Asymmetric polyamide nanofilms with highly ordered nanovoids for water purification

Tailor-made structure and morphology are critical to the highly permeable and selective polyamide membranes used for water purification. Here we report an asymmetric polyamide nanofilm having a two-layer structure, in which the lower is a spherical polyamide dendrimer porous layer, and the upper is a polyamide dense layer with highly ordered nanovoids structure. The dendrimer porous layer was covalently assembled in situ on the surface of the polysulfone (PSF) support by a diazotization-coupling reaction, and then the asymmetric polyamide nanofilm with highly ordered hollow nanostrips structure was formed by interfacial polymerization (IP) thereon. Tuning the number of the spherical dendrimer porous layers and IP time enabled control of the nanostrips morphology in the polyamide nanofilm. The asymmetric polyamide membrane exhibits a water flux of 3.7−4.3 times that of the traditional monolayer polyamide membrane, showing an improved divalent salt rejection rate (more than 99%), which thus surpasses the upper bound line of the permeability−selectivity performance of the existing various structural polyamide membranes. We estimate that this work might inspire the preparation of highly permeable and selective reverse osmosis (RO), organic solvent nanofiltration (OSNF) and pervaporation (PV) membranes.


Brunauer Emmett Teller (BET)
The mesoporosity and specific surface area of the G4D dendrimer and PSF support were identified by the pore volume as determined by the Brunauer Emmett Teller (BET) method (ASAP2020 specific surface area and pore analyzer), and the pore size distribution plot of the dendrimer powder was calculated by the quenched solid DFT using the adsorption branch.

Scanning electron microscopy observation (SEM)
Thin nanofilm composite membranes are consisted of three layers, an ultrathin polyamide top layer, a PSF support, and a nonwoven fabric. For accurately observing the cross-sectional morphology of the polyamide nanofilm, the nonwoven fabric was first peeled off by using adhesive tape. Then, the remaining PSF support with polyamide layer was soaked in DMF until the polyamide became fully transparent, indicating that the PSF support material was no longer present, and then washed with methanol. For the cross-sectional morphology, the polyamide layers without PSF support were deposited onto the coverslips by a floating method and fractured in liquid nitrogen for scanning electron microscopy observation (SEM, Hitachi SU8010). The samples were coated with gold before SEM analysis.

Transmission electron microscope (TEM)
The G4D dendrimer and G4D dendrimer nanoparticles formed by diazotization coupling reaction were deposited onto the copper meshes and investigated by transmission electron microscope (TEM, JEM-1200EX, JEOL). For the internal morphology of the traditional polyamide nanofilm and asymmetric polyamide nanofilm, the polyamide layers without PSF support were deposited onto the copper meshes and investigated by transmission electron microscope (TEM, JEM-1200EX, JEOL).

Atomic force microscope (AFM)
A scanning probe microscope (SPM-9700, SHIMADZU) was used to measure the surface morphology and the roughness of the polyamide membranes.

Attenuated total reflectance fourier transform infrared spectroscopy (ATR-FTIR) and X-ray photoelectron spectroscopy (XPS)
The chemical composition and structure of the PSF, PSF-G4D-1, PSF-G4D-2, PSF-G4D-3, and the fabricated polyamide membranes (including the traditional polyamide membrane and asymmetric polyamide membrane) were characterized by attenuated total reflectance fourier transform infrared spectroscopy (ATR-FTIR) and X-ray photoelectron spectroscopy (XPS), respectively. The chemical composition and elemental data obtained from XPS were analyzed and fitted using CasaXPS software.

UV-visible spectroscopy
UV-visible diffuse reflectance spectroscopy (UV-vis DRS) was conducted to character the chemical composition variation of the PSF, PSF-G4D-1, PSF-G4D-2 and PSF-G4D-3 support membranes. UV-vis spectra were conducted to monitor the diazotization coupling reaction in the solution.

