Bioinspired cellulose-based membranes in oily wastewater treatment

It is challenging to purify oily wastewater, which affects water-energy-food production. One promising method is membrane-based separation. This paper reviews the current research trend of applying cellulose as a membrane material that mimics one of three typical biostructures: superhydrophobic, underwater superoleophobic, and Janus surfaces. Nature has provided efficient and effective structures through the evolutionary process. This has inspired many researchers to create technologies that mimic nature’s structures or the fabrication process. Lotus leaves, fish scales, and Namib beetles are three representative structures with distinct functional and surface properties: superhydrophobic, underwater superoleophobic, and Janus surfaces. The characteristics of these structures have been widely studied and applied to membrane materials to improve their performance. One attractive membrane material is cellulose, whichhas been studied from the perspective of its biodegradability and sustainability. In this review, the principles, mechanisms, fabrication processes, and membrane performances are summarized and compared. The theory of wettability is also described to build a comprehensive understanding of the concept. Finally, future outlook is discussed to challenge the gap between laboratory and industrial applications.


Introduction
Water sanitization issues are related to the direct utilization of water itself, for example for drinking or household use,

H I G H L I G H T S
• Cellulose-based membrane separates oily wastewater mimicking the living things. • The three central surface mechanisms were reviewed. • Preparation, performance, and mechanism are critically evaluated. • First review of wettability based cellulose membrane as major material. • The current and future importance of the research are discussed. and food production and energy issues, otherwise known as the food-water-energy nexus (Scanlon et al., 2017;D'Odorico et al., 2018;Kibler et al., 2018). Polluted water endangers animals and decreases agricultural production, threatening the food stock. Dangerous chemicals dissolved in irrigation water may enter the plant body and eventually contaminate the food chain system. Even an improvement in the watering practice of a rice field can decrease agricultural methane as a source of greenhouse gas significantly (Runkle et al., 2019). Food production itself accounts for~70% of freshwater withdrawal and 90% of freshwater consumption (Siebert et al., 2010). Conversely, arbitrary food consumption, production processes, and global food chain delivery threaten the global surface water (D'Odorico et al., 2018). In regard to energy issues, turbine generators, water cooling systems, and oil and gas excavation consume water. The consumption of the energy sector is much less than the food sector. Globally~15% of water is withdrawn, and 11% is consumed (Scanlon et al., 2017). Conversely, water has the inherent capability to dissolve many types of materials causing water to be easily contaminated. Purifying this wastewater requires several separation steps, consuming a large amount of energy. However, increasing industrial activities, oil exploration, and oil spill accidents decrease the quality and increase the quantity of wastewater, and threaten the aquatic ecosystem. Currently, existing oily wastewater treatments, for example using an open lagoon or centrifugation, require a large amount of space and long residence times. Simultaneously, centrifugation requires intensive energy consumption before reaching the maximum allowable limit specified in the standard of biological oxygen demand (BOD)/ chemical oxygen demand (COD) (Nasution et al., 2018). The advantages and disadvantages of several separation methods are resumed in Table 1 (Rasouli et al., 2021).
The surface interaction-based membrane has been introduced to overcome the limitations mentioned above (Table 1). Surface interaction is based on selective wettability through a hydrophobic/hydrophilic interaction between the surface membrane and solute (Padaki et al., 2015). A wide range of surface interaction-based membrane materials and their fabrication techniques have been intensively studied, for instance, metal mesh (Yang et al., 2019a;Zhu et al., 2020), carbon nanotube (CNT) (Hsieh et al., 2016;Saththasivam et al., 2018), graphene oxide (GO) (Wang et al., 2015a), synthetic polymers (Miao et al., 2020;Wang et al., 2020a), biopolymers-derived carbon (Yue et al., 2018a), Carbon nanofiber (Noamani et al., 2019), and biopolymer, including chitosan (Li et al., 2019a), alginate (Li et al., 2017b), and cellulose (Halim et al., 2019). The surface interaction-based membrane shows higher separation performance than conventional membrane (Wu et al., 2018, Guo et al., 2019, Yi et al., 2019. Cellulose is the most abundant natural polymer, sustainable, and not in competition with the food supply. Currently, the primary source of commercial cellulose is extracted from wood or cotton. However, cellulose also can be obtained from various sources such as agricultural waste (Suhas et al., 2016;Thakur and Voicu, 2016). In 2013 itself, cellulose-based polymers are 61.8% of biobased polymers (Shaghaleh et al., 2018). The utilization of cellulose, either fully or partially in membrane application, attracts many researchers (Tan et al., 2020). Life cycle assessment shows that biobased polymers membrane combined with environmentally friendly processes reduce environmental impact (Foroughi et al., 2021;Yadav et al., 2021).
Surface interaction-based separation using cellulose membranes is divided into three categories: hydrophobic, hydrophilic, and Janus. The various sustainability and biodegradability characteristics of cellulose as a polymer are advantageous to cellulose products. Several reviews of membrane-based methods on wettability have been conducted. For instance, Miller et al. reviewed several techniques to engineer the surface wettability of common polymeric membranes (Miller et al., 2017;Zarghami et al., 2019). More generally, Wang et al. (2016c) reviewed environmental applications based on wettability characteristics of the surface. Applications include anti-biofouling, enhanced vapor condensation and distillation, and water harvesting. Reviews of underwater superoleophobic membranes and in-air superhydrophobic membranes from all kinds of materials have been published (Chu et al., 2015;Brown and Bhushan, 2016;Lv et al., 2021;Sam et al., 2021), including their design, fabrication, and separation mechanisms.
Furthermore, the theoretical principles of surface wettability and its biomimicking properties have been reviewed (Nishimoto and Bhushan, 2013;Liu et al., 2013Liu et al., , 2014Bellanger et al., 2014;Wang et al., 2015c;Zhang et al., 2017b). Ma et al. (2016) specifically reviewed two types of materials with their surface wettability modified: filtration-based membranes and absorption-based absorbents. A review of cellulose membranes, in general, was conducted by Thakur and Voicu (2016) and Tan et al. (2020). Even though reviews on cellulose-based materials exist (Teisala et al., 2014;Liu et al., 2016a;Li et al., 2017a), the focus remains on superhydrophobic membranes. Hence, we will review bio-mimicking technology research using cellulose as either a supporting or active material in its application as an oil-water separator. This review exclusively handles cellulose of which a significant proportion is comprised of supporting materials or which plays a significant role as an active material. The cellulose properties vary widely with their source. However, this variation is due to the differences in how cellulose polymers are packed and assembled. At the molecular level, it is common knowledge that cellulose is a polymer of glucose through 1,4 β-glycosidic bonds. This review deals with cellulose from all sources.
However, we excluded cellulosic materials applied only for matrix reinforcement and significantly modified cellulose by introducing functional groups, including nitrocellulose and cellulose acetate. We included the cellulose nanocrystal (CNC) and 2,2,6,6-tetramethylpiperidine-1-oxylradical (TEMPO) oxidized cellulose nanofiber, TOCNF in this review as the sulfate group on CNC or carboxylic group on TOCNF is not the main purpose of synthesis but solely a side effect. This review will discuss the theory of water-wettability of cellulosic materials followed by the principles, mechanisms, preparation, and performance of superhydrophobic, superoleophobic, and Janus-based membranes consecutively. Organisms and their plant or animal-mimicking surface characteristics will be discussed in each sub-topic, followed by a description of how fabrication methods, material selection, and performance evaluation were applied.

