Pilot scale nanofiltration membrane fabrication containing ionic co-monomers and halloysite nanotubes for textile dye filtration

The increasing population is one of the main factors affecting climate change globally, which is expected to exacerbate water scarcity soon (Dilaver et al. 2018). Water recovery has become a necessity in both domestic and industrial use for these reasons. As one of the most common water recovery methods, desalination systems provide the formation of concentrated liquids and are used in drinking water treatment processes for human use (Keskin et al. 2021a). Because of the generation of a large number of effluents, the textile industry is one of the most water-consuming industries, which seeks novel and efficient membrane technologies for the treatment of wastewater and to obtain high-quality recovered water for reuse (Ağtaş et al. 2020). The treatment of textile wastewater containing dyes is increasingly becoming more important since improperly treated wastewater containing dyes and other contaminants discharged to the environment can influence aquatic life and leads to pollution. Wastewater characteristics, especially color, organic contaminants and salinity are the most important parameters that need to be addressed before reusing recovered water from industrial wastewater (Mohanty & Purkait 2011).

Recently, membrane technologies have attracted great interest in the treatment of textile wastewater because of the narrow pore sizes of membranes that can eliminate dye compounds and generate high-quality permeate (Liu et al. 2017). Many studies have been published on the removal of dyes using membranes technologies (Keskin et al. 2021b). Qin et al. (2007) has reported the treatment of dye wastewater using three different nanofiltration membranes (Desal-5, NF-70 and TS40), which has resulted in greater than 99% of dye removal. Zuriaga-Agusti and Ellouze have similarly reported the use of membrane systems to remove the color from textile wastewater (Zuriaga-Agustí et al. 2010; Ellouze et al. 2012). Specifically, nanofiltration membrane systems are suitable for dye removal with optimum rejection and flux balance (Lau & Ismail 2009). Three different methods which are phase inversion (Hendrix et al. 2014; Zeng et al. 2016), post-treatment of a more porous support (Chen et al. 2007; Song et al. 2015) and interfacial polymerization for thin-film composite/nanocomposite (TFC/TFN) membranes (Gao et al. 2017; Karisma et al. 2018; Sutedja et al. 2018) are used for the production nanofiltration membranes. TFC/TFN membranes contain three layers, which are a non-woven, support layer and a thin, selective skin layer (active layer).

The incorporation of various nanoparticles into membranes is widely reported in the literature (Ormanci-Acar et al. 2020). The most commonly used nanoparticles include silver, titanium (Mollahosseini et al. 2012), titanium dioxide (Sotto et al. 2011), carbon nanotubes (CNT) (Choi et al. 2006) and halloysite nanotubes (HNTs) (Chen et al. 2012; Wang et al. 2014a; Zeng et al. 2016; Ormanci-Acar et al. 2018). The addition of silver, titanium and titanium dioxide nanoparticles into the solution used for the ultrafiltration membrane fabrication has resulted in increased permeability and decreased water contact angles for obtained membranes (Mollahosseini et al. 2012). Moreover, the addition of silica nanoparticles (SiNP) and CNT into the ultrafiltration membrane fabrication solution has resulted in an initially decreased permeability, which has increased with increasing SiNP and CNT concentrations in the solution (Keskin et al. 2021a). In addition, some studies have reported that the incorporation of zinc oxide (ZnO) results in enhancement in the ultrafiltration membrane permeability and a reduction in the ultrafiltration membrane fouling with increasing ZnO concentration (Leo et al. 2012). Lastly, as a result of the addition of some functional groups such as NH₂ functionalized multiwalled carbon nanotube (NH₂-MWCNT) on polyamide nanofiltration (NF) membrane, the mechanical strength of the nanofiltration membranes increases (Zarrabi et al. 2016).

