A major drawback in any membrane water treatment system is fouling of the membrane surface. Various aspects of mass transport lead to the attachment, accumulation or adsorption of particles onto membrane surfaces and pores, causing membrane fouling [47
]. Although membrane technology has advantages over conventional water treatment, membrane fouling continues to be a major operational problem [48
]. Fouling of the membrane can occur as a result of a variety of contaminants in the feed water including inorganic compounds, colloidal or particulate matter, dissolved organics, chemical reactants, and microorganisms and microbial products [49
]. However, it is considered difficult to predict the origin and extent of the effects of the simultaneous occurrence of these fouling mechanisms [47
]. Although fouling can be reduced by employing low-fouling membrane materials, optimising the system operation conditions, and cleaning membrane units, it is unavoidable when feed water contains certain materials. Membrane fouling results in the requirement of pre-treatment of source water and membrane chemical cleaning, which incurs additional costs and increases energy consumption. Therefore, minimisation of the fouling is the key to success and cost-effective membrane operation [52
]. The fouling mechanisms in FO systems may be different from and more complicated than those in the pressure-driven membrane processes [10
]. For instance, Luo et al. [54
] compared bacterial variation, contaminant removal and membrane fouling between an osmotic membrane bioreactor (OMBR)-RO and a conventional MBR-RO processes. Due to salinity build-up in the bioreactor, biomass characteristics and microbial community structure were altered, and soluble microbial product (SMP) and extracellular polymeric substances (EPS) in the mixed liquor increased in the OMBR. As a result, fouling was more apparent on the FO membrane in the OMBR compared to the microfiltration (MF) in the conventional MBR process where only minimal fouling occurred. However, the OMBR had a less adverse impact on the subsequent RO fouling, where a patchy fouling layer was observed in the OMBR-RO while a homogeneous cake layer was formed in the MBR-RO. The FO membrane effectively prevented foulant permeation into the draw solution, and reduced fouling in the downstream RO membrane during the OMBR-RO process, while substantial humic- and protein-like substance and inorganic salts were detected from the RO foulants in the conventional MBR. Understanding the fouling behaviour in the FO processes is particularly important because both sides of the FO membrane are in constant contact with impaired waters: the active layer with impaired water and the support layer with synthetic or natural DSs (e.g., seawater or brine). For instance, when a FO system was used to treat actual secondary effluent wastewater on the active layer and pre-filtered seawater from Red Sea was used as the DS, natural organic matter (NOM)-biopolymer fouling occurred on the active layer and transparent exopolymeric particle (TEP) fouling occurred on the support layer [55
]. Therefore, there is a critical need for a systematic understanding of membrane fouling behaviour and for the development of strategies for fouling mitigation.
3.1. Organic Fouling
Organic fouling occurs by organic macromolecules found in the feed water. It can be caused by hydrophobic, transphilic and hydrophilic fractions in the feed water. Some studies have reported that the hydrophobic fractions (e.g., humic acids) are the main source of fouling in membrane filtration, but other research has indicated that hydrophilic (e.g., polysaccharide) fractions are the major problem [56
]. Another study documented that the adsorption tendency of the polysaccharides on membrane surfaces is three times higher than that of humic acids [58
]. For wastewater treatment using the membrane bioreactor (MBR) application, organic fouling is the major issue where organics are the precursor of biofouling. This is the dominant foulant in MBR applications [59
]. It is also known that the concentration of the organic matter in seawater is relatively low (about 1–3 mg/L) and accordingly the portion of organic foulants is small in comparison with inorganic constituents. However, seawater organic matter is a more severe problem to be solved especially in SWRO, as it can also be converted to biofouling [60
]. For the prevention of biofouling in membrane-based water treatment processes, control of organic fouling is essential to sustain water purification without hampering the overall system performance. In general, hydrophilic, H-bond acceptor, non-H-bond-donor, and neutrally-charged membranes are resistant to organic fouling [61
], while hydrophobic and rougher membranes are prone to fouling by NOM and SMP [62
In FO processes, organic fouling has been extensively studied in the last decade [5
]. Mi and Elimelech investigated organic fouling in FO processes [8
]. They found a strong correlation between organic fouling and intermolecular adhesion force, indicating that foulant–foulant interaction plays an important role in determining the rate and extent of fouling. Atomic force microscopy (AFM) adhesion force measurement also demonstrated that a small percentage of adhesive sites on the membrane surface play an important role in organic fouling formation and in decreases in cleaning efficiency. It was found that permeation drag, hydrodynamic shear force, and calcium binding are the major factors governing organic fouling development. It was discovered that organic fouling (alginate as a model foulant) is reversible without chemical cleaning [5
