The selection of the membrane matrix and nanomaterials was based on the specific treatment requirements for high-salinity and high-hardness underground karst water. Priority was given to materials exhibiting superior hydrophilicity, antibacterial properties, and environmental compatibility. PES was selected as the membrane matrix due to its excellent mechanical strength, chemical stability, and pollution resistance [2]. Molybdenum disulfide (MoS₂) was chosen as the nanocore material because its layered structure provides nanochannels to enhance water flux, and its electronegativity aids in electrostatic repulsion of pollutants [22]. Tannic acid (TA) served as the biosource modifier; its abundant phenolic hydroxyl groups offer strong adhesion and reactivity, enabling functionalization through self-polymerization or coordination, aligning with green modification concepts [23]. To introduce antibacterial capabilities, polyethylenimine (PEI) and copper ions (Cu²⁺) were utilized, targeting bacterial cell membranes and metabolism, respectively.

Polyethersulfone (PES, industrial grade, Ultrason® E 6020 P) was supplied by BASF (Ludwigshafen, Germany). Scanning electron microscopy (SEM) grade MoS₂ nanoparticles (80 nm, purity $\ge$99.5%, Cat. No. 44888) were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Tannic acid (TA, analytically pure, Cat. No. 10019218) was obtained from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Polyethylenimine (PEI, MW 10,000, 99%, Cat. No. P816158) and Humic acid (fulvic acid content $>$90%, Cat. No. F809543) were purchased from Macklin Biochemical Co., Ltd. (Shanghai, China). Bovine serum albumin (BSA, purity $>$99%, fraction V, Cat. No. S14022) was supplied by Shanghai Yuanye Biotechnology Co., Ltd. (Shanghai, China). Copper sulfate pentahydrate (CuSO₄$\cdot$5H₂O, Cat. No. 10008218), Sodium hydroxide (NaOH, Cat. No. 10019718), and Hydrochloric acid (HCl, Cat. No. 10011018) were of analytical grade and procured from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). N,N-Dimethylacetamide (DMAc, Cat. No. D104969) and PEG (MW 2000, Cat. No. P103731) were obtained from Aladdin Biochemical Technology Co., Ltd. (Shanghai, China). Disodium hydrogen phosphate (Cat. No. S817748) and Potassium dihydrogen phosphate (purity $\ge$99%, Cat. No. P816688) were also purchased from Macklin Biochemical Co., Ltd (Shanghai, China). Deionized water (DW, 18.2 M$\mathrm{\Omega }$$\cdot$cm) was produced using a laboratory water purification system.

The preparation of antibacterial bionanomaterials adopts a step-by-step modification strategy to endow MoS₂ nanoparticles with biocompatibility and antibacterial activity through chemical modification. The specific steps:

Pretreatment of MoS₂ suspension: MoS₂ nanoparticles were initially washed with DW to remove impurities and dust attached to the surface, and then solid particles were separated by centrifugation. Next, 1.0 g of washed MoS₂ was redispersed in 100 mL Tris-HCl buffer solution, and 200 W power ultrasonic treatment was used for more than 30 minutes to break the agglomerates and disperse the particles evenly to improve the stability of the suspension. Afterwards, large particle agglomerates were removed by low-speed centrifugation for further purification to obtain a suspension with a relatively uniform particle size distribution. Finally, the concentration was measured, sealed, and stored. It needed to be lightly stirred or sonicated again before use to avoid particle sedimentation.

Coordination modification of TA/Cu²⁺: 1.0 g TA was dissolved in 100 mL DW, adding 0.5 g CuSO₄$\cdot$5H₂O and stirring for 30 minutes until completely dissolved. The MoS₂ suspension was slowly dropped into the mixed solution and reacted at 60 °C for 2 h while maintaining pH 8–9 (this pH range is designed to promote the deprotonation of the TA phenolic hydroxyl group, thereby enhancing its coordination ability with Cu²⁺ and subsequent electrostatic interactions). The polyhydroxyl structure of TA formed a coordination bond with Cu²⁺, and Cu²⁺ was adsorbed on the surface of MoS₂ through electrostatic interaction, improving the antibacterial performance. After the reaction, the suspension was transferred to a centrifuge tube, centrifuged at 10,000 rpm for 10 minutes, the supernatant was discarded, and then washed three times with DW to remove unreacted TA and Cu²⁺.

