High-performance PVDF ultrafiltration membranes modified with ZIF-8@g-C₃N₄ nanohybrids for enhanced humic acid removal and antifouling performance

Abstract

In this work, a novel metal–organic framework (MOF) hybrid, referred to as ZGCN (ZIF-8@g-C₃N₄), was incorporated into polyvinylidene fluoride (PVDF)-based ultrafiltration membranes fabricated via the phase inversion method. The ZGCN nanohybrid was synthesized through a two-step process: first, g-C₃N₄ was prepared by thermally condensing melamine at 550 °C, followed by the in situ growth of ZIF-8 nanocrystals on the g-C₃N₄ sheets under solvothermal conditions. The ZGCN nanoparticles were then dispersed in a 15 wt% PVDF casting solution and cast into asymmetric membranes using N-methyl-2-pyrrolidone (NMP) as the solvent. Comprehensive characterization including X-ray diffraction (XRD), scanning electron microscopy (SEM), Brunauer–Emmett–Teller (BET) surface area analysis, thermogravimetric analysis (TGA), and Fourier-transform infrared spectroscopy (FTIR) confirmed the incorporation and properties of the hybrid. The BET surface area of ZGCN reached 987 m²/g, indicating high porosity. SEM images showed a well-dispersed ZGCN phase within the membrane matrix and an enhanced pore structure. The modified membranes exhibited an initial pure water flux of 318.5 L·m⁻²·h⁻¹ (LMH), stabilizing at 172.8 LMH after 4 h of humic acid (HA) filtration at 2 bar, with a separation efficiency of 91.6%. In comparison, pristine PVDF membranes showed a lower initial flux of 201.3 LMH and HA rejection of only 76.4% under the same conditions. Moreover, the ZGCN-PVDF membranes demonstrated superior antifouling performance, with a flux recovery ratio (FRR) of 89.2%. These results highlight the synergistic effect of the MOF hybrid in enhancing membrane permeability, selectivity, and fouling resistance, offering a promising approach for environmentally friendly water purification applications.

Introduction

The growing global need for clean, safe water has resulted in more research projects aimed at finding new ways to eliminate natural organic matter (NOM), including humic substances, from surface and groundwater [1, 2]. Humic acid (HA), the most important part of NOM, consists of a mix of aromatic and aliphatic compounds that come from the breakdown of plants and animals by microbes [3]. HA is known to be one of the hardest contaminants to remove from water. In addition to the potential to form carcinogenic disinfection byproducts (DBPs), HA will foul filtration membranes, impart color and odors into drinking water, and in all these ways represent a water protection threat [4]. There has been some reported success in dealing with the problem through traditional treatment options, such as coagulation–flocculation, activated carbon adsorption, and advanced oxidation processes; however, they still have challenges such as high operating costs, limited selectivity, generation of secondary waste, and inconsistent regulatory compliance with changing water qualities [5, 6].

Filtration through membrane, in particular ultrafiltration (UF), is a technology that has attracted the attention of the researchers as a result of its simplicity of operation, small-installation space, and the ability to keep the molecules of macromolecular organic matter in solution [7, 8]. Nevertheless, membranes made of traditional polymers such as polysulfone (PS), polyethersulfone (PES), and polyvinylidene fluoride (PVDF) often bring with them problems of major fouling, a selective capacity that is not high enough, and the issue of permeability being in a trade-off with the rejection of the membranes [9, 10]. The main ways of solving the problem of the membrane are to be found in treatment by modification of the surface, blending with hydrophilic additives, and nanomaterial forming [11]. Out of all these, the utilize of metal organic frameworks (MOFs) is the most exciting one as it provides membranes with their adjustable pore size, the huge surface area, and the myriad of functions achievable by their bending with molecules [12, 13]. ZIF-8 is a zeolitic imidazolate framework known for its stability, both thermally and chemically. It also allows for molecular sieving effects [14–16]. However, ZIF-8 alone does not completely solve the fouling issue, especially during long-term operation [17]. Graphitic carbon nitride (g-C₃N₄) is a metal-free photocatalyst with several benefits. It can work in visible light, has a hydrophilic nature, and creates a self-cleaning surface when combined with an embedded membrane system.

The aim of this study is to synthesize and characterize a new ultrafiltration membrane based on polyvinylidene fluoride (PVDF) incorporated with a hybrid nanomaterial, ZIF-8@g-C₃N₄ (ZGCN), for enhancing the removal of humic acid (HA) from water. The study aims at addressing issues of low permeability, fouling, and separation performance of membrane filtration by combining the attributes of metal-organic frameworks (MOFs) and g-C₃N₄. Its main objective is to synthesize a high-performance membrane with increased water flux, high HA rejection levels, and better antifouling properties. It is therefore a perfect candidate for green water treatment.

Although MOF- and carbon nitride-modified PVDF membranes have attracted considerable attention for water treatment applications, many previous studies have focused on either a single functional filler or a specific membrane function. ZIF-8 has been incorporated into PVDF primarily to improve membrane hydrophilicity, permeability, and antifouling properties [18], whereas g-C₃N₄ has been blended with or immobilized on PVDF membranes mainly to exploit its visible-light-driven photocatalytic activity for pollutant degradation and self-cleaning applications [19]. More recently, MOF/g-C₃N₄ hybrid systems have also been integrated with PVDF membranes. For example, NH₂-MIL-101/oxygen-doped g-C₃N₄ heterojunction membranes [20] and Ti-MOF/g-C₃N₄ composite membranes [21] have been developed for photocatalytic and photo-Fenton-assisted applications. In these studies, the MOF/g-C₃N₄ components were introduced through surface-deposition or heterojunction-coating approaches applied after membrane formation, rather than through direct incorporation into the membrane casting solution. Furthermore, an alkali-etched ZIF-67/PVDF ultrafiltration membrane was reported to enhance membrane permeability and antifouling behavior [22], but it did not employ g-C₃N₄ as a co-filler.