Wettability
The wettability of a surface determines the efficacy and mechanisms of separation; whether it absorbs or repels oil. Surfaces with special wettability of certain liquids are referred to as a combination of "liquid + philic" if the surface attracts the liquid, and "liquid + phobic" if the surface repels the liquid. For example, hydrophobic, oleophobic, and hemophobic are used to refer to water, oil, and blood repelling surfaces, respectively. The Greek word "amphi" is also used to indicate that the surface acts the same way to both oil and water. However, amphiphobic is exaggerated as, in the air, oleophobic automatically becomes hydrophobic. The contact angle is the most simple and direct method to measure the wettability of a surface (Michel et al., 2001;Gupta et al., 2019). Young (1805) proposed that the contact angle of a liquid on a solid surface depends on the liquid surface tension, the surface energy of the solid, and the interfacial tension between the solid and liquid, as shown in Fig. 1(a). When the liquid is static, Newton's second law can be applied to obtain the contact angle using the equation shown in Eq. (1) (Young, 1805): where γ is the interfacial tension, and subscripts S, A, and L correspond to the solid, vapor, and liquid phases, respectively. The contacted surface area influences the interaction between the solid and liquid. Therefore, on a rough surface with a larger area, the interaction becomes stronger ( Fig. 1(b)). If r is roughness defined as the ratio between the total surface area to the projected surface area, the apparent contact angle equation is expressed by the Wenzel equation as shown in Eq.
(2) (Wenzel, 1949): where W is an apparent contact angle in the Wenzel state. The roughness of itself ranges from macroscopic to molecular scale (Youngblood and McCarthy, 1999;Huang and Gates, 2020). As the liquid droplet is smaller than the microscale groves, the drop will receive interfacial surface energy and roughness corresponding to the surface. Therefore, several attempts have been introduced to create nano to microscale roughness (Nosonovsky and Bhushan, 2007;Cho and Choi, 2008;Jeong et al., 2009;Cha et al., 2010;Li et al., 2013;Chen et al., 2015;Jiang et al., 2017;Sun et al., 2019). A void may exist between the solid surface and the liquid, thus decreases the solid-liquid contact area ( Fig. 1(c)). The composite condition increases the apparent contact angle mathematically expressed by the Cassie-Baxter equation (Cassie and Baxter, 1944), as shown in Eq. (3).
where CB is the apparent contact angle in the Cassie-Baxter state and f SL and f LA are the area fraction of the solid-liquid and liquid-air interfaces, respectively. As the void is filled with vapor, the Cassie-Baxter state is thermodynamically metastable. In contrast, Wenzel's state is considered to be stable (Murakami et al., 2014). However, some energy transition is applied to push the Cassie-Baxter state to the Wenzel state, as illustrated in Fig. 1(d). The summary of which equation to predict contact angle is depicted in Fig. 1(e). Figure 2 shows several characteristics of the surface. When a sessile drop of water comes into contact with a hydrophilic surface, it spreads out on the surface and sticks to it. Conversely, a superhydrophobic surface spontaneously makes a water droplet spherical and roll down. However, the superhydrophobic surface may have a high contact angle and be sticky, for example the surface of a rose petal (Bhushan and Nosonovsky, 2010). Some surfaces also have a low contact angle and are slippery, for example a Nepenthes pitcher (Wong et al., 2011;Semprebon et al., 2017) (Fig. 2(a)). The minimum angle of tilting for the droplet to roll is referred to as the roll-off or sliding angle ( S ), which is approximately similar to the hysteresis angle, or the difference between receding and advancing angles ( Fig. 2(b)). The receding angle ( R ) is the minimum angle at which the droplet volume decreases, while the advancing angle ( A ) is the maximum angle at which the droplet volume increases. The measurement of receding and advancing angles is shown in Fig. 2(c). The volume of an attached droplet increases for advancing contact angles and decreases for receding contact angles, maintaining a stable contact line but with an unstable contact angle. The droplet continues to reduce and advance, transferring the contact line and maintaining a stable contact angle, which is a true receding and advancing contact angle (Korhonen et al., 2013).
As membranes for oil-water separation are usually applied in underwater environments, modifications needed to be made to Eqs. (1)-(3) to accommodate for the interfacial tension between water, oil, and a solid membrane surface. Therefore, Young's equation was modified for water environments as follows (Eqs. (4)-(7)): where subscript O and W refer to oil and water, respectively. Equation (7) is a rearrangement of Eqs. (4)-(6) (Liu et al., 2009). As this equation is derived from Young's equation, it is only applicable to smooth, chemically homogeneous surfaces. However, several surface properties of materials change depending on the environment especially zeta potential for cellulose-derived materials (Cheng et al., 2017a). Due to the surface of oil droplets tend to be positively charged under water environment, the repellency increases significantly to the positively charged surface such as the surface of TEMPOoxidized cellulose nanofiber (Halim et al., 2020a). Depending on the surface characteristics, three approaches (an oleophobic, hydrophobic, or Janus membrane) were applied to separate an oil-water mixture or emulsion. Hydrophobic and oleophobic membranes refer to those with surfaces with water or oil contact angles above 90°. Whereas superhydrophobic and superoleophobic membranes refer to those with water or oil contact angles of 150°or higher.

Principles and mechanisms
The lotus leaf is a natural example of a superhydrophobic surface and has a wax layer on its surface Ensikat et al., 2011). The roughness of the lotus leaf on a microscopic scale enhances its hydrophobicity ( Fig. 3  (a)). Low sliding angles make it easy for water droplets to roll and sweep up all residues of dust on the leaf's surface, keeping the lotus leaf perpetually clean.
Superhydrophobic oily wastewater remediation technology mimics the inherent properties of the lotus leaf surface. The superhydrophobicity of the surface is the simplest of three major mechanisms. Water has a higher surface tension than most organic oils. Based on Young's equation, the water contact angle is almost always higher than the oil contact angle (Tuteja et al., 2007). A superhydrophobic surface is effective in the separation of water-in-oil systems. Oil will penetrate the membrane due to its higher affinity than water, as shown in Fig. 3(b). Based on this mechanism, superhydrophobic membranes are, therefore, an appropriate method for separating water-in-oil mixtures or emulsions. The food industry is currently dealing with palm oil effluents as one industry which uses water-in-oil systems.
3.2 Preparation of superhydrophobic membranes 3.2.1 Cellulose as a supporting material Due to cellulose's hydrophilic nature, as a supporting material it is modified to be non-polar, for example through esterification with acetate, or coated to become hydrophobic. In general, cellulosic materials, such as cellulose aerogel or sponge, are processed by coating, mixing, or combining them with nanoparticles or active materials to increase their roughness or to add various unique properties. Supporting materials with a high degree of roughness may not require a roughness enhancer. Cellulosic products that have been reported as supporting materials include paper (Wang et al., 2010;Cheng et al., 2017c;Zhang et al., 2020a;Xu et al., 2021), woven Cheng et al., 2017c;Zahid et al., 2017) and non-woven fabrics (Han et al., 2018), sponges (Peng et al., 2016a), and aerogel (Zhou et al., 2018;Li et al., 2019c). In general, cellulose as a supporting material is classified into three types: 3D porous structures (sponge or aerogel), fabrics (woven or non-woven), and paper ( Fig. 4(a)). The 3D porous superhydrophobic membranes purify wastewater by absorbing oil until it is saturated. Then, the oil will pass through the membrane pores. However, some cellulosic paper or cotton fabrics applied only as an absorbent, have also been reported to separate oil-water emulsions (Han et al., 2018;Li et al., 2019c;Zhang et al., 2020a).
Aerogel has longer permeating channels, but the fabrication process is complicated. The production process to produce paper as a filter material is simple. However, paper has low physical strength and is limited only to laboratory use, testing, and small instruments. On the other hand, both woven and non-woven cotton fabrics provide greater physical strength compared to paper or sponge membranes. SEM images of several types of membranes are shown in Figs. 4(b)-4(f).
