Ultrathin ZSM-5 zeolite nanosheet laminated membrane for high-flux desalination of concentrated brines

See allHide authors and affiliations

Science Advances  23 Nov 2018:
Vol. 4, no. 11, eaau8634
DOI: 10.1126/sciadv.aau8634


The tremendous potential of zeolite membranes for efficient molecular separation via size-exclusion effects is highly desired by the energy and chemical industries, but its practical realization has been hindered by nonselective permeation through intercrystalline spaces and high resistance to intracrystalline diffusion in the conventional zeolite membranes of randomly oriented polycrystalline structures. Here, we report the synthesis of ZSM-5 zeolite nanosheets with very large aspect ratios and nanometer-scale thickness in the preferred straight channel direction. We used these ZSM-5 nanosheets to fabricate ultrathin (<500 nm) laminated membranes on macroporous alumina substrates by a simple dip-coating process and subsequent consolidation via vapor-phase crystallization. This ultrathin b-oriented ZSM-5 membrane has demonstrated extraordinary water flux combined with high salt rejection in pervaporation desalination for brines containing up to 24 weight % of dissolved NaCl. The ZSM-5 nanosheets may also offer opportunities to developing high-performance battery ion separators, catalysts, adsorbents, and thin-film sensors.


The MFI-type zeolite, including its pure silica form (i.e., silicalite) and aluminum-containing structures (i.e., ZSM-5), has a three-dimensional interconnected channel system, with the nearly cylindrical (0.56 nm × 0.53 nm) straight channels running in the b axis and zigzag channels of slightly oval cross section (0.55 nm × 0.51 nm) laying in the a-c plane. These channel dimensions are well suited for many high-value and high-impact molecular separations, proton-selective ion conduction, and brine desalination by the extremely selective size exclusion and steric effects (13). MFI zeolite membranes, with the b axis oriented along the membrane thickness direction, are particularly desirable for reducing mass transport resistance because of their slightly larger opening and shortened diffusion length as compared to the channels in the a-c plane. However, zeolite membranes synthesized by the conventional in situ crystallization or seeded secondary growth methods have limited rooms for thickness reduction because the in-plane growth required for closing up the intercrystalline gaps is accompanied by considerable out-of-plane (thickness) growth when the common structure-directing agent (SDA) tetrapropylammonium ion (TPA+) is used. Meanwhile, the nonselective intercrystalline spaces, which are inevitable in the polycrystalline structures and can be enlarged during SDA removal by calcination, are shortcutting along the membrane thickness (4). The very recent successes in synthesizing MFI zeolite nanosheets of large aspect ratios have enabled the fabrication of zeolite membranes with minimized thickness, reduced intercrystalline boundaries at the surface, and controlled crystal orientation (5, 6). In the literature, high–aspect ratio flat nanosheets of silicalite were synthesized and used as primary building blocks to fabricate b-oriented thin membranes on macroporous alumina substrates. These ultrathin silicalite membranes were either of laminated multilayer structure prepared by filtration-deposition of dispersed nanosheets or in the structure of intergrown monolayer formed by secondary growth of sparsely coated nanosheets using a sacrificial silica layer (79). These ultrathin silicalite membranes have demonstrated extremely high selectivity (approaching theoretical values) and high permeance in separation of isomer vapor mixtures such as i-butane/n-butane and p-xylene/o-xylene. The high performances of these membranes in isomer separations are indicative of minimized bypassing through the nonselective intercrystalline spaces, the realization of size and shape selectivity, and reduced resistance to molecular diffusion in the straight channels along the small thickness. However, silicalite membranes, because of their hydrophobic and nonionic surface, are highly resistive to the transport of water molecules and hydrated protons (H3O+) in aqueous environments; on the contrary, ZSM-5 zeolite membranes with low Si/Al ratios have hydrophilic and ionic surfaces that can markedly enhance the permeability for water and proton (1014). Despite these useful findings and strong demands for high-performance water separation membranes and proton conductors, there has been a lack of success in synthesizing b-oriented ZSM-5 thin membranes because morphological control is rather difficult for polycrystalline zeolite membranes with high aluminum contents.

Here, we report the first synthesis of MFI zeolite nanosheets with ZSM-5 surfaces of low Si/Al ratios and the fabrication of an ultrathin b-oriented ZSM-5 membrane by dip coating and consolidating the laminated layer of nanosheets on a macroporous α-alumina substrate. Zeolite nanosheet laminated membranes were tested for pervaporation (PV) desalination of brines, with concentrations up to 24 weight % (wt %) of total dissolved salt (TDS).