For HA removal specifically, previous PVDF/MOF membranes have employed MOF-5, MOF-199, and MOF-801 systems) [23–25], while ZIF-8@HPAN has been reported using a non-PVDF support [26], without combining ZIF-8 and g-C₃N₄ as co-fillers. However, these HA-targeted studies did not combine a MOF with g-C₃N₄ as a co-filler, while the reported MOF/g-C₃N₄-PVDF hybrid membranes discussed above did not target HA removal. To the best of our knowledge, we found no reported study combining a preformed ZIF-8@g-C₃N₄ nanohybrid with direct casting-solution blending in PVDF for HA separation. In the present work, a solvothermally synthesized ZIF-8@g-C₃N₄ (ZGCN) nanohybrid was directly incorporated into PVDF through non-solvent-induced phase separation (NIPS). The novelty of this work lies in integrating the preformed ZIF-8@g-C₃N₄ nanohybrid architecture with direct PVDF blending and applying the resulting membrane to HA separation, rather than in the individual use of ZIF-8 or g-C₃N₄ alone.

Experimental

Materials

All chemicals used in this study were of analytical grade and used without further purification. Polyvinylidene fluoride (PVDF) powder with a molecular weight of approximately 275,000 g/mol and 99.5% purity was purchased from Arkema Inc. (France). N-Methyl-2-pyrrolidone (NMP, ≥ 99.5%), used as the solvent for membrane casting, was supplied by Merck KGaA (Germany). Zinc nitrate hexahydrate (Zn(NO₃)₂·6 H₂O, 98%) and 2-methylimidazole (99%), used as precursors for the fabrication of ZIF-8, were obtained from Sigma-Aldrich (USA). Melamine (99%), the precursor for the fabrication of graphitic carbon nitride (g-C₃N₄), was also supplied by Sigma-Aldrich. For membrane performance testing, humic acid sodium salt (technical grade) was used as a model organic foulant for initial screening and proof-of-concept studies and was purchased from Acros Organics (Belgium). The authors acknowledge that commercial humic acid standards represent a simplified model system compared to the complex mixture of humic and fulvic substances found in natural waters. This standardized model compound was selected to establish baseline performance metrics under controlled, reproducible conditions, enabling systematic comparison with existing literature. Future studies will extend this work to environmentally relevant NOM sources, including IHSS Suwannee River reference materials (Humic Acid Standard II and Fulvic Acid Standard I). Additionally, ethanol (≥ 99.9%) and deionized (DI) water were used for washing and membrane rinsing throughout the fabrication and testing processes. All reagents and solvents were stored in tightly sealed containers and used under ambient laboratory conditions unless otherwise specified.

Synthesis of g-C₃N₄ and ZIF-8@g-C₃N₄ (ZGCN) nanohybrids

The ZIF-8@g-C₃N₄ (ZGCN) hybrid (Fig. 1) was synthesized through two primary steps: the thermal preparation of g-C₃N₄ and the subsequent in situ growth of ZIF-8 on its surface. g-C₃N₄ was synthesized through the thermal condensation of 10 g of melamine in a covered alumina crucible. The crucible underwent heating in a muffle furnace at 550 °C for duration of 4 h, with a ramping rate of 5 °C/min in an air atmosphere. Following natural cooling to room temperature, the pale-yellow solid, with purity of ~90%, was pulverized into a fine powder with an agate mortar and pestle and subsequently stored in a desiccator.

Fig. 1.

Schematic diagram for the preparation of ZGCN @PVDF membrane.

To synthesize the ZIF-8@g-C₃N₄ (ZGCN) nanohybrid, 1 g of g-C₃N₄ powder was dispersed in 100 mL of ethanol via ultrasonication for 1 h to ensure uniform dispersion. In a separate container, 2.93 g of Zn(NO₃)₂•6 H₂O was dissolved in 50 mL of ethanol, referred to as Solution A, while 6.49 g of 2-methylimidazole was dissolved in 50 mL of ethanol, designated as Solution B. Solution A was introduced dropwise to the g-C₃N₄ dispersion while maintaining vigorous magnetic stirring, subsequently followed by the prompt addition of Solution B. The mixture was stirred continuously at room temperature for 24 h to facilitate the in situ nucleation and growth of ZIF-8 crystals on the surface of g-C₃N₄ sheets. The pale-gray precipitate was obtained through centrifugation at 8000 rpm for 10 min, subsequently washed three times with methanol to eliminate unreacted species, and dried overnight at 60 °C in a vacuum oven. The final product, designated ZGCN, was stored in airtight containers for subsequent use in membrane fabrication.

Synthesis of PVDF-ZGCN composite membrane

The PVDF-ZGCN composite membrane was fabricated (Fig. 1) by first dissolving PVDF powder in N-Methyl-2-pyrrolidone to form a homogeneous solution. The synthesized ZIF-8@g-C₃N₄ (ZGCN) nanohybrid was dispersed into the PVDF solution by ultrasonication at a concentration of 15% w/v for 30 min to achieve a homogeneous distribution of the nanohybrid in the solution. The composite solution was cast onto a flat glass substrate using a doctor blade with a fixed casting gap of 200 μm. The cast film was exposed to ambient air for 30 s to allow partial solvent evaporation prior to immersion. The film was then immersed in a coagulation bath of deionized water maintained at 25 °C to induce non-solvent-induced phase inversion, thereby solidifying the membrane. After phase inversion, the resulting membrane had a final thickness of approximately 100 μm, as measured using a digital micrometer at five random locations across the membrane surface. The membrane was then washed with fresh deionized water and dried under vacuum to remove any residual solvent.

Characterization techniques

Comprehensive physicochemical and structural characterization of the synthesized g-C₃N₄, ZIF-8, and ZGCN nanohybrids, as well as the fabricated membranes, was performed using multiple analytical techniques. The crystalline structures of the materials were analyzed by X-ray diffraction (XRD) using a Bruker D8 Advance diffractometer with Cu Kα radiation (λ = 1.5406 Å), operated at 40 kV and 40 mA in the 2θ range of 5°–80° with a scanning rate of 2°/min. Fourier-transform infrared (FTIR) spectra were recorded using a PerkinElmer Spectrum Two FTIR spectrometer in the range of 4000–400 cm⁻¹ via the KBr pellet method to identify functional groups and chemical interactions. The surface morphology and cross-sectional structures of the membranes were observed using scanning electron microscopy (SEM) performed on a JEOL JSM-7610 F field emission SEM operated at 5–10 kV, with membrane samples sputter-coated with a thin layer of gold under vacuum prior to imaging. Brunauer–Emmett–Teller (BET) surface area analysis was carried out using a Micromeritics ASAP 2020 system under liquid nitrogen adsorption conditions. Specific surface area, pore volume, and pore size distribution were derived from N₂ adsorption–desorption isotherms measured at −196 °C, after degassing samples at 150 °C for 6 h. Thermal stability of the materials was evaluated by thermogravimetric analysis (TGA) using a TA Instruments Q50 TGA analyzer. The samples were heated from room temperature to 800 °C at a heating rate of 10 °C/min under a nitrogen atmosphere with a flow rate of 50 mL/min. The hydrophilicity of membrane surfaces was determined via static water contact angle measurements using a Krüss DSA30 drop shape analyzer, where 5 µL of deionized water was deposited on membrane surfaces and the average contact angle was calculated from five measurements.

