Introduction

Freshwater scarcity has become a major concern due to population growth and industrialization over the past decades, with over 2 billion people living in water-stressed countries1. This is expected to escalate further in the upcoming years owing to climate changes and environmental pollution2. Added to this, stringent regulations regarding water quality standards have also compelled us to seek novel ways to tackle the issue of pure water scarcity3,4. This ever-increasing demand can be effectively balanced through seawater desalination and wastewater treatment. Among various technologies available for this, membrane-based technologies such as nanofiltration (NF) have been increasingly popular due to their lower energy requirement and compactness5,6. NF membranes have been used to remove contaminates such as textile dyes or as a pretreatment to reverse osmosis (RO). To date, researchers have explored several polymeric, ceramic, and other novel 2D materials for NF membrane fabrication7,8. Polymeric membranes stand out among various membranes due to their facile fabrication, flexibility, and dense structure, allowing molecular-level separation9. Numerous modifications in combination with inorganic materials have been proposed to enhance the properties of polymer-based membranes.

TFC membranes with a selective dense polyamide layer have become a center of attraction for lab-scale and commercial-scale membrane systems wherein a thin selective layer is fabricated on top of a porous support10. Contrary to the symmetric thicker membranes, TFC membranes are characterized by a thinner dense selective layer contributing to a relatively lower transport resistance, thereby achieving improved permeance. More importantly, each selective layer and support could be independently optimized to achieve desired performance. PA membranes with TFC structure are usually fabricated through interfacial polymerization (IP) reaction, which is spontaneous and challenging to control. This in-situ rapid reaction enables the formation of a highly cross-linked polymer layer with superior separation performance. However, as commercial NF membranes can only offer permeance up to 10 Lm−2h−1bar−1, tremendous efforts have been made in recent years to advance the performance of TFC membranes. Consequently, many researchers have explored the possibility of tuning the NF performance of such interfacially polymerized membranes by altering the support, intermediate, or even the selective layer through chemical and physical modifications11,12,13,14.

The most straightforward strategy to enhance the permeability of polyamide TFC membranes is to fabricate an ultrathin selective layer, as the transport resistance offered by the support layer is often negligible15,16. For instance, Junyong Zhu et al. fabricated polyamide-based membranes with a selective layer ranging from 49 nm down to 12 nm simply by varying the monomer concentrations15. Different from the conventional IP process, support-free polyamide synthesis was carried out where nanofilm formed at the solvent interface was subsequently transferred onto a chemically modified porous support, thereby achieving a fast permeance of 25.1 Lm-2h-1bar-1. Although controlling the PA layer thickness is an effective method to lower the transport resistance to achieve unprecedented permeance, utilization of such an ultrathin defect-free selective layer in a practical context is highly challenging.

Sufficient attempts have been made to develop composite membranes by incorporating nanomaterials in the selective layer to overcome the permeability-selectivity trade-off exhibited by PA TFC membranes. Fillers added in the aqueous or organic phase before the polymerization reaction imparts various functionalities in the selective layer, including improved hydrophilicity17,18,19, antifouling properties18,20, chlorine resistance19 and/or improved surface charge17,19 consequently achieving enhanced performance. However, weak compatibility of the inorganic fillers in the highly cross-linked PA layer often results in sacrificing the selectivity despite improved permeance being achieved21,22. Alternatively, modifying the support layer was also demonstrated to be an effective strategy by incorporating nanomaterial within the support layer without disturbing the selective layer23,24,25,26,27. Nevertheless, such modifications can only offer limited enhancement in the performance as the permeability and selectivity of TFC membranes is mostly governed by the characteristics of the selective layer.

Apart from the above-mentioned methods, enlarging the filtration surface area by tuning the morphology of the selective layer was also demonstrated to be an effective strategy to boost the performance of polyamide membranes. Inspired by the exciting results observed by Zhe Tan et al. on morphologically tuned polyamide nanofiltration membranes, several researchers explored the possibilities of introducing wrinkles in the selective layer to advance the nanofiltration performance28,29,30,31,32,33,34. Manipulating the interlayer is demonstrated to be an effective strategy for tuning the morphology of polyamide layers wherein the IP reaction is carried out on a templated support layer. While some researchers proposed subsequent removal of such interfacial templates/fillers to form crumpled nanofilm exhibiting excellent water permeance28,29, reproducibility of such lab-scale results at a commercial level is extremely challenging due to the requirement of an additional etching step. Alternatively, numerous studies proposed the fabrication of polyamide membranes without the elution of nanofillers, still maintaining wrinkles on the selective layer, promoting faster water permeability30,31,32,33,34. For example, Xuerui Gao and co-workers reported the fabrication of highly wrinkled NF membranes with enhanced permeability by nanoparticle-templated IP reaction31. The authors demonstrated the variations in the morphology of the wrinkles with respect to the varying shape and size of the nanoparticle incorporated. In a different study, crumpled polyamide membranes fabricated by templating porous TiO2 interlayer via vacuum filtration and subsequent polymerization was presented by Mingzhu Chi et al. 34. At optimized loading, water permeability peaked at 12.8 Lm-2h-1bar-1 which is 1.21 times higher than the neat TFC membranes. Although the above-mentioned studies clearly highlight the role of templating strategy in creating wrinkles on the selective layer to achieve enhanced performance, further attention should be given to (1) the type of fillers used for membrane fabrication; (2) the methodology of incorporation of fillers35. Firstly, templates used should be preferably porous to promote faster water transport which may otherwise lead to the blocking of substrate pores34,35. Secondly, mere physical incorporation of interlayer, as in most cases, may result in instabilities and subsequent peeling off the layer due to weak adhesion between the PA layer and the support36. Therefore, anchoring novel porous materials through appropriate bonding on the support layer can be effectively used for developing morphologically tuned TFCs with superior performance.

