Designing of a novel polyvinylidene fluoride/TiO₂/UiO-66-NH₂ membrane with photocatalytic antifouling properties using modified zirconium-based metal-organic framework

Population growth and industrialization lead to the increase of water consumption and exacerbation of water pollution (Shannon et al. 2008; Feng et al. 2016; Wei et al. 2019; Sabri et al. 2020). Large amounts of contaminated water produced from industry, agriculture, and households have raised a grave concern. Many techniques have been developed to purify and reuse the wastewater, among which (Gao et al. 2014; He et al. 2019) membrane-based ultrafiltration (UF) has been recognized as a safe, clean and economical method for separation of extensive contaminants in the wastewater. However, the membrane-based ultrafiltration is also associated with disadvantages, such as fouling effect caused by foulants in the wastewater and large amount of energy and water consumption required for clearing the foulants. Membrane modification by introducing foreign particles with photocatalysis is an efficient method to diminish the fouling effect (Hatat-Fraile et al. 2017; Zhang et al. 2018; Li et al. 2020). Despite the new problem of clogging of the membrane caused by foreign particles, the photocatalytic method remains environmentally and economically friendly for its sustainability in decomposing organic pollutants without generating chemical wastes. In summary, photocatalytic based membrane modification is superior in diminishing membrane fouling effect and improving membrane function (Wang et al. 2020).

TiO₂ has been widely used in photocatalytic systems owing to its high hydrophility, excellent light response and suitable energy band location, and highly active photocatalytic performance. However, the wide band gap (3.2 eV) property of TiO₂ made it only reactive to ultraviolet light, limiting its application. Some recent studies have reported the outperformance of MOFs (Metal-Organic Frameworks) materials in adsorption and photocatalysis (Liu et al. 2014; Kolesnyk et al. 2020; Zhang et al. 2020). Among many MOFs, UiO-66 is a newly developed photocatalyst material that can be used for the catalytic degradation of organic pollutants under ultraviolet and visible light. UiO-66-NH₂, a derivative of UiO-66, has large specific surface area, outstanding chemical and thermal stability, and satisfactory charge carrier ability, which made UiO-66-NH₂ a photo-generated charge carrier (Dong et al. 2010; Schaate et al. 2011; Wang et al. 2017; Li et al. 2019; Feng et al. 2020). In addition, UiO-66-NH₂ has a certain degree of hydrophilicity, which endows the membrane with anti-fouling property. Tran et al. (Tran & Chun-Cehieh 2020) designed a PVDF-g-PAA-TGO membrane by grafting TiO₂-GO onto a PVDF membrane and discovered that the membrane exhibited high photocatalytic activity, mainly because of the decreased recombination rate of electron-hole pairs under UV illumination. Kolesnyk et al. (2020) synthesized C₃N₄-modified membranes and revealed their anti-fouling property to various types of foulants. Being inspired by those findings, we proposed a complementary strategy – combining MOFs with TiO₂ could facilitate photo-generated electron transfer, enhance the photocatalytic activity and ultimately promote catalytic degradation of organics (Ma et al. 2017).

In this study, an ultrafiltration membrane system combining PVDF with TiO₂-modified zirconium-based MOF was established by using the phase inversion technique. The nanoporous MOFs were synthesized and embedded into the membrane (Chen et al. 2006; Liu et al. 2011; Li et al. 2014; Islam et al. 2015; Crock et al. 2017; Cao et al. 2020). Membrane characteristics such as properties of anti-fouling clogging and photocatalytic self-cleaning were measured. The anti-fouling performance of MOF-modified membrane in the purification of protein and dye wastewater was quantified (Venkatesh et al. 2016; Wu et al. 2016; Shaffer et al. 2019; Wang et al. 2019).

Major materials were: PVDF (Solef^® 87 6010, Solvay Advanced Polymers); L.L.C (Alpharetta GA, USA); RhB (Sinopharm Chemical Reagent Co.,Ltd); bovine serum albumin (BSA, MW = 58,000, Nanjing Oddfoni Biological Technology Co., Ltd, China). ZrCl₄(98%), 2-Aminoterephthalic Acid (98%), Polyvinylpyrrolidone (PVP, K30), HCl (36–38%, AR), TiO₂ (99.8%, AR), N,N-Dimethylacetamide (DMAc, AR), N,N-dimethylformamide (DMF, AR) and Ethanol (AR) were offered by Shanghai Titan Scientific Co., Ltd.