Sponges and aerogel as supporting materials are produced from cellulose by dissolving or dispersing cellulose followed by regeneration. The mixture of sodium hydroxide and urea (Peng et al., 2016a), NaClO 2, and acetic acid (Li et al., 2019c) is used as a cellulose solvent. Another review  describes in more detail cellulose solvents and dissolution. For the electrospinning fabrication method, the porous structure of a sponge is generated by a foaming agent, as the electrostatic force produces only two-dimensional cellulose fiber mats. A use of CaCO 3 was reported for template pore formation by CO 2 gas generation (Peng et al., 2016a;Meng et al., 2020).
Furthermore, the aerogel form provides the function of not only oil filtration but also oil or organic liquid absorption. The light and robust structure of aerogel are excellent properties for an oil absorbent. However, the fabrication method of aerogel is complicated and requires advanced instruments. Direct freeze-drying of a cellulose solution (Peng et al., 2016a;Li et al., 2019c;Meng et al., 2020), and electrospinning followed by gas-forming then freeze-drying have also been reported to produce aerogel (Li et al., 2019c). Solely solvent exchange failed to produce high enough porosity or surface area. N'methylenebisacrylamide (MBA) (He et al., 2021), or epichlorohydrin (Peng et al., 2016a), was used as a crosslinking agent between cellulose fiber.
Commercial cotton woven fabrics are commonly used due to their superiority of physical strength compared to paper and cellulosic aerogel. Non-woven-based superhydrophobic membranes are generally made from synthetic materials (Gu et al., 2017;Zhang et al., 2019aZhang et al., , 2019b or cellulose derivatives such as cellulose acetate (Arslan et al., 2016). Cellulose solvent limitations and the complicated process hinder the usage of non-woven based fabrics as a superhydrophobic membrane. Hence, nonwoven fabrics are usually used as superoleophobic membranes, where the cellulose membrane is fabricated by dissolving and regenerating it to obtain its nanostructure. The structure of woven cotton facilitates active material development by the coating chemicals' deep penetration of every single cotton fiber (Zhou et al., 2013). Hence, even physical damage caused by stretching does not decrease its superhydrophobicity. For example, polyaniline (PANI) coated cotton and 1H,1H,2H,2H-perfluorooctyltriethoxysilane (C 14 H 19 F 13 O 3 Si) (PTES) coated cotton demonstrate strong superhydrophobicity even after stretching or scratching (Zhou et al., 2013 The coating of superhydrophobic membranes consists of particles as rough enhancers and/or polymers to change the surface wettability from hydrophilic to hydrophobic ( Fig. 5(a)). Two types of coating materials are applied after regeneration or during dissolution, i.e., particles to enhance roughness or polymers as active materials. The hierarchical structure of aerogel on a nano-to micro-scale ensures positive results of surface hydrophobization. Meso-to micro-scale particles are deposited on the surface of the substrate to increase the roughness. The particles may be synthesized or commercial particles used, as depicted in Figs. 5(b) and 5(c). Common particles used are silicon oxide (SiO 2 ) (Han et al., 2018), Zinc Oxide (ZnO) Cheng et al., 2017c), Iron oxide (Fe 3 O 4 ) (Peng et al., 2016a;He et al., 2021), TiO 2 (Yang et al., 2019b;Wang et al., 2020b;Arumugam et al., 2021), AlOOH (Yue et al., 2018b) or Cu nanoparticles (Li et al., 2019c). These metal oxide nanoparticles increase the roughness and embody additional functions in the membrane. For instance, TiO 2 nanoparticles' capability to catalyze organic breakdown can be used to enhance the membrane's self-cleaning function under UV irradiation (Kettunen et al., 2011). TiO 2 also has the function of switching surface properties from hydrophobic to hydrophilic and vice versa. The high surface area of CNF aerogel facilitates weight gain of coatings on the surface. In comparison, CNF aerogel contains 8%-8.6% TiO 2 , while a CNF film and filter paper may contain only 0.3wt% and 0.9wt% TiO 2 , respectively. The TiO 2 -coated CNF aerogel showed hydrophobicity, but hydrophilicity after UVirradiation, and then changed back to hydrophobic after storage in the dark for two weeks (Kettunen et al., 2011). Meanwhile, Fe 3 O 4 is mainly used for its magnetic properties and anti-bacterial activities. A sponge coated with Fe 3 O 4 can be moved freely by magnetic force in order to absorb oil (Peng et al., 2016a). Although the particles increase the surface roughness and enhance the surface affinity, too high concentrations negatively affect the contact angle. The optimum fly ash concentration is reported to be 0.5 wt%. However, a higher concentration than 1.25 wt% decreases the contact angle from 152°to 115° . Higher concentrations decrease adhesion strength due to particle aggregation. The statistical approach has been employed to obtain optimal conditions of the particles deposited on the surface (Zhang et al., 2020b).
The nanoflower  or nanoneedle (Wang et al., 2017b) is usually more desirable, as it provides higher roughness and is better at trapping air in the void of the membrane surface than spherical forms. Using nanoparticle deposition on the aerogel surfaces, the surface area increases from 28.3 to 94.8 to 108.6-195.5 m 2 /g (Zhou et al., 2018). Interestingly, the aerogel successfully separated not only oil-water mixtures but also oil-in-water emulsions.
Several techniques, including dip coating (Cheng et al., 2017c), simple immersion (Wang et al., 2010;Xu et al., 2021), chemical precipitation (Peng et al., 2016a;Li et al., 2019c), and chemical vapor deposition (Kettunen et al., 2011;Zhou et al., 2013), were applied to deposit the nanoparticles or polymers. Dip coating and simple immersion usually use commercially-available particles (Khan et al., 2018) with different shapes and sizes for each product. Nanoparticles synthesized through chemical precipitation are prepared in situ with substrate preparation or by subsequent preparation. In high temperatures or high acid concentrations, the cellulose may be hydrolyzed. A neutral pH and low temperature are desirable during particle synthesis. During hydrothermal synthesis, the cotton fibers cracked and created a flower-like TiO 2 crystal on the fiber surface (Figs. 5(d)-5(f)) Li et al., 2015).

Polymer coating
Cellulose and most particles contain hydroxyl groups in their surface, with high surface energy. Therefore, active materials to decrease surface energy are applied. The link between nanomaterials and surface energy decreasing agents may be chemically crosslinked or simply a physical coating. Hydroxyl groups on the surface of nanomaterials are utilized as the active side (Peng et al., 2016a). To increase the linkages between materials (polymers or particles) and a cellulose fabric or cellulose fibers, several crosslinking agents, including Kymene 557H, which contain carboxyl groups , and epichlorohydrin (Peng et al., 2016a), were used. Otherwise, active materials, such as epoxidized soybean oil (ESO), also act as an adhesive between cellulose and ZnO (Cheng et al., 2017c(Cheng et al., , 2018b. Chemical crosslinking increases the stability of the surface hydrophobicity. The necklace-like CNC/SiO 2 (Huang et al., 2019a) shows stronger adhesive properties and a more robust Cassie-Bexter's state compared to CNF only (Huang et al., 2016).