Owing to the presence of unique intercrystallite pores and hydrophilic nature, Zeolite-Y with faujasite (FAU) structure and high Si/Al ratio has been utilized in several studies to fabricate membrane by itself or as a filler to form polymer composite membranes37,38,39,40,41,42. While particles with smaller dimensions are favored for a defect-free templated interfacial polymerization31, direct synthesis of Y-type zeolites with a high Si/Al ratio in smaller dimensions down to a few hundred nanometers is challenging due to unfavorable hydrothermal route43. To overcome this barrier, a facile ball milling approach was demonstrated previously in our group to produce smaller zeolites from commercial micro-zeolites using carbon nanostructure as a damping material41,43. This unique fabrication produces zeolites with high crystallinity and enhanced surface area. On the other hand, to address the issue of weak adhesion of the interlayer, suitable modifications should be adopted to enhance the stability in long-term operation, which is a major barrier for real-life applications. Dopamine, that possesses bio-inspired self-adhesiveness, could be a suitable candidate to build better binding of the interlayer to promote the stability of the TFC membrane14. In this study, we report a facile approach for the fabrication of wrinkled PA membranes on polydopamine (PDA) coated zeolite anchored on Polysulfone (PSf) support (Fig. 1). The variations in surface morphology and corresponding changes in membrane performance with zeolite loading have been carefully investigated. The novelty of this study lies in two aspects: (1) the unique method for size reduction of zeolites as explained previously; (2) facile membrane pretreatment to achieve tuned surface morphology which is different from most of the previously reported studies that rely on vacuum filtration. Therefore, the objective of this study is to develop a simple and efficient strategy for constructing robust wrinkled polyamide membranes, which may be potentially extended to any other inorganic templates.

Fig. 1: Membrane fabrication.
figure 1

a Illustration of membrane fabrication steps, b Interfacial polymerization reaction scheme.

Full size image

Results and discussions

Size reduction and PDA functionalization of zeolite-Y

Zeolite particles in the smaller dimension obtained after ball milling were subsequently characterized prior to their use for membrane fabrication. Firstly, morphology and particle size were examined from the SEM images presented in Fig. 2a–c. As can be seen from the images, particles exhibited a random morphology due to the vigorous ball milling; which is consistent with previously reported studies that employed a similar procedure for size reduction41,43. Notably, most particles of nZ were observed to be less than 500 nm with peaking around 200 nm contrary to the micron-sized pZ (Supplementary Fig. 2). X-ray diffraction, FTIR spectrum, pore size distribution, and nitrogen adsorption results are presented in Supplementary Fig. 3. XRD patterns of nZ were identical to that of pZ where peaks corresponding to the crystal planes of Y-type zeolites (111), (311), (331), (440), and (533) were preserved ensuring no crystal changes during size reduction. It is critical to investigate any changes in the porous structure of the zeolites due to the harsh grinding conditions and calcination steps. Therefore, nitrogen adsorption experiments were conducted. Adsorption isotherms and pore size distribution revealed matching trends with previous reports indicating that the pores were preserved during the process44,45. These results not only confirm the chemical and thermal stability of zeolites, but also the protection of the microstructural characteristics during templated ball milling and subsequent thermal treatment. Furthermore, FTIR analysis was acquired for nZ sample (Supplementary Fig. 3) and compared with pZ. It is worth noting that nZ exhibited identical peaks to that of the pZ, confirming that no chemical changes occurred during the CNS templated ball milling and subsequent calcination. FTIR spectrum of zeolites in both cases demonstrates typical zeolite-Y characteristic peaks at 527, 618, and 833 cm−1 along with the common zeolite peaks (1060 and 1210 cm−1)46,47.

Fig. 2: Characterization of zeolites.
figure 2

SEM image of a parent micro zeolite, zeolite after ball milling, b pristine, and c PDA functionalized. Scale bar 2 µm (d) Comparison of FTIR sepctrum of nZ and PDA-nZ. Photographs of aqueous solution of nZ and PDA-nZ is provided in the inset.