Membrane filtration performance (flux, rejection, and antifouling) was assessed using a dead-end stirred ultrafiltration cell (Sterlitech HP4750 model) with an effective membrane area of 14.6 cm², operated at 2 bar transmembrane pressure and room temperature. HA concentrations before and after filtration were estimated from UV–Vis absorbance measurements at 254 nm. This wavelength was selected because humic substances contain aromatic and conjugated chromophores that exhibit appreciable UV absorption in this region. UV254 is widely used as an operational optical surrogate for aromatic dissolved organic matter and humic-substance content in water-treatment and membrane filtration studies [27, 28]. In the present work, a defined synthetic HA sodium-salt solution was used as the sole organic foulant; therefore, UV254 was used as an indirect measure of HA concentration for comparing the feed and permeate solutions under controlled experimental conditions. It should be noted that UV254 is not specific exclusively to humic acids, as other UV-absorbing organic compounds may contribute to the measured absorbance. Thus, the UV254-based values reported here should be interpreted as an operational measure of HA concentration within the model-foulant system.

Application process: humic acid separation test

The separation performance of the fabricated membranes was evaluated through the filtration of humic acid (HA) solution under controlled conditions using a dead-end stirred ultrafiltration setup (Sterlitech HP4750 cell). Each membrane sample was cut into circular discs with an effective surface area of 14.6 cm² and pre-compacted with deionized (DI) water at 2 bar pressure for 30 min to stabilize the membrane structure and eliminate compaction effects during testing. Following compaction, the membranes were subjected to HA separation experiments using a synthetic humic acid solution prepared by dissolving 20 mg of HA sodium salt into 1 L of DI water, resulting in a concentration of 20 mg/L, adjusted to pH 7.0 ± 0.2 using 0.01 mol dm⁻³ NaOH or 0.01 mol dm⁻³ HCl, as needed.

Filtration tests were conducted at a constant transmembrane pressure of 2 bar and room temperature (25 ± 2 °C), with continuous stirring at 300 rpm to minimize concentration polarization. The permeate was collected at timed intervals for 4 h, and the instantaneous flux (J) was calculated using the equation:

1

where J is the flux in L·m⁻²·h⁻¹ (LMH), V is the permeate volume (L), A is the membrane effective area (m²), and t is the filtration time (h). The HA rejection (R %) was determined using UV–Vis spectrophotometry (Shimadzu UV-1800) by measuring the absorbance at 254 nm before and after filtration, and calculated as:

 

2

where C_f and C_p are the feed and permeate HA concentrations (mg/L), respectively.

After each test, the membrane was washed gently with DI water and subjected to a cleaning cycle, after which the flux recovery ratio (FRR %) was calculated to assess antifouling performance. FRR was derived from the ratio of the recovered pure water flux (after cleaning) to the initial pure water flux, expressed as a percentage. All measurements were repeated three times using freshly prepared HA solutions and independent membrane samples to ensure reproducibility, and the mean values were reported with standard deviations.

Statistical modeling and data analysis for membrane performance evaluation

To quantitatively assess and predict the behavior of the ZGCN–PVDF membranes during humic acid separation, statistical modeling was applied to describe the flux decline kinetics, rejection efficiency, and antifouling performance. Nonlinear regression and error analysis were carried out using IBM SPSS Statistics 26 to validate data consistency and model accuracy. Three principal models were utilized to describe flux behavior and fouling mechanisms: the exponential decay model, the Hermia fouling models, and the pseudo-second-order kinetic model for flux loss.

Flux decay modeling (exponential model)

The permeate flux decline over time was fitted using an exponential decay function:

3

where: J(t) is the flux at time t (LMH), J₀is the initial flux (LMH), k is the fouling rate constant (h⁻¹), and J ∞ is the steady-state flux (LMH).

Hermia’s fouling model analysis

To identify the dominant fouling mechanism, Hermia’s classical models were applied under constant pressure conditions using the following linearized forms [29]:

• Complete pore blocking:

4

• Intermediate blocking:

5

• Standard (pore constriction):

6

• Cake filtration:

7

Kinetic modeling of rejection efficiency (pseudo-second-order model)

To model HA rejection dynamics over time, the pseudo-second-order model was used:

8

where: q_t is the rejection at time t, q_e is the equilibrium rejection value (%) that the system approaches over time, and k₂ is the rate constant (min⁻¹), which determines the rate at which the rejection changes over time as the filtration progresses.

Error function and model fit validation

To confirm model reliability, multiple error functions were computed: Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), Chi-square test (χ²) and R² (coefficient of determination).

Reproducibility and statistical significance

All experimental trials were repeated in triplicate, and standard deviation (SD) and coefficient of variation were calculated.

Results and discussion

Characterization of nanomaterials and membranes

The crystalline structure of the fabricated materials was first evaluated by X-ray diffraction (XRD). As shown in Fig. 2a, the XRD pattern of the pure PVDF membrane displays signature peaks which confirm the arrangement of its crystalline structure. The major peak at a vicinity of 2θ = 19.9° indicates the presence of a crystalline phase, probably related to the α-phase of PVDF [30]. Such a phenomenon is typical for PVDF, as it is a material that can crystallize in a variety of phases with the α-phase being the one that is most stable at room temperature. The ZIF-8 material showed a few intense peaks at 7.3°, 10.4°, 12.7°, and 18.0°, which are characteristic for its sodalite structure [31]. The pattern of XRD for pure g-C₃N₄ is characterized by two broad peaks around 13.1° and 27.4° that were identified as the (100) and (002) planes, referring to the in-plane structural packing and interlayer stacking, respectively [32]. The XRD patterns of ZGCN reveals distinct features characteristic of both materials. In ZGCN, the g-C₃N₄ phase is marked by a single distinct peak at roughly 27° (2θ) that represents the (002) plane, the ordered crystalline structure of g-C₃N₄. The occurrence of this peak, which indicates the interlayer spacing, demonstrates its stability. Conversely, the ZIF-8@g-C₃N₄ XRD pattern illustrates many peaks as one can find from ZIF-8. These include the ones at 7.3°, 10.4°, 12.7°, and 18.0°. The peaks are used to both verify that the ZIF-8 has been incorporated with the g-C₃N₄ and confirm the crystalline nature of the two materials.

