1 Introduction
In recent years, membrane-based separation technologies have received increasing attention attributing to lots of advantages such as the low energy consumption, easy operation, and environmental friendliness. Till now, polymeric membranes still dominate the membrane separation industry market because they are economic with good processibility, easy operation and competitive separation performance. However, for most polymeric membranes, there is a Robeson bound where an intrinsic trade-off effect exists, i.e., highly permeable membranes generally with low selectivity and vice versa [1]. Membrane separation performance can be improved by reducing the thickness, where the ultimately achievable refinement in the thickness will be a one-atom-thick layer that can be achieved by twodimensional (2D) materials. Hence, ultrathin 2D membranes have the possibility to achieve high permeability while retaining high selectivity by precisely tuning pore size, surpassing the upper bound of traditional membranes. The emergence of novel 2D materials, such as graphene, provides a new opportunity for membrane development. Discovery of monocrystalline graphitic films by Geim’s group in 2004 has drawn widespread attention for 2D materials in some fields including the membrane separation [2]. Up till now, in addition to the earliest studied graphene, other 2D membranes have been prepared using building blocks of other materials such as zeolites, metal-organic frameworks (MOFs), covalent organic frameworks (COFs), graphene oxides (GOs), layered double hydroxides (LDHs), transition metal dichalcogenides (TMDs), MXenes and graphitic carbon nitrides (g-C3N4). Owing to the precise control of the interlayer nanochannel/sub-nanochannel or the pore size and reduction of mass transfer resistance, ultrathin membranes based on 2D nanosheets have become promising candidates for separation as the next-generation membrane, which could achieve high permeability while retaining high selectivity, surpassing the upper bound of traditional membranes.
2 Separation mechanism of 2D membranes
In the membrane separation process, ions or molecules pass through the membrane selectively where the membrane serves as the separation medium. In general, 2D membranes can be grouped into two types: porous nanosheets-based membranes and non-porous nanosheets-based membranes, usually named as stacking 2D lamellar membranes. The most conspicuous difference between these two types of 2D membranes is the separation mechanism and the mass transportation route.
Porous nanosheets-based membranes, such as 2D zeolite membranes, 2D MOF membranes, and 2D COF membranes, comprise nanosheets with intrinsic nano/subnano-porosity. For these porous nanosheets-based membranes, precise molecule sieving is primarily determined by the size of in-plane pores. Owing to the size of nanopores, porous nanosheets-based membranes only allow size-specific ions or molecules to pass through (Fig. 1(a)). According to the channel size and chemical environment, gas diffusion mechanisms of porous nanosheet membranes can be broadly divided into Knudsen diffusion, surface diffusion, molecular diffusion, and viscous flow [3]. The porous nanosheet membrane with regular nanopore structure and ultrathin thickness is a type of ideal separation membrane morphology, which could reduce mass transfer resistance and lead to ultrahigh permeance, while maintaining high selectivity resulting from the precise pore sieving. For example, Yang et al. successfully prepared several nanometer-thick Zn2(bim)4 molecular sieve nanosheet membrane with an excellent H2/CO2 separation factor of 291 accompanied by an H2 permeance of 2700 gas permeation unit (GPU), far surpassing the Robeson upper bound [4].
Stacking 2D lamellar membranes, such as 2D GO, LDH, MXene and TMD membranes, comprises nanosheets without inherent in-plane pores. The ions or molecules transportation and sieving are achieved by the controlled interlayer nanochannels/subnanochannels formed by stacked nanosheets. As shown in Fig. 1(b), the interlayer spacing between the neighboring nanosheets allows transport of small ions or molecules while blocks large ions or molecules. For the 2D laminar membranes, design and tuning the interlayer spacing are important for membrane separation performance, and significant achievements have been made in this field [5–7].
In general, there are two types of pathway for mass transportation in 2D membranes, including the inter-sheet pathway A, (referring to interlayer spacing between nanosheets) and the inner-sheet pathway B, (referring to in-plane pore-based transport pathway). For gas separation, the membrane permeability was calculated by the following equation [8]:
where h is the membrane thickness; e and t are the porosity and tortuosity for pores of pathways A and B, respectively; D and K are the gas diffusivity and adsorption equilibrium constant in pathways A or B, respectively. In particular, the tortuosity (t) of 2D membranes can be approximately calculated as the ratio of the lateral size to thickness of nanosheets, while the porosity (e) is related to the interlayer d-spacing, the thickness, and porosity of nanosheets.
For inter-sheet pathway A, the diffusion length can be estimated by the Nielsen transport model [9]:
where l is the diffusion length, h is the membrane thickness, L and W are the lateral dimensions and thickness of the platelet, respectively, and d is the effective distance between the platelets.
