, Zn and O were carried out over 170 points in a grid spectrum where each point was allowed to undergo a 60 s detection period.

Cross-ow permeation experiments of TFN-NF membranes.
Transport experiments of the Control-NF, Aq-ZnO:TFN-NF and Org-ZnO:TFN-NF classes of membranes were conducted in cross-ow test cells, using single-feed solutions of three different solutes consisting of varying combination of monovalent and bivalent cations and anions. Solutions of NaCl, MgCl 2 and Na 2 SO 4 , having concentrations iso-osmotic with 0.03 M NaCl, were used as feeds. The experiments were conducted at 10 bar transmembrane pressure using a pump with a feed ow rate of 10 L min À1 . Temperature of the feed solution was maintained at 25 C. Circular membrane samples having an effective membrane area of 15.1 cm 2 were properly washed in deionized water and placed in the test cell with their selective skin layers facing the incoming feed. All membranes were given sufficient time for stabilization before evaluating their performances.
Steady-state solute rejections of the membranes were determined by measuring the solute concentrations in the feed and permeate using a digital conductivity meter (TESTRONIX 15, Microlab, Mumbai, India). The percent solute rejection, R, was calculated using the following expression: where C p and C f are the bulk concentrations of the solutes in permeate and feed solutions, respectively. All membrane samples were prepared and tested in duplicate for their NF performances.
The corresponding permeate solvent ux (J W ) was determined by direct measurement of the permeate ow, V (mL), through the membrane of area A (cm 2 ), for a given time t (min), using expression (2), and presented in liters per meter 2 per day (LMD).

Analysis of textural and structural features of nano-ZnO
The X-ray diffraction patterns of nano-ZnO obtained by thermal decomposition of the acetylacetonate in n-octylamine at 110 C are shown in Fig. 2. The patterns can be indexed to hexagonal wurtzite lattice structure of the nano-ZnO. The indexing to X-ray patterns is attributed to (100), (002), (101), The FTIR spectrum of the as-synthesized ZnO nanocrystals, acquired in atmospheric conditions, is shown in Fig. 3. The most prominent band at 510 cm À1 and a weak band at 660 cm À1 are assigned to the stretching vibrations of Zn-O bonds in the tetrahedral and octahedral coordinations, respectively. 38 A Raman spectrum of the ZnO nanocrystals acquired at room temperature is shown in Fig. 4. The hexagonal wurtzite structure of ZnO belongs to the space group C 4 6v with two formulae units per primitive cell, where all atoms occupy C 3v sites. The Raman active zone-centre optical phonons predicted by group theory are A 1 + 2E 2 + E 1 . Of these, the phonons of A 1 and E 1 symmetry are polar and both Raman and IR active. Hence, they show different frequencies for the transverse optical (TO) and longitudinal optical (LO) phonons. A weak band at 407 cm À1 is assigned to E 1 TO phonon mode. The nonpolar optical phonon modes with symmetry E 2 , being Raman active only, exhibit dual frequencies where E 2 (low frequency) at 99 cm À1 is associated with oxygen atoms, and E 2 (high frequency) at 437 cm À1 is associated with the Zn sublattice, which are characteristic of the wurtzite hexagonal phase of ZnO. [39][40][41][42] The XRD, FTIR and Raman scattering measurements conrm the formation of ZnO nanocrystals in hexagonal wurtzite structure.   44 However, the weakest but relatively broadened shoulder peak at 532.5 eV with FWHM of 2.04 eV is due to chemisorbed oxygen contributed by surface hydroxyl groups. 45

Characterization of skin layers of TFN-NF membranes by XPS
Representative TFN-NF membranes, one each from the series made by introducing nano-ZnO through aqueous phase as well as organic phase (i.e., Aq-ZnO:TFN-NF-2 and Org-ZnO:TFN-NF-2), were employed for XPS analysis to investigate the electronic-and molecular-level interactions of the nanocomposite polyamide skin layers within the XPS probe depth (5-10 nm). Further, comparisons of XPS analysis of skin layers of these two membranes with that of the Control-NF were made, with respect to the variation in core-level binding energies and respective peak areas, to study the effect of nano-ZnO impregnation in polyamide matrices.