Innovative materials have been used to fabricate flat sheet NF membranes in recent years. Among them, the use of natural HNTs in membrane fabrication is attracting great attention (Ormanci-Acar et al. 2018). HNT is a naturally abundant material that has a hollow tubular shape. It is a unique clay mineral with superior chemical and physical properties, which consists of a hydrogen and oxygen double-layer structure (de Paiva et al. 2008). Besides such advantages, their non-toxic nature has yielded HNTs ideal nanoparticles in a diversity of applications in food, drug and petroleum refining industries on a full scale. Because of its unique properties, HNTs have also been used as an additive to prepare nanocomposite membranes. The existence of surface hydroxyl groups on HNTs enables the production of more hydrophilic membranes (Ormanci-Acar et al. 2021). HNTs have been reported to evolve the adsorption capacity and rejection performance of ultrafiltration membranes with increased stability and reusability (Keskin et al. 2021a). Yet, there exists a limited number of research studies on the incorporation of HNTs into nanofiltration membranes in the literature. In these studies, the phase transformation (Zhu et al. 2014; Zeng et al. 2016) and interfacial polymerization (Ormanci-Acar et al. 2018) methods have been reported for membrane fabrication. As a result, the overall permeability values apparently increased above 90% and dye rejection was successful over 95% due to the HNT additive in the membranes. In addition, there exist two research articles in the literature so far, reporting the fabrication of advanced osmosis membranes containing HNTs (Tang et al. 2009; Zeidler et al. 2013).

In our recent study, we have investigated the usage of a novel sulfonated diamine monomer, namely disodium-3-3'-disulfonate-bis [4-(3-aminophenol phenoxy) phenyl] sulfone (S-DADPS), in combination with piperazine (PIP) for the production of novel nanofiltration membranes by interfacial polymerization (Ormanci-Acar et al. 2020). The incorporation of S-DADPS increased the surface roughness and hydrophilicity of the resulting membranes. At the same time, the membrane surface has become more negative, while pure water, dye solution and salt solution flux values have increased without a noticeable decrease in yield (Ormanci-Acar et al. 2020). The significantly higher molar mass of S-DADPS (637 g mol⁻¹) according to PIP (86.14 g mol⁻¹) was supposed to yield larger space between crosslink points and free volume in the resulting TFC polyamide network, which resulted in increased membrane flux while the electrostatic effect caused by the charger sodium sulfonate groups (NaSO₃) of S-DADPS in the polyamide layer DADPS in combination with m-phenylene diamine (MPD) for the production of chlorine resistant reverse osmosis (RO) membranes (Xie et al. 2012). With the incorporation of S-DADPS into the polyamide matrix, the flux of resulting RO membranes also increased.

Limited number of articles in the literature describe the pilot-scale manufacturing of flat sheet membranes. Soroko et al. (2011) reported the continuous fabrication of flat sheet membranes on the pilot scale (Soroko et al. 2011). However, the main focus of this study was the composition of the membranes rather than the fabrication parameters on the pilot scale. In addition, in a study reported by Wu et al. (2016), porous polyethersulfone (PES) membranes were fabricated via thee phase inversion method both on the laboratory-scale and pilot scale for gas separation (Wu et al. 2016). Some obvious differences were observed especially in the pore size distribution of membranes when they scaled up the fabrication, and they reported the development of a PES membrane with desirable surface morphology. In another study (Chen et al. 1997), commercial NF membranes have been tested on the pilot scale for the treatment of textile wastewater. Recently, the investigation of the long-term use of commercial NF membranes was reported for wastewater treatment to identify causes of biofouling that may occur (Xu et al. 2020). It is critical to note that there exists a lack of further studies in the literature that report the pilot-scale fabrication and characterization of not only well-established but also emerging membranes.

In this study, NF membranes were manufactured by interfacial polymerization in pilot scales using either S-DADPS in the aqueous phase in combination with PIP and/or HNTs in the organic phase in combination with trimesoyl chloride (TMC) in comparison with a regular, control NF membrane fabricated via the interfacial polymerization of PIP and TMC. NF membranes were converted to modules prior to filtration tests. Following the conductivity, chemical oxygen demand (COD) and color analyses of fabricated membranes, their flux and rejection performance was determined using pure water, synthetic wastewater (SWW) containing 2,000 ppm MgSO₄ salt and 100 ppm Reactive Red 120 dye, and lastly industrial textile wastewater. The incorporation of S-DADPS and HNTs together into the polyamide matrix of NF membranes positively affected the flux values of resulting modules without any significant changes in rejection values for salt and dye species. Lastly, the resulting modules were found to be more effective in the rejection of salt and dye species from the SWW compared to the industrial textile wastewater. Within the scope of this study, both the pilot-scale membrane production was made and a comparison of real textile wastewater and SWW was given. Since there are a limited number of membrane applications produced at scale using HNT and sDADPS in the literature, this study is not intended for this purpose.