3 Preparation of 2D nanosheets and 2D membranes
A high-performance 2D membrane needs high-quality 2D nanosheets and the precise assembly of nanosheets. Hence, preparation strategies of 2D nanosheets and 2D membranes are essential issues to realize the membrane separation. Many synthetic methods to prepare 2D nanosheets have been developed, such as mechanical exfoliation, intercalation exfoliation, chemical vapor deposition, and wet-chemical syntheses. In general, there are two types of strategies to synthesize 2D nanosheets, i.e., the top-down route and the bottom-up route. The former refers to the disintegration of bulky materials into delaminated nanosheets and the latter refers to direct synthesis from the corresponding starting materials. Several reviews on the preparation of various 2D materials have compared the advantages and disadvantages of different methods to prepare 2D nanosheets [10–12].
Pressure-assisted filtration is the most widely used method to fabricate 2D membranes, which has distinct advantages over other methods due to its simple operation and low cost. Layer-by-layer assembly is the process of depositing multilayers with some interactions such as electrostatic interactions, hydrogen bonds and covalent interactions. Although high requirement for equipment, high-quality monolayer membranes can be obtained through layer-by-layer process. The deposition of nanosheets are achieved by vertically lifting the substrate in the LangmuirBlodgett process while the substrate was contacted face-to-face with the nanosheet layer at the interface in the LangmuirSchäfer process. In recent years, electrostatic deposition has been widely concerned on the fabrication of 2D membranes, although electrostatic deposition is only suitable for the assembly of charged 2D materials. Synthesis of 2D nanosheets and preparation methods for 2D membranes are also discussed.
4 Different types of 2D membranes
As discussed above, 2D membranes can be assembled by two types of nanosheets. One is non-porous 2D nanosheets, such as GOs [13–15], MXenes [16,17], TMDs [18,19]. In the 2D membranes constructed by non-porous nanosheets, the mass can be transferred through the interlayer spacing between nanosheets. The other type is porous 2D nanosheets such as zeolites [20,21], MOFs [4,22,23], COFs [24,25] and g-C3N4 [26,27]. In this section, the development progress and recent breakthrough of different types of 2D membranes are introduced to show high potentials for membrane separations.
4.1 2D membranes based on porous nanosheets
4.1.1 Zeolite membranes
Zeolites are microporous materials based on crystalline silica with the pore size in the range of 0.25 nm to more than 1 nm [28]. In 2009, Ryoo’s group achieved the single-unit-cell thick Mobil Five (MFI) nanosheets through the zeolite-SDA-functionalized surfactant [29,30]. As a new class of zeolite nanosheets, have received more and more attentions in terms of its property and application in membrane separation. In 2011, Tsapatsis’s group successfully prepared MFI-nanosheet membranes via simple filtration. Owing to non-selective gaps between the neighboring MFI nanosheets, these films did not show any separation performance for p-xylene and o-xylene [31]. To address this problem, they prepared 200 nm-thick MFI-nanosheet membranes via secondary growth approach on filtered oriented MFI nanosheet layers containing non-selective gaps, exhibiting promising molecular sieving property for p-xylene and o-xylene [32]. In 2013, Yoon’s group reported a new method of membrane growth named as “gel-free secondary growth” and obtained 200-nm-thick silicalite-1 membrane, exhibiting a good p-xylene/o-xylene separation factor of 1050 accompanied by a p-xylene permeance of 13 × 10–8 mol∙m–2∙s–1∙Pa–1 [33]. Then, Tsapatsis’s and Yoon’s groups changed the seeding procedure to avoid scale-up challenges arising because of manually rubbing the seed crystals on the support. They combined filter-coating seeding method and gel-less secondary growth to prepare oriented MFI membranes as thin as 100 nm. Compared to the silicalite-1 membrane (200 nm) reported by Yoon’s group, the obtained thinner membrane (100 nm) showed enhanced p-xylene permeance (2.4 × 10–7 mol∙m–2∙s–1∙Pa–1) but a reduced separation factor for p-xylene/o-xylene (500) [34].