The Zn-2p photoelectron spectra of nano-ZnO impregnated polyamide skin layer of Aq-ZnO:TFN-NF-2 and Org-ZnO:TFN-NF-2 are presented in Fig. 6b and c, respectively. In the case of Aq-ZnO:TFN-NF-2, the high-intensity peak at the low BE of 1020.7 eV (FWHM: 1.87 eV) corresponds to Zn-2p 3/2 and the peak at the higher BE with the lower intensity, attributed to Zn-2p 1/2 , appears at 1044 eV (FWHM: 3.64 eV) (see Table 1).
Similarly, for the Zn-2p photoelectron spectrum of Org-ZnO:TFN-NF-2, the peak corresponding to Zn-2p 3/2 appears at 1020.6 eV (FWHM: 2.75 eV) and a weaker peak, corresponding to Zn-2p 1/2 , appears at 1043.7 eV (FWHM: 3.94 eV). The value of the tting parameter, c 2 resides in the range of 1.12-2.25, which validates the fact of proper tting of the curves in the Zn-2p XPS spectra. No signicant deviations in BE values for Zn-2p 3/2 , as well as Zn-2p 1/2 of both Aq-ZnO:TFN-NF-2 and Org-ZnO:TFN-NF-2, from the corresponding BE values of the pure nano-ZnO, indicate that there is no redistribution of charge density in the Zn-O framework; hence, there is no occurrence of interatomic charge transfer between the nano-ZnO and the host polymeric material. In other words, ZnO does not form any chemical bonds with the host polymer matrix during the nanocomposite polymer formation. Thus, it is assumed that only the peripheral hydroxyl groups of nano-ZnO play a role in    ZnO:TFN-NF-2, respectively. The curve tting and deconvolution of each O-1s XPS spectrum yield two constituent spectra, which indicate the presence of oxygen in two different chemical states. The value of c 2 stays in a 1.00-1.45 range, which conrms the fact of proper tting of the O-1s curves in all spectra. The intense peak with low BE value of 530.4 eV (FWHM: 1.99 eV) for the Control-NF is assigned to physisorbed oxygen ( Table 1) (Table 1), correspond to the 2 nitrogen atoms of free N-H of polyamides and free 1 as well as 2 amines of the polymer backbone. For N-1s XPS spectra, the c 2 value resides in the 0.90-1.75 range, indicating a proper t. The peaks with lower intensity at higher BE of 400.6 eV for Control-NF, 400.0 eV for Aq-ZnO:TFN-NF-2, and 400.7 eV for Org-ZnO:TFN-NF-2 are resulted from 3 nitrogen atoms of amides and amines of the polymer backbone as well as nitrogen atoms of amines and amides that are involved in intermolecular H-bondings, either through inter-chain interaction or through the surface hydroxyl groups of nano-ZnO. The peak positions are concordant with reported data in the literature. 50 The chemical shis of the N-1s signals to higher BE values in the case of Control-NF as well as Org-ZnO:TFN-NF-2, as compared to Aq-ZnO:TFN-NF-2, are attributed to enhanced protonation of amine sites during the course of polycondensation reaction of the former two membranes. This may have resulted in an increase in the positive charge densities on N sites of these membranes. The peak area referred to free N-H sites in the polyamide corresponding to the Control-NF, gets decreased by 14.02% as nano-ZnO of 0.1% is incorporated in the skin surface of Aq-ZnO:TFN-NF-2. This is ascribed to the extensive association of the nano-ZnO with N-H of polyamides and 1 as well as 2 amines of polymer backbone, reducing the number of free amine sites. However, the peak area assigned to similar N-H sites corresponding to the Org-ZnO:TFN-NF-2 is increased only by 2.56% as compared to the Control-NF, indicating that there are less extensive H bondings and more free N-H sites. These changes in the areas of the major peaks for Aq-ZnO:TFN-NF-2 and Org-ZnO:TFN-NF-2 are compensated in the areas of the shoulder peaks at a higher BE. The peak areas of N and O are sensitive to the presence and variation of nano-ZnO in the polyamide skin layer, and hence may not be considered as absolute.

Spectral characterization of TFN-NF membranes by ATR FTIR
ATR FTIR spectrum of Control-NF and those of the TFN-NF membranes, made by introducing nano-ZnO either through the aqueous or organic phase, are presented in Fig. 7a and b, respectively. They show that introduction of progressively increasing amounts of nano-ZnO through either medium leads to signicant physicochemical changes in the skin layer polymers of the resulting nanocomposite membranes.
