Synthesis of polyethersulfone (PES)/GO-SiO₂ mixed matrix membranes for oily wastewater treatment

Water scarcity is increasing rapidly, creating significant stress and affecting millions of people worldwide. Several alternatives have been introduced to reduce this stress, such as seawater desalination (Shannon et al. 2008), water reuse (Wade Miller 2006) and wastewater treatment (Qu et al. 2013). In addition, more attention has been given recently to treating the industrial effluents from the food (Lefebvre & Moletta 2006; Cecconet et al. 2018), pharmaceutical (Jelic et al. 2011; Yang et al. 2017a, 2017b), textile (Khandegar & Saroha 2013; Serge Raoul et al. 2018), and oil refinery (Yu et al. 2017) sectors. Industrial effluents from oil refineries are usually the hardest to treat since they contain massive quantities of emulsified oil and suspended particles (Zhong et al. 2003). Conventional methods of treating oily wastewater include air flotation, gravity settling, coagulation and flocculation (Zhu et al. 2014). However, these methods are inefficient for treating emulsified oil/water mixtures, especially emulsions containing oil droplet sizes smaller than 20 µm (Zhu et al. 2014). Therefore, membrane-based separation technologies have been put forward as alternatives to these conventional methods (He & Jiang 2008).

Treating oily wastewater using membrane-based technologies was first investigated in the early 1970s. These techniques included microfiltration (MF) (Anderson et al. 1987), ultrafiltration (UF) (Christensen & Plaumann 1981; Lipp et al. 1988), reverse osmosis (RO) (Kutowy et al. 1981), and membrane distillation (Curtin 1984). The use of these technologies offers significant benefits, such as high removal efficiency and low power consumption. Nonetheless, these membranes suffer from low flux and fouling problems. This leads to decreases in the oil rejection, specifically when dealing with wastewater effluents with a high concentration of oil. Oil rejection in membrane processes is dependent mainly on two aspects: pore size and membrane wettability (Shi et al. 2013; Ma et al. 2016). In the first, the membrane blocks all the oil droplets with a diameter larger than the pores of the membrane. This involves applying a specific pressure that allows only the water to pass through the membrane while retaining the oil (Huang et al. 2018). The second aspect ensures that oil droplets do not become wet, which would prevent them from passing through the membrane. This depends on the membrane's oleophobicity and hydrophilicity properties (Zhang et al. 2016; Zhu et al. 2017a, 2017b). Also, the adhesion force of oil droplets on the membrane's surface can be reduced by increasing the surface hydrophilicity, thus improving the water flux and decreasing fouling susceptibility (Mansourizadeh & Javadi Azad 2014).

Recently, different membrane types have been investigated for improved flux, oil rejection, and fouling resistibility. These include, but are not limited to, ceramic (Das et al. 2017), polymeric (Mansourizadeh & Javadi Azad 2014), metallic (Yang et al. 2015), and carbon-based (Song et al. 2006; Sarfaraz et al. 2012) membranes. Although ceramic membranes have good oil rejection and mechanical strength, they are characterized by their extremely high cost and fabrication difficulties (Li 2007; Wu et al. 2015). By contrast, polymeric membranes are relatively cheap and easy to fabricate (Huang et al. 2018). The most common materials for these membranes are polyvinylidene fluoride (PVDF), polysulfone (PSF), and polyethersulfone (PES) (Padaki et al. 2015). PVDF membranes are very common in UF systems due to their great anti-oxidation ability, high mechanical and thermal properties, and chemical resistibility. However, they are very susceptible to fouling due to their hydrophobic nature. This is a significant limitation of this type of membrane in oily wastewater treatment (Loukidou & Zouboulis 2001). PSF membranes have natural hydrophobic properties and low mechanical strength (Ionita et al. 2014). PES based membranes show high oxidative, thermal and mechanical properties (McKeen 2012). They also have high chemical resistances (Brandt & Wiese 2003; McKeen 2012). The outstanding chemical and physical properties displayed by PES make it ideal for use in preparing asymmetric membranes with different pore sizes and structures. Although PES has a low hydrophilicity, it is higher than for common polymers used in membrane applications, such as PS, PVDF, polypropylene (PP) and polytetrafluoroethylene (PTFE).

