A review on RO membrane technology: Developments and challenges
The material properties of the membrane dominate the performance of a RO process. The emergence of nano-technology and biomimetic RO membranes as the futuristic tools is capable of revolutionizing the entire RO process. Hence the development of nano-structured membranes involving thin film nano-composite membranes, carbon-nanotube membranes and aquaporin-based membranes has been focussed in detail. The problems associated with a RO process such as scaling, brine disposal and boron removal are briefed and the measures adopted to address the same have been discussed.
The availability of clean water has become a global problem because of the continuously increasing costs of energy and increasing scarcity of water resources1. This problem has been exacerbated in recent years in the so-called century of water. By far, the domestic ro membrane process persists as the most reliable and cost-effective water desalination technique and numerous large-scale RO plants have been constructed around the world2,3. A wide range of polymers have shown potential for fabricating desalination membranes to be used in RO4. However, PA-based membranes tend to exhibit the best performance in terms of selectivity, flow, chemical stability and ease of large-scale fabrication. PA membrane technology was developed in the mid-70 s and has become the commercial benchmark in RO membranes5. In order to improve the membrane performances, the recent trend in polymer-based membrane research has been to investigate various types of nanocomposite films as an active layer of RO membrane, so-called nanocomposite membranes, in which these films are fabricated using a nanosized filler such as MWCNT, graphene, graphene oxide, silica, or zeolite6. In this regard, MWCNT·PA-based membranes have been prepared by several groups and in general, these membranes have exhibited some level of improved performance7,8,9,10,11,12. The advantages claimed for these membranes range from increased salt rejection, large fluxes, greater durability and even antimicrobial properties.
MWCNT synthesized by catalytic chemical vapour deposition13,14 have been widely studied due to their fascinating chemical and physical properties and among all nanocarbon materials, they can be mass-produced for commercially available applications, such as the electrode additives in high performance lithium ion batteries15. Interestingly, while the structure of the fully aromatic PA-based commercial ro membrane derived from m-phenylendiamine (MPD)-trimesoyl chloride (TMC) is constrained due to its stoichiometry; the addition of MWCNT can significantly vary their performance due to their unique features such as dispersability diameter, length, straightness and chemical functionalities, among many others. Therefore, although these past reports acknowledge the key role of MWCNT in aromatic PA nanocomposite membranes, still little attention has been devoted to the mechanisms related to the improvement of flow rate, selectivity and chlorine tolerance2. Carbon nanotubes inducing chlorine tolerance are particularly interesting because chlorine sensitivity has been recognized as a major drawback of PA-based RO membranes16,17. During long-term operation, chlorine is often added as a pre-treatment to reduce algae biofouling18 and is particularly needed for drinking water purification. Moreover, high-concentration short-term exposure to chlorine is also common during domestic nf membrane backwashing. For these reasons, several studies have been carried out and the degradation mechanism of aromatic PA membranes during chlorine exposure is relatively well-known19,20. Recently, our group demonstrated that the addition of MWCNT to rubber can considerably reduce the chlorine-induced degradation of the polymer matrix21. Although the degradation mechanism of rubber by chlorine is different from that of PA, particularly due to the lack of hydrolysis, covalent chlorination is a common problem for both polyamide and rubber. For rubber, we found that MWCNT effectively restricted the adsorption of chlorine within the polymer matrix, thus resulting in a limited exposure of the polymer to this reactive reagent and thereby decreasing the oxidative degradation. For these reasons, we believe MWCNT are not only promising composite fillers with chlorine protective properties, but might also help to provide mechanical robustness to PA-based RO membranes.