Polysulfone (PSf, Udel^® P-3500) was purchased from Solvay Specialty Polymers, USA. Polyvinylpyrrolidone (PVP) K30 (molecular weight (MW): 40,000 g mol⁻¹) and PVP K15 (MW:10,000 g mol⁻¹), anhydrous piperazine, TMC, anhydrous cyclohexane, 3-aminophenol, sodium hydroxide (NaOH), dimethyl sulfoxide (DMSO), chlorobenzene, isopropyl alcohol and Reactive Red 120 (MW:1,463 g mol⁻¹) dye were purchased from Sigma Aldrich Ltd N, N-dimethyl formamide (DMF) was purchased from Ak-Kim Chemicals, Turkey. HNTs were provided by Esan (Turkey). Disodium-3-3'-disulfon-4-4'-diclorodifenylsulphone (s-DCDPS) was bought from Akron Polymer Systems, USA. MgSO₄ was purchased from Tekkim Chemistry Ltd, Turkey.

For the classification of HNTs, surface area measurements and particle size analysis were done in our previous study (Ormanci-Acar et al. 2018; Keskin et al. 2021a). The specific surface area of HNTs was found 100.74 m² g⁻¹. Also, the average agglomeration size was measured under 5 μm in the supplied form, before the dispersion by ultrasonication. Additionally, the cation exchange capacity was 24.42 meq per 100 g for HNTs. S-DADPS monomer was synthesized by the nucleophilic aromatic substitution reaction as reported previously (Ormanci-Acar et al. 2020).

Three types of NF membranes were fabricated. An NF membrane (named as ‘pure PIP’), was fabricated via the interfacial polymerization, of 2% (w/v) of PIP in an aqueous phase solution with 0.2% (w/v) of TMC in an organic phase solution. The second one (named as ‘60% S-DADPS’) consists of a mixture of 0.8% (w/v) of pure PIP and 1.2% (w/v) of S-DADPS to prepare the aqueous solution and the same concentration was used for organic phase. The last one (named as ‘0.04% HNT’) was prepared by adding 0.2% TMC into the organic phase. On the other hand, 2% (w/v) PIP was used for the aqueous phase (Table 1). HNTs were dispersed at 90 W for 2 h by sonication method and stirred at 1,000 rpm during the fabrication process. All preparation and fabrication were conducted at room temperature; n-hexane was used as an organic solvent for the pure PIP and 60% S-DADPS membrane fabrication. However, cyclohexane was preferred for 0.04% HNT membrane fabrication as the polarity of cyclohexane is higher than hexane to avoid any complications during the dispersion of HNTs.

All fabricated membranes were stored in a cold room at +4 °C for 1 week, due to the elimination of all impurities. Zeta potential (via current potential) analysis was performed for all membrane surfaces with the Anton Paar Surpass electrokinetic analyzer. 0.001 M KCl at pH 8 was used as an electrolyte solution. To determine the hydrophilicity of the membranes, the contact angle of the membrane surface was measured by KSV Attension–Theta. It was calculated by taking at least 10 frames by dripping distilled water on the surface at 1-second intervals. Moreover, field emission scanning electron microscope (FESEM, FEI Quanta Feg 250) analysis was performed to determine the surface and cross-section morphologies of membranes.

SWW containing 2,000 ppm MgSO₄ salt and 100 ppm Setazol Red dye and the textile wastewater were filtered by modules under nine bars. After filtration, permeate was collected for measuring conductivity, COD and color rejection. Conductivity measurements were obtained by using a Hach HQ40D conductimeter. CN values were measured by a UV–Vis spectrophotometer using Hach Lange DR 5,000 COD values were determined by closed-reflux method as Standard Method UV wavelength was measured as 520 nm for Setazol Red. The characterization of textile wastewater was listed in Table 2.

Pilot-scale membranes were characterized and the results are given in Table 3. The contact angle was examined to understand the hydrophilicity of both the support layer and pure PIP, 60% S-DADPS and 0.04% HNT nanofiltration membranes. The contact angle of the support layer surface membrane was 80.6 ± 3.1°. As it is known the hydrophilicity of polyamide membrane is very high (Ormanci-Acar et al. 2018), the contact angle of nanofiltration membranes for the pilot scale was found to be 33.4 ± 0.6°, 42.3 ± 4.1° and 48.6 ± 2.4° for pure PIP, 60% S-DADPS and 0.04% HNT membranes, respectively.

