The main factor limiting FO development was the (ICP) that occurred in the SL, because it led to a decrease in water flux and increased operating costs. Similar to ECP, diluted ICP occurred in AL-FS mode and concentrated ICP occurred in AL-DS mode. Diluted ICP referred to the fact that the penetration of water molecules from FS into DS would make the concentration of DS decrease in SL, so that the DS concentration on membrane surface was less than that of the bulk solution, which resulted in the reduction of the effective osmotic pressure difference. Concentrated ICP meant that the FS was enriched in the porous SL, so that the concentration of the FS on the membrane surface was greater than that of the bulk solution, which led to the decrease of the effective osmotic pressure difference. Illustration of dilutive ICP (DICP) and concentrative ICP (CICP) in FO was exhibited in Fig. 3. Diluted ICP reduced the effective osmotic pressure difference greatly compared to concentrated ICP, which proved that diluted ICP was the main reason for the decrease of water flux. In order to reduce ICP and increase the water flux as much as possible, it was necessary to minimize the DS diffusion through the SL, among which the SL was the key factor. The ideal porous SL should maintain low reverse solute flux (J_s), high water flux (J_w), high porosity and good chemical stability etc. [31]

Classical solution-diffusion theory was adopted to evaluate the effect of ICP on the FO performance [32–33]. As can be seen in Fig. 3, compared to the CICP, DICP led to a much greater decrease in the effective osmotic pressure difference across the membrane. Therefore, the water flux under the effect of DICP could be shown in Eq. (1) [32].

where A and B represented water permeability coefficient and solute permeability coefficient for the membrane, respectively. And K was the resistivity coefficient reflecting the extent of ICP. The water flux for the CICP could be expressed as Eq. (2) [32] by using a similar approach.

For FO and PRO, a smaller K value indicated a smaller ICP, which resulted in high water flux. The K value was represented by the following Eq (3) [32].

Here, t, τ, ɛ and S represented the membrane thickness, tortuosity, porosity, and structural parameters, respectively. And D was the solute diffusion coefficient. Based on the above model, it could not only predict the water flux at the two membrane modes, but also evaluate the FO performance. Hydrophilic alumina nanoparticles were used to modify the polysulfone support layer, and the prepared membrane had high porosity and made S decrease from 1422 μm to 1028 μm, which increased the water flux and reduced ICP [34]. In order to reduce ICP, hydrophilic modification of the SL was usually carried out.

Reverse solute diffusion indicated that the solute infiltrated through the membrane from DS to FS in FO because of the solute concentration gradient, which would lead to the reduction of the osmotic pressure of the DS and cause membrane fouling. Besides, regular replenishment and regeneration of the DS would also increase the operating cost. At the same time, the penetration of solutes into the FS might change the chemical properties of the FS. FO membranes with the polyamide (PA) active layer and porous support layer were usually prepared to enhance the membrane selectivity (i.e., lower B/A ratio, higher J_w and lower J_s), thereby alleviating RSD [35]. Reverse solute diffusion in FO was inevitable and was affected by factors such as physicochemical properties of DS, membrane structure parameter, flow velocity and concentration polarization [36]. Therefore, in order to reduce RSD, it was necessary to select appropriate DS and add hydrophilic materials (GO, CNTs, and nanofibers etc.) into the support layer, intermediate layer, and active layer of the FO membrane for improving the membrane performance.

Membrane fouling was also one of the major challenges faced by the FO technology. The properties of membrane materials directly affected membrane fouling, so improving the structure and properties of membrane materials played an important role in alleviating membrane fouling, and the surface morphology of FO membrane was the main factor of affecting membrane fouling. Polyethylene glycol (PEG) block copolymer was grafted into the surface of the TFC-FO membrane, and the hydrophilic properties and anti-fouling ability of the modified membrane were enhanced [36]. In different application examples, the properties of pollutants that FO treated were different, and the types of membrane fouling were diverse. The better antifouling behavior could be attributed to the enhanced hydrophilicity and surface negative charge of membrane caused by the GO incorporation with the presence of functional groups on GO nanosheets, e.g., carboxyl, hydroxyl and epoxide groups [37]. Thus, the membrane materials could be modified in a targeted way to alleviate membrane fouling. The composition of the FS and DS could also affect the membrane fouling process. Strong interactions from the pollutants or between the membrane surface and pollutants in the FS could lead to the water flux loss and exacerbate membrane fouling [38]. In addition, higher velocity could also reduce the deposition capacity of pollutants on the membrane surface, thus contributing to the decrease of membrane fouling [39].