The ZnO nanoparticles, which are produced by synthesis in n-octylamine as in the present case, are known to have a tendency to chemisorb water molecules on their surfaces making them predominantly hydrophilic 51,52 and assuming a structure where the core material is enveloped in a sheath of hydroxyl groups. When this nano-ZnO, at varying concentrations of 0.05, 0.1 and 0.2 (w/v)%, is taken into the aqueous monomeric solution containing PEI, due to the alkaline nature of the solution, it remains in the form of nano-ZnO(OH) n , where "n" represents the number of surface-anchored hydroxyl groups depending on the pH of the medium. 53 In this phase, extensive noncovalent-type of secondary chemical interactions occur, mostly in the form of H-bondings, between surface hydroxyl groups of nano-ZnO and amine groups of the branched PEI monomer. These H-bondings preferentially engage the primary amine sites of PEI because, in addition to being more electropositive than the H of the secondary amines, the former sites are known to be oriented more towards the outside in a somewhat spherically coiled structure of the polymer. 54 When PEI, in the form of this molecular association, participates in an interfacial polycondensation reaction with TMC, the resulting cross-linked polyamide is assumed to form similar H-bondings involving peripheral OH of nano-ZnO and N-H as well as >C]O of amide groups of the polyamide. Thus, when nano-ZnO accommodates in the interstices of the nanocomposite polymer it promotes more H-bondings in the polyamide network structure of Aq-ZnO:TFN-NF-1 than that of the Control-NF. The ATR FTIR spectra of the nanocomposite membranes, presented in Fig. 7a, corroborate these facts where deviations are seen in both amide-I region, i.e., in carbonyl stretching frequency, n C]O and amide-A region, i.e., N-H stretching frequency, n N-H of the Aq-ZnO:TFN-NF-1 from those of the Control-NF. Thus, with incorporation of 0.05% nano-ZnO in Aq-ZnO:TFN-NF-1, the n C]O value shows a red shi from 1656 to 1647 cm À1 and n N-H from 3268 to 3264 cm À1 . However, a reverse trend is noticed, showing a blue shi for n C]O from 1647 to 1661 cm À1 and then to 1664 cm À1 as well as for n N-H , from 3264 to 3279 cm À1 and further to 3288 cm À1 , with the incremental addition of 0.1 and 0.2% nano-ZnO in making the respective membranes. This reversal of trend is assumed to happen because at this higher concentration the nanomaterials nd much less space to accommodate in the resulting polymer network, forcing a reorganization of the polymer chains and subsequent disruption of the intermolecular H-bondings. The blue shi in n C]O values may further reect the fact that with enhanced nanoparticle density in the nanocomposite polymer matrix, there is a decline of dipole-dipole interaction or n / p* interaction happening though delocalization of a lone pair (n) of the donor carbonyl groups of amide segments or amine groups of the PEI chains into the antibonding (p*) orbital of the acceptor carbonyl groups (C]O/C]O or C-N/C]O) of the adjacent polymer chains. This also supports the above mentioned occurrence of reorganization of the polymer chains affecting the conformationally sensitive amide-I region of amide segments. [55][56][57] In contrast to the aqueous phase, introduction of nano-ZnO in the polyamide matrix through the organic phase results in an initial blue shi in the n C]O from 1656 to 1663 cm À1 and n N-H from 3268 to 3285 cm À1 , for Org-ZnO:TFN-NF-1 as compared to the Control-NF (Fig. 7b). It is well known that there are extensive intermolecular H-bondings between the amide-I and amide-A regions as well as strong dipole-dipole interaction or n / p* interaction among the adjacent chains of polyamides in Control-NF. More conformationally free >C]O groups in Org-ZnO:TFN-NF-1, evident from its higher n C]O value than the Control-NF, implies that there are less extensive H-bondings in the nanocomposite polymer matrix. This is possible if the polymer chains of the former membrane are farther away from each other, with fewer >C]O groups present within the range of H-bonding distance with the adjacent H atom of either N-H of the amide group or peripheral OH of ZnO to experience extensive noncovalent interactions. The pathway of the nonpolar organic medium employed for introducing nano-ZnO may be responsible for this type of network structure of the nanocomposite polyamide, resulting in lesser extent of H-bondings for Org-ZnO:TFN-NF-1. This contrasts with the counterpart of aqueous phase, that is, Aq-ZnO:TFN-NF-1. However, with the increase in nanoparticle density in the organic phase from 0.05 to 0.1 (w/v)% and then to 0.2 (w/v)%, a red shi occurs from 1663 to 1656 cm À1 via 1658 cm À1 for n C]O and 3285 to 3260 cm À1 through 3274 cm À1 for n N-H . The former change can be ascribed to an increase in secondary chemical interactions within the nanocomposite polyamide matrix, which is assumed to happen because of more nano-ZnO distributed in the open interstices of the polyamide network promoting more H-bondings between the surface OH of the nano-ZnO and the nearby >C]O groups of the polyamide. However, the later change in n N-H may result mostly due to the involvement of more N-H sites in forming association with the liberated HCl during the course of reaction, which can further lead to progressive increase in positive charge densities of the membranes. The ATR FTIR results for both classes of membranes show that the medium of incorporating nanomaterial into TFN-NF membranes has signicant inuence over the nature of interaction within the polymer network and physicochemical features of the inorganic-polymer hybrid material.