Table 3

Characterization of membranes

Membrane module name .	Zeta potential at pH 8 .	Contact angle (°) .	SEM image of cross section .	SEM image of surface .
Support layer	−23.2	80.6 ± 3.1
Pure PIP	−61.3	33.4 ± 0.6
60% S-DADPS	−62.9	42.3 ± 4.1
0.04% HNT	−64.8	48.6 ± 2.4

Membrane module name .	Zeta potential at pH 8 .	Contact angle (°) .	SEM image of cross section .	SEM image of surface .
Support layer	−23.2	80.6 ± 3.1
Pure PIP	−61.3	33.4 ± 0.6
60% S-DADPS	−62.9	42.3 ± 4.1
0.04% HNT	−64.8	48.6 ± 2.4

View Large

The surface load of the produced membranes was determined by zeta potential analysis at pH 8. Due to the unique structure of the NF membrane, the separation capacity varies depending on the small pore diameters depending on the steric effect and the load on the pore surface (Donnan effect) (Donnan 1995; Seidel et al. 2001). Since NF membranes are generally known to have a negative charge (Van der Bruggen et al. 1999), when the zeta potential results are examined, it is seen that the surface loads of these membranes were found negative for all membranes. All results of zeta potential at pH 8 for all pilot-scale membranes are presented in Table 3. According to Table 3, the zeta potentials of all NF membranes were found very similar. The zeta potential of the support layer was −23.2 mV. Applying polyamide structure onto the support layer increased the negativity of the surface. Correspondingly the surface charge of pure PIP, 60% S-DADPS and 0.04% HNT membranes were −61.3, −62.9 and −64.8 mV, respectively. Moreover, using S-DADPS monomer and embedding HNT accelerated the surface negativity a little bit more. The sulfonic acid groups of S-DADPS could be increased the negativity of the surface beside carboxylic acid functional groups of polyamide structure. These groups are separated from their protons at natural pH and become more negative as the pH increases. Many studies say that pH change affects membrane load due to the separation of functional groups (Elimelech & Childress 1996; Van der Bruggen et al. 1999; Bellona et al. 2004; Tiraferri & Elimelech 2012). Hu et al. (2016) stated that SO₃H sulfonic acid groups in the sulfonated monomer increase negative groups of the membrane surface. We can say that the SO₃Na sulfonic acid groups in the S-DADPS monomer used in this study showed the same effect. The researcher searched the surface load of HNT, which was found to be remarkably negative because of the residual anionic silanol groups on the HNT surface (Wei et al. 2013). To increase the negative surface charge of thin-film nanocomposite membranes, it is expected that membrane contamination from textile wastewater containing negatively charged ions could be prevented.

The SEM images of pilot-scale membranes are presented in Table 3. The sponge-like structure of the support layer can be detected from cross-sectional images. The sponge-like structure has a benefit for use as a support for high pressured membranes like nanofiltration or RO owing to its high compaction resistance (Widjojo et al. 2013), smaller macro void formation and increased pore formation (Liu et al. 2003) and high permeability (Wang et al. 2014b).

According to SEM images, the sponge-like morphology of production PSf support membranes without any macro voids for all membranes. Also, the surface of the pilot-scale support layer was formed as porous, but the pore size of the support layer was found a little bit higher. The surface morphology and features of the formed polyamide active skin layer on the PSf support membrane were examined with the pure PIP membrane in the fabrication process. However, combining S-DADPS with PIP, changed the surface morphology, and rougher and similar surfaces were observed. On the other hand, embedding HNTs into a polyamide matrix did not cause a noticeable difference according to the pure PIP one.

The permeability of pure PIP, 60% S-DADPS and 0.04% HNT membrane modules were 2.65, 10.70 and 7.52 L m⁻² h⁻¹ bar⁻¹, respectively.

The performance of pilot-scale membranes were further tested on cotton dying textile wastewater after synthetic solution tests. As the characterization is presented in Table 2 textile wastewater has a complex nature due to the different contaminants that are used in the textile industry. In order to calculate the membrane performance, inlet and outlet wastewater samples were collected, and conductivity, COD, total oxygen demand (TOC) and color parameters were examined. The wastewater flux of pilot-scale membranes were presented in Figure 6. The flux performance of 60% S-DADPS was found to be higher than the other membranes like other filtration test results and, the flux value of 0.04% HNT membranes was higher than pure PIP membranes. The textile wastewater flux of pure PIP, 60% S-DADPS and 0.04% HNT were 22.42, 79.58 and 40.06 L m⁻² h⁻¹ for the pilot-scale system, respectively. The wastewater flux increased up to four times in 60% s-DADPS membrane compared to pure PIP membrane flux.