PEI cross-linking enhancement: The washed particles were redispersed in 50 mL of DW, 2.0 g of PEI (molecular weight 10,000) was added, the pH was adjusted to 8.5, and the reaction was performed with magnetic stirring at 60 °C for 2 h. PEI could combine with the TA-copper coordination layer (TA-Cu²⁺ layer) through multiple effects: First, the amine group in the PEI molecule forms a hydrogen bond with the phenolic hydroxyl group in the TA molecule that is not involved in coordination, enhancing the interface bonding force; Second, the positive charge carried by PEI electrostatically adsorbs with the negative charge generated by the ionization of phenolic hydroxyl groups in the TA-Cu²⁺ layer, further stabilizing the coating structure; In addition, the amine group of PEI may undergo a nucleophilic substitution reaction with the phenolic hydroxyl group of TA to form a covalent bond, and ultimately build a 3D network structure to improve material stability. After the reaction was completed, it (10,000 rpm for 10 min) was washed twice with DW and ethanol each to remove residual reagents.

Drying and preservation: The finally obtained MoS₂-TA@Cu²⁺ nanoparticles were dispersed in ethanol, treated with ultrasonics for 5 minutes to break the agglomerates, and then transferred to a vacuum drying oven (VDO) and dried at 60 °C and 0.1 mbar for 24 h. The dried particles were sealed and stored in a desiccator to avoid moisture absorption.

The non-solvent induced phase inversion method (NIPS) was utilized to prepare UF membranes [24, 25]. The process flow mainly includes the steps of film casting liquid preparation, scraping film forming, solidification bath phase separation, and post-processing, as shown in Fig. 1.

Fig. 1.

Membrane preparation process. PES, polyethersulfone; PEG, polyethylene glycol; DMAc, N,N-Dimethylacetamide.

Preparation of the film casting liquid: 15% PES and 8% PEG (molecular weight 2000) were weighed according to the total solid mass percentage. The solvent was DMAc. First, PES and PEG were added to DMAc and magnetically stirred in a constant 70 °C water bath for 8 h until they were completely dissolved to form a uniform and transparent base liquid. Subsequently, according to the target loading amount per unit membrane area (0, 5, 10, 15 g/m²), the corresponding mass of MoS₂-TA@Cu²⁺ modified nanoparticles was dispersed in a small amount of DMAc. Ultrasonic treatment with a power of 200 W was used for 30 minutes to break the agglomerates, and then slowly poured into the base liquid, and continued stirring for 4 hours to ensure good compatibility between the nanoparticles and the polymer matrix. Finally, the mixed solution was transferred to VDO and degassed at 40 °C and 0.1 mbar for 12 h to obtain a uniform and stable casting solution. Based on the nanoparticle loading, the un-Mf-M were recorded as PES, and the Mf-Ms were PES-MoS₂-TA@Cu²⁺-5, PES-MoS₂-TA@Cu²⁺-10, and PES-MoS₂-TA@Cu²⁺-15.

Scraping and solidification molding: The degassed casting liquid was poured evenly onto a clean glass plate, and a scraper with 200 µm thickness was used to scrape the film at a constant speed of 3 cm/s to form a liquid film with a uniform thickness. The scraped glass plate was immediately immersed in a DW coagulation bath at 25 °C to induce phase separation of the polymer through rapid exchange of solvent (DMAc) and non-solvent (water). During the coagulation process, DW needed to be replaced every 8 h for 24 h to completely remove residual DMAc and PEG. After solidification was completed, the film was peeled off from the glass plate to obtain a primary film with a finger-like pore structure.

Post-processing and performance stabilization: The primary membrane was soaked in DW for 48 h, during which the water was changed every 12 h to further clean the remaining impurities. Subsequently, the membrane was transferred to VDO and dried at 60 °C and 0.1 mbar for 6 h to remove moisture and avoid membrane pore collapse. The dried membrane needed to be prepressed with pure water before use: pure water was passed in at 0.15 MPa pressure for 30 min to stabilize the membrane structure and activate surface hydrophilic functional groups.

After the membrane sample was quenched by liquid nitrogen, a 5 nm thick platinum layer was sprayed on the surface. The surface and cross-sectional morphology were observed using a scanning electron microscope (Apreo 2, Thermo Fisher Scientific, Waltham, MA, USA) at an acceleration voltage of 10 kV.

The porosity ($\epsilon$) of the membranes was measured using the gravimetric method. The membrane samples were cut into standard sizes, soaked in distilled water for 24 h, and the wet weight ($W_w\mathrm{}$) was measured after wiping off the surface water with filter paper. Subsequently, the samples were dried in a vacuum oven at 60 °C for 24 h to obtain the dry weight ($W_d\mathrm{}$). The porosity was calculated using the following Eqn. 1:

(1)$$\epsilon =\frac{W_w-W_d}{{\rho }_w\times A\times l}\times 100\%$$

where ${\rho }_w$ is the density of water, $A$ is the membrane area, and $l$ is the membrane thickness. To ensure the reproducibility of the results, three independent replicate measurements were performed for each membrane sample, and the average value was reported.