Fig. 2.

a XRD patterns, b N₂ adsorption–desorption isotherms, c FTIR spectra and d TGA of PVDF, ZIF-8, g-C₃N₄, ZGCN, and ZGCN-PVDF membrane

The ZGCN-PVDF membrane XRD pattern indicates that the features mostly of the ZGCN and PVDF phases are present. It is confirmed from the data that the crystal structures of the two materials can be identified with both ZGCN and PVDF exhibiting their characteristic diffraction features. This means that the crystallinity of the two materials is not compromised by the mixing of ZGCN and the membrane has the same structural properties as the pure components.

Figure 2b illustrates the BET surface area analysis of PVDF, ZIF-8, g-C₃N₄, ZGCN, and the ZGCN-PVDF membrane. The ZGCN nanohybrid revealed a significant increase in surface area, with a measured BET surface area of 987 m²/g, much higher than that of pure g-C₃N₄ (61 m²/g) and ZIF-8 alone (1620 m²/g). The ZGCN–PVDF membrane showed a substantial surface area of 640 m²/g, which is significantly higher than the pure PVDF membrane (51 m²/g). The nitrogen adsorption–desorption isotherms exhibited Type IV behavior with an H1 hysteresis loop, confirming the mesoporous structure of the materials. Pore size distribution analysis of all samples indicated a dominant mesoporous region between 2 and 20 nm, aligning with the size exclusion properties necessary for effective humic acid (HA) removal. The increased surface area of the ZGCN nanohybrid, and ZGCN–PVDF membrane provides more adsorption and interaction sites, enhancing fouling resistance and HA rejection performance.

The IR spectrum of pure PVDF membranes (Fig. 2c) reveals the polymer’s molecular structure and functional attributes. The C-H bending vibration shows up as a strong peak around 1453 cm^− 1, and the C-F stretching vibration shows up as a strong peak around 1284 cm^− 1. This shows that the PVDF backbone has strong bonds between fluorine and carbon [33]. Additionally, there is also a clear peak at 840 cm^− 1 that is linked to the β-phase crystalline structure of PVDF. This structure is very important for its piezoelectric and electroactive properties [34].

The FTIR spectra of ZIF-8 (Fig. 2c) displays the principal vibrational characteristics anticipated for the ZIF-8 structure. The wide absorption band around 3400 cm^− 1 could be due to an O-H stretching vibration caused by water molecules that may have adsorbed or by a few hydroxyl groups that may have been left on the surface of the material [35]. The peak at about 1550 cm^− 1 is due to a C = N stretching vibration of the imidazolate linkers that come from the structure of ZIF-8 [36].

The FTIR spectrum of g-C₃N₄ (Fig. 2c) has a number of distinct peaks. The main peaks in the range of 3200 to 3500 cm^− 1 could be linked to the N-H stretching vibration, which could be caused by amine that is bound to the surface. The strong peaks at 1200 cm^− 1 and 1600 cm^− 1 are due to C-N stretching and C = N stretching vibrations, which show that there are triazine rings in the g-C₃N₄ structure that are the main chains [37]. The bands from 800 to 1200 cm^− 1 are the in-plane bending vibrations of C-H and C-N bonds, which are what make up the structure of graphitic carbon nitride.

The FTIR spectra of the ZGCN (Fig. 2c) composite reveal significant characteristics of both materials. ZIF-8 has a wide band around 3400 cm⁻¹ from O–H stretching, which is caused by water or hydroxyls on the surface, and a peak near 1550 cm⁻¹ from C = N stretching in the imidazolate linkers [38]. g-C₃N₄ has N–H stretching peaks between 3200 and 3500 cm⁻¹, strong C–N and C = N stretching peaks at 1200 and 1600 cm⁻¹, which show that there are triazine rings [37], and bands from 800 to 1200 cm⁻¹ that show that C–H and C–N bonds are bending in the plane [39]. These confirm that the composite has both ZIF-8 and g-C₃N₄ structures.

The IR spectrum of the ZGCN@PVDF membrane (Fig. 2c) shows the characteristic peaks of the pure components PVDF, ZIF-8, and g-C₃N₄ indicating that each material retains its molecular structure within the composite. This confirms the successful integration of all three materials without significant structural changes, validating the formation of a well-defined composite.

Thermal stability of the materials was assessed using thermogravimetric analysis (TGA), and the data are presented in Fig. 2d. The TGA curve of the PVDF membrane shows that it doesn’t lose much weight until it reaches about 300 °C, which means it is stable at lower temperatures. After this point, the polymer starts to break down, which means that it gets worse as the temperature rises. This stepwise thermal degradation shows that the membrane is stable up to 300 °C, after which it loses its integrity. This is important for its use in high-temperature situations.

The thermal stability and decomposition behavior of ZIF-8 (Fig. 2d) are demonstrated by its TGA curve. The curve is initially flat, signifying stability. A dramatic drop in weight occurs as the temperature rises (usually between 300 and 500 °C), indicating decomposition brought on by the loss of volatile components. The curve then flattens, indicating stabilization.

The graph (Fig. 2d) shows a significant weight loss at temperatures between 400 and 500 °C, which is normal for g-C₃N₄ because of the structure’s breakdown, which most likely involves the loss of volatile components or nitrogen-containing groups. Due to the material’s instability at high temperatures, this quick weight loss suggests that g-C₃N₄ decomposes significantly at higher temperatures. The slope of the curve between 400 °C and 600 °C indicates a significant mass change, which could be caused by the material’s structural change or the loss of amorphous carbon. After temperatures surpass 600 °C, the plot stabilizes, signifying that the majority of the volatile components have been liberated and a more thermally stable residue has been left behind.