In recent years, GO had been widely studied in membrane manufacturing and modification, and it could be obtained through oxidizing graphene by using strong acid. Its carbon skeleton was composed of sp²-bonded carbon atoms with single atom thickness and rich in oxygen groups such as hydroxyl, carboxyl and epoxy [40]. Therefore, GO could increase the hydrophilicity of membrane surface, had uniform dispersion in organic solvents and pure water and reacted with other chemicals easily, which was more favorable for membrane application [41]. GO nanosheets had been extensively studied in the field of FO membrane modification because of their high hydrophilicity, special 2D structure, and sufficient mechanical strength. An ultrathin porous support layer with high water flux and salt rejection rate could be prepared by vacuum filtration of the GO suspension [42]. Recent results suggested that GO nanosheets could be used to construct ultrathin interlayer on a porous support layer to improve the performance of TFC-FO membrane [43]. GO interlayer could provide a “smooth” reaction surface for the polyamide polymerization process, thereby enhancing the hydrophilicity of the FO membrane.

Wu et al. [37] firstly elaborated the preparation and performance of three types of GO-based FO membraned, and evaluated the advancements of GO-based FO membranes in improving water permeability, salt rejection and anti-fouling ability that were highly dependent on the modification methods. However, it was still difficult to fabricate an ultrathin GO lamellar FO membrane with enough mechanical strength and full utilization of specific advantages of GO nanosheets. And the results indicated that the dispersion property of GO nanosheets in a specific casting solution could be an important factor for the FO membrane’s performance based on the GO blending matrix, and the loading amount of nanosheets in composite membrane matrix should be carefully optimized. GO nanosheets were used to modify the hydrophilic polyacrylonitrile (PAN) electrospinning nanofiber support layer, and the polyamide (PA) active layer was prepared by interfacial polymerization of TMC and MPD, besides the structural morphology characteristics of the support layer and FO membrane with GO different loading were studied [44]. The final results showed that the highly hydrophilic GO nanosheets improved the hydrophilicity of the polyacrylonitrile electrospinning nanofiber support layer (PAN-ENs), which could reduce the structural parameter significantly and the ICP. Meanwhile, the load of GO nanosheets could adjust the amount of PA, and the number of PA formed in the support layer would decrease with the increasing load of GO. So, the water permeability of PA active layer became small, and the prepared membrane had higher water flux and lower reverse solute flux in FO and PRO mode. These membranes with adding GO to the support layer could be used to remove heavy metals from industrial wastewater, and the highest remove rates of P_b, C_d and C_r were reached at 99.9%, 99.7% and 98.3% respectively [45]. TFC-FO membrane that was resistant to biological fouling and had high water flux was fabricated by modifying the support layer, the polyamide active layer or both. It could be concluded that the membrane modified in the support layer could greatly increase the water flux due to the improved porosity and porous structure, and membrane modified by the active layer or both at the same time had good hydrophilicity, smooth surface, high salt rejection rate and anti-fouling ability [46].