The membrane samples were cut into 10 cm $\times$ 10 cm squares and fixed on the sample stage after drying. A contact angle meter (OCA 50, DataPhysics Instruments GmbH, Filderstadt, Germany) was employed to test the static water contact angle (WCA) of deionized water (1 µL) on the membrane surface. Each sample was tested at 5 different positions with a spacing of $\ge$5 mm, and the average value was taken.

The casting liquid was added dropwise to the glass slide and immediately immersed in a DW coagulation bath. A high-speed camera (Phantom v2512, Vision Research Inc., Wayne, NJ, USA) with a resolution of 1280 $\times$ 800 and a frame rate of 30,000 fps (exposure time 1.1 µs) was used to record the phase separation process. The gel front displacement was measured using ImageJ software (Version 1.53k, National Institutes of Health, Bethesda, MD, USA), and its average front velocity $\nu$ (µm/s) was calculated to evaluate the kinetics of the phase inversion process. The formula:

(2)$$k=\mathrm{\Delta }x/\mathrm{\Delta }t$$

where $\mathrm{\Delta }x$ is the displacement difference of the gel front (µm), and $\mathrm{\Delta }t$ is the corresponding time interval (s).

The membrane was installed in a cross-flow filtration device and prepressed at 0.1 MPa pressure for 30 min. The pure water permeation volume within 30 min was recorded, and the pure water flux ($\mathrm{PWF},J_w$) was calculated.

(3)$$J_w=V/(A\times t)$$

where $V$ represents the filtration volume (L), $A$ is the effective membrane surface area (m²), and $t$ represents the filtration time (h).

The MWCO of the membranes was determined by conducting ultrafiltration experiments using a series of PEG aqueous solutions with different molecular weights (e.g., 20, 40, 60, 100, and 200 kDa) at a concentration of 1.0 g/L. The filtration was performed under a transmembrane pressure of 0.1 MPa. The concentrations of PEG in the feed solution ($C_f$) and the permeate solution ($C_p$) were measured using a total organic carbon analyzer (Elab-TOC/DWT, Suzhou Elan Analytical Instrument Co., Ltd., Suzhou, China). The rejection rate ($R$) for each PEG molecular weight was calculated using the formula:

(4)$$R=(1\mathrm{–}{C}_p\mathrm{}/C_f\mathrm{})\times \mathrm{\hspace{0.25em}100}\%$$

The MWCO was defined as the molecular weight of the solute corresponding to a rejection rate of 90% in the solute rejection curve.

In this study, a mixed solution of humic acid and BSA was used as a pollutant model, and a cross-flow filtration device was used for testing. Before testing, the membrane samples were pre-pressed. Under the operating pressure of 0.05 MPa, the membrane PWF $J_0$ was continuously recorded for 30 min, and the mean was taken as the initial flux. The feed liquid was switched to a mixed solution of humic acid (50 mg/L) and BSA (50 mg/L). It was continuously filtered under the same pressure for 30 min, and the real-time flux $J_f$ was recorded. The operating pressure was adjusted to 0.1 MPa, the membrane surface was backwashed with DW for 10 min, and then adjusted back to 0.05 MPa, and the membrane PWF $J_w$ was remeasured. The four parameters of flux recovery rate:

(5)$$FRR=J_w/J_0\times \mathrm{\hspace{0.25em}100}\%$$

reversible fouling resistance:

(6)$$RFR=(J_w\mathrm{–}{J}_f)/J_0\times \mathrm{\hspace{0.25em}100}\%$$

irreversible fouling resistance:

(7)$$IFR=(1\mathrm{–}{J}_w/J_0)\times \mathrm{\hspace{0.25em}100}\%$$

and total fouling resistance:

(8)$$FR=(1\mathrm{–}{J}_f/J_0)\times \mathrm{\hspace{0.25em}100}\%$$

were used to comprehensively evaluate the anti-fouling performance (AFP) of the membrane.

The membrane sample was immersed in 1 $\times$ 10⁸ CFU/mL E. coli bacterial suspension. After culturing at 37 °C for 24 h, 100 µL of the suspension was spread on an Luria Broth (LB) agar plate, and the number of colonies was counted. The antibacterial rate ($S$) was calculated as:

(9)$$S=(N_0\mathrm{–}{N}_s)/N_0\times \mathrm{\hspace{0.25em}100}\%$$

where $N_0$ and $N_s$ are the number of colonies in the blank control group and sample group, respectively. All antibacterial tests were conducted in triplicate to minimize experimental error, and the data were presented as mean $\pm$ standard deviation.