]. Although the flux patterns in FO and RO modes were similar, the flux recovery rate after cleaning in FO process was much higher than for RO process. It is likely that the fouling layer formed (more related to inorganic scaling) on the FO membrane surface will be less compact due to the lack of hydraulic pressure (lower flux than pressurised membrane system; Figure 4
). Xie et al. [64
] supplemented these findings by comparing the alginate fouling characteristics between FO, pressure-assisted FO (PFO) and RO. Two possible mechanisms of fouling layer compaction have been proposed, namely permeation drag force and compression of foulants. The variations in the permeation drag force were eliminated by employing the identical initial water flux, and the fouling layer thickness, volume and density were identified. The fouling layer thickness decreased in the order of FO, PFO and RO while the volume and density increased from FO, PFO to RO. It was concluded that the applied hydraulic pressure contributed one factor to compression of fouling layers to a significant extent. The two possible compaction mechanisms may occur simultaneously and reinforce one another, resulting in irreversible, dense and compact fouling layers in RO in this study, while the drag force was the only applied compressive force in the case of FO. However, a contradicting study was also reported [65
]. Lack of applied pressure does not necessarily mean that the FO is an inherently low fouling membrane process. There could be other contributing factors to the lack of significant fouling such as the compensating effect of ICP acting under AL-FS orientation as mentioned earlier in Section 2
, and the low operating flux of the FO process could also induce the threshold/critical flux for less fouling.
She et al. [63
] systematically investigated membrane fouling in the PRO process. Alginic acid and humic acid were used as feed foulants and various draw solutes were used, including synthetic seawater and seawater brine, NaCl, CaCl2
, and MgCl2
. RSD of Ca2+
, and synthetic seawater brine caused significant flux decline, although the FS does not contain Ca2+
. Compared to Mg2+
has much faster interaction with alginate due to the faster RSD of Ca2+
ions, while Mg2+
ions strongly interact with humic acid. Increased DS concentration had an enhancing effect on the flux decline due to increased RSD and initial flux. Increased applied pressure showed decreased water flux but increased RSD. The decreased flux can mitigate fouling by the flux-dependent fouling mechanism, and increased RSD can exacerbate membrane fouling by the RSD-enhanced mechanism. The relative importance of these competing factors is strongly dependent on the types of draw solutes. Increased pressure mitigated alginate fouling for the NaCl DS, while facilitating alginate fouling for the CaCl2
DS, indicating that RSD enhanced fouling strongly interacted with feed foulants.
Parida and Ng [66
] also studied organic fouling in different membrane orientations. More severe fouling was observed in the PRO mode due to the smoother and denser membrane layer in FO mode, whereby the roughness of the selective and the porous support layers was 66 and 105 nm, respectively (Figure 5
). The loose structure of the porous support layer allowed the accumulation and deposition of foulants by the mechanism of direct interception and pore clogging, and facilitated further fouling development due to its rougher foulant surface and the foulant–foulant interaction. It was discovered that an organic loading of 50 mg/L—TOC (total organic carbon) or lower at a cross-flow rate of 50 cm/s caused minimal fouling in the FO mode during the 20 h of the fouling experiment. The presence of Ca2+
did not significantly deter the fouling resilience under this condition. It was suggested that FO mode is favourable when treating solutions with higher fouling/scaling tendencies (e.g., wastewater treatment) or higher salinity water (e.g., seawater desalination), while the PRO mode is to be preferred when using the solutions with lower fouling/scaling tendencies (e.g., brackish water desalination) or where intense concentration is necessary (e.g., power generation) [43
3.2. Inorganic Scaling
Inorganic scaling occurs when the concentration of sparingly soluble salts such as calcium sulfate, barium sulfate, and calcium carbonate in the feed water exceeds their solubility at high product water recovery and, as a result, precipitation of these salts may occur near or on the membrane surface, leading to severe membrane flux decline [67
]. Among the various scalants, calcium sulfate dihydrate (gypsum) and silica are the most common in seawater or brackish water desalination [67
]. Mi and Elemelech [68
] reported that flux decline rates in gypsum scaling experiments were practically the same in both FO and RO mode; however, more than 96% water flux recovery was shown in FO mode following a water rinse without chemical dosing. Flux recovery in RO mode was lower than in FO mode by 10%, which suggested that FO mode may provide the advantage of elimination of the need for a chemical agent for membrane cleaning. A similar study was conducted by the same authors [67