The dip-drying repetition process increases the contact angle and stability of the surface to surfactant pollution. Evaluation of the contact angles for SiO 2 and PDMS shows that the surface becomes relatively resistant to the surfactant impact even after repeating the dip-drying process 10 times. Higher dip-drying will increase the deposition of coating particles. The difference in contact angle before and after impact decreased from~20°for 10 dip-drying repetitions to~2°for 3 dip-drying repetitions (Han et al., 2018). Woven cotton coated with silica nanoparticles with an average diameter ranging from 7 to 40 nm and PDMS resin was capable of tolerating a hydrostatic head pressure of 2.56 kPa before leaking (Zahid et al., 2017). The surface coated with 1H,1H,2H,2H-perfluorooctyltriethoxysilane showed superhydrophobicity with a water contact angle varying between 142°to 162°, depending on the reaction temperature and time of TiO 2 . The coated fabric also demonstrated high laundering durability until five cycles. However, even surfaces with a high contact angle indicating relatively strong superhydrophobicity showed various sliding angles. The sliding angle decreased significantly to only 10°, and the adhesion strength remained at only 7 µN after a 20 h reaction time. This fact suggests a transition from Wenzel's to Cassie's state, as shown in Fig. 1(d).

Performance evaluation
The performance of a superhydrophobic membrane is evaluated by filtration performance and material performance ( Fig. 6(a)). Filtration performance measures flowrate, separation efficiency, and breakthrough pressure. Material performance measures adhesiveness, physical strength, and contact angle. Both filtration performance and material performance can be conducted under harsh conditions. The membrane is evaluated at least based on its oil and water contact angle, separation efficiency, and flow rate. Porous superhydrophobic surfaces will keep the air in their pores if immersed in water. Mirror-like observation by simply immersing the sample in water is used to confirm the existence of trapped air in the porous surface ( Fig. 6(b)) (Zhou et al., 2013;Han et al., 2016). The sample will show a silver appearance underwater. High water contact angles (>150°) and oil penetration as fast as possible are desirable. Figure 6(c) shows the separation process of superhydrophobic membranes. A photograph of polar liquid droplets on a superhydrophobic surface is shown in Fig. 6(d). Scanning electron microscopy (SEM) was applied to observe the surface morphology. Either Fourier transfer infrared (FTIR) or X-ray photoelectron spectroscopy (XPS) can be used to confirm the bonding between materials in the surface (Li et al., 2019c). Quantitatively, TG-DTA or dry matter content evaluates separation efficiency (Peng et al., 2016a). Flowrate depends on the performance of the membrane and oil viscosity. Higher viscosity generates lower flowrate (Cheng et al., 2017c). If the flow rate is very high, even only by gravitational force, the flowrate measurement is not necessary. However, if the flow rate is relatively low, it can be measured by simply weighing the filtrate over time divided by the membrane's surface area.
The stability and durability of a membrane are evaluated using several criteria. Immersion in solutions with extremely high or low pH (1-2 for strongly acidic and 12-13 for strongly alkaline solutions) or a high salt concentration (NaCl 3.5 wt%) for 1-72 h evaluates the durability of the membrane under harsh conditions . Adding a drop of a surfactant solution (4.1 mM SDS) (Zhou et al., 2013;Han et al., 2018) or heating at a temperature between 60°C and 120°C for 72 h (Zhou et al., 2013) evaluates the membrane surface's heat resistance. The dropping test using a surfactant aims to examine the influence of pollutants on the surface. Meanwhile, with water it aims to determine the possibility of transitioning from the Cassie-Baxter to the Wenzel state (Huang et al., 2019a). Stable surfaces only require drying to eliminate the effects of water dropping and recover their initial properties.
The detachment of the pollution evaluates a superhydrophobic membrane's self-cleaning ability by dissolving it in water. Methyl orange powder was sprinkled on the membrane surface, and then water was dropped on the surface. If the surface has a self-cleaning ability, all of the powder will be detached and carried away by the water leaving a clean surface . As a result, the surface's hydrophobicity decreased when the water molecules penetrated the surface and became attached to the mesostructure. Thus, the surface's affinity for sustainability was evaluated by subjecting it to high humidity (95% RH) for 72 h (Zhou et al., 2013) or immersing it in water for up to 7 days (Cheng et al., 2017c).
Various abrasion tests conducted with another cotton fabric, sandpaper for 30-600 scratch cycles at 50-260 g of weight applied (Zhou et al., 2013;Huang et al., 2015Huang et al., , 2019aWang et al., 2020b), finger wipes and knife scratches (Huang et al., 2019a), ultrasonication for 2 h, and laundering 30 times (Yang et al., 2019b) were conducted to evaluate the affinity of coated particles after physical damage. The resistance of the surface to UV irradiation was evaluated by exposure to UV light (ë = 254 nm) at a distance of 10 cm for 1-6 days Huang et al., 2019a). The material's biodegradability was evaluated for environmental concerns by measuring weight loss over time due to hydrolytic degradation in a phosphate buffer solution (pH 7.4, 37°C) (Cheng et al., 2017c). More than ten filtration cycles using the same filter were used to assess the membrane's continuity in separating the oil/water mixture/ emulsion. The aerogel was recovered after each separation process by washing it with an organic solvent such as ethanol Zhou et al., 2018) or a combination of alcohol and water (Zhou et al., 2013).
The summary of several superhydrophobic membranes is presented in Table 2. Several membranes are without roughness enhancer because the structure of supporting materials is rough enough. The non-cellulose based membranes are also without polymer coating due to nonpolar inherent properties of supporting materials.

Principles of superoleophobic membranes
Cellulose with abundant hydroxyl groups quickly absorbs water molecules into its structure. The simulation study shows that the hydrophilic functional groups are, in decreasing order, mixed-charged zwitterionic, hydroxyl, ethanolamine, oligo(ethylene glycol), amide, methyl groups (Cheng et al., 2017a). The hydrophilicity of cellulose surfaces provides underwater oleophobicity. In nature, fish scales (Waghmare et al., 2014;Halim et al., 2020a), clamshells (Liu et al., 2012), and mussels (Wang et al., 2015d) show underwater oleophobicity. Fish keep their skin surface clean even when living in harsh environments. The mucus layer of the fish skin is created by absorbing water by protein to produce skin moieties. This layer repels organic materials, including oil sticking to the fish's skin. The characteristics of fish skin inspired the production of a membrane that can repel organic materials, including oil and bacteria. This feature has the potential to create a self-cleaning membrane and immune to the organic fouling (Wu et al., 2018;Guo et al., 2019;Yi et al., 2019). The mucus layer in fish skin increases the oleophobicity, as indicated by the contact angle (Figs. 7(a)-7(c)). Fish scales also consist of micropapillae to increase their roughness, as shown in Fig. 7(d) (Liu et al., 2009;Waghmare et al., 2014). The composite electrospun cellulose with graphene oxide showed a corrugated structure similar to the micropapillae of a fish scale (Ao et al., 2017). In membrane application, the term mucus layer may refer to hydrogel (Rohrbach et al., 2014;Fan et al., 2015;Ao et al., 2018;Xie et al., 2020) or a liquid infused surface (Halim et al., 2020a;Ashrafi et al., 2021). Many hydrophilic surfaces lose their hydrophilicity and self-cleaning function due to a defect that occurs once a hydrophobic liquid comes into contact with the surface. Cellulose nanofiber demonstrated high stability even after oil defects . Figure 7(e) shows the mechanism of a superoleophobic membrane. First small droplets collide with each other or with large droplets to form even larger droplets. Sequentially, the droplets will be repelled once they touch the membrane surface. Therefore, superoleophobic membranes are suited to oil-in-water mixtures or emulsions.