ZGCN’s TGA (Fig. 2d) provides important thermal stability information for the composite material. The breakdown of ZIF-8 and the release of volatile components from the g-C₃N₄ matrix are responsible for the significant weight loss in the graph that occurs in the temperature range of roughly 350–450 °C. Given that the organic linkers in the ZIF-8 framework are known to break down at moderate temperatures, the abrupt drop signifies the material’s structural breakdown. The curve plateaus as the temperature rises above 500 °C, indicating that the composite material has achieved thermal stability. This behavior is common in hybrid systems, where the overall decomposition process is influenced by the stability of each component (in this case, g-C₃N₄ and ZIF-8). The presence of more thermally stable components, possibly including the g-C₃N₄ framework, is indicated by the last residue that remains.

The ZGCN-PVDF membrane’s TGA curve (Fig. 2d) shows a number of important thermal behaviors. Below 100 °C, there is initially a slight weight loss, most likely as a result of the volatile components and absorbed moisture being removed. Up to about 350 °C, the membrane remains stable; after that, a slow weight loss starts. This implies that the ZGCN composite and PVDF matrix have undergone thermal degradation, with the latter aiding in the decomposition process as a whole. Between 350 °C and 500 °C, the most significant weight loss happens. This temperature range is typical for the breakdown of polymeric materials like PVDF and may also be where ZGCN breaks down at high temperatures. The notable decrease at about 500 °C suggests that the bonds between PVDF and the composite are weakening or breaking, which may indicate the membrane’s thermal limit.

Water contact angle measurements were used to assess the membrane surface’s hydrophilicity; the findings are compiled in Table 1. With a comparatively high contact angle of 87.6°, the pristine PVDF membrane demonstrated its inherent hydrophobicity. The contact angle dropped to 62.3° after adding 1 weight% ZGCN, indicating a marked increase in hydrophilicity. This improvement is probably the result of the ZGCN hybrid’s high surface energy and polar functional groups, which improve water affinity and increase water flux while lowering fouling.

Table 1.

Water contact angle values of pristine and ZGCN–modified PVDF membranes indicating enhanced surface hydrophilicity

Membrane type	Water Contact Angle (°)
Pristine PVDF	87.6
PVDF +15 wt% ZGCN	62.3

Morphology of PVDF membrane, ZIF-8, g-C₃N₄, ZGCN and ZGCN-PVDF membrane

The surface of the PVDF membrane has a porous, uneven structure as shown in Fig. 3a. This kind of uneven pore shape makes it likely that the pores were made in a way that makes both round and long, column-like holes. Some pores also look like they have dents in them, which make a network of pores that are connected to each other. Porous membranes often have irregular pore shapes, which give them longer flow paths and better particle trapping ability. This makes them better for filtering applications. The membranes are better at keeping out particles of different sizes and properties because the surface is rough and the pores are not perfectly round.

Fig. 3.

SEM of a PVDF, b ZIF-8, c g-C₃N₄, d ZGCN, ande ZGCN-PVDF membrane

The SEM of ZIF-8 (Fig. 3b) shows crystalline particles that are clearly defined and mostly hexagonal and polyhedral in shape. The framework structure of ZIF-8 gives subparticular aggregates their core-shell structure and sharp edges. While the SEM image of g-C₃N₄ (Fig. 3c) displays angular, faceted particles with a crystalline structure. Despite the presence of some smaller fragments, the size of these polyhedral particles is generally uniform. The material’s surface area and reactivity are increased by the sharp edges and clustered arrangement, which imply a self-assembled structure.

ZGCN composites’ SEM image (Fig. 3d) reveals a highly porous structure with interconnected networks of tiny particles. The ZIF-8 component, which is renowned for its microporous qualities, is represented by the surface’s network of voids and cavities. Some of the pores seem to be filled in and coated by the g-C₃N₄, suggesting that the two materials have successfully integrated.

The modified membrane (Fig. 3e) showed a more open porous structure with visible clusters of nanoparticles dispersed throughout the cross-section after ZGCN nanoparticles were added. The nucleation effect of ZGCN and increased thermodynamic instability during phase inversion are responsible for these morphological changes, which encourage the formation of macrovoids and increased porosity. Improved water permeability may be a result of the modified membranes’ increased pore density and roughness, as shown by surface images.

Application performance: humic acid separation

The filtration performance of the fabricated membranes was evaluated using a synthetic humic acid (HA) solution (20 mg/L) under a constant pressure of 2 bars. As shown in Fig. 4, the pristine PVDF membrane exhibited an initial pure water flux of 201.3 L·m⁻²·h⁻¹, which decreased sharply to 110.7 L·m⁻²·h⁻¹ after 4 h of continuous HA filtration. In contrast, the ZGCN–PVDF membrane demonstrated a significantly higher initial flux of 318.5 L·m⁻²·h⁻¹, which stabilized at 172.8 L·m⁻²·h⁻¹ after 4 h. This enhanced flux is attributed to the improved membrane porosity and hydrophilicity induced by the ZIF-8@g-C₃N₄ hybrid. The open pore structure facilitated rapid water transport, while the hydrophilic surface reduced hydraulic resistance during operation. The relatively slow flux decline in the modified membrane indicates reduced fouling tendency, consistent with the synergistic effects of g-C₃N₄’s antifouling behavior and ZIF-8’s molecular sieving function.

Fig. 4.

Time-dependent permeate flux profiles of pristine PVDF and ZGCN–PVDF membranes during 4-h humic acid filtration at 2 bar pressure

The rejection efficiency of the membranes was assessed by measuring HA concentrations in the feed and permeate streams using UV–Vis spectroscopy at 254 nm. As presented in Fig. 5, the pristine PVDF membrane achieved an average HA rejection of 76.4%, which is limited by its relatively large pores and low surface affinity for HA molecules. On the other hand, the ZGCN–PVDF membrane exhibited significantly higher HA removal, with an average rejection of 91.6%. This improvement is attributed to the dual action of physical sieving by the ZIF-8 framework and electrostatic interaction between the negatively charged HA molecules and the ZGCN-functionalized membrane surface. The presence of polar nitrogen-containing groups in g-C₃N₄ may also contribute to stronger adsorption interactions.