A solvothermal method was used. Typically, 1.0 mmol of ZrCl₄ and 1.5 mmol of 2-aminoterephthalic were added into 30 mL of DMF solution. After stirring for 30 min at room temperature, the solution was transferred into a 100 mL Teflon-lined autoclave for solvothermal treatment (353 K for 24 h). After cooling to ambient temperature, the slurry was centrifuged. Precipitates were washed with DMF and absolute ethanol 3 times, before drying (393 K for 12 h in a vacuum oven) to be UiO-66-NH₂ product.

TiUN was synthesized using a similar procedure described above, except replacing pure DMF with TiO₂ suspension.

The procedures for preparing PVDF/TiUN membrane are illustrated in Figure 1. For composing the membrane casting solution, PVDF (16 wt.%) and PVP (2 wt.%) were configured at the same percentage and the filler loadings were different. Casting solution constituents were mixed using a blending method, with 5 different combinations described in Table 1. The samples were stirred at 60 °C for 12 h to form a homogeneous solution. The casting solution was immediately casted onto a glass substrate and then immersed in a coagulation bath (deionized water with 0.5 g nanoparticles). The membrane was stored in deionized water after detaching from the glass substrate. To ensure complete phase separation, membranes were immersed into freshwater for around 24 h.

Scanning electron microscopy (SEM, S-3400N, Japan), total reflection Fourier infrared spectrometer (FTIR, VERTEX-70, Germany), X-ray diffractometer (XRD, D/max 2200PC, Japan) and X-ray photoelectron spectrometer (XPS, Escalab 250Xi, USA) were used to affirm the topography and composition of nanoparticles.

In this equation, M_w/M_d denotes wet/dry weights of the membrane; A denotes membrane effective area (m²); L is the membrane thickness (m) and ρ is water density (0.998 g/cm³).

In this equation, η denotes water viscosity (8.9 × 10⁻⁴ Pa·s), Q denotes permeated water volume per time unit (m³/s) and Δp denotes operation pressure (bar).

Q denotes permeation volume (L), A denotes membrane area (m²) and Δt denotes the permeation time (h).

C_P and C_F denote BSA concentrations (mg/L) in permeation and feeding solutions respectively.

In the photo-catalytic and adsorption experiment, different types of membrane (3 × 3 cm × 5 pieces for each type) were added to 100 mL RhB solution (10 mg/L, natural pH). For the photocatalytic experiment, membranes were irradiated with visible light in a photocatalytic device (long arc xenon lamp, 500 W; 25 °C; 2 h). The solution was kept in the dark for 30 min to reach adsorption equilibrium before irradiation. The RhB concentration in the feeding and permeation solution was determined by spectrophotometer at 554 nm (UV-vis spectroscopy, HACH, DR6000, USA).

The self-cleaning experiment was conducted in a similar configuration with the antifouling test described above. Each membrane was continuously pressurized for 30 min at 0.1 MPa. After that, RhB solution (300 mg·L⁻¹) was forced to permeate through the membrane at the same operating pressure. In the first round of tests, the fouled membrane was immersed in distilled water (15 min). In the second round of tests, the fouled membrane was placed in water and exposed to ultraviolet light for 15 min to degrade the residual RhB. The self-cleaning property of PVDF/TiUN membrane was evaluated by a separating test of RhB aqueous solution.

Measurements of functional groups and molecular structures by FTIR spectroscopy are shown in Figure 2(a). The FTIR spectrum measurements of UiO-66-NH₂ revealed the characteristic peaks at ∼3444, ∼1635, ∼1574, ∼1434 and ∼1384 cm⁻¹, which were corresponded respectively to N-H stretching, C=O stretching vibration, N-H bending vibration, N-H bending vibration and C-N stretching vibration. The characteristic peaks between 600 and 800 cm⁻¹ became weaker with the addition of TiO₂, which were attributed to the Zr-O₂ cluster (Islam et al. 2017; Ma et al. 2017). The FTIR spectrum of TiUN was similar to that of UiO-66-NH₂, indicating that the combination of TiO₂ has limited effect on the intrinsic property of MOFs.