Analysis of skin surface morphology of TFN-NF membranes by SEM
The compatibility of nano-ZnO as a constituent of the monomeric aqueous amine phase is evident in the clear homogenous nano-ZnO/PEI/water solution (Fig. 8b). In contrast, the turbid suspensions of nano-ZnO/water solution (Fig. 8a) and of nano-ZnO/TMC/n-hexane solution (Fig. 8c) indicate the physicochemical incompatibility of the nanomaterials in the medium. Electrodynamic interactions between the nano-ZnO and the two respective dispersion media, water and n-hexane, play signicant roles in determining interparticle interactions. Thus, because van der Waals pair interaction energy is proportionally dependent on a material property, the Hamaker constant (which varies with the contrast between the dielectric constants of nanoparticles and the dispersion medium), and inversely dependent on the separation distance between the particles (which differs according to concentration and surrounding chemical environment of the particle), it becomes a measure of stability for such dispersions. Higher values of the Hamaker constant for nano-ZnO in water (26 Â 10 À21 J) as compared to that in n-hexane (19 Â 10 À21 J) reects the higher degree of incompatibility of nano-ZnO in water than in n-hexane since a higher value of the parameter implies greater interparticle attractive force and, hence, more prone is the system to come closer and induce occulation. 58 However, in the PEI environment of the aqueous monomeric solution, the occulation of nano-ZnO gets restricted due to a reduction in the attractive van der Waals inter-particle potential through accumulation of PEI chains on their surfaces. The steric repulsion between the chains of branched PEI molecules on the exterior of sheathed nano-ZnO also helps in stabilizing the solution, emphasizing the role of the former in enhancing the physicochemical compatibility of the latter in the solution. 59 Scanning electron micrographs of the skin layer as well as cross-sections of the Aq-ZnO:TFN-NF class of membranes are presented in Fig. 9a-c. Skin layer images show that the particles of nano-ZnO are well accommodated and evenly distributed inside the nanocomposite polyamide matrix of the membrane, formed through an interfacial polycondensation process. With increased concentration of the nano-ZnO from 0.05 to 0.2%, their density on the membranes' skin layers increase, which is supposed to perturb the polymeric chains through stretching and induce deformation in the network structure. As the nano-ZnO domains grow with increased concentration, they arrange themselves to form some long-range ordering and patterns indicating that extensive cohesive interactions occur between the nanomaterials themselves as well as between the nanomaterials and polymer chains. This pattern formation presumably occurs through multiple-point secondary chemical interactions between the nano-ZnO and the polyamide network which lead to a signicant loss of translational entropy for the particles as well as conformational entropy for the polymeric chains. At higher nano-ZnO density of 0.2%, therefore, there are formation of a greater quantity of smaller coils of the nanomaterials in the polymer matrix to compensate for the entropy loss. 60 In contrast to the aqueous amine medium, the nano-ZnO is chemically incompatible in solution of n-hexane containing TMC and results in the formation of an unstable dispersion (Fig. 8c). The skin surface as well as cross-sectional SEM images of the Org-ZnO:TFN-NF class of membranes, prepared by introduction of nano-ZnO via the organic phase, are presented in Fig. 10a-c. Skin layer images show that the nanomaterials are not evenly distributed in the polymer matrix but rather present in the form of discrete clusters whose number and size increase with rising concentration of the nanomaterials in the organic phase. This indicates that the nanomaterials are unable to cohesively accommodate themselves within the polymer matrix when they are introduced through the organic phase. The orders and patterns found in the membranes of the Aq-ZnO:TFN-NF series are signicantly missing. Since formation of the thin skin layer of the TFN-NF membrane takes place through cross-over of the PEI monomers from the aqueous phase to the organic side and across the polymer barrier (Fig. 1b), in a diffusion-controlled manner by the in situ interfacial polycondensation route, 61 the nanomaterials do not nd adequate opportunities to encounter a compatible polar environment to get evenly distributed before being embedded in the  These observations through SEM studies emphasize the fact that the dynamics of skin layer formation through the interfacial polycondensation process, distribution of the nanomaterials within the skin layer and the resultant skin layer morphology of the nano-ZnO impregnated TFN-NF membranes are greatly inuenced by the type of precursor media employed for dispersion of the nano-ZnO.