Preparation of superolephobic membranes
Superoleophobic membranes generally require prewetting before use. However, in some non-cellulose membranes no prewetting was reported to be necessary (Peng et al., 2020). Different prewetting conditions produced different surface properties (Fan et al., 2018). However, some membranes maintained good hydrophilicity even in oil environments (Wang et al., 2015b). The prewetting process is necessary to determine the interaction between water molecules and membrane surfaces. Due to cellulose's natural oleophobic properties, cellulose requires less modification of its surface compared to superoleophobic and Janus membranes. The primary modifications include increasing surface roughness, crosslinking between cellulose molecules, cellulose and nanomaterials, or cellulose with supporting materials, and decreasing the membrane's pore size. Particles that increase the nanostructure roughness of the surface, such as BaSO 4 particles, were precipitated in the surface of the cellulose fiber .
The cellulose could act as either a supporting material or an active material ( Fig. 8(a)). For cellulose as a supporting material, the fabrication and preparation are relatively similar to superhydrophobic membranes. Cellulose as an active material mainly occurs in the form of a hydrogel. Paper-like (Rohrbach et al., 2014), aerogel (He et al., 2016;Fu et al., 2020) and sponge-like (Wang et al., 2015b;Halim et al., 2019), woven (Dai et al., 2019), and non-woven membranes (Almeida et al., 2020) were reported. Non-cellulose supporting materials provide relatively higher physical strength compared to cellulose filter paper or sponges. A wide range of mesh from nylon (Lu et al., 2014;Cheng et al., 2017b) to metal mesh (Ao et al., 2018;Xie et al., 2020) were reported. However, all biodegradable organic materials are preferred. A 3D printing of cellulose acetate mesh followed by cellulose regeneration using an ethanol/NaOH solution was reported by Koh et al. (Koh et al., 2019). The pore size and structure are easily engineered.
Moreover, production using the 2D printing of CNC on a non-woven cellulose acetate makes the pattern, size, and CNC ratio easy to control in order to optimize oil retention (Almeida et al., 2020). Figure 8(b) shows an SEM image of a superoleophobic membrane for cellulose as an active material. Here, cellulose hydrogel coats a metal mesh. Figure 8(c) shows cellulose acting as both a supporting material and as an active material. Figure 8(d) shows cellulose as a supporting material with coated particles, and Fig. 8(e) shows a cellulose sponge.
The quantity of cellulose deposited to the supporting materials influences the hydrophilicity and oleophobicity of the membrane. The nylon's contact angle changed from 88.7 without a CNF usage to 175.9 with a CNF usage of 0.73 g/m 2 , due to the cellulose area coverage. After all supporting material is covered, the contact angle is independent of the CNF concentration (Cheng et al., 2017b). The chitosan forms a self-assembly structure that produces a micrometer dot on the surface during composition with chitosan. A decrease in XRD intensity indicates decreasing crystallinity and hydrogen bonding. Exposing more hydroxyl groups to the surface is verified by increasing the OH peak by FTIR (Peng et al., 2016b). Roughness to form a water cushion, repellency of polar liquids, and the membrane surface's zeta potential enhance its underwater oleophobicity. Underwater, the oil droplets are slightly negative (Roger and Cabane, 2012). Negative zeta potential of the membrane surface will enhance its oleophobicity. The TEMPO-oxidized cellulose showed a high underwater oil contact angle compared to the Mechanical CNF due to higher negative zeta potential (Halim et al., 2019). The surface charge of the membrane interacts with the surfactant and determines the separation efficiency. Due to the charge of the membrane being affected by the environment's pH, feed plays a key role. Anionic surfactants, such as sodium dodecyl benzyl sulfonate (SDBS), were the least stable due to their higher surface tension and positive zeta potential. Cationic surfactants, including hexadecyltrimethylammonium bromide (CTAB), and neural surfactants, such as polysorbate 80 (Tween 80), demonstrated higher stability. By controlling the pH, the membrane's zeta potential can be adjusted to positive or negative to selectively increase the filtration performance of a particular surfactant (Kim et al., 2019).
Similar to the hydrophobic membranes, electrospinning was also applied to the oleophobic membranes followed by freeze-drying to fabricate aerogel. A thin layer of cellulose membranes was reported on different sources of cellulose nanofiber (Mautner et al., 2014), cellulose nanocrystal (Cheng et al., 2017b), and cellulose solution in a urea/ NaOH system (Zhou et al., 2014;Peng et al., 2016b). During heating, crystallization occurred in the cellulose solution, expelling the cellulose to form a nanosheet. The nanomembrane shows a transparent and chiral nematic substance, due to the cholesteric self-assembly of CNC (Cheng et al., 2017b). The nanopaper of CNF fabricated by paper making is affected by the role of zeta potential in coagulation and formation (Mautner et al., 2014). The thickness of the paper was a linear function of the concentration of the quantity of cellulose suspension. At the same time, thickness reversely affects the flowrate almost in linear function (Zhou et al., 2014;Cheng et al., 2017b). Cellulose sponge was fabricated by routine dissolution and regeneration with the template of Na 3 PO 4 (Halim et al., 2019, 2020b) or Na 2 SO 4 $10H 2 O (Wang et al., 2015b). Reinforced fiber can be added to increase mechanical strength. Sandwiching of nanoporous layers and a microporous layer was also reported (Wang et al., 2015b).
Additional polymers increase oleophobicity by increasing roughness. PEI modified CNF aerogel followed by grafting with PDMAEMA increases the surface area from 4.30 to 12.72 m 2 /g due to the brushiness of the polymer increasing the roughness of the surface . Coating and additives also produce additional features, including switchability and absorption ability. Switchable oleophobicity of cellulose aerogel can also be tuned by nanoparticles TiO 2 using UV radiation, as explained in the previous section (Kettunen et al., 2011). Another method is generating CO 2 gas-sensitive switchable surfaces by coating with PDMAEMA . PDMAEMA is in the deprotonated state in the absence of CO 2 , while CO 2 exposure protonates the brush polymer and acquires a positive charge. The flux of oil (hydrophobic state) is four times higher than the water flux (oleophobic state) due to effective porosity reduction; however, the separation efficiency is almost one-ninth of the oleophobic state. PDMAEMA reduces the porosity by bonding with H 2 O. As cellulose has an abundance of hydroxyl group, the membrane is also used as a dye absorbent (methylene blue, rhodamine B, and gentian violet) Huang et al., 2019b). By composting with graphene oxide, the dye removal efficiency is up to 97% with flux from 170 to 300 L/m 2 /h . The cationic dye is absorbed by an electrostatic interaction between the dye and oxygen group of graphene oxide (GO). Therefore, a desorption agent, for example HCl, is used to renew the membrane. A composite is also necessary to elevate the efficacy of the membrane's performance. The CNF membrane without graphene oxide was incapable of separating an oil-water mixture. The graphene oxide affected the microstructure and wettability of the membrane (Ao et al., 2017).
For the coating of active materials, the cellulose solution was coated onto the supporting materials followed by drying. The catalyst performed chemical crosslinking to increase the adhesion between cellulose and the supporting material or nanoparticles (Rohrbach et al., 2014;Xu et al., 2019). Types of crosslink and catalyst materials included citric acid and glutaraldehyde (Rohrbach et al., 2014), guar gum and NaIO 4 (Dai et al., 2019). Epichlorohydrin (Huang et al., 2019b) and polyamide amine-epichlorohydrin (He et al., 2016) were also used as crosslinking materials. Crosslinks typically used are esterification between aldehyde, and alcohol groups between coated materials, coated materials and cellulose, or between cellulose. The polyamide amine-epichlorohydrin waterinsoluble prevents detachment of fibers during rewetting (He et al., 2016). The concentration of guar gum and NaIO 4 plays a significant role in the success of a reaction. Under 1 wt% of GG, no reaction happens for£0.0022 and≥2.2 g NaIO 4 /g GG. Under 0.022 g NaIO 4 /g GG, no reaction happens for 0.5wt% and 1wt% GG. Crosslinking only by annealing at 150°C was reported by Almeida et al. (Almeida et al., 2020). The annealing process changes the color of CNC to be darker. Crosslinking agents provide more crosslinking density between materials, which enhances the separation efficiency and improves physical strength (Huang et al., 2019b).