Taking the underground karst water in the Fengcheng area of Jiangxi Province as the feed water sample, a turbidimeter (HK-288, Beijing Huakeyi Technology Co., Ltd., Beijing, China) was used to measure the turbidity value before and after membrane filtration. Before testing, the water samples were pretreated via a 0.45 µm microporous filter membrane (MFM) to remove large particle interference. Each test was repeated three times to take the average. Turbidity removal rate ($C$) formula:

(10)$$C=(1\mathrm{–}{c}_1/c_2)\times \mathrm{\hspace{0.25em}100}\%$$

where $c_1$ and $c_2$ are the turbidity of the permeate and the feed water, respectively.

According to the ASTM D4189 standard method, a silt density index (SDI) meter (Model MRSDI3325-82905, Roddy Technology, Claremont, CA, USA) equipped with a 0.45 µm mixed cellulose ester MFM was used for testing. First, using underground karst water in the Fengcheng area of Jiangxi Province as the feed liquid, under a constant pressure of 0.21 MPa, the time $t_1$ for the first filtration of a 500 mL water sample was recorded. Then, the pressure remained unchanged, and after continuing to filter for 15 min, the time $t_2$ for filtering another 500 mL of water sample was recorded again. Finally, the SDI values were calculated according to the formula:

(11)$$SDI=(1\mathrm{–}{t}_1/t_2)\times \frac{100}{15}$$

The SDI reduction value was the difference between the raw water SDI and the filtrate SDI.

A total organic carbon (TOC) analyzer (Elab-TOC/DWT, Suzhou Elan Analytical Instrument Co., Ltd., Suzhou, China) was used to measure the TOC concentration of water samples before and after membrane filtration. Before testing, the water samples were filtered through a 0.45 µm filter membrane to eliminate the interference of suspended organic matter, and each sample was measured three times in parallel. The TOC removal rate ($O$) was calculated as:

(12)$$O=(1\mathrm{–}{o}_1/o_2)\times \mathrm{\hspace{0.25em}100}\%$$

where $o_1$ and $o_2$ are the TOC concentration of the filtrate and the raw water, respectively.

Static scaling experiment: The membrane samples were cut into 2 cm $\times$ 2 cm pieces, washed with DW, vacuum dried, and then immersed in supersaturated CaCO₃ and CaSO₄ solutions, respectively, in a constant temperature water bath at 25 °C for 72 h. After removing the membrane, DW was utilized to gently rinse loose crystals not attached to the surface, followed by vacuum drying to constant weight. An analytical balance (accuracy 0.1 mg) was used to record the membrane mass change to obtain the amount of fouling per unit area.

Dynamic filtration anti-scaling experiment: The membrane sample was installed in a cross-flow filtration device, and an aqueous solution of humic acid (mixed concentration of 50 mg/L in underground karst water) was continuously filtered three times under 0.1 MPa operating pressure and 1.5 L/h cross-flow rate. After each filtration cycle, a backwash was performed, and the PWF of different membrane materials was measured.

First, the quality of underground karst water taken from the Permian Changxing Formation and Triassic Qinglong Formation limestone aquifers in the Fengcheng area of Jiangxi was analyzed, as shown in Table 1. It showed a typical HCO₃-SO₄-Ca-Mg water quality type, with a total hardness as high as 380–490 mg/L, which was a typical hard water.

SEM was utilized to characterize the cross-sectional morphology of the UF membrane before and after modification with MoS₂-TA@Cu²⁺ nanoparticles at different loadings (0, 5, 10, 15 g/m²), as shown in Fig. 2. As shown in Fig. 2A, the unmodified PES membrane exhibits a typical asymmetric structure, consisting of a dense skin layer and a support layer dominated by finger-like macrovoids. While this fundamental asymmetric configuration was preserved across all modified membranes (Fig. 2B–D)—ensuring adequate mechanical stability—a closer inspection reveals certain modifications in the pore architecture. In the pristine PES membrane (Fig. 2A), the macrovoids are relatively narrow and isolated, separated by thick polymer walls. However, upon the incorporation of nanoparticles, these macrovoids showed moderate widening. Specifically, at the highest loading of 15 g/m² (Fig. 2D), the polymer skeletal walls between the voids appeared slightly thinner, contributing to improved local connectivity compared to the thick-walled structure of the pristine membrane. This phenomenon can be attributed to the role of nanoparticles as hydrophilic pore-forming agents, which accelerated the exchange rate of solvent (DMAc) and non-solvent (water) during the phase transformation process. This accelerated instantaneous liquid-liquid phase separation process facilitated the development of macrovoids in the support layer, enhancing the pore connectivity and providing a structural basis for the subsequently observed increase in water flux.