Desalination of high-salinity brines (TDS > 10 wt %) is key to achieving efficient reclamation and “zero liquid discharge” of environmentally damaging waters produced from industrial activities such as oil and gas drilling, power plant desulfurization and cooling, and seawater and in-land brackish water desalination (1519). The interest in high-salinity brine desalination is growing rapidly because of the imminent needs for treatment, beneficial use, and safe handling of formation waters generated by saline basin CO2 geological storage, which often contain >10 wt % of TDS (20). However, existing technologies such as the multieffect flash (MEF) and membrane reverse osmosis (RO) desalination are inefficient for treating concentrated brines (2025). The high TDS lowers water vapor pressure and hence requires increased temperature for MEF operation that causes high-energy consumption and severe equipment corrosion and scaling problems (26). The high TDS also generates extreme osmotic pressures that cannot be handled by existing RO technologies due to membrane compaction and intensified salt contamination (27). Moreover, for membrane desalination with liquid water permeates, such as the RO and forward osmosis processes, the salt rejection usually decreases with increasing feed TDS because ion crossover is driven by the transmembrane concentration gradient and influenced by surface contamination. The porous membrane PV desalination is thermal driven, with vapor permeate that, in principle, can avoid diffusive transport of the nonvolatile salts into the water vapor product. However, the dissolved salts can diffuse through the liquid-filled membrane pores and crystallize after dewatering to block pores and form fine particles to enter the permeate vapor stream. The pore sizes of MFI zeolites are slightly smaller than the kinetic diameter (dk) of hydrated metal ions (dk > 0.6 nm) but much bigger than the dk of water molecules (~0.27 nm) (2). Thus, MFI zeolite membranes, with their exceptional hydrothermal and chemical stabilities, have the potential to obtain perfect salt rejection by ion-sieving effect in PV desalination of high TDS brines. However, conventional polycrystalline zeolite membranes are relatively thick and inherently contain nanometer-scale intercrystalline pores. Consequently, in PV desalination of high-salinity brines, conventional MFI zeolite membranes exhibited low water flux because of the long distance for intracrystalline diffusion of water and limited salt rejection due to ion transport through the intercrystalline passages (11, 28, 29). We hypothesize that the ultrathin ZSM-5 nanosheet laminated membrane can markedly shorten the length for intracrystalline water diffusion and substantially lengthen the intercrystalline ion passages to simultaneously enhance water flux and salt rejection in PV desalination of concentrated brines.


ZSM-5 nanosheet synthesis and characterization

The ZSM-5 zeolite nanosheets were prepared by a multistep procedure modified from the methodology recently developed by Jeon and coworkers (6). The original method involves two major steps that are suitable for synthesizing silicalite nanosheets: First, silicalite nanoparticle seeds of ~30 nm in average size, as shown in fig. S1A, are obtained from a pure silica precursor solution containing tetrapropylammonium hydroxide (TPAOH) as SDA, and second, silicalite nanosheets are formed by secondary growth of the silicalite nanoseeds in a tetraethoxysilane (TEOS)–derived siliceous precursor using bis-1,5(tripropyl ammonium) pentamethylene diiodide (dC5) as SDA. The silicalite nanosheets shown in Fig. 1 (A and D) obtained by this two-step process are morphologically almost identical to those synthesized by the same procedure in (6, 7). The silicalite nanosheets in Fig. 1A formed by 3 days of secondary growth were found to be 1.8 to 2.5 μm in lateral lengths along the a and c axes and ~4.7 nm in thickness along the b axis throughout the flat area surrounding a ~300-nm-diameter and ~40-nm-tall center that evolved from the nanoseed (Fig. 1, B and C). The nanometer thickness of the zeolite sheets was also evidenced by the visibility of the underlying nanosheet in Fig. 1A due to the secondary electrons even under a moderate electron beam energy of 10 keV. When the duration of the secondary growth was extended to 4 days, the silicalite nanosheets grew further to reach lateral dimensions of about 2.2 and 3.5 μm in the a-c plane (Fig. 1D) and a center height of more than 130 nm, but the flat area thickness remained virtually unchanged (Fig. 1, E and F). The flat fringe areas of the large nanosheets could be readily fractured into smaller flakes and detached from the thick core by sonication in liquid media.

Fig. 1 Microscopic characterizations of the silicalite nanosheets.

Scanning electron microscopy (SEM) picture (A), atomic force microscopy (AFM) image (B), and AFM height profiles of silicalite nanosheets formed in 3 days (C). SEM picture (D), AFM image (E), and AFM height profiles of silicalite nanosheets formed in 4 days (F).

The silicalite nanosheets, obtained by ultrasonication, thorough cleaning, and centrifugal separation, were reasonably uniform in size and free of debris from the center core according to SEM observations in Fig. 2A. However, when the precursor for secondary growth of the nanoseeds had a small amount of aluminum source (AlO2), e.g., with a starting composition of 80 TEOS:1.6 NaAlO2:3.75 dC5:20 KOH:9500 H2O (i.e., Si/Al ratio = 50), the zeolite products from the 4-day seeded secondary growth under the same conditions contained crystals mostly in the common coffin shape with only a few nanosheets, as can be seen in Fig. 2B. To obtain MFI nanosheets with ZSM-5 surfaces, we carried out the 4-day secondary growth of the nanoseeds in a specially designed reaction vessel schematically depicted in fig. S1B, in which a scheduled change of precursor composition was made during the hydrothermal process. The uninterrupted secondary growth process included 3.5 days of reaction with the pure silica precursor, followed by 0.5 days of reaction in the precursor with the addition of aluminum source (i.e., NaAlO2 solution). The last 0.5 days of secondary growth in precursors of different NaAlO2 contents have led to zeolite products shown in Fig. 2 (C to F), which are samples of flat flakes obtained through the ultrasonication cleaning and centrifugal separation. Zeolite crystals with large flat areas of uniform thickness were successfully achieved when the NaAlO2-added precursors for the last 0.5 days had overall Si/Al ratios of 100, 50, and 25. Figure 2 (C to E) shows the flat nanosheets separated from the crystals obtained from precursors with overall Si/Al ratios of 100, 50, and 25, respectively. However, when the overall Si/Al ratio in the precursor was lowered to 10, the product in Fig. 2F had a large amount of particulate crystals grown on the nanosheet surfaces. This result can be explained by the fact that, after 3.5 days of hydrothermal reaction, silica content in the liquid phase was mostly consumed for growth of the nanosheets. Formation of new crystals in the remaining dilute solution became difficult when the amounts of NaAlO2 added for a Si/Al ratio of ≥25 were not enough to sufficiently increase the reactant concentration. The significantly more NaAlO2 added for a Si/Al ratio of 10 apparently made the aluminosilicate reactant concentration high enough for inducing nucleation and growth of new crystals. The energy dispersive x-ray spectroscopy (EDS) examination revealed that the overall Si/Al ratios of the zeolite nanosheets that resulted from precursors with overall Si/Al ratios of 100, 50, and 25 were 74 ± 12, 44 ± 2, and 30 ± 2, respectively; meanwhile, the aluminum content in the silicalite nanosheets was undetectable by EDS measurements, as expected (details are provided in fig. S2).