The removal efficiencies of conductivity, COD, color and TOC are given in Figure 7. According to Figure 7, it was observed that all membrane removal efficiencies of conductivity were higher than 47% for the pilot scale. The pure PIP membrane's efficiency was higher than the other membranes. Furthermore, it was observed that all membranes removal efficiencies of COD were higher than 47%. Also, all the membrane's removal efficiencies show similar results. On the other hand, color removal efficiencies were found to be higher than 80%. Also, pure PIP and 0.04% HNT membranes showed similar results for the removal efficiency of color but the 60% S-DADPS membrane's efficiency was lower than other membranes. When the zeta potential results of the membranes are analyzed, it was observed that the analyzes of membranes containing pure PIP and 0.04% HNT were close to each other and their surfaces were more negative than those of the S-DADPS membrane. As the TOC results show, examined removal efficiencies of all membranes were higher than 85%. Also, all the membrane's removal efficiencies show similar results for both systems.

To sum up, the removal results of real textile wastewater were found to be more realistic than synthetic solution experiments. Because of the complex characteristic of it, apparently, it contains also monovalent ions besides divalent ions, the conductivity rejection values of textile wastewater were found lower than SWW. A similar increase was observed with the data COD; textile wastewater's rejection values were found higher than SWW values. It is assumed that the real textile wastewater contains more and complex organic contaminants as the TOC and color rejection values stay stable.

This section is assigned for comparing the recent studies reported for synthetic and real textile wastewater treatment using a pilot-scale nanofiltration membrane. When the studies are examined, experiments with commercial membranes are generally reported. In studies using NF membrane, differences in membrane performance were observed according to the wastewater source.

Bes-Piá et al. (2009) applied an activated sludge process and three different NF membranes to treat textile wastewater. Since the activated sludge process does not provide the treated wastewater quality required for reuse in the textile factory, The NF unit was added to the system as a final treatment process. Removal efficiencies were calculated as 94% for COD, 50% for conductivity removal and 60% for salt retention (Bes-Piá et al. 2009). In a similar study, NF membrane, the COD removal efficiency was determined as 90%, color removal 90% and conductivity removal 80%. Considering this study, it was found appropriate to obtain high-quality permeate in the production processes of the textile industry and to reuse the treated water (Aouni et al. 2012). The differences in performance between these two studies are due to the pore sizes of the membranes. The treatment of textile wastewater by NF and RO processes has been investigated. In the study, when the textile wastewater is treated with NF, both COD and color removal performance are observed over 95%. Also, when the membrane contamination experiments in the study were examined, it was observed that the irreversible contamination was 23% for the NF membrane for 5 days (Kurt et al. 2012).

In this study, thin-film composite/nanocomposite membranes were manufactured by using pure piperazine, S-DADPS monomer and HNT additive on a porous support layer at a pilot scale. It is aimed that increasing membrane water/wastewater fluxes and removal efficiencies in the context of using innovative materials. The water, synthetic and textile wastewater performance of pilot-scale membranes were tested in this study. It was observed that the sulfonic acid groups hosted by the innovative S-DADPS monomer increased the performance of the NF membrane without considerable rejection lost in pilot-scale membranes. Also, the use of HNT in the production of NF membranes accelerated the flux performance of membranes.

When the synthetic solution and real textile wastewater performance were compared, the pilot-scale membrane performances were found sufficient. As a result of higher fluxes, the performance of pilot-scale membranes was affected slightly, especially they were tested with real textile wastewater. However, the decrease in conductivity was also related to variable proportions of monovalent and divalent ions in real textile wastewater.

Pilot and full-scale systems were examined once sufficient information was obtained from laboratory-scale applications. According to the literature investigations, there were fewer pilot and full-scale studies undertaken in contrast to laboratory-scale applications (Keskin et al. 2021b). It has been proven that the 60% s-DADPS membrane provides up to four times improvement in wastewater flux and simultaneously, there is no loss in COD, conductivity and color parameters. In addition, it has been determined that 60% s-DADPS NF membrane is a cost-friendly choice on a pilot scale with the enhancement in flux, which directly affects the operating cost. It is recommended to carry out further studies on pre-treatment alternatives in order to improve the effluent quality.

This work was supported by The Scientific and Technological Research Council of Turkey (TUBITAK) under project numbers of 113Y350/113Y371.

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

The authors declare there is no conflict.