Fig. 2.

Cross-sectional SEM images of Mf-Ms with different loading amounts of MoS₂-TA@Cu²⁺ nanoparticles. (A) PES. (B) PES-MoS₂-TA@Cu²⁺-5. (C) PES-MoS₂-TA@Cu²⁺-10. (D) PES-MoS₂-TA@Cu²⁺-15. SEM, scanning electron microscopy; Mf-Ms, modified membranes; TA, tannic acid. Scale bar = 50 µm.

Fig. 3 shows the changes in porosity, MWCO, and PWF of the Mf-M. In Fig. 3A, the porosity of the Mf-M increased from 71.6% of the PES membrane to 82.4% of the PES-MoS₂-TA@Cu²⁺-15 membrane. It is noteworthy that the porosity exhibited a sharp jump at the initial loading (0 to 5 g/m²), while the growth rate slowed down noticeably as the loading further increased to 10 and 15 g/m². This phenomenon is attributed to the competitive mechanism between thermodynamic instability and rheological hindrance. Initially, the addition of hydrophilic nanoparticles significantly lowers the thermodynamic barrier, triggering rapid instantaneous demixing and creating abundant voids. However, at higher loadings ($>$5 g/m²), the increased viscosity of the casting solution kinetically retards the solvent/non-solvent exchange rate, thereby limiting the further linear expansion of void volume.

Fig. 3.

Changes in membrane properties of the Mf-Ms. (A) Porosity. (B) MWCO. (C) PWF. PWF, pure water flux; Mf-Ms, modified membranes; MWCO, molecular weight cutoff. Data are presented as mean $\pm$ standard deviation (n = 3).

Correspondingly, the PWF of the membrane also increased significantly (Fig. 3C), from 257.23 L/m²$\cdot$h of the PES membrane to 385.44 L/m²$\cdot$h of the PES-MoS₂-TA@Cu²⁺-15 membrane. Regarding the specific flux behavior at high loadings (10 and 15 g/m²), although the porosity growth plateaued, the water flux continued to improve. This can be explained by the evolution of pore connectivity. At these higher loadings, slight micro-agglomeration of nanoparticles may occur, which acts as larger “porogens” to form wider and more interconnected finger-like macrovoids (as confirmed by the SEM image in Fig. 2D). These expanded channels reduce the hydraulic resistance, ensuring high permeability even when the porosity increase is marginal.

In addition, the change in MWCO of the UF membrane is exhibited in Fig. 3B. Compared with the PES membrane, the MWCO of the MoS₂-TA@Cu²⁺ Mf-M increased moderately from 127 kDa to 169 kDa. This phenomenon seems to be contrary to the expected separation performance, but its underlying mechanism can be explained as follows: MoS₂-TA@Cu²⁺ nanoparticles acted as hydrophilic “porogens” to accelerate phase transformation and form larger supporting pores (Fig. 2D). At the same time, they may also cause microscopic disturbances in the stacking of polymer chains in the dense cortex of the membrane surface. This perturbation may lead to a slight increase in the mean value of the effective pore size distribution of the cortex, which is manifested as an increase in MWCO. Therefore, the Mf-M achieved a balance between a large increase in water flux and a moderate adjustment of selectivity, rather than a simple performance degradation.

Fig. 4 presents the hydrophilicity test results of MoS₂-TA@Cu²⁺ nanoparticle-modified UF membrane. In Fig. 4A, the contact angle of the unmodified PES film was 69°. As the amount of MoS₂-TA@Cu²⁺ nanoparticles gradually grew from 0 to 15 g/m², the contact angle of the modified film showed a downward trend, and the contact angle of the PES-MoS₂-TA@Cu²⁺-15 film dropped to 56°, indicating that the hydrophilicity of the film gradually increases. This is because the introduction of MoS₂-TA@Cu²⁺ nanoparticles increases the polar groups on the membrane surface and improves the water wettability of the membrane. Fig. 4B shows the gel displacement curve of the cast film liquid. The gel displacement of all casting solutions gradually increased over time. Among them, the cast liquid gel displacement of the PES-MoS₂-TA@Cu²⁺-15 membrane increased most significantly, reaching 110 µm at 60 seconds, which was significantly higher than the 95 µm of the PES membrane. This demonstrated that as the loading amount of MoS₂-TA@Cu²⁺ nanoparticles increased, the gelation rate of the casting solution accelerated and the gel displacement increased. This change may be due to the introduction of nanoparticles, changing the phase separation process of the casting liquid and promoting the formation and development of gel. Faster gel rates and larger gel displacements contributed to a more uniform and denser membrane structure, which may have a positive impact on membrane performance.