] and the trends of water flux decline in silica scaling experiments were found to be similar in both FO and RO mode; however, almost 100% water flux recovery was shown in FO mode, and only 80% water flux recovery was obtained in RO mode. This indicates that although different driving forces (osmotic gradient and hydraulic pressure) did not result in different flux decline rates, they however may affect the structure or density of the formed scaling layer. When calcium ions are present with alginate (main component of polysaccharides) in the water, more pronounced flux decline is reported due to the formation of a cake/gel layer as described in the earlier subsection (Section 3.1
In our recent pilot scale study on the fouling characterisation of a FO-RO system treating high fouling potential brackish surface water, inorganic scaling was found to be difficult to remove after the combination of physical and chemical cleaning [69
]. Detailed characterisation revealed that it is likely gypsum and organic components presented in the FS could form the gel layer (calcium bridging) and enhance the fouling layer rigidity. A comparable study was published where seawater was treated using a spiral-wound FO (SWFO) module [70
]. The fouling consisted mainly of scale-like foulants surrounded by biopolymeric substances. The silica scaling was caused by the polymerisation of dissolved silica. This silica scaling facilitated the deposition of NOM, as well as biopolymers. However, most NOM foulants can be easily removed while silica scaling is difficult to be physically removed. Work on a fertiliser-drawn FO (FDFO) process evaluated different types of draw solutes as a potential fertiliser while treating synthetic brackish groundwater for possible application of the process for irrigation [71
]. Diammonium phosphate (DAP, (NH4
) was observed to cause the most scaling among tested fertilisers as draw solutes due to their significant RSD phenomenon. Contrary to the previously mentioned studies, the flux decline caused by scaling was completely removed after physical cleaning with higher cross-flow velocity than the cross-flow velocity used for FDFO operation.
Biofouling is defined as the bacterial adherence with growth forming a biofilm, causing a membrane performance decline exceeding 10–15% of the start-up values under the applied operational conditions. At variations larger than 10–15%, corrective actions are recommended and guarantees are restricted by the manufacturers of membrane elements [72
]. Biofouling causes significant technical problems and influences the system performance such as by increasing necessary operational pressure, bringing about membrane flux decline, causing membrane biodegradation leading to increased salt passage, and raising energy requirements. These eventually result not only in higher operating and maintenance costs but in a shortening of membrane lifetime. Produced water quality is also lowered.
In any membrane technology, biofouling control is considered as a major challenge because all other types of fouling are fairly readily avoided by either chemical and physical pre-treatment (viz. various inhibitors for inorganic scaling and physical pre-treatments for particulate fouling) [48
]. However, biofouling formation only requires a few colonies to be developed with microorganisms present in all water systems which tend to adhere to surfaces and multiply on any surface in contact with the water even in an oligotrophic environment [74
]. In addition, biofouling is a complicated process in which many factors can influence each other [72
]. Biofouling is influenced by membrane surface properties (roughness, hydrophobicity, electrokinetic charge, and pore size), feed water chemistry (temperature, pH, ionic strength, nutrients, pollutants, and osmotic pressure) and also by microbial properties (size, cell surface hydrophobicity, and charge) [76
]. Table 1
summarises the factors affecting microbial attachment to a solid surface.
Two major strategies are used to prevent biofilm formation. These are physical pre-treatment by MF or ultrafiltration (UF) of feed water and dosing with a biocide such as chlorine, respectively. Chlorination is considered as standard practice to control biofouling. However, it may generate harmful products such as trihalomethanes (THMs) and other potential carcinogens. In addition, chlorine shortens the membrane lifetime due to degradation and this leads to cost problems [74
]. Physical pre-treatment applications have been extended to nutrient removal and modification of membrane surfaces to lower fouling occurrence [80
]. Pre-treatment of the feed water may minimise microbial growth in the fouling layer. However, once attached on the membrane surface, microorganisms can grow and increase the amount of EPS [74
]. Spiral-wound membrane elements, which are the most widely used in a variety of water treatment industries, make it difficult to remove the fouling layer due to their mechanical design (spacers, narrow feed channels, relatively low cross flow). Moreover, these spiral-wound designs directly support the accumulation of microorganisms and growth of biofilms [82
It has been known that the EPS and SMP are the main fouling factors in membrane systems [83
]. It is now recognised that TEPs play an important role in the process of aquatic biofilm formation [84
], particularly in the early stage of biofilm development. Bar-Zeev et al. [84
] introduced the new term ‘proto-biofilms’ to refer to TEPs with microbial outgrowth and colonisation. The authors found that these were the main sources of the early biofilm formation, particularly under the seawater condition.