Evaluation of skin surface topographies of the TFN-NF membranes by AFM
The 3D AFM images as well as the respective height histograms of TFN-NF membranes, prepared by introducing nano-ZnO through the aqueous phase are presented in Fig. 11. They show that the surface topographies of the membranes change signicantly from a relatively smooth prole for the membrane Aq-ZnO:TFN-NF-1 to progressively rougher surfaces for membranes made with impregnation of more and more nanomaterials. Further, pattern formations on membrane surfaces, a feature observed in the respective SEM images as well, justify the occurrence of long-range association of the nanomaterials in the polymer matrix. The 3D AFM images of Aq-ZnO:TFN-NF class of membranes, read along with the corresponding height histograms, show that the peaks are more or less uniform in height, having a very narrow distribution of 18-28 nm for Aq-ZnO:TFN-NF-1. However, in the case of Aq-ZnO:TFN-NF-2, the surface is seen to be populated with a larger number of peaks whose heights are signicantly greater than those of Aq-ZnO:TFN-NF-1, with most occurring at 20-130 nm, contributing to a much broader distribution pattern. The greater area of the height histogram of Aq-ZnO:TFN-NF-2 also indicates a higher peakedness of this membrane compared to Aq-ZnO:TFN-NF-1, and consequently, an increased membrane surface area of the former membrane compared to the latter. With a still higher concentration of the nanomaterial in Aq-ZnO:TFN-NF-3, the peak heights of the membrane are far from uniform; their distribution is much wider and ranged between 10 to 270 nm. A signicant tail in the height histogram suggests the presence of a few isolated high peaks with heights between 150 to 270 nm. As compared to Aq-ZnO:TFN-NF-2, where a number of narrow peaks are distributed over the entire surface of the membrane, the presence of higher concentration of nano-ZnO makes some of the isolated high peaks in Aq-ZnO:TFN-NF-3 quite broad.
The computed amplitude parameters of the nanocomposite membranes, presented in Table 2, show that the average roughness (R a ) of the membranes increases from 1.48 to 17.59 nm and further to 31.69 nm as concentration of nanomaterials in the aqueous monomer solution is increased. The corresponding root mean square roughness (R q ) of the membrane increases from 2.09 to 21.96 nm and nally to 41.96 nm. This is in accordance with the topographical features discussed in response to their 3D images. R 10z , which is the difference in heights between the average of the ve highest peaks and the ve lowest valleys relative to the mean plane, denotes the contrast between the heights and depths of extreme peaks and valleys that may be occasionally present on the surface of the membrane. The lowest R 10z value of 9.96 nm for Aq-ZnO:TFN-NF-1 results from a surface having peaks and valleys of less contrast in heights and depths. However, impregnation of a still higher amount of nanomaterials leads to higher values of R 10z for Aq-ZnO:TFN-NF-2 (39.89 nm) and Aq-ZnO:TFN-NF-3 (101.57 nm), which indicates the occurrence of greater contrast in heights and depths due to the onset of isolated high peak and low valley formation on the membrane surface.