Full size image

PDA functionalization has been carried out to attain improved adhesion between the zeolite particles and the membrane support. Successful functionalization was evident visibly from the color of the aqueous solution. The solution color changed from light yellow to dark brown after functionalization with PDA (inset of Fig. 2d). This obvious color change is due to the self-polymerization of dopamine molecules to PDA covering the surface of zeolite particles. The SEM images after PDA functionalization exhibited similar particle sizes, although some aggregations could be also observed within the sample, as seen in Fig. 2c. The successful adhesion of PDA onto the zeolites was also confirmed through FTIR investigations presented in Fig. 2d. While zeolite peaks were dominated in the FTIR spectrum of PDA-nZ, the presence of minor peaks at 1270, 1420, and 1620 cm−1 could be associated with the vibrations from C=O, N-H, and C-N groups confirming PDA modification14,48. Additionally, the broad peak observed in PDA-nZ around 3420 cm−1, possibly due to the O-H and N-H groups, further confirms the wrapping of PDA49.

Membrane characterization

To investigate the surface chemistry of the as-prepared membranes, FTIR spectra were captured, which is shown in Fig. 3. To confirm the formation of the PA layer, a comparison with pristine PSf support is presented. All obtained spectra had obvious peaks from the PSf support that included peaks at 1490 and 1585 cm−1 (aromatic rings), 1240 cm−1 (ether bond), and 2968 cm−1 (methyl groups)50. TFC membranes displayed additional peaks evidencing the successful formation PA layer. For instance, peaks around 1442 and 1660 cm−1 corresponds to the stretching vibration of C=O of the amide groups51. Different from pristine PSf, broad peaks around 3410 cm−1 was also seen, which could be attributed to the residual hydroxyl groups in the TMC. However, no significant difference was observed between the spectra obtained for different PA membranes; possibly due to the smaller loading of zeolites and the masking effect from the stronger signals of the PA layer and PSf membrane support. Nevertheless, this indicates that the addition of the zeolite layer had no significant impact on the chemical structure of the PA layer.

Fig. 3: FTIR spectra.
figure 3

Comparison of FTIR spectra of all synthesized TFC membranes and PSf support for a range of a 4000 cm−1 to 100 cm−1 and b 2000 cm−1 to 1000 cm−1.

Full size image

Surface SEM of all supports prior to the fabrication of the PA selective layer has been presented in Supplementary Fig. 4 to examine the distribution of PDA-nZ on the membrane support. PSf support layer appeared to be highly porous with pores around 100 nm. Moreover, significant aggregation of zeolite particles was observed, particularly exposing membranes at a higher concentration that would probably impact the overall membrane performance. PDA modification of zeolites facilitated proper anchoring of particles onto the membrane support while pristine particles failed to remain on the surface after gentle rinsing with water, as evident from the SEM images provided in Supplementary Fig. 5. Further, to analyze the impact of zeolites on the membrane support, SEM surface images of all the TFC membranes were captured and are presented in Fig. 4. After the IP reaction, the membrane surface appeared to be denser, indicating the successful formation of PA layer over the highly porous PSf layer. Such a defect-free dense layer formed by the cross-linking reaction of PIP and TMC is favorable for achieving efficient separation performance. Besides the typical nodular structure exhibited by the PA layer, no specific morphology was observed on the dense layer for M-0 (Fig. 4a). However, the presence of zeolites altered the morphology of the PA layer significantly, especially for the case M-100 and M-150. For the case of M-50, although a noticeable amount of zeolite was observed to be distributed on the surface, its impact on the PA layer was less significant. While smaller wrinkles started to appear surrounding the zeolites on the membrane surface for M-50, at higher loadings, wrinkles started growing, covering larger areas of the membrane. For instance, M-100 and M-150 (Fig. 4c, d) exhibited highly crumpled morphology in the PA layer as the wrinkles extended to the nearby zeolite clusters/particles. The absence of larger clusters of zeolites within the proximity has led to minor wrinkles that are limited around the zeolites for the case of M-50, as visible in the SEM image.

Fig. 4: Membrane characterization.
figure 4

SEM and AFM characterization of surface morphology of all TFC membranes. Surface SEM images of a M-0, b M-50, c M-100, and d M-150 with scale bar 500 nm. Three-dimensional AFM images were obtained for e M-0, f M-50, g M-100, and h M-150. Cross-sectional SEM images of i M-0, j M-50, k M-100, and l M-150 with scale bar 2 µm.

Full size image

As observed in the above SEM images, the morphology of membranes with zeolite interlayer comprises a combination of hilly structures and macro wrinkles beside the typical smoother nodular structure observed in PA membranes. To identify whether these variations are actually produced due to the underlying zeolites, EDS mapping was performed. As shown in Supplementary Fig. 6, the Si-enriched area obtained by EDS analysis matched well with the hilly area observed on the membrane surface. This confirms that the elevated areas and, subsequently, macro wrinkles on the membrane surface have been formed due to the presence of zeolites in the interlayer. The morphology of the PA layer is greatly affected by the distribution of amine monomer and its diffusion towards the organic phase resulting in an IP reaction. While the pristine PSf supports led to a smooth PA layer, the crumpled morphology observed in the presence of zeolites can be attributed to two factors: (i) rugged reaction interface and (ii) interfacial instabilities28,31,33,35. Firstly, IP reaction occurs instantly at the aqueous/organic interface, due to which the morphology of the formed PA layer often mimics the initial water layer containing the amine monomer31. Consequently, the presence of hydrophilic zeolites as the interlayer renders a rough reaction interface yielding a crumpled morphology. Secondly, the hydrophilic properties of zeolites not only favor the enrichment of amine monomers within its vicinity but also restricts the diffusion of PIP molecules into the organic phase. As a result, the uneven distribution and controlled diffusion provided diffusion-driven instabilities causing a crumpled morphology28.