Performance of superoleophobic membranes
The performance evaluation criteria for superoleophobic membranes are similar with those for superhydrophobic membranes except for the evaluation of performance conditions. For instance, instead of using water as a contact angle measurement in the air, the superoleophobic membrane uses oil underwater to evaluate the contact angle. This underwater contact angle of the surface is the first characteristic to be investigated ( Fig. 9(a)). Otherwise, the strength of the membrane is evaluated by mechanical compression Halim et al., 2019) or tensile strength (Ahmed et al., 2014). For mechanical compression, a compression strain of up to 85% was reported Halim et al., 2019). For tensile evaluation, the stretching rate is 1 mm/min, and the stressstrain curve was recorded (Ahmed et al., 2014). A polymer coating of polyvinyl alcohol increases filter paper's physical strength by providing robust hydrogen bonding and protecting the acidic or alkaline condition . Cellulose coating on a PVDF-HFP nanofiber membrane also increases the mechanical tensile strength from 5.5 MPa to 8.6 MPa with the elastic modulus increasing from 17.4 MPa to 54 MPa with 15% cellulose content (Ahmed et al., 2014). Coating with paraffin wax shows an improvement in elastic modulus from 521.3 to 1.2 GPa. In contrast, the elongation break decreases from 9.4% to 4.1%. The improvement is assumed to be due to the physical crosslink between fiber and paraffin wax. However, coating with PDMS-b-PEO does not demonstrate any improvement in mechanical strength (Paul et al., 2016).
PDMAEMA grafting increases the compression stress from 30 kPa for PEI modified CNF aerogel to 61 kPa when compressed to 50% of its thickness due to an increase in density from 37.17 mg/cm 3 to 57.58 mg/m 3 . The compression strength of cellulose hydrogel 7.9 to 14.45 MPa depends on the regeneration conditions. Low temperature slows the cellulose regeneration producing more compact and homogenous stacking of the cellulose chain. Poor solvents also cause an irregular and bulky structure .
Higher cellulose content decreases the tensile strength due to high content; the cellulose is very thick. During drying, shrinking can take place and cracks appear. With a lower content, the cellulose glues the fiber, increasing the stiffness of the membrane. H + in the acidic condition decreases the mechanical strength by decomposition of the cellulose polymer. Ion H + and OHalso break the hydrogen bonding of cellulose. Underwater, the annealed CNC coating captures the microdroplets. The microdroplets coalesce with each other to form a larger oil droplet, which is removed from the surface (Almeida et al., 2020). Polymer coating of polydopamine acts as a glue between cellulose and BaSO 4 particles .
Other parameters that affected separation performance include the pore size of the membrane, tortuosity, and breakthrough pressure. These three parameters correlated with each other. The membrane requires a slightly lower pore diameter than the droplet diameter to prevent the oil droplet from passing through the membrane or entering the membrane's pores (Schutzius et al., 2017). The coating also decreases the pore size, providing not only surface repellency but also a size-based barrier. By decreasing the pore size, the membrane's performance enhances from only separating oil-water mixtures to separating oil-water emulsions (Wang et al., 2017a). During filtration, the repellency pushes against the applied pressure. The maximum pressure before the droplet penetrates the membrane is called the breakthrough pressure, expressed as Eq. (8): wherein P, γ OW , d, and are the breakthrough pressure, interfacial surface tension between oil and water, the diameter of a pore, and OCA against water, respectively. As filtration force is limited to gravity, P is equivalent to gravitational pressure P g ¼ gh, where is the density of oil, g is the gravitational acceleration (9.81 m/s 2 ), and h is the height of the oil-water mixture in the filter holder during the experiment ( Fig. 9(b)). From the equation, the maximum diameter of the filter pore can be calculated. The typical surfactant-free oil droplet emulsion is larger than 20 µm; therefore, a membrane with lower pore size of 20 µm will block the oil and the separation efficiency will depend on the membrane pore size (Ahmed et al., 2014). For instance, 3D printed cellulose using hexadecane as an oil model showed that at a diameter lower than 280 µm, the separation efficiency was higher than 91% but decreased to 74% at a diameter of 375 µm and 35% at a diameter of 545 µm (Koh et al., 2019). A cellulose nanofiber membrane with a pore size of approximately 21.6 nm was reported to separate a surfactant stabilized oil emulsion with an oil diameter of 119 nm (Zhuang et al., 2020). However, the pore size is not the only parameter to determine the separation efficiency. Tortuosity and flow dynamics that initiating the oil droplet collision increase the separation efficiency. However, even the viscosity of the oil determines the flow dynamics; viscosity shows no correlation with separation efficiency, as shown by xylene with a viscosity of 0.6 cP, which has a separation efficiency of only 94%. However, both PDMS with a high viscosity of 97 cP, and cyclohexane with a low viscosity of 1 cP show a high separation efficiency of 99% (Koh et al., 2019). Large pore size increases flux; however, the separation efficiency and breakthrough pressure decreases significantly. The oil intrusion evaluation observes the membrane's maximum static pressure to retain oil ( Fig. 9(b)). The electrospinning cellulose composited with graphene oxide had a pressure of 2.56 kPa for a membrane thickness of~100 µm (Ao et al., 2017) and 1.94 kPa for cellulose hydrogel coated mesh (Ao et al., 2018). The colander from cellulose-coated nylon mesh also showed a high breakthrough pressure of more than 1.26 kPa (Lu et al., 2014). The breakthrough pressure for cellulose coated with polydopamine and BaSO 4 was 1.25 kPa .
Not all coating methods decrease the porosity of the membrane. Paraffin wax decreases the porosity from 53% to 43%. However, PDMS-b-PEO does not have a significant effect on porosity (only decreasing up to 53.5%). In contrast, a combination of paraffin wax and PDMS-b-PEO decreases the porosity to 41% (Paul et al., 2016). The oil coalesces during separation also increasing the separation efficiency by increasing the diameter of the oil droplets. With a flow velocity of 6.7 Â 10 -4 m/s, the minimum diameter for oil droplet coalescences and upward flow is 89 µm (He et al., 2016). Nanoporous membranes are fabricated by freeze-drying without a forming agent, whereas microporous membranes are fabricated with a forming agent. Microporous membranes are only capable of separating oil-water mixtures, whereas nanoporous membranes are capable of separating oil-water emulsions (Wang et al., 2015b). Nanoporous membranes can be measured by the molecular cut-off of polyethylene oxide for several molecular weights of 100 kDa, 300 kDa, and 600 kDa (Cheng et al., 2017b). There is a trade-off between separation efficiency and flow rate. Increasing separation efficiency by decreasing pore size worsens the flow rate. In the same manner, a higher oil droplet diameter also decreases the flow rate by blocking the membrane surface. The flow rate decreases from 9.16 to 1.25 L/m 2 /h/ bar by increasing the oil droplet diameter from 119 to 152 nm (Zhuang et al., 2020). Freeze-drying cellulose aerogel at a concentration of 0.0197 g/mL has a deionized flow rate of 27022 L/m 2 /h with oil rejection being 56.3%, while cellulose aerogel at a concentration of 0.0337 g/mL has a deionized flow rate of 2405 L/m 2 /h, and oil rejection of 98.6% (He et al., 2016).