Fig. 5.

HA rejection efficiency (%) of pristine PVDF and ZGCN–PVDF membranes under constant pressure filtration

These results confirm the enhanced selectivity and potential for NOM removal using ZGCN-functionalized membranes. Antifouling performance was evaluated based on flux recovery ratio (FRR) after a physical rinsing step with deionized water. The pristine PVDF membrane exhibited an FRR of only 63.4%, indicating severe irreversible fouling. In contrast, the ZGCN–PVDF membrane demonstrated a significantly improved FRR of 89.2%, as shown in Table 2. The enhanced surface hydrophilicity that lowers the initial adhesion of HA and the photocatalytic activity of g-C₃N₄ under ambient light conditions which aids in the degradation of adsorbed foulants account for the superior antifouling behavior. These properties are critical for enhancing long-term operational stability and reducing the frequency of membrane cleaning cycles in practical applications.

Table 2.

Flux recovery ratio (FRR) and antifouling performance comparison between pristine and ZGCN–modified PVDF membranes

Membrane type	Flux recovery ratio (FRR) (%)	Antifouling performance
Pristine PVDF	63.4	Severe irreversible fouling, low recovery due to large pores and low surface affinity for HA.
ZGCN–PVDF	89.2	Superior antifouling due to photocatalytic activity and enhanced hydrophilicity, leading to higher recovery and reduced fouling.

It should be noted that the humic acid sodium salt employed in this study represents a standardized model foulant and does not fully reproduce the complexity of natural organic matter (NOM) in environmental waters. Natural NOM is a heterogeneous mixture that may contain humic and fulvic substances and other organic components with different molecular characteristics and fouling behavior. This model compound remains widely used in membrane fouling studies, which allowed the present results to be evaluated under controlled conditions and compared with prior literature. Therefore, the rejection and antifouling results obtained in the present study should be interpreted within the context of the model-foulant system used. Although the standardized humic acid model is useful for controlled membrane evaluation, further investigation using environmentally relevant NOM would provide a more comprehensive assessment of membrane performance under realistic water-treatment conditions.

Statistical modeling and data analysis for membrane performance evaluation

Flux decay modeling (exponential model)

The exponential flux decay model provided an excellent fit for both membranes (Table 3). For pristine PVDF, the model gave an initial flux of about 189.9 LMH with a fouling rate constant of 0.148 h⁻¹ (R² = 0.942). For the ZGCN–PVDF membrane, the initial flux increased significantly to 332.7 LMH, with a slightly higher fouling constant of 0.157 h⁻¹ and an even stronger model fit (R² = 0.991). These results highlight that ZGCN incorporation greatly improves the starting permeability while maintaining a similar exponential fouling trend as pristine PVDF.

Table 3.

Exponential flux decay model results of PVDF and ZGCN-PVDF membrane

Membrane	Intercept (a)	Time Coefficient (b)	R²	Equation (ln J)	Initial Flux (J₀, LMH)	Decay Constant (k, h⁻¹)	Final Model (J)
PVDF	5.245	−0.148	0.942	ln(J) = 5.245–0.148·t	189.9	0.148	J(t) = 189.9·e⁻⁰·¹⁴⁸t
ZGCN–PVDF membrane	5.808	−0.157	0.991	ln(J) = 5.808–0.157·t	332.7	0.157	J(t) = 332.7·e⁻⁰·¹⁵⁷t

Hermia’s fouling model analysis

Table 4; Fig. 6 represents the application of Hermia’s fouling models to the flux decline data of pristine PVDF and ZGCN–PVDF membranes. For both membranes, the intermediate blocking model provided the best fit, with high correlation coefficients (R² = 0.969 for PVDF and R² = 0.984 for ZGCN–PVDF). This indicates that in both cases, fouling was dominated by partial pore blockage and gradual constriction rather than complete blocking or extensive cake layer formation.

Table 4.

Fouling model parameters and regression coefficients for pristine PVDF and ZGCN–PVDF membranes

Membrane	Complete pore blocking		intermediate blocking model		Standard (pore constriction)		Cake filtration
	R 2	Kcb (h− 1)	R 2	Kib (h− 1)	R 2	Ksb (h− 1)	R 2	Kcl (h− 1)
PVDF	0.939	0.149	0.969	0.00102	0.939	0.149	0.957	1.44 × 10− 5
ZGCN-PVDF	0.969	0.152	0.984	0.00066	0.969	0.152	0.910	5.92 × 10− 6

Fig. 6.

a Complete pore blocking, b intermediate blocking model, c Standard (pore constriction) and d Cake filtration of PVDF and ZGCN-PVDF membrane

The slightly lower K_ib value for the ZGCN–PVDF membrane (0.00066 h^− 1) compared to PVDF (0.00102 h^− 1) suggests a slower fouling rate, confirming the beneficial effect of ZGCN incorporation in mitigating pore blockage. Although the cake filtration and complete blocking models showed moderate correlation, their R² values were consistently lower, reinforcing that the intermediate blocking mechanism was predominant for both membranes.

Kinetic modeling of rejection efficiency (pseudo-second-order model)

The rejection data for both pristine PVDF and ZGCN–PVDF membranes were analyzed using the pseudo-second-order (PSO) model (Fig. 7). A linear relationship was obtained when plotting t/q_t against t, with both membranes showing excellent agreement with the model (R² =0.999). For the pristine PVDF membrane, the equilibrium rejection capacity (q_e) was calculated as 74.6% with a rate constant (k₂) of 0.0134 min⁻¹. In contrast, the ZGCN–PVDF membrane achieved a higher qe of 91.0% and a k₂ value of 0.0048 min⁻¹. These findings confirm that the rejection of humic acid follows chemisorption-controlled kinetics, where surface interactions dominate the process. The enhanced q_e observed for the ZGCN–PVDF membrane highlights its improved ability to retain humic acid compared to pristine PVDF, reflecting the beneficial role of ZGCN incorporation in boosting rejection performance while maintaining strong kinetic agreement with the PSO model.

Fig. 7.