Furthermore, the effect of zeolite interlayer on the surface roughness and area has been examined through AFM analysis which is summarized in Fig. 5a. As expected, the root mean square roughness (Rq) and average roughness (Ra) of M-0 were the lowest due to its relatively smoother surface morphology in comparison to the remaining. Moreover, the incorporation of zeolite interlayer promoted the surface roughness proportionally, where the highest Rq and Ra were observed for M-150, which is consistent with the SEM images. A rougher surface is favorable for achieving a higher effective surface area which is considered an effective strategy to enhance the membrane permeance. To investigate this, actual surface areas of all membranes at different zeolite loading were calculated using AFM software. In this analysis, the projected area of each sample was kept the same, i.e., 100 µm2 (10 µm × 10 µm). Accordingly, the percentage difference in the surface area observed (though AFM analysis) with that of the projected sample area has been compared for each case in Fig. 5b. While pristine M-0 expressed a negligible change in the area, M-100 displayed the highest value achieving 21.2% enhancement in the effective surface area; evidencing the potential of wrinkles in improving the surface area. However, the surface area exhibited by M-150 was noticeably lower than M-100 despite higher zeolite content in the interlayer, which may be due to the aggregation of zeolites. As evident from SEM images, a larger number of wrinkles were present for M-100 elevating its effective surface area available for filtration. On the other hand, a further increase in the presence of zeolites has caused severe aggregation, significantly reducing the number of wrinkles in M-150 that eventually sacrificed the effective surface area. Furthermore, cross-sectional images revealed decreased selective layer thickness for modified supports, which could be attributed to the hydrophilicity of the PDA-modified zeolites. More PIP molecules get entrapped in the vicinity of hydrophilic particles limiting the diffusion of amine monomers towards the organic phase, causing a thinner layer52.

Fig. 5: Membrane characterization.
figure 5

a Surface roughness, b change in surface area (expressed as the percentage difference in calculated area and projected area), c contact angle, and d zeta potential of all the TFC membrane prepared. Error bars represent the standard deviation of the acquired data.

Full size image

The hydrophilicity of the membrane surface is of great concern as it directly influences the water permeance of NF membranes. To investigate this, the water contact angle (WCA) for all the membranes along with PSf support was compared, which is displayed in Fig. 5c. As expected, all PA membranes exhibited significantly lower contact angles with respect to the PSf membrane support, which was obvious due to the formation of amide bonds via interfacial polymerization. Further, the WCA of zeolite-incorporated TFCs were lower than pristine PA membranes; which can be attributed to the presence of hydrophilic PDA functionalized zeolites in the composite membranes. While pristine PA membrane recorded an average WCA of 53.5°, M-50, M-100, and M-150 exhibited a lower contact angle of 40.3°, 37.33°, and 43.33° respectively. The WCA decreased significantly after the formation of PA layer by around 21% compared to PSf support due to the contributions from polyamide functional groups. The incorporation of PDA-nZ has further brought down the WCA significantly, where the lowest contact angle was observed for M-100. Comparing the values obtained for M-0 and M-50, it is worth noting that, WCA drastically reduced further upon incorporation PDA-nZ in the TFC membrane. Although, it is widely acknowledged that the incorporation of hydrophilic materials in the PA matrix can effectively enhance the hydrophilicity, further increase in PDA-nZ only provided a slight enhancement in the hydrophilicity possibly due to the effect from morphological changes and the limited influence of underlying zeolite particles on the surface properties. Moreover, the WCA for M-150 was considerably higher than M-100, possibly due to the roughness arising from the severe aggregation and uneven distribution of particles, as revealed in the SEM images33.

Apart from this, the surface charge of the membrane, which has a significant impact on the membrane selectivity, has been characterized by measuring the surface zeta potential provided in Fig. 5d. All measurements were taken at neutral pH at which the membrane testing has been carried out. Irrespective of the amount of zeolites, all of the fabricated membranes displayed negative zeta potential ranging from −21 to −38 mV. This can be ascribed not only to the carboxyl groups formed by the hydrolysis of unreacted acyl groups but also due to the ionization of functional groups on the polyamide layer. The addition of hydrophilic PDA-nZ can potentially have two opposing effects concerning the surface charge and, subsequently, on the zeta potential values. First, a higher amount of zeolites can contribute to elevated surface charge due to the charge contribution from the zeolites themselves. On the other hand, improved hydrophilicity by the addition of PDA-nZ can absorb more PIP molecules onto the surface, leading to better polymerization reaction and subsequently lower residual TMC; reducing the charge contribution from hydrolysis. However, obtained values indicate that charge contribution from zeolite particles had a predominant effect such that, the zeta potential values appeared to be more negative with the introduction of zeolites which might be beneficial for rejecting negatively charged solutes. There was a gradual increase in the absolute zeta potential values from M-0 to M-100, while M-150 exhibited a slightly lower value than M-100, possibly due to the aggregation of zeolite particles at higher loading that led to a decrease in effective surface area.