The oleophobic membranes' performance and stability were evaluated based on their cleaning ability, water contact angle, ability to withstand harsh conditions, and physical damage to the surface. Stability in chemically harsh or corrosive conditions can be evaluated by filtration using an acid solution of 0.1-1 M HCl (Ao et al., 2017;Cheng et al., 2017b;Yang et al., 2020), pH = 1 (Koh et al., 2019) or 0.1 M muriatic acid (Lu et al., 2014), alkaline 0.1-1 M NaOH (Lu et al., 2014;Ao et al., 2017;Cheng et al., 2017b;Yang et al., 2020) or pH = 11-13 (Dai et al., 2019;Koh et al., 2019), or high salt concentration of 0.1-1 M (Lu et al., 2014;Cheng et al., 2017b;Yang et al., 2020) or 5 wt%-10 wt% NaCl (Peng et al., 2016b;Ao et al., 2017;Dai et al., 2019), or seawater (Koh et al., 2019). Furthermore, the membranes were exposed to harsh conditions for different durations of time. Artificial harsh condition evaluation involves the immersion of a membrane in saturated salt for 24 h (Ao et al., 2018), in an acidic condition of 2-8 M H 2 SO 4 for 10 min-30 days , in an alkaline condition of NaOH for 10 min-30 days , and in water for 90 days (Lu et al., 2014). The alkaline condition was able to coalesce the oil to form a stable oil emulsion even without a surfactant due to the oil droplets being negatively charged (Koh et al., 2019).
Hydrophilicity and self-cleaning performance under oil are evaluated by measuring the water contact angle under oil (Cheng et al., 2017b) or under oil after one day of immersion (Koh et al., 2019). Separation efficiency is qualitatively evaluated using the Tyndall effect (Cheng et al., 2017b) or microscope observation (Wang et al., 2017a) (Figs. 9(c) and 9(d)). Quantitative evaluation is conducted using TGA or drying (Halim et al., 2019). Type of oily wastewater affects the separation efficiency by its wastewater content such as pH or oil droplet size. Yi et al.
Stability under thermocycling is evaluated through five cycles of freezing at -18°C and thawing at room temperature, or by heating to 60°C and cooling to room temperature (Dai et al., 2019) to observe the water leaching out of irreversible phase change. Mild temperature evaluation is conducted through the use of a solution at a temperature of 10°C-80°C (Cheng et al., 2017b), or sonication at a power of 180 W for 30 min to evaluate the coating bonding Koh et al., 2019). Sandpaper abrasion with 100 g loading and adhesive tape with 100 N/m adhesive strength was used (Huang et al., 2019b). Resistance to UV light was evaluated at a 365 nm wavelength (Huang et al., 2019b).
The antifouling and reusability of the membrane (recovery rates) can be observed using 10-60 cycle filtrations and evaluating the flowrate or separation efficiency during cycling. Superoleophobic membrane shows antifouling properties by maintaining high flux and separation efficiency. If the coating is in hydrogel form, the rheology measurement can be applied under a frequency of 10 rad/s and strain at 1%. For clear appearance purposes, the water is dyed with methylene blue (Huang et al., 2019b), while the oil is dyed with Sudan III (Huang et al., 2019b) or red oil O (Rohrbach et al., 2014). Table 3 summaries several superoleophobic membranes. Generally, cellulose materials, either as supporting material or as active material has rough enough surface.
5 Janus surfaces 5.1 Principles of Janus membranes Janus materials have two surfaces with different properties in reference to the two-faced Roman god, Janus. This means Janus materials have two sets of opposite properties; Hydrophobic on one side and hydrophilic on the other side. In nature, plants and creatures that have two opposite properties include Salvinia (Barthlott et al., 2010) and Namib beetles (Parker and Lawrence, 2001). The eggbeater-like fur of Salvinia plants has a hydrophobic coating on the stalk and a hydrophilic coating on the terminal of the fur. This structure stabilizes the air-water interface, especially during rippling conditions (Figs. 10(a)-10(d)). The Namib beetle has a hydrophilic side to collect water from fog and a hydrophobic side to direct the water to the beetle's mouth (Parker and Lawrence, 2001;Zhai et al., 2006) (Figs. 10(e) and 10(f)). Therefore, a Janus membrane has superhydrophilic on one side and has superhydrophobic on another side as shown in Figs. 10(g) and 10(h).
Janus membranes have a hydrophilic surface on one side   Peng et al., 2020 Hexadecane/, decane/, hexane/, toluene/water mixtures 32.5 L/m 2 /s~9 8% Gravity and a hydrophobic surface on the other side. Janus membranes are categorized into three types, as depicted in Fig. 11(a) (Yang et al., 2016). Therefore, the separation mechanism of a Janus membrane is influenced by its type, functioning through demulsification and rejection employed during oil-water separation.
If we define A as the hydrophobic layer and B as the hydrophilic layer, for the "A and B" types, the A and B layers have a similar thickness. A third layer may exist between the A and B layers. Membranes of this type can separate either oil-in-water or water-in-oil emulsions, depending on which surface is in contact with the emulsion, as shown in Fig. 11(b). The hydrophilic face is suited to separating oil-in-water emulsions, and vise-versa; the hydrophobic face is suited to separating water-in-oil (Yue et al., 2018c;Lv et al., 2019). If the hydrophobic layer faces the feed, the membrane is able to separate water in oil. When the hydrophilic layer faces the feed, the membrane is able to separate oil in water emulsions. Therefore, Janus membranes are switchable depending on the type of wastewater. This principle of the "A and B" type is similar to superhydrophobic membranes or superoleophobic membranes, as previously discussed in parts 3 and 4, respectively. The "A and B" type has been reported by some researchers (Gore et al., 2016;Gupta and Kandasubramanian, 2017;Gore and Kandasubramanian, 2018;Yue et al., 2018c;Lv et al., 2019;Hu et al., 2020).
"A on B" and "B on A" mean the A layer is thinner than the B layer, and the B layer is thinner than the A layer, respectively. For A on B (Fig. 12(a)), the superhydrophobic surface retained oil (Tian et al., 2014). If moving in a positive direction, when water droplets contact the hydrophobic side, they spontaneously penetrate the surface due to its low breakthrough pressure. However, if the direction is reversed (Fig. 12(b)), when water droplets contact the hydrophilic side, they tend to spread and are blocked by the hydrophobic surface. This phenomenon shows that breakthrough pressure is affected by which side faces the wastewater. Furthermore, breakthrough pressure is also affected by the fluorine content as a hydrophobic polymer attached. Tian et al. (2014) have reported on the A on B type.
For B on A (Fig. 12(c)), water and oil droplets pass through the hydrophilic side. Yet, the water is retained in the superhydrophobic side, meaning only oil passes through. Wang et al. (Wang et al., 2016d have reported on the B on A type. The hydrophilic side of the Janus membrane disrupts the emulsion's stability by ionizing water, initiating coalescence of the oil droplets, and transporting them to the hydrophobic side (Wang et al., 2016d;2016e). The hydrophobic side selectively permeates the oil. This mechanism allows the membrane to separate oil in water emulsions, although the membrane's pore size is larger than the oil droplets (Wang et al., 2016d). In terms of oil-water mixtures, the membrane could separate the mixture regardless of which side faces the feed (Wang et al., 2016e). If the hydrophilic side faces the feed, the water and oil pass through the hydrophilic side, but the water is retained on the hydrophobic side. If the hydrophobic side faces the feed, water is retained, and only oil passes through the membrane. However, the membrane could separate oil-in-water emulsions if only the hydrophilic side faces the feed. When the hydrophobic surface faces the feed, no liquid is allowed to permeate the membrane, as shown in Fig. 12(d).