Fig. 4.

Membrane hydrophilicity and film formation kinetics. (A) Static WCA. (B) Curve of cast film liquid gel displacement with time. WCA, water contact angle.

A three-cycle filtration test was conducted using a mixed solution of humic acid and BSA as simulated pollutants to evaluate the dynamic anti-fouling stability. The normalized flux changes are presented in Fig. 5. All membranes experienced a typical flux decay during the filtration stage due to the accumulation of foulants. However, the unmodified PES membrane suffered the most severe attenuation, with the normalized flux dropping to 0.58 in the third cycle, indicating serious irreversible fouling caused by hydrophobic adsorption of proteins and organic matter.

Fig. 5.

Normalized flux changes of the membrane in the three-cycle filtration experiment.

In contrast, the Mf-M modified with MoS₂-TA@Cu²⁺ demonstrated significantly enhanced fouling resistance. Specifically, for the PES-MoS₂-TA@Cu²⁺-10 membrane, the normalized flux in the third cycle was maintained at 0.815, representing a 40.5% improvement over the pristine PES membrane. This superior performance can be attributed to two synergistic mechanisms: First, the abundant hydrophilic hydroxyl groups from TA and MoS₂ formed a tightly bound hydration layer on the membrane surface, which acts as a physical barrier to prevent the direct contact and adsorption of foulants (steric hindrance effect) [22]. Second, the electronegativity of the modified surface induces a strong electrostatic repulsion against negatively charged foulants (humic acid and BSA) at neutral pH, thereby mitigating the deposition rate [15]. Compared to similar PES-based modification studies where flux recovery rate (FRR) values typically range from 60% to 75% [3], the FRR of 81.5% achieved in this work highlights the exceptional reusability and operational stability of the MoS₂-TA@Cu²⁺ nanomodified membrane.

Fig. 6A shows the changes in membrane fouling parameters with the loading of MoS₂-TA@Cu²⁺ nanoparticles when a mixed solution of 50 mg/L humic acid and 50 mg/L BSA is used as a model pollutant. As the loading amount of MoS₂-TA@Cu²⁺ nanoparticles increased, the fouling parameters of the membrane changed significantly. Among them, RFR and IFR first decreased and then increased, while FRR first increased and then decreased. Fig. 6B shows the antibacterial performance test results. As the loading amount of MoS₂-TA@Cu²⁺ nanoparticles increased, the antibacterial rate of the membrane significantly increased. When the loading capacity was 15 g/m², the antibacterial rate reached 95%, which was significantly higher than that of the unmodified PES membrane. This indicated that the addition of MoS₂-TA@Cu²⁺ nanoparticles could effectively enhance the antibacterial performance of the membrane, reduce the attachment and growth of microorganisms on the membrane surface, thereby reducing the risk of membrane fouling. Meanwhile, combined with changes in membrane fouling parameters, appropriately loaded MoS₂-TA@Cu²⁺ nanoparticles could improve the AFP of the membrane while improving the antibacterial performance of the membrane.

Fig. 6.

Anti-fouling and antibacterial properties of membranes. (A) Fouling resistance parameters (RFR, IFR, FR) and FRR of different membranes. (B) Antibacterial rates of different membranes against E. coli. RFR, reversible fouling resistance; IFR, irreversible fouling resistance; FR, fouling resistance; FRR, flux recovery rate; E. coli, Escherichia coli. Error bars represent the standard deviation of three independent replicates.

According to Fig. 7, due to the lack of antibacterial components in the unmodified PES membrane, obvious bacterial proliferation occurred on and around the membrane surface, and no inhibitory zone was formed. The MoS₂-TA@Cu²⁺ Mf-M showed a dose-dependent antibacterial effect: when the nanoparticle loading reached 5 g/m², the diameter of the inhibition zone was 7.2 mm. When the loading capacity increased to 10 g/m² and 15 g/m², the inhibition zone expanded to 9.1 mm and 11.5 mm. This significant antibacterial performance can be attributed to the sustained release of Cu²⁺, which effectively disrupts microbial cell structures, thus endowing the Mf-M with excellent anti-biofouling properties.

Fig. 7.

Inhibition zone test results of different membranes. (A) PES. (B) PES-MoS₂-TA@Cu²⁺-5. (C) PES-MoS₂-TA@Cu²⁺-10. (D) PES-MoS₂-TA@Cu²⁺-15.