In general, top-down and bottom-up are two types of strategies to synthesize 2D nanosheets. The above-mentioned method for the synthesis of MFI nanosheets belongs to top-down strategy, which is hard for the preparation of high-aspect-ratio nanosheets. In 2017, Tsapatsis’s group made a breakthrough in the synthesis strategy of zeolite nanosheets. They successfully synthesized the high-aspect-ratio MFI nanosheets, which were used to prepare a membrane via a gel-free secondary growth. Compared to membranes made from exfoliated nanosheets, the obtained membrane possessed a high p-xylene permeance of 0.56 × 10–6 mol∙m–2∙s–1∙Pa–1, with a p-xylene/o-xylene separation factor of 2000 [35]. Then, Tsapatsis’s group applied a novel floating particle coating strategy to prepare MFI nanosheet seed layers on directly synthesized MFI nanosheets (Fig. 2(a)). Compared to the LangmuirSchaefer strategy, the floating particle coating strategy can produce higher-density zeolite nanosheet seed coatings on porous supports (Fig. 2(b)), resulting in MFI membrane with an excellent selectivity of p-xylene/o-xylene over 10000 and a maximum p-xylene permeance of 2.9 × 10–7 mol∙m–2∙s–1∙Pa–1 via a gel-free secondary growth of the MFI nanosheet coating [20]. Furthermore, Cao et al. prepared the ZSM-5 zeolite nanosheets with very large aspect ratios and nanometer-scale thickness by the bottom up method and then prepared a 2D ZSM-5 membrane by dip coating followed with consolidation by the vapor-phase crystallization on a porous a-Al2O3 support (Fig. 2(c)). The obtained membrane showed a high salt rejection rate (>99.8%) with a water flux of 3.69 L∙m–2∙h–1 for the pervaporation desalination of 24 wt% NaCl solution [21]. Recently, Min et al. made a progress on the preparation of 2D MFI membranes on a-alumina hollow fiber supports for large-scale molecular sieving applications via vacuum filtration and two sequential hydrothermal treatments. The obtained membrane exhibited a separation factor of 58 for n-butane/i-butane accompanied by a n-butane permeance of 1923 GPU [36].
In addition to 2D MFI zeolite membranes, other 2D zeolite nanosheets such as MCM-22 [37,38], AMH-3 [39,40], and JDF-L1[41–43] were also assembled into membranes for separation applications. Recently, Wei et al. designed a sodalite zeolite membrane via the hydrothermal crystallization of sodalite zeolite on a a-Al2O3 support with the predeposition of MCM-22 flakes. The obtained sodalite zeolite membrane demonstrated a permeation flux of 4.4 kg∙m–2∙h–1 accompanied by a good pervaporation separation of water/ethanol (>10000) [44]. Besides the preparation of pure 2D zeolite membranes, 2D zeolite nanosheets could be applied to nanocomposite membranes to solve trade-off effect [39–43,45]. For example, Galve et al. prepared a new type of mixed matrix membranes (MMMs) by combining two types of porous fillers (MCM-41 and JDF-L1). The obtained membrane demonstrated an H2/CH4 selectivity of 32 accompanied by an H2 permeability of 440 Barrer [43].
Fig.2 (a) Illustration of the floating-particle coating method to prepare MFI nanosheet monolayer seed coating; (b) SEM images of MFI nanosheet seed coatings on porous substrate via the LangmuirSchaefer (left) and the floating-particle coating (right). Reprinted with permission from Ref. [20], copyright 2018 Wiley VCH; (c) Microscopic characterizations of the silicalite nanosheets. Reprinted with permission from Ref. [21], copyright 2018 American Association for the Advancement of Science. |
4.1.2 MOF membranes
MOF nanosheets, as an important group of 2D materials, are promising candidates for membrane separation. 2D MOF membranes have drawn widespread attention for separation attributing to their advantages such as ultrathin thickness, high specific area, and tunable molecular structures. Recently, Peng et al. summarized a series of research studies about the MOF nanosheet-based membranes for separation [46]. Traditional MOF materials used for membranes mainly possess 3D framework; however, in 2014, Peng et al. made a great breakthrough in the field of 2D MOF membranes. The challenge in preparing high-performance 2D-MOF membranes with exfoliated MOF nanosheets from layered MOFs is retaining the morphological and structural integrity of MOF nanosheets. To address such issue, they successfully prepared Zn2(bim)4 nanosheets with a thickness of 1.12 nm and lateral dimension of 1.5 mm by low speed wet ball-milling and ultrasonic exfoliation. Another challenge is in preparing high-performance ultrathin 2D-MOF membranes without defects. They used a hot-drop coating method, where the dispersed Zn2(bim)4 nanosheets were assembled on a heated porous a-Al2O3 support (Fig. 3(a)). The resulting ultrathin membrane possessed a good H2/CO2 separation factor of 291, but an H2 permeance of 2700 GPU [4]. The H2 permeance of this resulting membrane is higher than other membranes for H2/CO2 separation and the H2/CO2 separation factor of this membrane is also quite high. Then, Peng et al. adopted similar method to prepare sub-10 nm-thick Zn2(bim)3 membranes, displaying an H2/CO2 separation factor of 166 accompanied by an H2 permeance of 8 × 10–7 mol∙m–2∙s–1∙Pa–1[22]. In 2017, Wang et al. reported the freeze-thaw method to fabricate MAMS-1 nanosheets, where the MAMS-1 crystals were freezed in liquid nitrogen bath followed by treatment using hot water to thaw. MAMS-1 crystals were exfoliated into layered nanosheets by the shear force produced in repeated freeze-thaw cycles. In addition, they used a hot-drop coating to fabricate MAMS-1 membranes. The obtained 40-nm membrane exhibited an H2/CO2 selectivity of 235±14 accompanied by an H2 permeance of 553±228 GPU [23]. Very recently, Zhang’s group successfully prepared exfoliated water-stable Al-MOF nanosheets. The ultrathin Al-MOF membrane formed by vacuum filtration showed a high water flux of 2.22 mol∙m–2∙h–1∙bar–1 and rejection rates of nearly 100% for inorganic ions, better than the most reported 2D laminar membranes on the water/ion selectivity (Fig. 3(b)), making 2D MOF membranes as promising candidates for ion separation [47].