Preparation of Janus membranes
There are not many publications on using Janus membranes from cellulose for oil-water applications. Figure 13  summarizes the general fabrication method of Janus membranes. Fabrication of Janus membranes is tedious, involving several steps, including coating, dissolution, and polymerization. Therefore, in situ fabrication reduces time and cost.
The A and B type was the most common type of Janus membrane reported. Membranes were fabricated by electrospinning on one side of cotton (Gore et al., 2016;Gore and Kandasubramanian, 2018), single-faced coating using a Meyer rod (Gupta and Kandasubramanian, 2017), floated deposition (Hu et al., 2020), sequential surface modification (Lv et al., 2019), and vacuum filtration (Yue et al., 2018c). Sodium laurate (Yue et al., 2018c) or stearic acid (Lv et al., 2019) were used for the hydrophobic side, and MnO 2 (Yue et al., 2018c) or immobilized Ag (Lv et al., 2019) were used for the hydrophilic surface. ZnO (Yue et al., 2018c), Ag (Lv et al., 2019), nanoclay (Gore and Kandasubramanian, 2018) particles were used as roughness enhancers. Subsequently, the cellulose fabric was immersed in the Ag solution and stearic acid while one side was protected by attaching tape. The hydroxyl group on the surface of the ZnO underwent a crosslinking reaction with sodium laurate. The ZnO was coated by in situ hydrothermal precipitation while an MnO 2 nanowire was fabricated hydrothermally, followed by deposition on one side using a vacuum.
The A on B type was fabricated using vapor treatment (Tian et al., 2014). In this system, the membrane is made of woven cotton vaporized with 1H,1H,2H,2H-perfluoro octyl trichlorosilane (POTS) (Tian et al., 2014). A POTS was placed inside of a crystallizing dish then covered with cotton fabric with a distance of~6 cm between the POTS and cotton fabric. A glass plate was used to press the cotton fabric. The POTS vapor then diffused and reacted with the hydroxyl group of cotton by a salinization reaction.
The B on A type was fabricated using single-faced photo crosslinking (Wang et al., 2016e) and segment immigration (Wang et al., 2016d). The cotton fabric was coated with PDMAEMA (Poly(N,N dimethyl aminoethyl methacrylate)) and PDMS (poly(dimethylsiloxane)) to create a hydrophilic and hydrophobic side (Wang et al., 2016e). First, the cotton was coated with diblock copolymer PDMS-b-PCEA, with PCEA referring to poly(2-cinnamoyloxyethyl acrylate). The polymer was then crosslinked using irradiation, followed by uncross-linked polymer extraction with hot tetrahydrofuran (THF). The fabric was then coated with trimethoxysilyl propyl methacrylate (TMSPMA) and PDMAEMA (Wang et al., 2016e). During segment immigration, the cotton was coated with a diblock copolymer of P1-b-Am. P1 refers to P(MMA-s-HEMA)-g-PDMS, and Am refers to PDMAEMA. MMA and HEMA refer to methyl methacrylate and 2-hydroxyethyl methacrylate, respectively. The hydrophilic hairy brush PDMAEMA would rise to the top of the PDMS when in contact with water, creating a Janus structure (Wang et al., 2016d).

Performance of Janus membranes
Similar to the two previous membranes, the performance of Janus membranes is evaluated based on several criteria, including wettability, separation efficiency, flow rate, and how they maintain their wettability, separation efficiency, and flow rate when exposed to various different conditions. The previous methods used for superhydrophobic and superoleophobic membranes are also applicable for the Janus membrane. For instance, to evaluate the membrane's performance under harsh conditions, a saline solution with a NaCl concentration of 10%-40% (Gore and Kandasubramanian, 2018), at a low temperature ( -20°C to 0°C) (Gupta and Kandasubramanian, 2017;Gore and Kandasubramanian, 2018) was used. A and B types are capable of separating oil-water mixtures (Gore et al., 2016;Gupta and Kandasubramanian, 2017;Gore and Kandasubramanian, 2018;Hu et al., 2020), oil-in-water (Yue et al., 2018c;Lv et al., 2019), or water-in-oil (Yue et al., 2018c;Lv et al., 2019), depending on the materials and fabrication methods used. The A on B type, and B on A type are more common than the A and B type, especially to separate oil-in-water or a water-in-oil emulsions. B on A has a high water flux of 1.5 Â 10 3 L/m 2 /h and a separation efficiency of almost 100%. Hexadecane, toluene, and chlorobenzene were used as oil models at concentrations ranging from 1.0% to 30% (Wang et al., 2016e). This high flux may be due to the pore Fig. 12 Separation mechanism of A on B and B on A with A as hydrophobic and B as hydrophilic. A on B is suited to separating waterin-oil emulsions when the A layer faces the feed (a) and no penetration occurs when the B layer faces the feed (b). B on A type is suited to separating oil-in-water emulsions when the B layer faces the feed (c) but no penetration occurs when the A layer faces the feed (d).
size of the filter being higher than that of the oil droplets (Wang et al., 2016d). Table 4 summaries the several Janus membranes.

Summary and outlook
Three kinds of membranes were summarized based on their surface wettability. Superhydrophobic membranes penetrate oil and reject water. Therefore, superhydrophobic membranes are suited to water-in-oil systems. Superhydrophobic membrane suits when oil collecting is preferred, for instance, oil collecting of oil spill accident or high oil content wastewater. Oil contaminants will be adsorbed leaving cleaner water. The superhydrophobic membrane provides an efficient and continuous process compares to the aerogel adsorbent. However, most industrial oily wastewater is an oil-in-water emulsion or light oil-water mixture so that oil collecting is insufficient and water collecting is preferred.
Superoleophobic membranes reject oil and penetrate water. Therefore, superoleophobic membranes are suited to oil-in-water systems such as household wastewater, refinery process, food processing industries, petrochemical industries, tannery industries, or steel processing industries. The organic oil contaminant could be fats, kerosene, diesel, gasoline, or other types of hydrocarbons. In terms of high viscous oil contaminants, the superoleophobic membrane will decrease the pressure drop because low viscous water will pass through the membrane. Furthermore, due to the water mostly has a higher density than oil, the separation system can be arranged to utilize gravitational force alone. The oleophobic characteristic also repels the non-polar substance minimizing fouling tendency.
However, both superhydrophobic and superoleophobic membranes require smaller or comparable pore size to droplet diameter. Therefore, the high flow rate is a trade-off for low separation efficiency. Janus membranes can separate oil droplets even when the pore size is larger than the oil droplets. Janus membranes tend to be more effective at separation, and the flowrate is comparable to superhydrophobic and superoleophobic cellulose-based membranes. However, further experimentation is necessary due to the lack of literature and theoretical knowledge on cellulose-based Janus membranes. Comparison to the commercial and actual oily wastewater are summarized in Table 5. In general, superhydrophobic, superoleophobic and Janus membranes show relatively higher flux and separation efficiency.
The laboratory-scale and homemade oily wastewater model show successful results, but the membranes have not been fully implemented in actual oily wastewater treatment. Many challenges must be overcome to fill the gap between theory and experimentation at the laboratory scale and industrial application, such as the physical strength of materials, the biodegradability of materials, leaching from used nanoparticles, reusability, modules fabrication, and actual oily wastewater treatment. Even though cellulose is biodegradable, additive materials used to advance its functionality are still non-biodegradable. The life cycle assessment has to be conducted especially for composite membrane. Nanoparticle leaching must be considered, especially when using metal or potentially toxic chemicals. Reusability and membrane fouling are related to the wettability of the membrane. Current research conduct testing for approximately hundreds of filtration times. Batch and non-module separation system are used for evaluation on a laboratory scale. However, on an industrial scale, module and continuous systems are Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.