Pseudo second order of PVDF and ZGCN-PVDF membrane

Error function and model fit validation

The residual analysis also showed the RMSE of the ZGCN–PVDF composite membrane to be 5.76 and that of the pure PVDF membrane to be 4.87 (Table 5). We computed the average of the square of the residuals (23.74 for PVDF and 33.18 for ZGCN–PVDF) and then took their square roots to arrive at these values. The overall magnitude of the observed vs. predicted flux value difference is what the RMSE actually measures. The low RMSE values show that the fouling model used is good and can successfully show how the flux changes for both membranes in the experiments. That of the ZGCN–PVDF membrane was somewhat higher, but the difference was still low enough to be acceptable. This shows that the model could successfully show how the rejection and fouling rates changed in both cases. The model was even more dependable as the mean absolute percentage error (MAPE) for pure PVDF membranes was 2.86% and for ZGCN–PVDF membranes was 2.35%. These results showed even more clearly that there is a strong correlation between experimental and predicted values.

Table 5.

Descriptive statistics of squared residuals for PVDF and ZGCN–PVDF membranes (RMSE calculation)

Descriptive statistics
	N	Minimum	Maximum	Mean	Std. Deviation
PVDF	5	0.00	73.96	23.7380	31.01111
ZGCNPVDF	5	0.00	77.44	33.1760	28.30532

The chi-square (χ²) test also indicated that the intermediate blocking model was an effective predictor, and the differences between observed and predicted flux were minimal (0.83 for PVDF and 0.79 for ZGCN–PVDF). The modified membrane is stronger and will not be contaminated as easily. The χ² value of ZGCN–PVDF membrane is lower than the other membranes. This shows that it follows the model better. The models eventually explained almost all the variation in experimental data. R² values for HA rejection and flux decline were 0.991 and 0.984, respectively. In total, these statistical tests reveal that the fouling and rejection models are capable of predicting well and work satisfactorily for both clean membranes and treated membranes.

Reproducibility and statistical significance

One-way ANOVA was used to test statistical significance and reproducibility of flux performance (Table 6). Pure PVDF membranes reported a mean flux of 144.4 LMH, while the ZGCN–PVDF membranes reported a significantly higher mean flux of 249.3 LMH. Levene’s test for homogeneity of variances was not significant (p = 0.206), confirming equal variances assumption. The ANOVA analysis showed there was a statistically significant difference between the two membranes (F(1,8) = 11.11, p = 0.010), and it indicates that the application of ZGCN showed a great improvement in the permeate flux performance. This confirms that the improvement found here is not by chance but is a statistically significant effect of the membrane modification.

Table 6.

One-way ANOVA for flux (PVDF and ZGCN–PVDF)

Test/Group	Mean (LMH)	Std. deviation	F (df = 1,8)	p-value
Pristine PVDF	144.40	36.81
ZGCN–PVDF	249.26	59.95
Levene’s Test				0.206
ANOVA (Flux)			11.11	0.010

The effect of ZGCN incorporation on HA rejection was statistically assessed using one-way ANOVA (Table 7). The pristine PVDF membranes achieved an average rejection of 74.2%, while the ZGCN–PVDF membranes showed a significantly higher average rejection of 90.5%. Levene’s test for homogeneity of variances was not significant (p = 0.178), confirming equal variance between groups. The ANOVA test demonstrated a highly significant difference between the two membrane types (F(1,8) = 425.29, p < 0.001), clearly indicating that ZGCN incorporation substantially improved HA rejection efficiency. These results confirm that the performance enhancement is statistically robust and not due to experimental variability.

Table 7.

One-Way ANOVA for HA rejection (PVDF and ZGCN–PVDF)

Test/Group	Mean (%)	Std. deviation	F (df = 1,8)	p-value
Pristine PVDF	74.18	1.57
ZGCN–PVDF	90.52	0.82
Levene’s Test				0.178
ANOVA (Rejection)			425.29	< 0.001

Discussion of mechanism

The significant enhancement in both flux and rejection efficiency observed for the ZGCN–PVDF membranes (Fig. 8) can be attributed to the synergistic mechanism arising from the structural and functional integration of ZIF-8 and g-C₃N₄ within the polymer matrix. The modification of the membrane surface and substructure with the ZGCN nanohybrid leads to simultaneous improvements in porosity, hydrophilicity, and interfacial interaction with target contaminants, ultimately optimizing membrane performance.

Fig. 8.

Mechansim of ZGCN–PVDF membranes

Firstly, the incorporation of ZIF-8 contributes primarily to size-selective molecular sieving. With a pore aperture of approximately 3.4 Å, ZIF-8 imparts a selective barrier that enhances humic acid rejection while allowing water molecules to pass through efficiently. Its highly crystalline and porous structure increases the effective filtration area and reduces resistance to water transport, contributing to the higher initial flux observed in modified membranes. Moreover, the presence of ZIF-8 nanoparticles promotes macrovoid formation during the phase inversion process due to their nucleation effect, as confirmed by SEM imaging, which further supports the observed permeability enhancement.

Secondly, g-C₃N₄ provides unique benefits associated with its photocatalytic and surface-active properties. Its polar nitrogen-rich structure enhances membrane hydrophilicity, as reflected by the substantial reduction in water contact angle measurements (from 87.6° to 62.3°). This increased surface wettability improves water uptake and decreases the likelihood of hydrophobic fouling. Additionally, g-C₃N₄ acts as a photocatalyst under visible light, generating reactive oxygen species (ROS) that can degrade adsorbed organic matter on the membrane surface. Although photocatalytic tests were not the focus of this work, the passive antifouling behavior inferred from high flux recovery ratios (FRR) is consistent with this intrinsic self-cleaning capacity.

The combined ZIF-8@g-C₃N₄ nanohybrid ensures a uniform distribution of active sites and promotes interfacial compatibility within the PVDF matrix. This prevents nanoparticle agglomeration and maintains the mechanical integrity of the membrane. FTIR and XRD results confirm that the hybridization of the two components does not lead to chemical degradation but rather preserves their individual functionalities in a physically integrated structure.

Fouling resistance and long-term operational stability are governed by both surface characteristics and molecular interactions with foulants. The reduction in irreversible fouling, demonstrated by a high FRR of 89.2%, suggests that the ZGCN-modified surface exhibits weaker interactions with HA molecules. This could be due to the increased negative surface charge and reduced hydrophobic interaction potential, both of which hinder the adhesion of negatively charged humic acid macromolecules. The Hermia model analysis further supports this conclusion by identifying intermediate blocking as the dominant fouling mechanism, indicating that foulant adsorption is minimized and does not result in complete pore occlusion.