Many successful methods have been used to increase the hydrophilicity of PES membranes: surface modification, addition of hydrophilic additives or nanoparticles and so on (Susanto & Ulbricht 2009; Ahmad et al. 2013; Zhao et al. 2013). The incorporation of nanoparticles (such as zeolites, carbon nanotubes (CNTs), titanium dioxide (TiO₂), zinc oxide (ZnO), graphene oxide (GO), and silica (SiO₂)) into the polymer solution results in the synthesis of mixed matrix membranes (MMMs) (Qadir et al. 2017). Compared to their pure polymeric membrane counterparts, the addition of these nanoparticles generally leads to higher oil rejection, water permeability or both (Ng et al. 2013). GO nanoparticles have been studied for their incorporation in membrane technologies due to their hydrophilic nature and abundance of functional groups on their surface leading to increased permeability and ease of surface modification (Abdel-Karim et al. 2018). Zhou et al. fabricated a GO/halloysite nanotube (HNT) membrane via a vacuum-assisted filtration process and using PES as the substrate support (Zhu et al. 2018). The membrane was tested for the separation of oil from oil-in-water emulsions. The addition of both HNT and GO nanoparticles resulted in high oil rejection (99%) compared to the PES support, which displayed a low oil removal efficiency. The GO/HNT membrane also showed excellent fouling resistance properties as well as improved hydrophilicity and water flux. SiO₂ nanoparticles have also been investigated in membrane fabrication because of their chemical and thermal stability, high surface area and non-toxicity (Ng et al. 2013). Their incorporation in membranes has led to enhanced hydrophilicity, increased pore size leading to higher flux, and improved anti-fouling performance (Shen et al. 2011; Li et al. 2015).

Most studies of the application of membranes for oil removal have been tested using synthetic oil-in-water emulsions. In addition, most graphene-based membranes fabricated for the treatment of oil-in-water emulsions have been prepared using a coating method as opposed to blending the nanomaterials directly in the dope solution. The former method produced membranes that are susceptible to leaching. To the best of our knowledge, no previous studies have reported on the fabrication of PES MMM incorporating GO and SiO₂ (i.e. PES/GO-SiO₂) in the polymer solution for the treatment of raw oily wastewater. Therefore, the main objective of this research study was to synthesize a novel PES/GO-SiO₂ MMM for the treatment of oily wastewater.

PES (average Mw. 75,000) was purchased from Prakash Chemicals Pvt. Limited. Polyvinylpyrrolidone (PVP; average Mw. 40,000) and dimethylacetamide (DMAc) (Mw 87.12 g/mol, ≥99% purity) were purchased from Sigma Aldrich. SiO₂ nanoparticles (with an average diameter of 20 nm) were supplied from EPRUI Nanoparticles & Microspheres Co. Ltd. GO (with a diameter in the range of 1.5–5.5 μm) was purchased from US Research Nanomaterials, Inc. Deionized (DI) water with a resistivity of 15 MΩ·cm at 25 °C was used. All solvents and materials were used as purchased without further purification. Raw oily wastewater was obtained from a local petroleum refinery in Abu Dhabi (UAE). The characteristics of the raw sample were 11, 6.3 mS/cm, 15 NTU, 361.8 mg/L, 40 mg/L, 44 mg/L, and 44 mg/L of pH, conductivity, turbidity, total organic carbon (TOC), Ca²⁺, K⁺, and Mg²⁺, respectively. The sample was used without further modification or purification as feed for the membranes.

A solution of 1 mg/mL GO in DI water was sonicated using a Branson 1,510 Ultrasonic Cleaner at 40 KHz frequency in a water bath for 30 min followed by overnight stirring. Similarly, an SiO₂ dispersion (in the ratio of GO:SiO₂ 2:1) was prepared by sonication followed by overnight stirring. The SiO₂ particle solution was added to the GO dispersion and stirred at 40 °C for 3 hours. The final solution was centrifuged (using a HERMLE Labortechnik Z 326 K centrifuge) at 6,000 rpm for 15 min. Finally, the obtained product was dried in a Memmert UF55 oven at 50 °C and crushed to obtain a powder.

PES based membranes were prepared via the phase inversion method. PES/GO, PES/SiO₂ and PES/GO-SiO₂ membranes were each prepared by using a loading concentration of 1.0 wt% of the respective nanoparticle to the polymer. The corresponding nanoparticles were dispersed in DMAc and ultrasonicated in a water bath for 30 min. PVP (4%) was dissolved in the above solutions followed by the addition of PES (16%) and stirred for 24 hours at 60 °C. The dope solution was set aside for 24 hours to remove trapped air bubbles (i.e. membrane degassing). The solution was subsequently cast on a polyester membrane support on clean glass at a thickness of 200 μm. The glass plate was immersed horizontally into deionized water at a temperature of 25 °C for 24 hours. Finally, the membranes were washed with DI and stored for use. A control PES membrane was also prepared using the same method for comparison.