Membrane performance

The nanofiltration performances of all TFC membranes were evaluated based on the pure water permeance and their rejection towards various inorganic salts operating at a transmembrane pressure of 6 bar. Firstly, to identify the best membrane in terms of the permeance and selectivity offered, we compared the pure water permeance (PWP) and the rejection towards MgCl2 for each case which is shown in Fig. 6a. While pristine TFC membrane offered a permeance of 7.1 ± 0.46 Lm−2 h−1 bar−1, an increase in pure water permeance to 14.4 ± 1.5 Lm−2 h−1 bar−1, 22.44 ± 2.09 Lm−2 h−1 bar−1, and 17.7 ± 1.31 Lm−2 h−1 bar−1was observed for M-50, M-100, and M-150 respectively. It is evident that the zeolite incorporation promoted the permeation of water across the membrane, which can be attributed to the variations in morphology and surface characteristics induced by the zeolite interlayer. However, larger aggregation of zeolites in the interlayer led to the blocking of substrate pores that eventually impacted the overall permeance hence, M-150 offered a lower permeance relative to M-100. Furthermore, it is observed that the rejection increased progressively and reached above 90% for M-100, and thereafter a drastic drop to 75.33% for M-150, possibly due to the large aggregates that resulted in defects in the selective layer. From the obtained results, it is evident that M-100 had the best performance in terms of permeance and MgCl2 selectivity, which is consistent with characterization results presented before.

Fig. 6: Membrane Performance.
figure 6

a Pure water permeance and rejection obtained using 1000 ppm MgCl2 for different TFC membranes, b PEG rejection performance, c Salt solution permeance and rejection towards NaCl, MgCl2, MgSO4, and Na2SO4 using M-100, d Long-term performance of M-100 and e Comparison of our membrane performance with recently published PIP/TMC based TFC membranes tested using 1000 ppm Na2SO4 feed solution12,14,22,23,28,34,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79. Error bars represent the standard deviation of the acquired data.

Full size image

To evaluate the molecular weight cut-off (MWCO) and further elucidate the mechanism of separation, a series of filtration experiments were carried out using PEG solutions. Figure 6b represents rejection exhibited by different membranes towards PEG molecules with molecular weights ranging from 200–2000 Da. The MWCO were 268, 258, 217, and 440 Da for M-0, M-50, M-100, and M-150 respectively. M-150 displayed the highest MWCO of 440 Da, which could be attributed to a defective PA layer due to the presence of zeolite aggregation. Alongside, there is a noticeable reduction in the MWCO for M-100 compared to pristine M-0. An increase in the surface charge may not have any significant contribution to the rejection of PEG molecules due to their neutral charge. Therefore, improved cross-linking degree and the possibility of additional size exclusion offered by the underlying zeolites would have contributed positively to achieving a lower MWCO for M-100.

Having identified the optimum membrane fabrication parameters, a series of filtration experiments were carried out to further evaluate the efficiency and mechanism of separation of M-100 toward aqueous solutions of NaCl, MgSO4, and Na2SO4 which are presented in Fig. 6c. In addition, Fig. 6c demonstrates the permeance achieved for different salt solutions where a significant reduction in the flux was observed compared to the pure water transport across the membrane. For example, the M-100 membrane displayed a decrease of around 8–28%, where a larger decline was observed for divalent salts, which is consistent with the literature53,54. Matching with typical PA TFC membranes, our membranes exhibited significantly higher rejection towards divalent salts. For instance, while the rejection of NaCl was just 48.6%, rejection above 90% was observed for MgCl2, MgSO4, and Na2SO4. Furthermore, a comparison of membrane performance fabricated with parent micron zeolites (M-m100) at the optimized condition has been also investigated, which is added in Fig. 6a. As evident from the result, the permeance and rejection of M-m100 were considerably lower than M-100. In particular, M-m100 offered permeance and rejection of 15.33 ± 1.8 Lm−2 h−1 bar−1 and 77.33%, respectively. This significant reduction in flux compared to M-100 can be attributed to the pore blocking due to larger zeolites while low rejection can be explained by the potential defects mimicking the results obtained for aggregated zeolites in M-150.

In addition to superior separation performance, the stability of the membranes to maintain their performance when operated over an extended duration is a key parameter considering its practical application. To investigate this, the long-term performance of M-100 was monitored using an MgCl2 solution over a period shown in Fig. 6d. Except for the initial fluctuation due to compaction, the membrane could successfully sustain its permeance and rejection throughout the experiment highlighting its capability for long-term operation. This stable long-term result not only demonstrates the compatibility of zeolite interlayer but also validates the fabrication procedure consisting of PDA modification to improve the structural stability of the fabricated membranes. Lastly, Fig. 6e presents a comparison of the results obtained in this work with commercially available NF membranes (NF270 and NF90) and some of the recently published PIP-based PA TFC membranes tested using 1000 ppm Na2SO4 feed solution. Our membrane exhibits superior performance considering both permeance and rejection.