It is worth exploring in depth why the membrane with a loading capacity of 10 g/m² shows the best anti-fouling stability, while the membrane with a loading capacity of 15 g/m² has better separation performance. It is speculated that this performance trade-off is closely related to the dispersion state of nanoparticles in the polymer matrix and the resulting changes in membrane structure. At lower loadings (such as 10 g/m²), MoS₂-TA@Cu²⁺ nanoparticles may be dispersed in the membrane matrix in a more uniform and isolated state, maximizing the role of their hydrophilic and antibacterial groups, thereby giving the membrane excellent anti-fouling adhesion capabilities. However, when the loading amount is further increased to 15 g/m², although the macroscopic hydrophilicity continued to increase, the possibility of local micro-agglomeration of the nanoparticles increased. Such microaggregates played a dual role in membrane formation and use. On the one hand, as “porogens”, they induced the formation of larger finger-like pores during the phase transformation process (Fig. 2D), significantly increasing the porosity and PWF of the membrane, which is reflected in the improvement of separation performance. On the other hand, the increased surface roughness caused by these agglomerates would become an “anchor point” for preferential adsorption of pollutants in long-term operation, which in turn slightly weakened its anti-pollution stability in dynamic tests. Therefore, 10 g/m² was considered the optimal loading to achieve a balance between anti-fouling and separation performance.

To further highlight the superiority of the MoS₂-TA@Cu²⁺ modified membrane, the comprehensive performance of the optimized membrane in this study was compared with other reported PES-based nanocomposite ultrafiltration membranes. As summarized in Table 2 (Ref. [2, 3, 4, 6]), the MoS₂-TA@Cu²⁺ modified membrane exhibits a highly competitive overall performance. Although Liang et al. [2] achieved higher flux, their membrane required a complex polyvinylidene fluoride (PVDF)/PES blend and silver nitrate reduction process. Compared to the pure PES-based membranes modified with inorganic nanoparticles [3, 4, 6], our work demonstrates a superior balance. Specifically, the modified membranes demonstrated a superior balance of properties through tunable nanoparticle loading. The membrane with 15 g/m² loading achieved a peak pure water flux of 385.4 L/m²$\cdot$h, which is nearly 7 times higher than that of the WO_2.89-modified membrane (54.9 L/m²$\cdot$h) [3] and 1.5–1.8 times higher than the MAX phase [4] and TiO₂/MXene [6] membranes. Meanwhile, the 10 g/m² loading optimized the anti-fouling stability, achieving a high FRR of 81.5% solely through physical backwashing without requiring additional UV energy consumption [6], validating the robust passive anti-fouling capability of the MoS₂-TA@Cu²⁺ surface layer.

Table 3 shows the comparison of turbidity removal rate (TRR), SDI reduction value, and TOC removal rate. The TRR of the Mf-M significantly surpassed that of the unmodified membrane, and increased with the increase of MoS₂-TA@Cu²⁺ loading. The removal rate of the PES-MoS₂-TA@Cu²⁺-15 membrane reached 92.9%, which was 19.1% higher than that of the 73.8% of the unmodified membrane (relative increase of 25.9%). This is related to the increased porosity and enhanced surface hydrophilicity of the Mf-M: a richer pore structure can effectively intercept colloidal particles, and the hydrophilic surface reduces particle adsorption through the hydration layer, thereby improving the turbidity removal effect. The SDI reduction value of the Mf-M outperformed that of the unmodified membrane. The SDI reduction value of the PES-MoS₂-TA@Cu²⁺-15 membrane reached 5.3, and the SDI of the filtrate was reduced to 1.2, which fully met the requirements for reverse osmosis pretreatment and reduced the risk of contamination in the subsequent desalination process. The TOC removal rate of the Mf-M increased with the increase in nanoparticle loading. The removal rate of the PES-MoS₂-TA@Cu²⁺-15 membrane reached 65.7%, which was 25.7% higher than the 40.0% of the unmodified membrane (relative increase of 64.3%). The reason is that the layered structure of MoS₂ and the phenolic hydroxyl group of TA can synergistically intercept organic matter through adsorption. At the same time, the optimization of membrane pore size can effectively block the penetration of macromolecular organic matter, thereby reducing the TOC content in water. In summary, MoS₂-TA@Cu²⁺ nanomodification significantly improved the UF membrane’s ability to remove turbidity, colloids, and organic matter in underground karst water, and the SDI reduction effect was particularly prominent. This showed that it could be used as an efficient pretreatment process for high-salinity HHKW to provide high-quality feed water for subsequent deep desalination.