Figure 1 shows the SEM images of the top surface of the membranes. The images taken confirmed the formation of pores and porosity in the membranes fabricated. Moreover, as can be seen in Figure 1(c) and 1(d), after the addition of GO, an increase in the pore size and porosity was observed. This was further confirmed by the porosity and pore size calculations presented in the ‘Surface wettability, porosity and pore size distribution’ section. In Figure 1(d), the formation of macrovoids was clearly visible. This could be explained by the synergistic effect of the GO/SiO₂ leading to increased hydrophilicity of the nanocomposite as reported in the literature (Tewari 2015). This increase in hydrophilicity increases the exchange rate between the solvent and non-solvent in the coagulation process, resulting in the formation of macrovoids or increased porosity (Zinadini et al. 2014).

Figure 2(a) shows the FT-IR spectra for both GO and SiO₂ nanoparticles used in membrane synthesis. The broad absorption band at 3,370 cm⁻¹ is attributed to the presence of OH and/or COOH functional groups within the GO structure (Goncalves et al. 2009; Rattana et al. 2012; Kumar et al. 2013). The GO nanoparticles also display peaks at 1,725 cm⁻¹ and 1,613 cm⁻¹ corresponding to the stretching vibration of C=C carbonyl C=O groups, respectively. C-O stretching of epoxy groups (1,200 cm⁻¹) and C-O stretching of alkoxy group (1,045 cm⁻¹) were also observed (Goncalves et al. 2009; Rattana et al. 2012; Kumar et al. 2013). The FT-IR spectra confirmed the structure of GO and the presence of various oxygen-containing functional groups such as hydroxyl, epoxy, carboxyl, carbonyl on GO.

The FT-IR spectra of SiO₂ nanoparticles also confirmed the structure of the SiO₂ nanoparticles. The peaks observed at 1,075 cm⁻¹ and 800 cm⁻¹ are attributed to the asymmetric and symmetric stretching vibrations of Si-O-Si respectively (Sanaeishoar et al. 2015). Figure 2(b) shows a comparison of the FT-IR spectra of the different types of membrane fabricated: PES (control), PES/GO, PES/SiO₂, and PES/GO-SiO₂. The Si-O-Si peak is clear at 1,060 cm⁻¹ in both the PES/SiO₂ and PES/GO-SiO₂ MMMs, confirming the presence of SiO₂. However, the peaks attributed to the bonds in the GO structure are not clearly identified in the FT-IR spectra due to the interference of the PES bonds. Raman spectroscopy was therefore performed to confirm the presence of GO in the fabricated membranes, as shown in Figure 2(c).

Figure 2(c) shows the Raman spectra of the PES, PES/GO, and PES/GO-SiO₂ membranes. The GO nanoparticles used in this research study are also shown in Figure 2(c). Two distinctive peaks in the GO nanoparticles and GO-based membranes are noticeable: the D and G bands occurring at 1,350 cm⁻¹ and 1,605 cm⁻¹ respectively. The G band represents the ordered sp² bonded carbon atoms (C-C) present in crystalline graphite-like structures, while the D band represents the disordered sp³ carbon structure indicative of the disruption within the hexagonal graphitic lattice due to internal structural defects and dangling bonds (Chen et al. 2010; Abdel-Karim et al. 2018). These peaks were absent in the PES membrane and confirm the presence of GO in the GO-based membranes.

Mechanical properties can be influenced by the incorporation of nanoparticles in membrane matrices. These depend on the interaction between the polymer chains and nanoparticles as well as the dispersion uniformity and loading concentration of the nanoparticles in the membrane (Namvar-Mahboub & Pakizeh 2013). Generally, higher tensile strengths are due to a more uniform dispersion of the nanoparticles in the dope solution and/or better compatibility with the polymer chains (Namvar-Mahboub & Pakizeh 2013). The ultimate tensile strength and % elongation (strain at break) of the pristine PES, PES/SiO₂, PES/GO and PES/GO-SiO₂ were 19 (30%), 15 (25%), 11 (23%) and 18 (24%) MPa, respectively. The addition of SiO₂ and GO nanoparticles separately in the membrane matrix led to a moderate decrease in the tensile strength and a slight decrease in ductility. However, the GO-SiO₂ nanocomposite had a negligible effect on tensile strength in comparison to the pristine PES membrane and a slight decrease in ductility. This means that the addition of the GO-SiO₂ nanocomposite has a better impact on tensile strength than the addition of either GO or SiO₂ nanoparticles separately. This could indicate that the synergistic effects of the GO-SiO₂ nanocomposite led to a more uniform dispersion in the dope solution and were more covalently bonded to the PES matrix than the separate incorporation of GO and SiO₂ nanoparticles.