The XRD pattern of the synthesized UiO-66-NH₂ shown in Figure 2(b) was consistent with previous reports, indicating that UiO-66-NH₂ was prepared with high quality. The reflexion angles at ∼7.4, ∼8.6, ∼14.6, ∼22.4, ∼25.6 correspond to (111), (200), (222), (400), (511) planes, respectively (Starck et al. 2012). In addition, the XRD pattern of TiUN and UiO-66-NH₂ matched well, no other distinct diffraction peak was found, indicating that the crystallinity and phase composition of MOF did not change with TiO₂ loading. We found no characteristic peaks for TiO₂, which could be attributed to its low concentration. In general, the result of XRD suggested that structures of UiO-66-NH₂ did not change during the synthesis process (Tambat et al. 2018).

As illustrated in Figure 3, the pristine UiO-66-NH₂ exhibited a smooth regular octahedron with particle sizes ranged from 20 to 50 nm. The particle size of nano-TiO₂ was ∼5 nm. TiO₂ nanoparticles were dispersed on the surface of UiO-66-NH₂ (Li et al.; Zhang et al. 2019). We also found that the size of TiUN was uniform, which may be beneficial to acquiring high-performance modified membranes.

Figure 4(a)–4(c) illustrate the XPS patterns of the PVDF, PVDF/UN and PVDF/TiUN membranes, in which the N 1s and Zr 3d signal the UiO-66-NH₂ in the PVDF/TiUN, which is consistent with a previous report (Li et al.; Rajati et al. 2019). In comparison with PVDF/UN membrane, PVDF/TiUN membrane showed new signals of Ti 2p. In the modified membranes, a peak signaling O-Ti was observed at 530.08 eV (Liu et al. 2018a), which illustrates that TiUN had been mixed with the PVDF membrane successfully.

We further tested the chemical composition of the prepared membranes by FTIR. Functional groups in the PVDF/UN and PVDF/TiUN membranes were analyzed with FTIR spectroscopy. Results are shown in Figure 4(d). In the spectrum band of PVDF membrane, peak 1,406 cm⁻¹ attributes to CF₂ and CH₂ deformation vibration. The peaks at 1,070 cm⁻¹ and 1,185 cm⁻¹ can be attributed to C-C and CF₂ stretching vibration, respectively. Peaks at 400–800 cm⁻¹ are attributed to the PVDF characteristic vibration. For the PVDF/UN and PVDF/TiUN membrane, the peak at 1,267 cm⁻¹ is the shear stretching of the unique C-N structure in aromatic amine, and the N-H shear stretching vibration in the absorption spectra appears at 1,455 cm⁻¹ (Shen et al. 2013; Ma et al. 2017; Liu et al. 2018b).

As shown in Figure 5(a), the cross-sectional images of the membranes showed cavity structures in the sub-layer and asymmetric structures within the dense skin layer. The porosity distribution of the modified membranes (M2, M3, M4 and M5) was more heterogeneous than bare membrane (M1). Moreover, the skin layer in the PVDF/TiUN membrane was thicker than pristine PVDF membrane, which is beneficial to substrate selectivity (Zheng et al. 2013). We also noted significant changes in structural texture when the membranes were loaded with different protocol (e.g. M3, M5). This observation could be explained by nanoporous filler not being able to disperse sufficiently in higher loading condition within casting solution. The distribution of nanoparticles in the membranes can be seen in Supplementary Figure 1.

The surface properties of PVDF, PVDF/UN and PVDF/TiUN membranes were measured by AFM. As shown in Figure 5(b), highly rough surfaces provided a foulant trap, thus causing membrane fouling and reducing flux. As summarized in Table 2, the unmodified membrane had a relatively smooth surface with an average roughness of 15.1 nm. Nanoparticle addition decreased surface smoothness of both PVDF/UN and PVDF/TiUN membrane. We found that higher filler loading resulted in a higher roughness of membrane, see M3 & M5 (0.5 wt.%) v.s M2 and M4 (0.1 wt.%), for example. This phenomenon could be explained by, during phase inversion, MOF particles enriching onto the top layer of the membrane, particularly in higher filler loading condition (Dehghankar et al. 2020). The AFM results, FRR and fouling resistance data indicate that M4 (Sa = 9.1 nm) is the optimal membrane that we have configured (Figure 7(c)).