In Fig. 8A, as the loading amount of MoS₂-TA@Cu²⁺ increased, the amount of fouling on the membrane surface showed an obvious downward trend, whether in supersaturated CaCO₃ solution or supersaturated CaSO₄ solution. In the CaCO₃ solution, the scaling amount of the unmodified PES membrane was 0.92 mg/cm². When the loading reached 15 g/m², the scaling amount was reduced to 0.32 mg/cm² corresponding to a reduction rate of approximately 65.2%. In the CaSO₄ solution, the initial scaling amount was 0.78 mg/cm², which dropped to about 0.29 mg/cm² at 15 g/m². This showed that MoS₂-TA@Cu²⁺ nanoparticles could effectively inhibit inorganic scaling under static conditions. In Fig. 8B, as the filtration volume increased, the normalized flux attenuation of the unmodified PES membrane was the most significant. When the filtration volume reached 1200 mL, its normalized flux had dropped to 0.1. In contrast, the specific flux attenuation of the Mf-M with MoS₂-TA@Cu²⁺ was significantly slowed down. The PES-MoS₂-TA@Cu²⁺-15 membrane exhibited the strongest resistance to flux attenuation, and the normalized flux remained around 0.5 at the end of filtration. This fully proved that MoS₂-TA@Cu²⁺ nanoparticles could also improve the AFP of the membrane during dynamic filtration.

Fig. 8.

Membrane AFP test results. (A) The amount of fouling per unit area in the static fouling experiment. (B) The change of specific flux with filtration volume in the dynamic filtration experiment. AFP, anti-fouling performance.

A core finding of this study is that the membrane with a loading capacity of 10 g/m² exhibits optimal anti-fouling recoverability and long-term stability, whereas the membrane with a loading capacity of 15 g/m² is superior in permeability and instantaneous separation efficiency. This performance trade-off can be attributed to the dispersion state of the nanoparticles in the polymer matrix and their dual regulatory role on the membrane structure and surface properties.

On the one hand, acting as efficient “porogens”, especially at a high loading of 15 g/m², the local micro-agglomeration of nanoparticles induces the formation of larger and more interconnected finger-like pores during the phase transformation process (macro pore structure in Fig. 2D). This significantly increases the overall porosity and PWF of the membrane, resulting in optimal separation performance.

On the other hand, these microaggregates also alter the micromorphology of the membrane surface. At a lower loading (10 g/m²), the nanoparticles are dispersed more uniformly, maximizing the exposure of hydrophilic and antibacterial groups, thereby endowing the membrane with excellent anti-fouling adhesion capabilities. However, at higher loading (15 g/m²), microagglomerates may increase surface roughness at the nanometer scale. These raised structures can act as “anchoring sites” for the preferential adsorption of pollutants during long-term operation. Although the macroscopic hydrophilicity remains strong, the presence of these “anchors” slightly increases the difficulty of foulant detachment during backwashing, which explains why the FRR (76.9%) at this loading is slightly lower than the optimal value achieved at 10 g/m².

Therefore, 10 g/m² is considered the optimal loading amount to achieve a balance between anti-pollution and separation performance, maximizing the anti-pollution potential while ensuring excellent separation efficiency.

While the elucidation of this tunable trade-off mechanism provides a precise theoretical guideline for balancing membrane permeability and stability in controlled environments, the transition from laboratory synthesis to practical industrial application still faces several constraints that must be acknowledged. First, the experiments were conducted on a laboratory scale using dead-end and cross-flow filtration modules. While these results provide a strong proof-of-concept, they may not fully predict the hydrodynamic behavior and fouling distribution in industrial-scale spiral-wound membrane modules. Second, although the 72-hour static scaling and three-cycle dynamic filtration tests confirmed the short-term stability of the membrane, the long-term durability ($>$1000 hours) of the modification layer and the potential leaching of Cu²⁺ under fluctuating hydraulic conditions require further longitudinal assessment. Third, this study focused specifically on high-hardness karst water; the applicability of this membrane to other complex water matrices containing high concentrations of oils or synthetic dyes remains to be verified. Fourth, regarding membrane characterization, this study primarily relied on SEM and static contact angle measurements. While sufficient for explaining the macroscopic performance trade-offs, the lack of quantitative nanoscale roughness data (typically obtained via Atomic Force Microscopy, AFM) limits the precise correlation between micro-topography and pollutant adhesion. Future research will address these gaps by conducting pilot-scale studies, employing advanced surface characterization techniques, and extending the evaluation to diverse water quality scenarios.