Fig. 2 Zeolite products obtained by secondary growth of the silicalite nanoseeds under different conditions.

(A) Four days in Al-free precursor, (B) 4 days in Al-containing precursor, and last 0.5 days in precursor with Si/Al ratios of (C) 100, (D) 50, (E) 25, and (F) 10.

The ZSM-5 nanosheets obtained from the precursor with an overall Si/Al ratio of 25 (i.e., those shown in Fig. 2E) were used for membrane fabrication. The surface of these particular zeolite nanosheets is expected to have high hydrophilicity that resulted from the large amounts of framework aluminum and extraframework ion compensators (Na+) (30). The nanosheets from the precursors with Si/Al ratios of 50 and 25 exhibited microscopically smooth surfaces with almost no nanometer-scale holes, which were observed in large numbers in the edge areas of the nanosheets obtained from precursors of higher Si/Al ratios (100 and ∞; Fig. 2, A and C). The morphological differences between the nanosheets of various Si/Al ratios in Fig. 2 suggest changes of growth behavior by the Al contents in precursor. The nanosheets from the precursor with a Si/Al ratio of 25 were further characterized by x-ray diffraction (XRD), SEM, AFM, and high-resolution transmission electron microscopy (TEM) techniques. The XRD patterns in Fig. 3A verified that the 30-nm-diameter spherical silicalite nanoparticles and the randomly packed silicalite and ZSM-5 nanosheets were of pure MFI-type crystal structure. The films of multilayer zeolite nanosheets coated on the glass surfaces, for instance, the silicalite and ZSM-5 coatings shown in Fig. 2 (A and E), were proven to be oriented with out-of-plane (i.e., thickness) direction in the b axis according to the sole appearance of [020] peak in the XRD spectra. Compared to the silicalite nanosheets in Fig. 1D, the ZSM-5 nanosheets grown in the solution with a Si/Al ratio of 25 had obviously less lateral growth, as shown by a typical single sheet in Fig. 3B. According to the AFM measurements in Fig. 3 (C and D), the central core of the ZSM-5 nanosheet was more than 130 nm in height, which was similar to that of the silicalite nanosheet in Fig. 1F; however, the flat area of the ZSM-5 nanosheet had a uniform thickness of ~6.7 nm (Fig. 3D), which was apparently thicker than the flat areas of silicalite nanosheets (~4.7 nm). The microscopic observations and EDS elemental analysis results suggest that the zeolite nanosheet growth was less preferentially directed along the a and c axes when the incorporation of [AlO2] into the framework created highly energetic sites in the crystal surface to enable growth in all facets with no substantial difference in kinetic rates. However, the specific mechanisms, thermodynamics, and kinetics of zeolite crystallization are much complicated in the templated solutions and currently remain as challenging research subjects in the field. It is generally understood that siliceous zeolite crystallization occurs by interconnection of [≡Si–OH] tetrahedron corner terminals in precursor with those in developing crystal surfaces. The thermodynamics of the oxolation-type reaction is determined by bond dissociation energies (BDEs) of Si–O–H and Al–O–H in reactants and those in products H2O, [≡Si–O–Si≡], and [≡Si–O–Al(Na+)≡]. The BDE is greater and the pKa (where Ka is the acid dissociation constant) is lower for Si–OH than for Al–OH, which could make the formation of [≡Si–O–Al(Na+)≡] more favorable than that of [≡Si–O–Si≡] when OH (in high pH) is used as a mineralizer (31). Thus, the addition of aluminum source into the pure silica precursor inevitably affects the zeolite growth mode.

Fig. 3 Characterizations of the ZSM-5 nanosheets from precursor with an overall Si/Al ratio of 25.

(A) XRD patterns of the randomly packed ZSM-5 nanosheets and nanosheet layered film on glass in comparison with the patterns of the MFI powder standard, the silicalite nanoseeds, the randomly packed silicalite nanosheets, and silicalite nanosheet layer deposited on glass. (B) SEM image of a typical ZSM-5 nanosheet, (C) AFM tapping mode image of the ZSM-5 nanosheets on silicon wafer, and (D) AFM height profiling for a complete ZSM-5 nanosheet (lines 1 and 2) and a separated piece of the flat area (lines 3 and 4).