In various intermediate layers, hydrophilic polydopamine (PDA) was a kind of bio-glue and could adhere onto a variety of materials [47], of which −NH₂ group could react with the −COCl functional group in TMC during the PA preparation process, making a strong chemical bond force between PA and PDA [48]. It was proved that the support layer of polysulfone modified by PDA-GO interlayer had a good water flux. Adding hydrophilic materials that could be mixed in organic and aqueous phase was another effective method for modifying PA during the IP process except introducing an interlayer [49]. Wu et al. [50] introduced GO into the active layer of thin-layer nanocomposite FO membrane and found that the water flux of modified FO membrane increased by 41% compared with the original membrane, and the pollutants on the surface of the active layer were easier to clean, and the addition of GO significantly improved the anti-fouling performance of the membrane. GO was crosslinked between m-xylylenediamine (MXDA) and TMC to form the active layer, and although the modified FO membrane’s water flux was enhanced, the membrane selectivity decreased by using Na₂SO₄ as the DS [51], which restricted the selection of DS to some extent. The lateral size of GO sheets was an important factor affecting membrane performance, and adding GO into the active layer made the water flux increase while the reverse solute flux also enhanced [52]. Akther et al. [53] incorporated the AQN that was formed by the hybridization of GO and aquaporin Z (AQPZ) into the PA layer (seen in Fig. 4) and concluded that the water flux greatly increased when the reverse solute flux was reduced because of the influence of AQN. Adding GO into the active layer could reduce the membrane surface pollution and improve the water permeability of the membrane. Meanwhile, modifying the FO membrane by combining GO with other materials could synergistically enhance the membrane’s selectivity.

In order to avoid the negative effects of hydrophilic nanofibers that could expand after contacting with the water, polyvinylidene fluoride (PVDF) with PDA coating was selected as the support layer, which was highly hydrophilic and not easy to absorb water. PDA-GO intermediate layer was prepared on a superhydrophilic nanofiber support layer of TFC-FO membrane [54]. PDA-GO could also provide transport nanochannels for water, reduce the transport resistance of the PDA layer, and further improve the performance of the TFC-FO membrane [55]. In the IP process, the PDA-GO nanosheets that were added to the aqueous solution were combined with PDA-GO interlayer together to modify the membrane, so as to analyze the effect of PDA-GO on the modified TFC-FO membrane. It was finally demonstrated that the addition of PDA-GO to the middle layer or PA active layer alleviated the formation of PA layer defects effectively when nanofibers were loaded on the TFC-FO membrane, thus improving the water flux and selectivity of the membrane [54]. Adding PDA-GO simultaneously to the middle layer and the polyamide active layer could further increase the water flux and selectivity of TFC-FO membrane. Different interlayer materials such as organic interlayers (polyphenols, ion polymers, polymer organic acids, and other organic materials) and nanomaterial interlayers (nanoparticles, one-dimensional nanomaterials, and two-dimensional nanomaterials) were used to precisely control of the IP process, which were helpful to regulate the structure and fabricate a thin, dense and defect-free PA active layer [56]. Therefore, preparing TFC-FO of high water flux on the hydrophilic nanofiber support layer was feasible and could effectively alleviate ICP. It was demonstrated that adding PDA-GO simultaneously in the middle layer and PA layer could prevent PA from forming at the deep pore in the support layer. Recent advances on modified TFC-FO membranes were shown in Table 1. FO membrane modified by GO could obtain the desired water flux, and it also reduced ICP. However, the dosage of GO was a factor that needed to be accurately controlled. If the GO loading amount was too large, the pores inside the SL may be blocked. Thus, it could hinder the formation of amide bonds in the PA active layer and result in the decline of FO membrane performance.

However, there were some problems when applying it to different membrane layers such as the support layer, interlayer, active layer. When GO was used as an intermediate layer, the preparation of ultra-thin GO layer with sufficient mechanical strength was a difficult problem. And the decrease of FO membrane selectivity was a major defect when it was adopted as the active layer. For future studies, whether GO was hybridized or used alone in FO membrane modification, the optimal area of GO flake could be found in advance, which might be an effective way to greatly improve membrane selectivity. In addition, when GO was used as a support layer, it was difficult to control the amount of GO nanosheets, which led to agglomeration and reduced the mechanical strength of the membrane. But plasma modification could be utilized to prompt the dispersibility of GO materials.

CNTs itself had poor hydrophilicity and wide pore distribution, and the selectivity of the FO membrane prepared after its modification was also improved. However, the major challenge currently faced was the poor compatibility between CNTs and polymers. Due to complex structure and high pore density, CNTs were easy to agglomerate [63], which would reduce the water flux. Thus, the modification of CNTs could reduce the agglomerate phenomenon effectively.