Studies of biofouling in FO have been actively conducted recently, yet the understanding of this phenomenon is limited compared to other types of fouling. Further understanding of biofilm formation both in pressure-driven membrane systems and osmotically-driven membrane systems may help to develop strategies to control biofouling. Yoon et al. [85
] studied biofouling occurrence by using the model bacterium Pseudomonas aeruginosa
PA01 GFP in the FO process in comparison with the RO process and its control. They reported that biofouling is less significant in the FO process than the RO process. However, physical cleaning was not effective to overcome the water flux decline due to biofouling in the FO process, while chlorination effectively controlled the biofouling. Another approach of phosphate limitation in the FS to prevent microbial growth and biofilm formation was studied in a FO system treating wastewater [86
]. They confirmed that limiting phosphate in the feed water is an effective way of inhibiting the water flux decline and biofouling development. In our previous study [14
], significant biofilm deposition was observed on the FO membrane surface for nutrients-spiked brackish surface water filtration. Acute flux decline occurred at the initial stage of the experiment and significant decline was also observed at the end of the 24-h fouling experiment due to the formation of conditioning layers by organic foulants and the EPS biopolymer, respectively. The fouling experiments lasted for 40 h, and it was found that the declined flux was fully reversible by hydraulic rinsing using pure water only.
More recently, Kwan et al. [87
] investigated the biofouling mechanisms in FO in comparison with the RO process using identical hydrodynamic conditions. The water flux decline was significantly lower in FO (~10%) than RO (~30%). Distinct differences in biofilm structure were observed. Biofilm characterisation using confocal laser scanning microscopy (CLSM) revealed that the FO biofilm exhibited a loosely organised thick layer (~50 μm) with prominent mushroom-shaped structure (~77 μm) while maintaining the initial conditions of membrane structure (e.g., finger like support layer structure) as shown in Figure 6
a. This biofilm structure imparts low hydraulic resistance to water flow and low CP due to enhanced back transport of solutes to the bulk solution.
In contrast to FO, the RO membrane was compacted over the embedded supporting woven mesh due to the applied pressure, which was ~14.5 bar in this study. The biofilm formed on the RO membrane displayed a similar configuration that followed the shape of the membrane surface (Figure 6
b). The biofilm exhibited a tight cell organisation (~29 μm), embedded in the polysaccharide layer. Interestingly, the live and dead cell components were slightly larger in FO, while polysaccharide concentration was higher in RO. The live to dead cell ratios for FO and RO were 0.72 and 0.52, respectively, which highlights the viability of the FO biofilm. This study concluded that applied hydraulic pressure creates a distinct difference in biofilm structure and facilitates hydraulic resistance, possibly intensifying the biofilm-enhanced osmotic pressure. This adversely affects membrane water flux (Figure 6
c). FO biofouling showed advantages in terms of membrane cleaning and filtration of high fouling potential feed waters.
It should be noted that the combined fouling occurs simultaneously. A number of studies have investigated the effect of combined fouling in FO processes and reported their synergetic detrimental effects on the membrane performance [35
]. Zhang and co-workers investigated the combined effect of organic fouling and inorganic scaling in a forward osmosis membrane bioreactor (FOMBR) [89
]. The fouling in a FOMBR treating wastewater was governed by the coupled influence of biofouling and inorganic scaling in the AL-DS orientation, while AL-FS offered stable flux when combined with intermittent cleaning using tap water, suggesting that the AL-FS orientation should be recommended for FOMBR. Liu and Mi [88
] also confirmed a synergetic effect of alginate fouling and gypsum scaling on FO performance. The aggravated gypsum scaling in the presence of alginate molecules caused more severe flux decline than the algebraic sum of flux declines by the individual foulant. In a detailed investigation of the effects of alginate on the kinetics of gypsum crystal growth, they revealed that alginate molecules act as nuclei for heterogeneous crystallisation of gypsum, resulting in a combined network of gypsum and alginate fouling. The alginate molecules shortened the nucleation time so that it increased the rate of gypsum growth and changed the morphology of gypsum crystals, also increasing their size.