In summary, the ZGCN nanohybrid facilitates a multi-functional modification strategy that simultaneously addresses permeability, selectivity, and fouling—three critical performance bottlenecks in ultrafiltration membranes. By carefully balancing the structural features of MOFs and the surface activity of g-C₃N₄, the resulting membranes achieve superior performance in HA separation, offering a promising pathway for sustainable water treatment technologies.

Summary of membrane performance and discussion

Table 8 shows a full comparison of how well ZGCN–PVDF and pristine PVDF membranes work. The ZGCN-modified membrane rejected HA almost 20% better (91.6% vs. 76.4%) and had an initial water flux that was 58.2% higher (318.5 LMH vs. 201.3 LMH). Improved hydrophilicity is directly related to higher water permeability and lower fouling susceptibility, as shown by the contact angle dropping from 87.6° to 62.3°. The flux recovery ratio (FRR) also showed that the antifouling performance had gotten better; it went from 63.4% to 89.2%. The surface activity of g-C₃N₄ and the molecular sieving of ZIF-8 work together to improve water transport while keeping foulants from sticking. The ZGCN–PVDF membrane is the best choice for removing humic acid from water because it has the perfect balance of permeability, selectivity, and resistance to fouling.

Table 8.

Comparative performance of pristine and ZGCN–modified PVDF membranes in HA separation

Parameter	Pristine PVDF membrane	ZGCN–PVDF membrane
Initial pure water flux (LMH)	201.3 ± 4.5	318.5 ± 6.2
Final flux after 4 h (LMH)	110.7 ± 3.1	172.8 ± 4.7
HA rejection (%)	76.4 ± 2.2	91.6 ± 1.9
Water contact angle (°)	87.6 ± 1.7	62.3 ± 1.5
Flux recovery ratio (FRR, %)	63.4 ± 2.5	89.2 ± 2.1

Comparison with recently reported PVDF/MOF-based membranes

The performance of the ZGCN-PVDF membrane was compared with recently reported PVDF/MOF-based membranes (Table 9). The ZGCN-PVDF membrane exhibited a relatively high water flux of 318.5 LMH, exceeding those of PVDF/Ag@MOF-0.5 (236.5 LMH), and PVDF/MOF-199 (187.75 LMH). It achieved 91.6% HA rejection and 89.2% FRR at only 2 bar, demonstrating a favorable balance between permeability, separation efficiency, and antifouling performance. Although some reported membranes showed higher HA rejection or FRR, the ZGCN-PVDF membrane offers competitive overall performance at a relatively low operating pressure. These results confirm the potential of the ZGCN nanohybrid for enhancing PVDF membranes for HA-containing water treatment.

Table 9.

Comparison of the ZIF-8@g-C₃N₄/PVDF (ZGCN-PVDF) membrane with recently reported PVDF/MOF-based membranes for humic acid removal

Membrane composition	Configuration/Modification	Water flux (LMH)	HA rejection (%)	FRR (%)	Operating pressure	References
PVDF/Ag@MOF-0.5	Ag-loaded MOF-5 (0.5 wt%) incorporated into PVDF by phase inversion	236.5 (PWF	96.5	92.4 (HA)	NR	[25]
PVDF/MOF-199 (3 wt%)	MOF-199 (Cu-BTC) blended into PVDF via phase inversion	187.75	94.3	96.76 (HA)	345 kPa (~3.45 bar)	[23]
PVDF/MOF-801 (2 wt%)	MOF-801 incorporated into PVDF membrane matrix	NR	88	85.2 (HA)	NR	[24]
ZIF-8@g-C₃N₄/PVDF (this work)	Direct blending of preformed ZGCN nanohybrid into PVDF by NIPS	318.5	91.6	89.2	2 bar	This work

Conclusion

This research effectively synthesized an innovative ZIF-8@g-C₃N₄ (ZGCN) hybrid nanomaterial and integrated it into PVDF membranes to enhance the elimination of humic acid (HA) from water. The membrane worked much better when ZGCN was added to the polymer matrix. This was because ZIF-8’s ability to sieve molecules combined with g-C₃N₄’s ability to repel water and keep things from sticking to it. The modified membrane had a much higher initial pure water flux (318.5 LMH) and a higher sustained flux after 4 h of use (172.8 LMH) than the pristine PVDF membrane (201.3 LMH and 110.7 LMH, respectively). The efficiency of HA rejection went up from 76.4% to 91.6%, and the water contact angle went down from 87.6° to 62.3°. This means that the material is more hydrophilic. The antifouling has gotten better because the flux recovery ratio went from 63.4% to 89.2%. Characterization results showed that ZIF-8 and g-C₃N₄ were successfully combined without changing their structural properties. The ZGCN-modified membrane showed consistent performance with great reproducibility and statistical reliability. Using Hermia’s fouling model showed that intermediate blocking was the main mechanism, which is in line with the hybrid membrane’s ability to resist irreversible fouling. In general, the new ZGCN–PVDF membrane is a promising way to get rid of natural organic matter (NOM) in water treatment systems. Future research should therefore focus on validation using IHSS Suwannee River humic and fulvic acid standards, as well as NOM isolated directly from representative natural water sources. Further studies should also investigate long-term stability under continuous-flow conditions, fabrication scale-up, and membrane performance in complex water matrices containing multicomponent foulants and microbial loads.

Author contributions

Ayman K. El-Sawaf conceived the study, designed the methodology, performed the experiments, analyzed the data, and contributed to writing – review & editing. Amal A. Nassar contributed to the methodology, supervised the research, validated the results, and contributed to writing – review & editing. A. Taha contributed to conceptualization, methodology, software, data interpretation, visualization, writing – original draft, and writing – review & editing, and served as the corresponding author. Mahmoud F. Mubarak contributed to conceptualization, methodology, software, experimental work, characterization, data analysis, validation of the results, writing – original draft, and writing – review & editing. All authors read and approved the final manuscript.

Funding

The authors extend their appreciation to PrinceSattam bin Abdulaziz University for funding this research work through project number (PSAU/2026/01/45511).

Data availability

Data will be made available by the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable. This article does not contain any studies involving human participants or animals performed by any of the authors.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.