Role of zeolite interlayer on the membrane performance

The above-presented results clearly highlight the influence of the zeolite interlayer on the characteristics of the PA layer that eventually influenced the overall performance of the composite membranes. Membrane permeance is not only influenced by the properties of the selective layer but also by the resistance of the support layer. PSf supports used in this study remain unchanged in all cases, leaving behind the variations in the PA layer as the main contributor to the improved membrane permeance. Incorporation of zeolites that led to a tuned morphology and surface characteristics rendered enhanced permeance with respect to pristine TFC membranes. For instance, membrane permeance tripled for the case of M-100 in comparison to M-0, clearly emphasizing the role of the zeolite interlayer. This improved performance can be attributed not only to the enhanced hydrophilicity but also to the highly wrinkled membrane surface. Compared with the smooth pristine membranes, such a crumpled morphology can positively contribute towards the water permeation due to the enhanced surface area and introduction of interfacial voids11. Firstly, it is widely acknowledged that higher effective surface area is often accompanied by an enhancement in the permeance, although this is not the only deciding factor in the flux enhancement. For example, regardless of the thickness of the selective layer, crumpled morphology can enhance the water permeance theoretically by a factor of 1.57–235. Secondly, the presence of interfacial voids in the wrinkles and tent-like structure formed by the presence of zeolite interlayer can provide additional optimized water transport pathways, creating shortcuts to the permeate side11. Water transport across conventional TFC membranes experiences significant constraints from the funnel effect caused by the hindrance from the substrate35,55. Consequently, water molecules have to travel a longer distance than the thickness of the PA layer, as illustrated in Fig. 7a, b; experiencing considerably higher transport resistance and deviating further from the ideal permeance of free-standing PA layers. The presence of macro-wrinkles and tent-like structures over multiple substrate pores can potentially shorten the transport resistance to a great extent35,56,57. Cross-sectional SEM image (Fig. 7c) at a lower magnification than presented earlier clearly revealed the presence of several such tent-like structures on the selective layer for M-100. Different from some of the previously reported crumpled structures achieved after etching away the templates, our composite membrane still contained zeolite particles that would reduce the availability of interfacial voids. However, the porous nature of the zeolites would limit this effect to a great extent, as observed in ref. 34, and more importantly, zeolite hindrance was negatively influenced only when the loading was high enough to cause severe aggregation, as observed in M-150. Furthermore, a non-homogeneity of the PA layer was observed at higher magnification, as shown in Fig. 7d, where smoother PA film was observed at zeolite-rich regions contrary to the larger nodular structure. A clear distinction of this non-uniformity could be explained by the hydrophilicity of the PDA-modified interlayer that resulted in inhibited diffusion of amine monomers into the organic phase. As a result, the selective layer in the zeolite-rich areas was smoother and thinner than the rest of the polyamide regions; consequently, higher zeolite loading was proportionally accompanied by more such regions, thus elevating the overall membrane permeance35. These observations conclude that the synergistic effect from the morphological characteristics and surface hydrophilicity due to the zeolite interlayer rendered enhanced water transport across the composite membranes.

Fig. 7: Illustration of water transport length.
figure 7

a smooth and b crumpled TFC membranes. The effective distance to be traveled by water molecules (Deff) can be higher than the actual membrane thickness (D), while there is more opportunity for water molecules to permeate at the same thickness as the membrane, consequently reducing overall membrane resistance. c Cross-sectional image of M-100 with tent-like structures marked red. d Higher magnification of the boundary of zeolite-rich regions in M-100 demonstrating inhomogeneity in the membrane surface. Scale bar 1 µm.

Full size image

Apart from the enhancement in the permeance, results revealed that the composite membranes offered improved selectivity upon the incorporation of zeolite interlayer. In general, all prepared membranes exhibited higher rejection towards divalent salts highlighting the role of steric hindrance. For example, divalent salts characterized by larger hydrated radius experience higher steric hindrance and can be easily rejected by the membrane compared to monovalent Na+ and Cl- ions. Consequently, even our best-performing membrane could only offer rejection below 50% for NaCl. Analyzing the rejection obtained for MgCl2 for various membranes, we can identify a noticeable enhancement in the separation performance by the introduction of the zeolite interlayer. One of the reasons could be the enhancement of surface charge, as evidenced in the zeta potential measurement that contributed to the electrostatic repulsion between the negatively charged membrane surface and the oncoming negatively charged ions. Results obtained for PEG rejection also evidence the possibility of additional size sieving from the underlying zeolites. For instance, the distribution of zeolites as an interlayer rather than in the aqueous phase facilitated coverage of the larger pores on the support layer with finer intercrystallite characteristic pores, increasing the possibility of size exclusion11. Besides this, variations in cross-linking intensity would have caused better rejection, but validating this requires further investigations on the variations in polyamide cross-linking degree. It is also worth noting that, while creating a zeolite interlayer improved the membrane rejection by less than 10% (based on MgCl2), membrane permeance was enhanced by three times (based on pure water permeance). Therefore, we can reasonably conclude that the zeolite incorporation has a pronounced effect on the membrane flux than its selectivity.