The ZSM-5 nanosheets were examined by high-resolution TEM using samples deposited on the TEM grids and glass substrate, respectively. The sample films were sparsely coated by drying a thin layer of suspension containing 0.05 wt % of nanosheets. The top views of thin films on the grids in Fig. 4 (A and C) show areas of both nanosheet monolayer and nanosheet overlaid layer. The high-magnification image in Fig. 4B and the electron diffraction patterns in Fig. 4C reveal the crystal structure of [010] (b axis) orientation along the thickness direction. The cross-sectional image in Fig. 4D for a glass-supported sample prepared by the focused ion beam (FIB) technique shows side dimensions of the zeolite nanosheets in the range of 10 to 20 nm, which are significantly greater than the thicknesses measured by AFM likely because of nanosheet bending under stresses in the TEM sample. The TEM observation of crystalline structure in the cross sections of zeolite nanosheets was unsuccessful because of the lattice structure damages caused by the high-energy beams of ions and electrons during FIB preparation and TEM examination, respectively.

Fig. 4 TEM images of the ZSM-5 nanosheets.

(A and B) Top views showing crystal structure perpendicular to the surface in the a-c plane, (C) electron diffraction patterns (insets) in single sheet and overlaid areas, and (D) cross section of a film with overlaid nanosheets on glass surface.

Membrane fabrication and characterization

The silicalite nanosheets shown in Fig. 2A and ZSM-5 nanosheets shown in Fig. 2E were dispersed in deionized (DI) water and peptized by HNO3 to form stable translucent sols containing 0.2 wt % of nanosheets and 0.1 wt % of the drying control agent (DCA) hydroxyl propyl cellulose (HPC). The zeolite membranes were fabricated on the homemade macroporous α-alumina discs by the conventional dip-coating technique, followed by drying and firing at 500°C for 6 hours in air. The silicalite and ZSM-5 zeolite nanosheets were found to be thermally stable at up to 1000°C, which is similar to their counterpart crystals of regular shapes (fig. S3A). Thus, attempts were made to consolidate and densify the supported membranes of laminated zeolite nanosheets by firing at a high temperature of 900°C for 8 hours. The final zeolite membranes were first evaluated for their structural integrity by permeation of an ambient pressure H2/CO2 equimolar gas mixture at room temperature and high temperature, respectively, which is one of the effective means for MFI zeolite membrane quality assessment (32). The membranes fired at 900°C exhibited room temperature H2/CO2 separation factor (αH2/CO2) of >1.5 with CO2 permeance (Pm,CO2) of ~1.5 × 10−6 mol m−2 s−1 Pa−1, which was close to that found on the bare substrate (Pm,CO2 ≈ 2.1 × 10−6 mol m−2 s−1 Pa−1). In addition, the zeolite layer was found to swell and peel off after being immersed in water for less than an hour (fig. S3B). MFI zeolite membranes with minimized intercrystalline spaces are expected to exhibit αH2/CO2 of <1 at room temperature because CO2 is preferentially adsorbed into the zeolitic pores to block the nonadsorbing H2 from entering; on the contrary, membranes containing an appreciable amount of intercrystalline pores show αH2/CO2 of >1 (up to Knudsen factor of 4.7) because gas permeation through the nanometer-scale nonzeolitic pores is dominated by the gaseous diffusion mechanism, with permeability far greater than molecular diffusion in the intracrystalline pores (32). It is obvious that the high-temperature calcination was ineffective for densifying and consolidating the supported membranes of laminated zeolite nanosheets. Therefore, in this work, the internanosheet porosity in the laminated membrane layer, after firing at 500°C, was loaded with a dilute precursor solution of 3.33 SiO2:0.033 NaAlO2:0.35 NaOH:1.0 TPAOH:176 H2O by contacting its surface with the solution for 10 s. The same solution with the exclusion of NaAlO2 was used for treating the laminated silicalite membrane. This precursor-loaded membrane disc was subsequently treated in the vapor phase of 1 M TPAOH solution at 190°C for 6 hours under autogenous pressure. This particular precursor composition was experimentally identified to avoid either excessive or insufficient crystal formation in the laminated zeolite membrane (figs. S4 and S5). Crystallization of the precursor loaded in the membrane layer not only reduced the internanosheet spaces but also strongly interlinked the nanosheets and bonded the entire zeolite membrane layer to the substrate, as schematically illustrated in Fig. 5A. The XRD patterns of the resultant membranes in Fig. 5B indicated that the precursor was converted into a very small amount of zeolite crystals in random orientations, which were, however, hardly seen by SEM examinations of the membrane surfaces and cross sections in Fig. 5 (C to F). The membranes after consolidation by vapor-phase crystallization were found to remain in excellent b orientation, with average thicknesses of about 315 nm (±65 nm) and 410 nm (±90 nm) for the silicalite and ZSM-5 nanosheet laminated membranes, respectively.

Fig. 5 Structures and EDS elemental mapping of the zeolite nanosheet laminated membranes on the macroporous α-alumina supports.