Results and Discussion
We prepared aromatic PA membranes using a support consisting of a porous polysulfone layer deposited on a polypropylene nonwoven. These support membranes were soaked sequentially in MPD and TMC solutions, to synthesize the aromatic PA membrane by interfacial polymerization. In order to incorporate MWCNT into the PA membrane, an anionically stabilized dispersion of MWCNT (Supplementary Fig. S1) was mixed with the MPD solution and the synthesis was conducted similarly. Figure 1a shows an image of the resulting membranes, with and without MWCNT. The black color developed in the membrane prepared using surfactant dispersed MWCNT is characteristic of the high carbon nanotube content of the present membrane (Fig. 1a). Thermogravimetry of the active layer (Supplementary Fig. S2) of the black color membrane indicates that it contains ca. 15.5 wt. % of MWCNT, which is at least 150 times higher than previously reported MWCNT-filled RO PA membranes7,8,12. The SEM image showing the surface morphology of the membrane is typical for the interfacial PA polymerization22, consisting of the multi-layered ridge-and-valley (Fig. 1b); the morphology of this membrane clearly changed after the addition of MWCNT (Fig. 1c). The thickness of the membranes was measured using SEM (Supplementary Fig. S3). The addition of MWCNT did not modify the thickness of the active layer and both samples were approximately 100 nm thick. However, water contact angle measurements showed a slight increase in wettability upon addition of MWCNT to the PA membrane (Supplementary Fig. S4). Notably, no MWCNT were visible on the surface, thus indicating that they were perfectly embedded within the PA matrix, a key factor needed for avoiding MWCNT leakage during operation. Flow permeation rates, as indicated below and SEM images confirmed that the membranes can be produced pinhole-free in a reproducible way. After the membrane was dried for SEM studies, cracks were generated by manual deformation of the membrane (Fig. 1d) and MWCNT embedded, parallel along the membrane surface, were observed bridging the fracture within the polymer matrix. The apparent diameter of these nanotubes are ca. 20 nm, which is about two times larger than the pristine nanotubes (Fig. S1a). These facts suggest that the nanotubes must be coated with polymer to achieve a good matrix-nanotube adhesion. In order to support our proposed structure consisting of a polymeric network with aromatic moieties in parallel arrangement to the MWCNT walls, we performed theoretical simulations of the monomer molecules orientation in the vicinity of a carbon nanotube surface, see Supplementary Fig. S5. Here, four different cases, consisting of two geometrical configurations, are demonstrated: horizontal and vertical alignments with respect to the MWCNT surface (modelled as a graphene surface), for both monomers (MPD and TMC). The results indicate a clear energetic preference for the horizontal arrangements of these molecules interacting with sp2 hybridized carbon networks; these preferences are related to π-π stacking and are known to be common for aromatic compounds on sp2 hybridized carbon surfaces. Similarly, Fig. S5b shows a simulation of 50 MPD molecules absorbed on a graphene surface and it can be seen that the molecules adopt a similar geometrical orientation after relaxation (Fig. S5c). In order to rule out curvature effects, we carried the simulations using a (10,10) single-walled carbon nanotube (Fig. S5d), which evidently has a higher curvature than the 10 nm diameter MWCNT experimentally used in the membrane fabrication. It can be seen on Fig. S5e that after relaxation, the aromatic ring of the MPD molecules lies parallel to the carbon nanotube surface. We confirmed the strong affinity of MPD with MWCNT by filtering the solution and carrying out UV-Vis spectroscopy. We found that 16.7% of the MPD monomer remained attached to the MWCNT. These MPD functionalized MWCNT were polymerized in TMC solution. Supplementary Fig. S6a shows a homogeneous PA coating on the MWCNT. Supplementary Fig. S6b depicts a higher resolution image showing a coating of about 5 nm thick on the MWCNT surface. We used fast Fourier transformation (FFT) of the HRTEM images to analyze the orientation of the PA network and it is clear that PA regions that do not contain MWCNT, show an anisotropic molecular arranged structure (Supplementary Fig. S6c), whereas the PA coating the nanotubes show a preferential orientation of PA molecules along the MWCNT surface (Supplementary Fig. S6d). These experiments strongly support a templating effect caused by MWCNT. To assess the distribution of the MWCNT within the membrane, a Raman mapping of the characteristic D- and G- bands of MWCNT was conducted (see Fig. 1e,f). Through all the studied areas only the D- and G- peaks could be observed, indicating a homogenous mixture and a high content of MWCNT, which is not common in these type of nanocomposites, because the MWCNT are prone to aggregation even when loading at low concentrations. Commercial NF membrane exhibited a lower contact angle; however in this case, the presence of wetting additives or a surface treatment is likely responsible for this phenomena. The method used to synthesize the MWCNT·PA nanocomposite relies on the transport of the MWCNT to the organic/aqueous interface during polymerization23. Indeed, the presence of a limited amount of anionic surfactant has been recently reported to improve PA membrane formation, resulting in better performance24. This is most likely due to a reduction of the oil/water interfacial tension, a process that in our case is also promoted by the small amount of surfactant that provides amphiphilicity to the nanotubes It is important to emphasize that we did not used covalent functionalization of MWCNT, in contrast to some previous reports8,11.