The 3D AFM images and height histograms of the Org-ZnO:TFN-NF class of membranes, prepared by introduction of nano-ZnO through the organic phase, presented in Fig. 12, show different types of surface topographies than the membranes prepared by introduction of the nanomaterials through the aqueous phase. The occurrence of long-range patterns due to distribution of nano-ZnO in the polymer nanocomposite, a characteristic of the latter class of membranes, is signicantly absent. The nanomaterials, though found present throughout the membranes' skin surfaces in the form of discrete clusters, are not evenly distributed. The 3D images and the respective height histograms for Org-ZnO:TFN-NF-1 and Org-ZnO:TFN-NF-2 show that the number of peaks are fewer and their heights are much lower than for respective membranes in the aqueous phase. For the Org-ZnO:TFN-NF-1, the peak heights on the membrane surface are between 20 to 45 nm and, for Org-ZnO:TFN-NF-2, between 10 to 45 nm. However, for Org-ZnO:TFN-NF-3, the peak heights range between 100 to 500 nm, which is due to the presence of occasional high peaks on the surface of the membrane caused by entrapment of larger-sized clusters of nano-ZnO in the polyamide network.
The R a values of the membranes, listed in Table 2, show that they are 4.07 and 5.58 nm, respectively, when 0.05 and 0.1% of the nano-ZnO are used. This nominal increase in roughness values is due to the absence of any signicant number of peaks on the membrane surface, unlike their counterparts of the aqueous phase membranes. However, the R a value increases to 49.09 nm in the membrane Org-ZnO:TFN-NF-3, which is believed to be due to the uneven distribution of the entrapped clusters of nanomaterials in the polymer matrix of the membrane. The variation in surface topography due to the effect of impregnation of nano-ZnO at progressively higher concentrations is also manifested through the change in R q and R 10z values.
The AFM studies have brought out the fact that surface roughness features such as formation and distribution, as well as variation of peaks and valleys, are highly inuenced by the nature of the medium through which nanomaterials are impregnated into the skin layer of TFN-NF membranes.

Elemental analysis and mapping of TFN-NF membranes by EDX
The skin layer elemental mappings of both Aq-ZnO:TFN-NF and Org-ZnO:TFN-NF class of membranes presented in Fig. 13 and 14, respectively, show a differential distribution of the elements (N, Zn and O) within the bulk of the skin layer, with an increasing amount of nanomaterial impregnation. Elemental analysis of the Aq-ZnO:TFN-NF class of membranes by EDX, presented in Table 3 reveals that when the nano-ZnO concentration is increased from 0.05 to 0.1% and then to 0.2%, the relative nanoparticle density in the skin layer   However, the increase in respective values of the Org-ZnO:TFN-NF class of membranes is much higher. The nano-ZnO density increases from 3.05 AE 0.59 weight% (i.e., 0.82 atomic%) to 9.16 AE 0.20 weight% (i.e., 2.33 atomic%), and then to as high as 16.57 AE 0.29 weight% (i.e., 4.40 atomic%) with the introduction of 0.05, 0.1 and 0.2% of nano-ZnO, respectively. Similarly, the relative atomic ratios of the unique elements Zn and N of these membranes increase from 1 : 34.21 to 1 : 11.69, and nally to 1 : 5.93, indicating incorporation of progressively greater quantities of nano-ZnO in the nanocomposite polyamide skin layer. The determined elemental ratios of the unique elements may not refer only to composition of the skin layer ($500 nm), but also imply the contribution of these elements residing underneath the polyamide layer, i.e., the polysulfone layer ($3 mm).
Impregnations of larger quantities of nano-ZnO in the skin layer of the latter class of membranes have also been observed during respective SEM and AFM analyses. This occurs because during progress of the heterogeneous type interfacial polycondensation reaction and successive formation of the nascent polymer layer, the migrating reactant PEI met the reactant of the organic phase, i.e., TMC (with or without nano-ZnO, as the case may be) in the organic phase ( Fig. 1) close to the boundary between the mutually immiscible solvents. In this situation, when the nanomaterials are present in the organic phase, greater quantities of nano-ZnO are likely in the reaction zone during the process of polycondensation reaction. The unstable dispersion of the nanomaterial in the organic medium may further be responsible for some deposition and subsequent entrapment of the nano-ZnO in the skin layer of the membrane. But, when the nano-ZnO becomes a constituent of the aqueous amine phase, the monomer along with nano-ZnO has to cross the barrier of the nascent polyamide lm and approach the reactants of the organic phase to undergo further polymerization. The transport of nano-ZnO in the reaction zone may not be facilitated because the rate of diffusion across the nascent polymeric barrier for the bulky-sized nano-ZnO associated PEI of the aqueous amine phase becomes comparatively slow. This may be responsible for the lower quantity of nano-ZnO in the skin layer polyamide of the Aq-ZnO:TFN membrane class.