(A) Schematic showing consolidation of the multilayer nanosheet lamina by vapor-phase crystallization. (B) XRD patterns of the dip-coated and vapor phase–consolidated zeolite membranes on alumina supports. (C and D) Cross-sectional SEM images of the silicalite and ZSM-5 membranes. (E and F) Surface SEM images of the vapor phase–treated silicalite and ZSM-5 membranes. (G) Si/Al mapping across the silicalite membrane. (H) Si/Al mapping across the ZSM-5 membrane.

When the vapor-phase treatment was performed using precursors of sufficient concentrations, both the silicalite and ZSM-5 nanosheet laminated membranes were able to achieve room temperature αH2/CO2 of ~0.75 while maintaining high CO2 permeance of ~1.1 × 10−6 mol m−2 s−1 Pa−1 for the equimolar H2/CO2 gas mixture. This evidenced the effective reduction of the nanometer-scale internanosheet spaces and the predominance of gas transport by the adsorption-diffusion mechanism through the intracrystalline pores. The zeolite membrane layers showed no signs of swelling or peeling even after contacting with high-salinity brines for several weeks. The results of EDS elemental mapping for the cross sections of alumina-supported zeolite membranes show virtually no aluminum in the silicalite membrane layer (Fig. 5G) and a notable amount of aluminum distributed throughout the ZSM-5 membrane layer (Fig. 5H).

PV desalination test

The consolidated silicalite and ZSM-5 thin membranes were then tested for PV desalination of brines, which contained up to 24 wt % of dissolved NaCl. The experimental setup for PV desalination measurements is schematically shown in fig. S6. The PV operating temperature was 80°C, which is considered a desirable temperature for PV desalination by zeolite membranes (28). For both the silicalite and ZSM-5 thin membranes, the effects of size exclusion of the hydrated metal ions [Na+·(H2O)nh] at the 0.56-nm zeolitic pore openings enabled high salt rejection rate (rs = 1 − Cs,p/Cs,f) of >99.5% even for feed with salt concentrations up to 24 wt % because the hydration number “nh” of metal ions and hence the sizes of the hydrated metal ions, i.e., Na+·(H2O)nh (dk > 0.6 nm) in solutions, are essentially independent of salt concentration and temperature (3336). The membranes achieved unprecedented high water flux (FH2O) in combination with high rs for the extremely concentrated brines, as presented in Fig. 6A (23, 24, 37, 38). When N2 sweeping was used in the permeate side, the ZSM-5 membrane exhibited FH2O of 6.28 and 3.69 liters m−2 hour−1 for the 3 and 24 wt % TDS solutions, respectively, with rs of >99.8%. The extraordinary water flux was a result of the very small thicknesses (<500 nm) and the orientation of b-axis straight channels along the small thickness. The FH2O values of the ZSM-5 membrane were >35% higher than those of the silicalite membrane, although the former was slightly thicker than the latter. The greater FH2O on the ZSM-5 membrane may be attributed to its ionic surface, which is more hydrophilic (water contact angle, ~0°) than the nonionic silicalite surface (water contact angle, ~12°). The surface properties of the MFI zeolite strongly influence water adsorptivity (i.e., solubility CH2O) and diffusivity (DH2O), which determine water permeability (Pb,H2O = CH2O · DH2O). The CH2O values could differ by orders of magnitude between the defect-free silicalite surface (CH2O,silic), silicalite surface containing defects of silanol nests (CH2O,silanol), and ZSM-5 surface with extraframework ions [AlO2]·M+ (M+ = Na+, K+ …) (CH2O,ion) in the order of CH2O,ion >> CH2O,silanol >> CH2O,silicalite (30). Furthermore, the water molecular diffusivity in MFI zeolites does not follow a simple Arrhenius dependence over a broad range of temperature and exhibits a behavior change from liquid-like to vapor-like at around 80°C (39). On the other hand, the DH2O was found to be greater in nonionic silicalite than in ZSM-5 zeolites because of the stronger interactions between water molecules and extraframework metal ions (Na+) (40). The recent work of Fasano et al. (41) also revealed that external surface defects, such as molecular species attached to framework terminals at the zeolitic pore entrances, could be another critical factor profoundly affecting the overall molecular transport diffusivity. The influences of these factors on the PV desalination performance of MFI zeolite membranes are complex, and further improvement of water flux may be made through effective control of surface defects and optimizations of the zeolite chemistry and PV operating conditions.

Fig. 6 Results of PV desalination as a function of NaCl concentration in feed brines.

(A) Water flux and salt rejection, (B) water permeance, (C) salt transfer and solid deposition in the conventional ZSM-5 membrane (ZSM5-C), and (D) prevention of salt transfer in the ZSM-5 nanosheet laminated membrane. SILM-S, silicalite membrane under N2 sweeping; ZSM5-S, ZSM-5 membrane under N2 sweeping; ZSM5-V, ZSM5 membrane under vacuum on permeate side; ZSM5-C, ZSM-5 membrane made by in situ crystallization.