Zeta potential
Surface charges of the traditional and asymmetric polyamide membrane were determined with an Anton Paar SurPass solid surface analyzer.

Contact angle
Contact angle was measured under room temperature (25°C) using a drop shape Analyzer-DSA30 (KRÜSS, Germany) in the sessile drop mode for characterizing the surface hydrophilicity and surface energy of the membrane surface.

Synthesis steps of 3,5-bis (N-trifluoro acetamido) benzoic acid
Trifluoroacetic anhydride (78.6 mmol) was added to a 30 mL THF solution containing 3,5-diaminobenzoic acid (24 mmol) at 0°C under nitrogen, and stirred at that temperature for 15 min. Subsequently, the system was stirred for 3 h at 25°C oil bath.
Then, water (30 mL) was added and continued to be stirred for 6h at that temperature.
The resulted mixture was extracted with the ethyl acetate to obtain the organic layer and aqueous layer. The organic layer was washed with water 3-4 times, and dried with anhydrous magnesium sulfate overnight. Afterwards, the resulted filtrate was evaporated to give a purple powdery solid, and recrystallized from acetonitrile, filtrated and gave purple solid particles. The product was dried at 120°C for 12 h to give pale purple solid particles (6.1 g, a yield of 79%).

Synthesis steps of 3,5-bis (trifluoro acetamido) benzoyl chloride
A 150 mL of thionyl chloride solution containing 3,5-bis (N-trifluoro acetamido) benzoic acid (46.6 mmol) was refluxed for 6 h at 120°C. The thionyl chloride is distilled off and the residue was dissolved in tetrachloroethane at 100°C, Subsequently, cooled to room temperature to precipitate a purple solid powder. The solid was washed with nhexane three times to give a brown powder. Afterwards, the power was recrystallized with dichloromethane, dried at 60°C to give slightly yellow power (12.1 g, a yield of 72%).

Synthesis steps of G1 dendrimer (G1D)
3,5-bis (N-trifluoro acetamido) benzoic chloride (22 mmol) was added to a 10 mL of NMP solution containing P-phenylenediamine (10 mmol), stirred for 15 min at 0°C, and subsequently stirred for 1 h at 25°C oil bath. The water (50 µL) was added to the system and reacted for 1.5 h at 50°C. Then, hydrazine hydrate (120 mmol) was added dropwise and continued to be stirred for 1.5 h at that temperature. The reaction solution was poured into a 100 mL of 2 wt% NaHCO3 solution, stirred for 30 min, filtered under suction, and dried to give light gray solid of G1D (7.4 g, yield 98%).

Synthesis steps of G2 dendrimer (G2D)
3,5-bis (N-trifluoro acetamido) benzoic chloride (22 mmol) was added to a 20 mL of NMP solution containing G1 (5.0 mmol), stirred for 15 min at 0°C, and subsequently stirred for 1 h at 25°C oil bath. The water (50 µL) was added to the system and reacted for 1.5 h at 50°C. Then, hydrazine hydrate (120 mmol) was added dropwise and continued to be stirred for 1.5 h at that temperature. The reaction solution was poured into a 100 mL of 2 wt% NaHCO3 solution, stirred for 30 min, filtered under suction, and dried to give light gray solid of G2D (4.4 g, yield 97%).

Synthesis steps of G3 dendrimer (G3D)
3,5-bis (N-trifluoro acetamido) benzoic chloride (22 mmol) was added to a 27 mL of NMP solution containing G2 (2.5 mmol), stirred for 15 min at 0°C, and subsequently stirred for 1 h at 25°C oil bath. The water (50 µL) was added to the system and reacted for 1.5 h at 50°C. Then, hydrazine hydrate (120 mmol) was added dropwise and continued to be stirred for 3.5 h at that temperature. The reaction solution was poured into a 133 mL of 2 wt% NaHCO3 solution, stirred for 30 min, filtered and dried to give light gray solid of G3D (4.7 g, yield 94%).