Hydrophilicity is the most important feature of a membrane and is crucial for flux and antifouling properties. We discovered that modified membranes showed higher hydrophilicity in comparison with the bare membrane, as indicated by smaller contact angles (Figure 7).

The high hydrophilicity of the membrane can be explained by two mechanisms. Firstly, the amine groups in MOF strongly attract water molecules through hydrogen bonding. Secondly, MOFs provide a hydration layer that rejects hydrophobic components on the surface of the modified membranes. This hydrophilicity endows the membrane with foulant resistance (antifouling improvement) (Shen et al. 2014). As expected, TiO₂ provides extra hydrophilic and endows modified-MOF membrane with smaller contact angles (compared to unmodified-MOF membranes). We noted that the hydrophilicity of modified-MOF membrane was positively correlated with the filler content of the matrix (in a range of 0.1–0.5 wt. %), as indicated by gradual reduction of contact angles.

Figure 7(a) shows the result of PWF measurements of the nanocomposite membranes. We found that the addition of modified-MOF improved membrane hydrophilicity and PWF. Addition of UiO-66-NH₂ (0.1 wt.%) increased PWF, compared with non-addition in the bare membrane (M4 vs. M1). We also noted that PWF decreased with the addition of the unmodified MOF. However, surprisingly, WCA increased as well (Figure 6). This result can be explained by short amine functional group in the MOFs tending to form into cages, because of rich hydrogen bonding. But the cage formation depends on disparity and/or homogeneity, such as agglomeration, of the fillers, which resulted in different flux properties of the MOF-based membrane (modified MOF vs. unmodified MOF) (Li et al.). Furthermore, during phase inversion, fillers with smaller size tend to migrate to the surface layer of the membrane, thus forming a dense texture in the top layer and porous texture in the sub layer. This mechanism could also explain the reduction of membrane flux (Bui et al. 2021).

The fouling behaviors of membranes were examined, and results are shown in Figure 7(b). To evaluate the impact of foulant on the membrane, in the first test we observed that lower additive loading (M2, M4) resulted in a steady reduction trend of foulant effect in the modified membrane, in comparison with the control membrane (M1). The second step was the BSA solution test. Working as a mimetic protein solution, BSA solution exhibited very limited flux for all membranes. In the third step, the recovering test, M4 and M5 possessed the highest PWF.

The dynamic fouling results (flux recovery ratio, FRR) for prepared membranes are summarized in Figure 7(c). As illustrated, membranes containing TiUN had higher FRR, reflecting higher reversible protein adsorption. Basically, this reversibility was attributed to the hydrophilicity of MMMs. Higher surface roughness resulted in depositing of protein molecules irreversibly into surface pores. Higher percentage of TiUN in the membrane caused an increment of surface roughness and thus lower FRR. M4 (0.1 wt% TiUN) sample showed the highest fouling resistance, which can be attributed to surface roughness and hydrophilicity.

Three parameters, reversible fouling (R_r), irreversible fouling (R_ir) and total fouling (R_t) were used to characterize antifouling properties of the membranes. R_r is the fouling part that could be washed out through hydraulic cleaning. R_ir is the fouling part that could not be removed with only hydraulic cleaning. R_t reflects the degree of total flux. Values for R_r and R_ir were shown in Figure 7(d). Lower R_ir value as well as higher FRR indicates better antifouling properties of the membranes. Unfilled PVDF membrane is less hydrophilic and with less surface charge, thus it had a great irreversible fouling ratio (26.88%, exceeding half of 47.67% of total fouling). Among the modified membranes, 0.5 wt.% addition of MOF showed the highest irreversible fouling property. Membrane with addition of 0.1 wt.% PVDF/TiUN had the highest FRR value (95.37%) and the lowest R_ir value (4.63%). In summary, surface roughness parameters, R_r, R_ir and FRR were improved for MOF modified membranes. Compared with the reported literature, the anti-pollution ability of M4 is higher (Bui et al. 2021).