When the membrane PV was conducted by vacuuming the permeate side for product removal, the laminated ZSM-5 membrane was able to obtain FH2O values of 10.4 and 6.4 liters m−2 hour−1 for the 3 and 24 wt % of feed solutions, respectively, with rs of >99.5%. These FH2O values represent more than 66% increases from those measured under N2 sweeping in the permeate side. Such an increase of FH2O by vacuum operation was likely due to the elimination of resistance from N2 counter diffusion to water vapor transport in the small pores of the thick substrate. However, the vacuum operation requires much higher standard in eliminating microdefects in both the membrane layer and membrane module seals because bubble formation and eruption of the liquid solution could happen in the defect channels when the permeate side total pressure (Pperm) falls far below the equilibrium vapor pressure (Embedded Image) of the feed brine (as illustrated in fig. S7). This may accelerate the solution penetration through defects to cause unusually high FH2O but markedly lowered rs by particle and droplet entrainment especially when dealing with high-concentration feeds.

The conventional silicalite and ZSM-5 membranes made by the seeded secondary growth method have been used for PV desalination in the literature. These membranes experienced serious loss of ion rejection for feed brines with >3 wt % of TDS, which was attributed to salt transfer through the intercrystalline pores to precipitate on the permeate side (11, 28). The conventional Al-rich ZSM-5 membrane was able to reach FH2O of 11.5 liters m−2 hour−1 by PV at 75°C, which was three to four times that of a similarly prepared silicalite membrane, but the salt (NaCl) rejection rs dropped markedly to <60% for a feed concentration of 7.5 wt % (28). The poor performance in PV desalination of high-salinity brines suggests the existence of large amount of intercrystalline spaces in the membrane, as also evidenced by a N2 permeance of >10−6 mol m−2 s−1 Pa−1, which is unusually high for its kind. In the present work, a ZSM-5 zeolite membrane (Si/Al ratio ~ 50) made by the conventional in situ crystallization method was also tested for comparison. This ZSM-5 membrane was about 2 μm thick (shown by the cross-sectional SEM image in fig. S8A) and deemed of reasonably good quality based on its room temperature αH2/CO2 of ~0.54 and Fm,CO2 of ~1.0 × 10−7 mol m−2 s−1 Pa−1. Under the same PV conditions using N2 sweeping, this ZSM-5 membrane (denoted as ZSM5-C for data in Fig. 6A) had rs of 99.8% for the 3.0 wt % TDS feed but with a FH2O of 1.22 liters m−2 hour−1, which is less than one-fifth that of the laminated ZSM-5 membrane. The performance of the ZSM5-C membrane for low TDS solution was comparable to literature reports on similar membranes of good quality (11). However, for the 24 wt % TDS brine, FH2O of the ZSM5-C membrane diminished quickly to unmeasurable level in just 6 hours, suggesting complete block of membrane pores by precipitated salt. Examinations of this membrane using an optical microscope and SEM-EDS techniques found large quantities of salt deposits blocking the substrate pores near the zeolite membrane layer (fig. S8B) and appearing on the external surface (fig. S9A) after PV operation for the brine with 24 wt % of TDS. Figure 6C illustrates the proposed salt transfer through the liquid-filled intercrystalline pores to deposit on the permeate side. This conventional ZSM-5 membrane was unable to operate by vacuuming the permeate side because the accelerated salt deposition and pore blocking made it difficult to collect a reliable sample in 10 hours.

The large aspect ratios of zeolite nanosheets in the laminated membranes significantly reduced the intercrystalline (i.e., internanosheet) boundary openings in the surface and considerably lengthened internanosheet pathways to minimize the penetration of liquid brine to the permeate side surface. As illustrated in Fig. 6D, in contrast to the conventional ZSM-5 membrane, although liquid brine could also enter the nanometer-scale internanosheet spaces, the water flow rate (Fw,in) of penetrating liquid (brine) in these small spaces may be outmatched by the flow rate of water (Fw,out) diffusing through the covering 6.7-nm-thick nanosheet. Molecular dynamics simulation has predicted that the siliceous zeolite nanosheets of similar pore sizes could completely reject salts and provide water permeance of orders of magnitude higher than the regular membranes (42). In addition, the precipitation of salt at the surfaces of the internanosheet entrances may be facilitated by the ZSM-5 ionic surface under fast dewatering rate. The salt attachment and crystallization could be inhibited on nonionic or nonpolarizable and hydrophobic surfaces, such as smooth carbon nanofibers, which have been recently used to construct nanoporous (dp ~ 30 nm) membranes for high-flux membrane distillation at <5 wt % of TDS (43). As a result, the incipient salt precipitation is likely to occur near the entrances of internanosheet spaces at the membrane surface, which effectively prevents liquid solution from penetrating deeper into the laminated layer. The ZSM-5 nanosheet laminated membrane, after a month of operation for measuring the data presented in Fig. 6A, was stored in DI water for more than 2 months and retested for PV of pure water and the brine with 24 wt % of TDS at 80°C using N2 sweeping. The results are provided in fig. S10, which demonstrated excellent reproducibility of FH2O and rs as compared to the initially measured values in Fig. 6A. Furthermore, the high FH2O (~3.5 liters m−2 hour−1) and rs (~99.9%) showed no appreciable decline over 60 hours of continued PV desalination for the feed with 24 wt % of TDS (fig. S10). The permeate side membrane surface was carefully scanned with an optical microscope over the entire area, and no salt deposits were found on the permeate side surface (fig. S9B). Thus, the nanosheet laminated membranes were able to maintain high levels of rs for PV desalination of concentrated brines even under vacuum pressure operation. The very high FH2O, however, appeared to cause concentration polarization, which reduced the actual water vapor pressure at the feed side membrane surface (Embedded Image). Therefore, the apparent water permeance Pm,H2O(= FH2Op) in Fig. 6B shows a declining trend with increasing feed TDS, because the driving force Δp(Embedded Image is the permeate side vapor pressure) is overestimated when the vapor pressure calculated from the bulk concentration Embedded Image is used instead of Embedded Image. In practical applications, higher feed flow velocity and baffles may be used to reduce concentration polarization and further enhance the water flux.