Synthesis steps of G4 dendrimer (G4D)
3,5-bis (N-trifluoro acetamido) benzoic chloride (22 mmol) was added to a 27 mL of NMP solution containing G3 (1.3 mmol), stirred for 15 min at 0°C, and subsequently stirred for 1 h at 25°C oil bath. The water (50 µL) was added to the system and reacted for 1.5 h at 50°C. Then, hydrazine hydrate (120 mmol) was added dropwise and continued to be stirred for 3.5 h at that temperature. The reaction solution was poured into a 133 mL of 2 wt% NaHCO3 solution, stirred for 30 min, filtered and dried to give light gray solid of G4D (4.8 g, yield 92%). Figure 1. Synthesis steps of dendrimer G4D.

Fabrication of PVDF and PI supports.
P84 casting solution was prepared by dissolving 18 wt% of P84, 4 wt% of PVP K60 and 6 wt% of PEG 600 in DMF under stirring at 70°C for 12 h, and then degassed for a further 12 h at 45°C to eliminate any air bubble trapped in the solutions. After cooled to room temperature, the dope solutions were cast on non-woven by using a casting knife with a fixed thickness of 110 μm. Then the fresh scraped film solution was allowed to parallel immersion into a precipitation water bath at room temperature. The PVDF casting solution with 20 wt% of PVDF, 4 wt% of PVP K60 and 6 wt% of PEG 600 was dissolved in DMF under stirring at 70°C for 12 h, and then conducted the similar process to fabricate the PVDF support. All fabricated supports were washed with distilled water for 3 h to remove any residual solvent for further use.

Preparation of asymmetric polyamide RO membrane.
We used m-Phenylenediamine (MPD) to synthesize the asymmetric polyamide RO membrane. Specifically, the PSF and PSF-G4D-1 substrates were firstly immersed into the amine solution (2 w/v% MPD, 1.1 w/v% TEA and 2.3 w/v% CSA) for 4 min, and subsequently the extra amine solution on the support surface was blown off with an air knife. Then, the above support substrate was contacted with the TMC/n-octane solution (0.12 w/v%) for 40s to form the traditional or asymmetric polyamide nanofilm, and the organic phase solution on the surface was instantly blown off with an air knife. The resulted polyamide membranes were finally dried at 60°C for 2-5 min and stored in DI water until use.

Membrane performance test
For the polyamide RO membranes, desalination performance of the prepared polyamide membrane was determined with different salt solutions in a cross-flow system with an effective test area (A) of 19.3 cm 2 . The NaCl concentration in the feed solution was 2 g L -1 . The desalination performance tests were conducted at an osmotic pressure of 1.55 MPa and a temperature of 25°C. The performance data were determined after the water flux and the conductivity reached a steady state.
The water flux (kg m -2 h -1 ) was calculated from the weight of the permeate (M) for a specified time, as given by the following equation: The salt rejection was determined from the conductivity of the feed solution (Cf) and the permeate (Cp). Hence, the salt/ion rejection can be calculated from the following equation:

Density measurement
Densities of polyamide dendrimer porous layer, traditional polyamide nanofilm and asymmetric polyamide nanofilm were measured and calculated by the ellipsometry (J. A. Woollam Co., Lincoln, NE) and QCM (Q-Sense, Explorer, Biolin Scientific). We For 3,5-bis (trifluoro acetamido) benzoic acid, the C-F absorption peak appeared at 1164 cm -1 , the N-H absorption peak is at 3304 cm -1 . For 3,5-diaminobenzoic acid, the N-H absorption peaks are located at 3350 cm -1 and 3432 cm -1 . Changes in these characteristic peaks indicate that the amino protection reaction has been completed, that is, 3,5-bis (trifluoro acetamido) benzoic acid with a protecting group CF3 was formed. is indeed connected to the 3,5-bis (trifluoro acetamido) benzoic acid chloride. And after deprotection reaction, the number of terminal amine groups of the first generation has been doubled. It is also worth noting that there is no characteristic absorption peak of the C-F group in the infrared spectrum of the G2D products, which also indicates that the deprotection reaction is carried out thoroughly. is indeed connected to the 3,5-bis (trifluoro acetamido) benzoic acid chloride. And after deprotection reaction, the number of terminal amine groups of the G2D has been doubled. It is also worth noting that there is no characteristic absorption peak of the C-F group in the infrared spectrum of the G3D products, which also indicates that the deprotection reaction is carried out thoroughly.
Supplementary Figure 7 shows the 1 H NMR chart of 3,5-bis (N-trifluoro acetamido) benzoic acid. Among them, the peak at 8.14 ppm is the proton peak adjacent to the carboxyl group on the benzene ring, the peak at 8.42 ppm is the proton peak at the para position of the carboxyl group on the benzene ring, and the peak at 11.58 ppm is the proton peak of trifluoro acetamido. We have drawn the chemical route to present the mechanism that how the dendrimers were anchored covalently on PSF support. The specific mechanism is summarized in Supplementary Figure 19. Generally, the anchoring covalently dendrimer on PSF support process are 5 steps, and the following is the details on mechanism 2 ：  The cross-sectional morphology of the pristine PI support and the PI-G4D support have no significant variation.  The atomic composition assessed by XPS measurements shows that due to the formation of dendrimer porous layer, the N content of the PSF-G4D-1 support is higher than that of the pristine PSF support. In addition, based on its lower O/N ratio, the crosslinking degree of the asymmetric PA nanofilm (63.2%) is higher than that of the traditional PA nanofilm (32.6%), which is conducive to enhance the rejection of salt.  As shown in Supplementary Figure 42, the isoelectric point (IEP) of the asymmetric polyamide nanofilms is lower than that of the traditional polyamide nanofilm, of which the former is 3.81, and the latter is 3.88. The surface zeta potential value of the asymmetric nanofilm is thus less negative than that of the traditional nanofilm in the range of pH 3.8 to 10. For example, the related zeta potential value at pH=7 is -55.9

Supplementary
mV for traditional nanofilm and -42.7 mV for the asymmetric nanofilm, respectively.
From the XPS results, the lower −COOH content (enhanced crosslinking degree) in the semi-aromatic polyamide nanofilms is mainly responsible for the decreased surface negative charge. Generally, charge exclusion and size exclusion contribute significantly to the separation performance of the fabricated IP polyamide membranes 5,6 . For the original semi-aromatic polyamide membranes with surface negative charge, sulfate rejection was always greater than that of chlorinated salt. This high sulfate rejection can be explained by the high negative surface charge of the original polyamide membranes.
In the case of anions, SO4 2with a valence of -2 will experience greater electronegative repulsion from the negatively charged membrane surface as opposed to Cl -. In addition, SO4 2has larger hydrated radius and lower ionic diffusivity than Cl -(Supplementary   is looser and larger than that of the single polyamide nanofilm (T-TMC-MPD). The magnified TEM image in Supplementary Figure 47d demonstrates that the voids in the ridge is larger, and have no over-stacking, which produces a better water transport path.
In addition, SEM images in Supplementary Figure 47e and 47f clearly show that the cross-sectional morphology and thickness of the resulted polyamide nanofilms. The single polyamide dense layer is ~100 nm, while that of the asymmetric polyamide nanofilm is ~120 nm. A further observation reveals that the asymmetric polyamide nanofilm has two layers, the upper is the dense layer with a thickness of ~60 nm, and the bottom layer is the dendrimer porous layer with an average thickness of ~60 nm.
Overall, the resulted asymmetric polyamide nanofilm has no over-stacking, showing a larger ridge-and-valley structure and thinner dense active layer, which can decrease the mass transfer resistance for water and improve solute separation efficiency.    As shown in Supplementary Figures 50 and 51 We determined the Cl -/SO4 2selectivity of the asymmetric polyamide membrane in case of mixed salt solution. As shown in Supplementary