In summary, a facile approach towards fabricating surface-crumpled TFC NF membranes was developed by designing a robust interlayer modification. The ball milling approach was utilized to produce smaller zeolites from commercial micro-zeolites using carbon nanostructure as a damping material to fabricate a defect-free PA layer. PDA-modified zeolites were used as an interlayer to impart highly wrinkled membrane morphology upon interfacial polymerization of PIP and TMC. IP membranes prepared with PDA-modified zeolite revealed a wrinkled layer with interfacial voids. At optimum conditions, a 21.2% enhancement in the effective surface area was achieved for the PA membrane upon interlayer modification. As a result, the membrane exhibited remarkable enhancement in the pure water permeance with more than 3 folds. Permeance increase was also ascribed to the enhancement in membrane hydrophilicity and the presence of interfacial voids. Additionally, results revealed that the PDA-modified zeolite PA membranes offered improved selectivity making MgCl2 rejection at 91.3% that could be attributed to the enhanced surface charge and presence of zeolite pores. The modified membranes maintained good performance for extended filtration time, demonstrating their robustness. The presented work showcased a strategy of facile substrate modification that can be extended to a variety of fillers for finer tuning of IP surface morphology. Further investigations are required to investigate the stability of crumpled polyamide layers in cross-flow operations targeting practical applications.

Methods

Materials

PSf pellets (average Mw ~35,000 g mol−1), polyvinylpyrrolidone (PVP) powder (Mw ~55000 g mol−1), n-Methyl-2-pyrrolidone (NMP), piperazine (PIP) (99%), 1,3,5-Benzenetricarbonyl chloride (TMC), dopamine hydrochloride (DA), tris-(hydroxymethyl) aminomethane (Tris), hydrochloric acid (HCl), sodium chloride (NaCl), magnesium chloride (MgCl2), magnesium sulfate (MgSO4), sodium sulfate (Na2SO4), and ethanol were purchased from Sigma-Aldrich. Micron-sized zeolite-Y (CBV 720) and carbon nanostructures (CNS) were supplied respectively by Zeolyst International and Applied Nanostructured Solutions LLC. All chemicals were used as received without any further purifications. Aqueous monomer and salt solutions were prepared using Milli-Q water.

Size reduction of zeolite-Y

As received, micron-sized commercial zeolites underwent a wet ball milling process which was previously developed by our group43. Zeolites and CNS (3:1 ratio by mass) were initially mixed thoroughly in a solution containing an equal amount of ethanol and deionized (DI) water for a few minutes to form a paste-like suspension. Thereafter, the thick paste was transferred to a zirconia jar and underwent ball milling using zirconia balls at 1000 rpm for a duration of 1 h using E-max high-energy ball mill machine from Retsch, Germany. Ball-milled product was centrifuged for 10 min at 4000 rpm to collect the bottom product while the top product containing finer particles was discarded. The collected product was dried in an oven overnight, followed by thermal treatment at 610 °C for 5 h to remove CNS. Obtained ball-milled zeolite products were stored under ambient conditions for characterization and membrane fabrication. Commercial micron-sized parent zeolite and zeolites obtained after size reduction through ball milling are labeled as pZ and nZ in the subsequent sections.

Preparation of PDA-modified zeolites

PDA-modified zeolites PDA-nZ were prepared following some previous reports14,48. Firstly, 500 mg of nZ was well dispersed in 500 mL tris-HCl buffer solution (10 mM, pH = 8.5) with the help of a probe sonicator (Q Sonica Ultrasonic Processor). About 2500 mg of DA was added to the evenly dispersed zeolite solution, followed by stirring at 25 °C for 5 h. The final product was collected by removing residuals through centrifugation and dried in an oven at 60 °C obtaining black powders. PDA-nZ solution at desired concentration was obtained simply by dispersing it in Milli-Q water with the aid of sonication.

Fabrication of TFC PA membranes

Fabrication of membrane support

Firstly, PSf membrane supports were fabricated through non-solvent-induced phase inversion following our previous report12. Initially, 17% of PSf and 3% PVP were dissolved in NMP by stirring at 60 °C overnight to form a homogeneous polymer casting solution. The solution was allowed to cool down to room temperature and subsequently cast on a glass plate using a semi-automatic casting machine (PMI Porous Materials Inc. Model BT FS- TC, US) at a constant shear rate of 200 s−1 and a casting thickness of 200 µm. After staying for 10 s, the glass plate containing nascent polymer film was immersed in a DI water bath to induce phase inversion. Subsequently, as-prepared membrane supports were stored in DI water for three days, replacing the water bath every 12 hours to minimize the effect of NMP and PVP.