In summary, ZSM-5 zeolite nanosheets with ~1 μm of average lateral dimensions in the a-c plane and ~6.7 nm of thickness in the b axis have been synthesized for the first time by a continuous seeded secondary growth method, in which a scheduled change of precursor composition was made during the process. The ZSM-5 nanosheets of very large aspect ratios (lateral length-to-thickness ratio, >100) were used to fabricate ultrathin (<500 nm) laminated membranes on macroporous alumina substrates by the simple dip-coating method. This zeolite nanosheet laminated membrane was densified and consolidated by vapor-phase crystallization of a minimal amount of dilute precursor preloaded in the internanosheet spaces. The consolidated thin membrane, with the b-axis straight channels oriented in the thickness direction, has demonstrated unprecedented combination of high water flux and salt rejection rates in PV desalination of brines containing up to 24 wt % of dissolved NaCl. Both the silicalite and ZSM-5 nanosheet laminated membranes showed no structural and performance degradations after PV operation at 80°C for over 1 month. The nanosheet laminated architecture of the ultrathin membrane may offer practical solutions to the two major issues associated with conventional polycrystalline zeolite membranes, which include the inherent intercrystalline spaces along the thickness that undermine the separation selectivity and the long intracrystalline pathways in randomly oriented thick membranes for molecular diffusion that severely limit the membrane flux. Computational work in the literature indicated that large rooms exist for water flux improvement by optimizations of membrane surface properties and reduction of thickness of the nanosheet laminated membrane, which is achievable by improving the membrane synthesis process (6, 7, 41, 42). The ZSM-5 nanosheets of various Si/Al ratios as well as the methodology for synthesizing these nanosheets may also offer new opportunities to developing high-performance battery ion separators, catalysts, adsorbents, and thin-film sensors.


Zeolite nanosheet syntheses

The ZSM-5 nanosheet synthesis procedure included two steps of hydrothermal crystallization processes that were modified from the methodology reported by Jeon et al. (6). The first step was to generate silicalite nanoparticle seeds from a starting precursor with molar composition of 10 SiO2:2.4 TPAOH:0.87 NaOH:114 H2O. The hydrothermal reaction was conducted under static condition at 50°C for 6 days. The resultant liquid phase was recovered by filtration and subsequently underwent further hydrothermal reaction at 140°C for 2 days to form silicalite nanoparticles of ~30 nm in average size, which were retrieved by centrifugal separation. The second step was to form MFI nanosheets of very large aspect ratios by secondary growth of the silicalite nanoseeds in a precursor sol containing homemade dC5 as SDA. The precursor had an overall molar composition of 80 TEOS:3.75 dC5:20 KOH:9500 H2O. The dC5 was synthesized by exhaustive alkylation of 1,5-diaminopentane with 1-iodopropane and subsequently purified by the multistep solvent extraction and washing procedure. The protocol used was the same as that detailed in the literature (6, 44). The silicalite nanosheets were obtained by a 4-day hydrothermal reaction in the Al-free precursor that was essentially the same as that reported in (6). For synthesis of the MFI zeolite nanosheets with ZSM-5 surfaces, a controlled amount of 1 M NaAlO2 solution was added into the reaction mixture just 0.5 days before the end of the 4-day treatment in the second step of seeded secondary growth. The as-synthesized nanosheets were separated by centrifugation and intensive ultrasonication for 2 hours to obtain nanosheets of uniform thickness and ~1 μm of average lateral lengths.

Zeolite membrane fabrication

The recovered silicalite and ZSM-5 nanosheets were redispersed in DI water by alternated stirring and sonication and peptized by HNO3 at pH of 3 to 3.5 for 2 hours. The suspension was then added with HPC solution under rigorous stirring to form a stable translucent sol, which had 0.2 wt % of zeolite nanosheets and 0.1 wt % of HPC. The HPC was used as binder or DCA for preventing peeling of the multilayer nanosheet coating during drying and calcination. The zeolite membrane layer was dip coated on a homemade α-alumina disc, which was 1.2 mm thick and 2.5 cm in diameter with average pore size and porosity of ~100 nm and 33 ± 3%, respectively. The surface of the disc for coating the membrane was finely polished with #800 sandpaper. Dip coating of the zeolite film was accomplished by bringing the suspension surface to the horizontally held disc surface and contacting for 5 s. The dip-coated membrane layer was dried in an oven at 40°C for overnight and then calcined at 500°C in air for 6 hours to remove the SDA and strengthen the film consisting of multiple layers of zeolite nanosheets. The dip-coating and calcination processes were repeated once to ensure the elimination of pinholes or uncovered spots in the zeolite film. After firing at 500°C, the zeolite membrane layer was loaded with a dilute zeolite synthesis solution by dipping the zeolite-coated surface into the solution for 10 s. The precursor-loaded membrane disc was subsequently treated in the vapor of 1 M TPAOH solution at 190°C for 6 hours under autogenous pressure. The vapor phase–treated membrane was dried and calcined in air at 500°C for 6 hours. The heating and cooling rates used in the calcination were 0.5°C/min. A ZSM-5 membrane was also synthesized on the porous alumina disc by the conventional in situ crystallization method following a previously reported procedure (32). The precursor had a molar composition of 0.055 SiO2:0.0058 NaOH:0.017 TPAOH:0.92 H2O, and hydrothermal reaction was performed at 180°C for 4.5 hours under autogenous pressure. The resultant MFI zeolite membrane was washed, dried, and then calcined in air at 500°C for 8 hours for SDA removal. During the hydrothermal treatment, alumina dissolved slightly into the highly alkaline solution. Thus, both the zeolite crystals formed on the substrate and those formed in the liquid phase contained framework aluminum, which was experimentally quantified.