In order to make full use of the advantages of FO, the problem of ICP must be solved [64]. CNTs were grafted on the membrane surface [65]. Ideally, when the nanomaterials were incorporated into the active layer of the thin-film composite membrane, the uniform distribution of nanomaterials were conducive to membrane preparation. FO membrane was modified by aminated CNTs to overcome ICP. The results showed that the composite nanomaterial thin-film membrane (ACNT-TFN) had high water flux and flux stability, and could operate for 6 hours and kept steady. It could be concluded that aminated CNTs introduced oxygen-containing functional groups which enhanced its surface hydrophilicity [57]. Besides, amino functional groups were introduced into CNTs, which could reduce the intermolecular Van der Waals force of CNTs and prevent agglomeration inside CNTs. It could be explained that CNTs as one-dimensional artificial water channels (AWCs) had sub-nanopore weakening of H-bonding for water transport mechanisms, with transport rates ranging from 1.9×10⁹ to 2.3×10¹⁰ water molecules per second [66]. The addition of tert-Butylamine (TBA) in the formation of PA active layer was considered as a new method for the fabrication of ACNT-TFN membrane and functionalized CNTs. Amination of CNTs could also adjust the polarity, which further enhanced the water flux and salt rejection rate of the ACNT-TFN membrane. In addition, the hydrophilicity of CNTs was further enhanced by modifying them with TBA. The amide functional group reduced the water contact angle of membrane surface, and a decreasing surface roughness also enhanced surface hydrophilicity of TFN membrane [67]. Thus, the overall surface hydrophilicity of the membrane was significantly increased. Hydrophilic groups such as sulfonic acid groups and amino groups were usually introduced into CNTs for membrane modification to improve FO performance because CNTs and polymers tended to agglomerate, thus blocking the internal pores of the support layer.

A single-wall carbon nanotubes (SWCNTs) interlayer modified by using dopamine-polyethyleneimine (DA-PEI) to regulate PA formed in IP reaction, which could be seen in Fig. 5 [10]. Since DA-PEI contained abundant −NH₂ groups, the positive charge of the SWCNTs interlayer gradually increased with the prolonged time of DA-PEI modification. The results showed that the hydrophilicity of the membrane with SWCNTs interlayer was stronger than that of the membrane without the middle layer, which was conducive to the formation of thinner, more uniform, and denser PA layer. So, the prepared TFC-FO membrane exhibited excellent performance in FO test, with lower reverse solute flux and higher water flux. SiO₂-ACNTs obtained by hydrolysis of tetraethyl orthosilicate (TEOS) and aminated carbon nanotubes (ACNTs) were grafted into the PVDF support layer to prepare FO membrane [68]. After adding the SiO₂-ACNTs, the membrane surface morphology was optimized, pore density was increased, pore size distribution was uniform, hydrophilicity was enhanced, strong amide bond was formed on the PVDF support layer, and the performance of FO membrane was greatly improved. Zhao et al. [69] attached carboxylated CNTs to the surface of a PVDF substrate, which increased the water flux and also demonstrated a certain capacity for flux recovery. This confirmed that the three-dimensional network structure of the CNTs interlayer could not only provide more space for water transport but also act as a buffer zone for solutions to facilitate the exchange of internal solutions. An excessive thickness of the CNTs intermediate layer could severely affect the IP process, while a thinner CNTs intermediate layer had minimal impact. Therefore, it is a challenging quantity to control. When CNTs were used to modify the active layer of FO membranes, although the water flux increased, the fouling degree and reverse salt flux also enhanced accordingly. To improve the overall performance of FO membranes, CNTs could be hybridized with various nanomaterials. Lee et al. [70] hybridized zeolitic imidazolate framework-8 (ZIF-8) with oxidized CNTs and made the desalting ability of the modified FO membranes increasing by modifying the active layer. Similarly, Golgoli et al. [71] hybridized carboxylated MWCNTs with aminated Zr₆O₄(OH)(BDC)₆ (UIO-66) material to act on the active layer, significantly improving the anti-fouling ability, water flux and flux recovery rate of the membrane. CNTs could easily be functionalized with various functional groups and also be hybridized with other materials such as MOFs to modify FO membranes as composite materials for enhancing the membrane properties.