Fabrication of PA layer

Wet membrane supports were held on a plate and frame with a dense surface exposed to air. Supports were exposed to PDA-nZ (50 mL) with the desired amount of PDA-nZ (50, 100, and 150 mg) for 30 min. Subsequently, the excess solution was removed with the help of an air gun and a gentle rinse with Milli-Q water. Thereafter, the PA layer was fabricated by the reaction of PIP and TMC solution. Membrane supports were contacted with PIP aqueous solution (2%) for 1 min, followed by the removal of excess solution with the aid of an air gun and filter paper. Following this, the TMC/ hexane (0.35%) was contacted for a short duration of 10 s to form the PA layer through a spontaneous interfacial reaction. Finally, as formed TFC PA membrane was cleaned with hexane to remove unreacted monomers and dried in the oven at 60°C for 3 min before storing in Milli-Q water. The interfacial polymerization reaction is presented in Fig. 1b. TFC membranes were prepared in varying concentrations of PDA-nZ solution to obtain the optimum membrane fabrication parameters. To have a comparison, pristine PA membranes without any zeolite modification were also prepared in a similar fashion. TFC membranes were labeled as M-0, M-50, M-100, and M-150 depending upon the amount of PDA-nZ added in the dipping solution used for membrane production.

Characterization

The morphology of zeolite particles was observed using a scanning electron microscope (SEM) (FEI Quanta 450 FEG) at a voltage of 10 kV. For this, zeolite solution at a dilute concentration was dispersed homogeneously with the aid of sonication and subsequently deposited on a silicon grid. Samples were dried completely, followed by metal coating before imaging. Changes in the crystal structure of zeolites due to ball milling were analyzed by X-ray diffraction (XRD) patterns recorded in the 2θ range of 2−80° at room temperature using X-ray Powder Diffraction (Malvern™, Empyrean 2, Malvern, UK) at a scan rate of 2° min−1. To investigate chemical groups on the zeolite particles, Fourier transform infrared (FTIR) spectra were captured in a range of 500 to 4000 cm−1 using a spectrometer (Thermo scientific™, Nicolet iS5 with iD7 ATR accessory). Further, to examine the pore size distribution after ball-milling, samples were initially degassed at 200 \(^\circ {\rm{C}}\) for 12 h to ensure the removal of moisture and any gases adsorbed; followed by an N2 adsorption experiment (NOVA®4200e Quantachrome Instruments).

All supports and membranes were air-dried before characterization. Changes in membrane hydrophilicity were investigated by checking the surface water contact angle using the sessile drop method (DSA100, Krüss GmbH, Hamburg, Germany). An average of at least three different locations was reported here to ensure the reliability of the values presented. The membrane surface and cross-section were examined by capturing an SEM image (FEI Quanta 450 FEG) at a voltage of 10 kV after gold coating for 20 s. The presence of zeolites on the membrane surface was confirmed by energy dispersive spectroscopy (EDS) integrated with the same SEM instrument. EDS analysis was carried out using beams irradiated at 10 kV voltage with a working distance of 10 mm and a take-off angle of 36°. The surface chemical functional groups of all the fabricated membranes were identified by FTIR analysis, as described previously. Variations in the surface charge of the membranes due to the presence of zeolites were investigated by evaluating the surface zeta potential at neutral pH using ZetaSizer (ZEN3600, Malvern Panalytical, UK) with polystyrene tracer solution. Finally, the MWCO of membranes were determined by evaluating its rejection towards a series of different 200 ppm PEG solutions (200 to 2000 Da). The MWCO is defined as the molecular weight of the solute, which is rejected by 90%. For this, the concentration of feed and permeate were monitored through the total organic carbon (TOC) analyzer (Sievers InnovOx TOC Analyzers).

Nanofiltration performance

The nanofiltration performance of all synthesized membranes was investigated using a benchtop dead-end filtration cell (supplied by Sterlitech, USA) schematically illustrated in Supplementary Fig. 1. All permeance and separation values were calculated from filtration experiments conducted at a pressure difference of 6 bar with an effective filtration area of 3.14 cm2. To obtain reliable results, an average of steady-state values obtained for three different membranes have been reported here, with error bars representing the standard deviation. The feed solution underwent continuous magnetic stirring during testing to minimize the effect of concentration polarization. Membrane separation performance was evaluated by assessing the ability of membranes to reject various aqueous monovalent and divalent salt solutions at a feed concentration of 1000 ppm. Permeance and rejection values were calculated using the following equations, where the concentrations of salt solutions were estimated by measuring the conductivity of the solution.

$${Permeance}=\frac{V}{A.\Delta t.\Delta P}$$

where V is the volume collected at the permeate side, \(\Delta {\rm{t}}\) is the time needed to collect, A is the membrane testing area, and \(\Delta {\rm{P}}\) is the transmembrane pressure during the testing.

$${Rejection}\,( \% )=\left(1-\frac{{C}_{P}}{{C}_{f}}\right)\times 100$$

where Cp and Cf are the conductivity of permeate and feed, respectively.