Evaluation of transport characteristics of TFN-NF membranes
The solute rejection behaviors of the Aq-ZnO:TFN-NF class of membranes as well as that of the Control-NF (see Fig. 15a) show This journal is © The Royal Society of Chemistry 2015 that the TFN-NF membranes give much higher rejections for MgCl 2 than for NaCl and Na 2 SO 4 , which follows the general trend MgCl 2 > NaCl $ Na 2 SO 4 . This implies that the membranes are inherently positively charged, which is characteristic of membranes derived using PEI. With greater concentration of nano-ZnO in the membranes' skin layers there is a decline in the differential rejection among these three solutes, more specically between MgCl 2 and Na 2 SO 4 . Their solute rejection ratios of 92. inuences the porous nature of the nanocomposite membranes, may have contributed to an overall decline of SR values for all solutes. In these positively charged membranes, the maximum solute rejection for MgCl 2 is predominantly governed by Donnan exclusion: due to the presence of multivalent cation (Mg 2+ ) with a higher positive charge density, the solute experiences maximum electrostatic repulsion by the charged membranes. 62-64 A lower solute rejection of NaCl results from a weaker Donnan exclusion of the monovalent Na + ions by the membrane. However, the presence of bivalent anions in Na 2 SO 4 facilitates its transport through the membrane, further reducing the respective solute rejection.
A similar order of solute rejection pattern, MgCl 2 > NaCl $ Na 2 SO 4 , for the Org-ZnO:TFN-NF class of membranes, however,  comes with a superior differential rejection ability between MgCl 2 and Na 2 SO 4 , as compared to the Aq-ZnO:TFN-NF class of membranes (Fig. 15b). This signies a higher positive charge of the Org-ZnO:TFN-NF class of membranes, the origin of which has been discussed during characterization of the skin layers of the membranes via FTIR as well as XPS. Consequently, the solute rejection ratios for MgCl 2  Both classes of TFN-NF membranes showed signicant increases in their solvent uxes, with the progressive impregnation of nano-ZnO in making the membranes. Solvent uxes for the representative solute MgCl 2 increase from 520 LMD for Control-NF to 600 LMD for Aq-ZnO:TFN-NF-1, 960 LMD for Aq-ZnO:TFN-NF-2 and 920 LMD for Aq-ZnO:TFN-NF-3 (Fig. 16a). Similarly, the solvent uxes increased from 520 LMD for Control-NF to 560 LMD for Org-ZnO:TFN-NF-1, 620 LMD for Org-ZnO:TFN-NF-2 and nally to 660 LMD for Org-ZnO:TFN-NF-3 (Fig. 16b). The increase of solvent uxes in the TFN class of membranes has been widely reported in the literature: the improvement is attributed to the presence of nanomaterials in the interfacial reaction zone, which inuences the kinetics of the interfacial polycondensation reaction by factors such as change in miscibility of the aqueous and organic phases through released heat of hydration of the nanomaterials, change in the diffusion rate of monomers that consequently alters the cross-linking density of the polymer network, and the hydrophilic/hydrophobic nature of the membrane surfaces. [65][66][67] For the Org-ZnO:TFN-NF class of membranes, the free PEI molecules (since aqueous amine medium of this class is devoid of the nano-ZnO) experience greater mobility and diffusivity to migrate from aqueous to the organic phase, across the nascent polymer layer, which ensures an enhanced rate of polymerization in the reaction zone. Also, the heat of hydration of nano-ZnO, which was present in considerable density in the organic phase, could induce increased local temperature in the reaction zone as the nanoparticles encounter the hydrated PEI molecules. This thermal impact can additionally contribute to improving the miscibility of the two heterogeneous phases locally, and subsequently facilitate the reaction by enhancing the rate of diffusion of the PEI monomers into the organic phase. 65 Such thermal effects may have some inuence in the exothermic type polycondensation reaction to reduce reactivity. Thus, it can be assumed that impregnation of more nano-ZnO  This journal is © The Royal Society of Chemistry 2015