Microscopic characterizations

SEM and EDS examinations were performed using a FEI Scios DualBeam microscope equipped with AMETEK Octane Super EDAX. The cross-sectional sample for TEM test was cut from the nanosheet film deposited on a glass chip with FIB using a FEI Helios NanoLab 660 DualBeam microscope. TEM examinations were performed using a JEOL 2010F field-emission electron microscope. To measure the thicknesses of zeolite nanosheets, the nanosheets were sparsely deposited on silicon wafers and analyzed using a Veeco Dimension 3100 AFM using height imaging/profiling under tapping mode. The water contact angles were measured for continuous zeolite films coated on 165-μm-thick glass chips using a Sigma 700 force tensiometer.

XRD examination

The zeolite particles, films, and supported membranes were examined by XRD using the PANalytical X’Pert Pro diffractometer with Cu Kα radiation (λ = 1.5406 Å). The standard patterns of MFI-type zeolite powders were taken from (45).

Membrane PV measurement

The membrane PV desalination measurements were carried out with an apparatus schematically shown in fig. S6. The disc membrane was mounted in a Teflon permeation cell using silicon rubber O-ring seals. The active membrane area after excluding the O-ring impression was 2.54 cm2. The tubing and connectors used were all made of Teflon or silicon rubber in the entire system except for the inlet sweep gas line and preheating coil, which used 1/8-inch stainless steel tubes. During the PV operation, the feed brine was flowed over the zeolite membrane surface at a velocity of 0.6 cm/s (at a volumetric flow rate of 63 cm3/min) and the substrate surface (permeate side) was swept by N2 gas at a velocity of 1.9 cm/s (at 200 standard temperature and pressure cm3/min) under atmospheric pressure. The tube and membrane cell outside of the water bath were well insulated, and the actual flow temperatures were verified using thermocouples inserted into the tubes at the inlets and outlets of the membrane cell. When operating with vacuum on the permeate side, the sweep gas inlet was blocked by switching the three-way valve and the permeate chamber pressure was maintained at around 0.01 bar. The vacuum pump was connected to the exit of the cold trap. The sodium ion concentration of the permeate water was determined with a PerkinElmer AAnalyst 300 atomic absorption spectrophotometer.


Supplementary material for this article is available at

Fig. S1. Silicalite nanoseeds and the nanosheet synthesis vessel.

Fig. S2. EDS elemental analyses for the zeolite nanosheets.

Fig. S3. XRD and SEM tests for thermally treated silicalite membrane.

Fig. S4. SEM and XRD tests for vapor phase–treated silicalite membranes.

Fig. S5. SEM and XRD tests for vapor phase–treated ZSM-5 membranes.

Fig. S6. The membrane PV desalination apparatus.

Fig. S7. Salt transfer in conventional ZSM-5 membrane under vacuum.

Fig. S8. SEM and EDS analyses of conventional ZSM-5 membrane after PV.

Fig. S9. Substrate surface of ZSM-5 membranes after PV.

Fig. S10. Stability of the ZSM-5 nanosheet laminated membrane.

Note S1. Chemicals and materials used.

Note S2. Zeolite nanosheet sample preparation for EDS tests.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank A. Angelopoulos for helping with the contact angle measurements and G. Sorial for providing the atomic absorption spectroscopic analyses. Funding: This research was financially supported, in part, by the U.S. Department of Energy through NETL (grant no. DE-FE0026435) and the Ohio Development Service Agency (grant no. OOECDO-D-17-13). Author contributions: The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. Z.C. led the experiments on zeolite nanosheet synthesis, membrane fabrication, characterizations, and PV desalination. S.Z. worked on membrane synthesis and PV desalination apparatus establishment and measurements. Z.X. and X.G. were involved in membrane preparation, microscopic characterizations, and membrane cell design and construction. A.A. prepared the substrates and contributed to the membrane evaluations by gas permeation. J.D. conceived the concept of the ZSM-5 nanosheet membrane and directed the entire research effort. Z.C. and J.D. did the main part of the data analysis and interpretation. S.Y. developed the method and reactor for synthesizing MFI crystals with the ZSM-5 surface. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from J.D. (Junhang.dong{at}

Stay Connected to Science Advances

Navigate This Article