The RhB degradation and adsorption profiles of M1 membrane, UiO-66-NH₂ membrane and TiUN membrane is shown in Supplementary Figure 2. As illustrated in Supplementary Figure 2a, both TiUN membrane and UN membrane exhibited adsorbability, while bare membrane showed poor adsorbability. This result could be explained by UiO-66-NH₂ providing π-π interaction, electrostatic interaction or hydrogen bonding, which is consistent with a previous study that reported robust ability in RhB separation (Zhang et al. 2019). TiUN membrane had good degradation efficiency on RhB while M1 membrane had almost no degradation efficiency. This can be attributed to the enhanced absorption of visible light provided by TiO₂ in the membrane. UiO-66-NH₂ has lowest empty molecular orbital and highest occupied molecular orbital, which can transfer electrons and holes and reduce the band gap width of TiO₂, thus expanding the spectral response range. The synergistic effect of TiO₂ and UiO-66-NH₂ enables TiO₂ to generate more photo-generated electrons (PGE) and holes (PGH), promotes the transfer of PGE and inhibits the recombination of PGE and PGH in light condition, resulting in better photocatalytic performance (Li et al. 2014; Wu et al. 2016; Shaffer et al. 2019). These experimental results demonstrated the outperformance of photocatalytic ability of the modified membrane.

The time-dependent fluxes of M1, M2 and M4 membranes for water and RhB aqueous solution were measured. The result is shown in Figure 8(a) and 8(b). We observed significant drop of flux for all membranes when water was replaced by RhB aqueous solution. This loss of flux is mainly attributed to the deposition or adsorption of RhB on the membrane surface/pores. The flux of all membranes was partly recovered to normal level after hydraulic cleaning. Meanwhile, M4 showed a complete recovery of flux after undergoing photocatalytic self-cleaning process, in contrast to limited recovery of M1 and M2 under the same conditioning. The superior photocatalytic self-cleaning property enables PVDF/TiUN membrane to be promising in water remediation application. The principle of the self-cleaning mechanism is shown in Supplementary Figure 4.

To quantify the separation ability of membranes, we examined the removal effect on RhB of PVDF membrane, PVDF/UN membrane and PVDF/TiUN membrane. As illustrated in Figure 8(c), the removal rate of PVDF/UN membrane and PVDF/TiUN membrane was higher than that of bare membrane. The result could be attributed to the screening and adsorption provided by UiO-66-NH₂ (Yang et al. 2019).

The antifouling property on RhB was also examined for bare membrane, PVDF/UN membrane and PVDF/TiUN membrane, under light exposure or dark conditions. As shown in Figure 8(d), in dark conditions, bare membrane showed higher FRR than PVDF/UN membrane and PVDF/TiUN membrane. However, in light exposure conditions, both PVDF/UN membrane and PVDF/TiUN membrane showed higher FRR than bare membrane. Notably, the PVDF/TiUN membrane exhibited an FRR value of 97.2%.

Long-term removal effect of M4 membrane on RhB is shown in Figure 8(e). Under light condition, the FRR% of M4 membrane can reach nearly 100% in the first 3 cycles but decreased to about 85% after the fifth cycle. Under no-light condition, the FRR% of M4 membrane dropped monotonically along all cycles. The flux was almost maintained in light condition while it decreased in the no-light condition (and eventually fell to 50 L m⁻² h⁻¹). This experiment demonstrated that the PVDF/TiUN membrane exhibited stable flux and FRR% after even 10 cycles of separation, showing its superior recycle ability. In consistents with the recycle ability, the structure of membranes showed substantial stability (Supplementary Figure 3).

In this report, we adopted a novel method for fabricating TiUN nanohybrid PVDF membranes, which exhibit higher water flux, antifouling, and self-cleaning capacity. The maximum pure water flux of TiUN-0.1 was 265.56 L/(m²·h), while the BSA rejection rate, WCA and FRR were 98.14%, 59.03̊and 95.37% respectively. Furthermore, PVDF/TiUN membrane exhibited superior self-cleaning property as indicated by full recovery of water flux after photocatalytic process. The photocatalytic property, alongside with hydrophilic property endows the TiUN membrane with robust antifouling capacity and superior separation performance. Thus, we offer here a strategy for designing and constructing antifouling membranes with great potential in wastewater remediation, by adopting Zr-based MOFs.

The authors gratefully acknowledge the financial support received from the National Science and Technology Support Project of China (2014BAB07B01 and 2015BAB09B01) and the Key Technology R&D Program of Shanghai Committee of Science and Technology in China (14231201503).

All relevant data are included in the paper or its Supplementary Information.