Oil/water separations are generally dependent on two important mechanisms: the ‘size-sieving’ effect and surface wettability (Chen & Xu 2013; Chu et al. 2015). These are determined through pore size and contact angle measurements. Contact angle measurements indicate the hydrophilic/hydrophobic nature of the membrane. For oil-in-water emulsions, hydrophilic/oleophobic membranes are preferred as they allow water to pass through while rejecting oil (Chen & Xu 2013). These membranes have a much lower tendency to fouling when compared to hydrophobic membranes. Thus, lower contact angle values are desired. Another mechanism that the membrane relies on in oil/water separation is the ‘sieving’ effect. Thus, membranes with pore sizes smaller than the emulsified oil droplets are desired (Chu et al. 2015). Most membranes used for the separation of oil in water generally lie in the MF to UF range. Generally, the pore sizes of MF membranes span from 0.1 to 1 μm, while those of UF membranes span from 0.01 to 1 μm. Pore size measurements carried out on the PES, PES/SiO₂, PES/GO, PES/GO-SiO₂ membranes found them to be 0.181, 0.158, 0.137, and 0.112 μm, respectively. These results revealed that the fabricated membranes were all in the lower range of MF membranes and thus closer to the UF range. Figure 3(a) shows that the addition of SiO₂ and/or GO nanoparticles in the PES matrices decreased the contact angle and improved hydrophilicity. The contact angle decreased from 85° in the pristine PES membrane to 58° in the PES/GO-SiO₂ MMM. As a result, the PES/GO-SiO₂ membrane displayed the highest water flux: 2,561 L/m² h (LMH) (Figure 4). Generally, as membrane hydrophilicity increases, water flux also increases, due to the strong affinity of the membrane towards water molecules (Ahmad et al. 2013). However, despite this, the PES/SiO₂ membrane displayed much lower flux (718 LMH) in comparison to all other membranes, as can be seen from Figure 4. This is because hydrophilicity is not the sole parameter in determining water flux.

The addition of nanoparticles tends to form a more viscous dope solution, leading to delayed membrane formation in the coagulation bath. This was confirmed by viscosity tests performed on the fabricated membrane, from which the viscosity of the PES/SiO₂ displayed the highest increase (i.e. 8,623 mPa·s compared to 3,395, 8,474, and 5,373 mPa·s for the PES, PES/GO, PES/GO-SiO₂, respectively). The viscosity of the PES/GO dope solution also increased significantly in comparison to the pristine PES dope solution. However, although the viscosity of the PES/GO-SiO₂ dope solution was higher than the PES dope solution, as expected, it was roughly 1.5 times lower than the PES/GO and PES/SiO₂ dope solutions. This could be explained by the fact that at this percentage, the dispersion of SiO₂ and GO nanoparticles separately in the dope solution caused agglomeration and thus led to higher viscosity, whereas the GO-SiO₂ nanocomposite was more uniformly dispersed and hence led to lower viscosity in the dope solution. This conclusion was also confirmed through the interpretation of the tensile strength measurements above. In the case of the PES/SiO₂ membrane, the increase in viscosity led to reduced macrovoid structure formation which reduced water permeability. Hence, the viscous nature of the SiO₂ dope solution could have overcome the inherent hydrophilic properties of the SiO₂ nanoparticles leading to decreased water flux. In contrast, the PES/GO membrane displayed high flux despite the increase in the viscosity of the dope solution. As shown in Figure 3(b), the porosity of the PES-based membranes lay in the range of 31 to 40%. The PES/SiO₂ membrane had the lowest porosity of 31%. The porosity increased with the addition of GO, as can be observed from Figure 3(b). This could be attributed to the hydrophilic nature of GO, due to its functional groups, which increases the exchange rate between the solvent and non-solvent in the coagulation both, thus increasing membrane porosity (Zinadini et al. 2014). The highest porosity of 40% was obtained for the PES/GO-SiO₂ nanohybrid membrane. The porosity values were in line with the water flux data of the membranes, in which PES/SiO₂ had the lowest water flux while the PES/GO-SiO₂ composite had the highest, as shown in Figure 4.