The results showed that the hydrophilicity of TFC-FO membrane could be improved by introducing an interlayer into FO membrane to form a loose PA active layer. And except GO [72] and CNTs [69], halloysite nanotubes (HNTs) [73], metal oxide nanoparticles [30], metal-organic frameworks (MOFs) [71], zwitterionic materials [71], natural proteins [71] and zeolite [74] were used as modified materials to prepare FO membrane, which also had good hydrophilic and low reverse solute flux, and improved the anti-fouling properties of membranes. Although the water flux of the prepared membrane was enhanced, the membrane selectivity was not improved, which was related to the pore size of these materials. GO's pores were too small for water molecules to pass through, and HNTs and CNTs had much lower salt rejection rate due to their large pores (>1.0 nm). The ideal pore size for the preparation of interlayer materials was about 0.8~0.9 nm, which had a high salt rejection rate and made it easier for water molecules to pass through [75]. In addition, pore properties such as hydrophilicity and charge also affected permeability. Therefore, although the development of intermediate layer materials was the main way to prepare high performance FO membrane, the cost-effectiveness and large-scale production of the modified membranes had yet to be reported on a commercial scale [71].

Sulfonated CNTs was used to fabricate the SCNTs-TFC membrane that showed good electrochemical activity [58]. And after electrochemical cleaning of the contaminated membrane for 30 min at a voltage of 2.5 V, the removal rate of pollutants reached 95.3% and the water flux recovered to 85.0% of original value. Hydrophilic sulfonated crystal COFs with desired pore size was prepared and used as intermediate layer to fabricated high performance TFC-FO membrane, and schematic illustration of the selective FO work of PA/SCOF/TMC-nylon membrane was exhibited in Fig. 6 [11]. It was shown that the hydrophilicity of the FO membrane modified by COFs was improved and the pore size was decreased from 1.20 nm to 0.8 nm. So, the reverse solute flux was reduced as the water flux was increased, which reduced S and alleviated ICP. Ethylenediamine (NPED) monomer was combined with MPD to prepare a novel TFC-FO membrane loaded polyacrylonitrile (PAN) [76]. With the increasing content of NPED, the surface of TFC-PAN membrane became smooth, and the ability of anti-fouling, hydrophilicity and water flux were increased. But the salt rejection rate decreased due to the decomposition of NPED in water. A new 4-aminophenyl sulfone (APS) monomer was synthesized instead of MPD to react with TMC for forming PA layer [77]. It could be concluded that the crosslinking degree of APS and TMC was higher and the generated active layer was more stable. Compared with the active layer generated by MPD and TMC, the water flux of the original PA membrane was higher than that of the APS-TMC membrane by using NH₄HCO₃ as the DS under both PRO and FO modes.

Adding hydrophilic sulfonated polyether ketone polymer (sPEEK) to the casting solution could form a porous structure within the support layer, thereby enhancing the water flux [78]. And the water flux of TFC-FO membrane containing 5% sPEEK was enhanced in both PRO and FO modes by using 0.5 M NaCl and deionized water as DS and FS respectively. Hydrophilic SiO₂ was added to the PA active layer and the electrospun nylon (nylon-6) (N6) support layer to make the nanocomposite membrane (TFN) which had high pollution resistance and water flux [79]. Moreover, the flux recovery was 98% after membrane cleaning, and the structural stability of the prepared TFN membranes were also enhanced.

In conclusion, the modification of FO membrane played a crucial role for the overall development of FO technology. In future studies, researchers would focus on FO membrane modification from three aspects such as the SL, AL and interlayer modification. Adding hydrophilic materials such as GO and CNTs etc. further improved FO membrane performance, which would be a focus of future research. It was urgent need to develop high performance Carbon-based FO membranes for addressing the globally rigorous issues of water shortage and energy demand.