Figure 4 shows the water flux and oil rejection of the fabricated membranes. All membranes showed considerably high values of water flux, ranging from ∼700 to 2,561 LMH. The control PES membrane exhibited a high water flux of 2,511 LMH. The addition of SiO₂ decreased the water flux significantly while the addition of GO alone maintained the same water flux. With the addition of the GO-SiO₂ nanohybrid, the water flux increased slightly to 2,561 LMH. Table 1 shows a comparison of pure water fluxes obtained for GO-SiO₂ MMM in other work. Moreover, the oil rejection in the MMM increased compared to the pristine PES membrane. The highest oil rejection was obtained for the PES/GO membrane at 30%, while the PES/GO-SiO₂ membrane achieved an 18% oil rejection. It is worth noting that the size of the oil droplets in the oily wastewater was >1 μm. This could confirm the rejection of oil particles due to size exclusion as all membranes' pore size diameters were <1 μm.

Besides size exclusion and hydrophilicity mechanisms, the repulsion and attraction forces due to the surface charge of the membrane play an important role in oil rejection (Abadikhah et al. 2018). As reported above, the oily wastewater sample contains Ca, Mg, and K ions. These are positively charged ions, in contrast with the emulsified oil, which is negatively charged (Alotaibi & Nasr-El-Din 2011). PES membranes have been reported extensively in the literature to be negatively charged (Salinas-Rodriguez et al. 2015; Li et al. 2016a, 2016b). Zeta potential was measured at pH 11 (which is the pH of the raw wastewater sample) on the MMMs to determine the influence of each nanoparticle or nanocomposite on the surface charge of the membrane. The results confirmed that all membranes fabricated were negatively charged. Therefore, with the interference of the positively charged ions in the oily wastewater, oil rejection can be significantly impacted. The PES/GO-SiO₂ membrane showed the highest negatively charged surface of −52.1 ± 1.1 mV due to the synergistic effects of both the negatively charged particles of GO and SiO₂. The PES/GO and PES/SiO₂, on the other hand, had zeta potential values of −41.2 ± 1.2 and −38.1 ± 1.1 mV, respectively. The slight increase in the zeta potential of the PES/GO membrane, compared to the PES/SiO₂ membrane, was attributed to the different epoxy, hydroxyl, carbonyl, and carboxyl functional groups attached to the surface of the GO. The increase in the zeta potential reflects a higher membrane surface charge negativity, which would result in a significant promotion of the Van der Waals electrostatic repulsion forces with the negatively charged oil droplets. On the other hand, the electrostatic attraction forces between the positively charged ions present in the wastewater and the membrane surface would also be prompted, suggesting a higher rate of ion or oil droplet adsorption onto the surface of the membrane.

PES/GO-SiO₂ MMMs were fabricated using the phase inversion technique. The addition of these nanoparticles led to increased hydrophilicity. The contact angle of the control PES membrane (85°) decreased steadily upon the addition of the nanoparticles, reaching a minimum value of 58° for the PES/GO-SiO₂ MMM. Tensile strength and viscosity measurements carried out on the membranes and dope solutions respectively showed that the GO-SiO₂ nanocomposite was more uniformly dispersed in the membrane matrix compared to the separate incorporation of the GO and SiO₂ nanoparticles. Moreover, all membranes displayed high water flux, with the PES/GO-SiO₂ nanohybrid membrane displaying the highest water flux of 2,561 LMH.

These membranes were also tested on raw oily wastewater and it was found that the modified membranes showed an increase in oil removal efficiency compared to the control PES membrane. The highest removal efficiencies were obtained for the PES/GO and PES/SiO₂ membranes at 32 and 26%, respectively, in contrast with the control PES membrane which exhibited a 13% oil rejection. Despite an improvement in overall properties of the PES/GO-SiO₂ MMM, the nanocomposite membrane showed an 18% oil removal efficiency. The difference in oil rejection values across the membranes was attributed to the increasing negative charges on the surface of the modified membranes which interfered with the positive ions present in the oily wastewater sample. These conclusions stress the importance of studying the impact of nanoparticles on the surface charge of the membrane in treating raw oily wastewater which contains a multitude of positive and negative ions.

This publication is based upon work supported by the Khalifa University under Award No. RC2-2018-009.