Besides the top-down method, Rodenas et al. successfully prepared highly crystalline and intact CuBDC nanosheets via a bottom-up synthesis strategy in 2015. Furthermore, these CuBDC nanosheets were used as fillers to prepare CuBDC/polyimide (PI) MMMs. The resulting ns-CuBDC/PI membrane exhibited a higher CO2/CH4 selectivity (88.2±1.3) than the polymeric membrane and the b-CuBDC/PI membrane [48]. Inspired by their work, many reports about MOF nanosheet-based MMMs have caught public attention [49–51]. Yang et al. used GO to repair the defects between the junctions of CuBDC nanosheets, and the obtained CuBDC-GO composite membranes showed an H2 permeance of 9.6 × 10–7 mol∙m–2∙s–1∙Pa–1 with an H2/CO2 separation selectivity of 95.1 [52]. Zhong et al. prepared two types of 2D ZIF-L membranes with different orientations (b-oriented and c-oriented) and demonstrated the important effect of controlling crystal orientation on the ZIF membrane separation performance. The b-oriented and c-oriented ZIF-L membranes were both prepared by the secondary growth method. However, the seeding procedures were different. The b-oriented ZIF-L membrane had a secondary growth from a randomly oriented seed layer by dip-coating while the c-oriented ZIF-L membrane was prepared by secondary growth from a c-oriented ZIF-L crystal layer by vacuum filtration with polyethyleneimine (Fig. 3(c)). Owing to different paths of the molecule transport, the obtained c-oriented ZIF-L membrane with a smaller pore had a positive effect on the gas separation [53]. In order to improve separation performance, 2D Zn-TCP(Fe) nanosheets, obtained from a direct synthesis, were cross-linked by polycations followed by a simple vacuum filtration to prepare 2D MOF membranes. The obtained 2D Zn-TCP(Fe) membrane exhibited an excellent water permeance of 4243 L∙m–2∙h–1∙bar–1 accompanied by a rejection rate of 98.2% for methyl red [54].
Fig.3 (a) SEM images of (I) the top view and (II) cross-section view of an ultrathin Zn2(bim)4 nanosheet on a-Al2O3 support. Reprinted with permission from Ref. [4], copyright 2014 American Association for the Advancement of Science; (b) Comparison of the Al-MOF membrane for ion separation with other representative 2D laminar membranes. Reprinted with permission from Ref. [47], copyright 2020 American Association for the Advancement of Science; (c) Illustration of the fabrication of b-oriented ZIF-L film and c-oriented ZIF-L film. Reprinted with permission from Ref. [53], copyright 2015 The Royal Society of Chemistry. |
In addition to the preparation of 2D MOF membranes on porous flat or disc-shaped substrates, Zhang’s group prepared the 2D ZIF membranes on porous tubes for possible industrial applications. They reported the formation of Zn2(bim)4 nanosheet tubular membranes via ZnO self-conversion growth in a GO confined space, and the obtained oriented membrane exhibited an H2 permeance of 1.4 × 10–7 mol∙m–2∙s–1∙Pa–1 accompanied by a selectivity of 106 for H2/CO2 (Fig. 4(a)) [55]. Then, they advanced the preparation of 2D ZIF membranes on porous alumina tubes via ZnO self-conversion with ammonia assistance. The obtained 50-nm-thick Zn2(bim)4 nanosheet membrane exhibited an H2 permeance of 2.04 × 10–7 mol∙m–2∙s–1∙Pa–1 accompanied by selectivities of 53, 67, and 90 for H2/CO2, H2/N2 and H2/CH4, respectively [56]. More recently, the same group achieved the direct growth of the 57-nm-thick Co2(bim)4 nanosheet tubular membrane with an H2 permeance of 17.2 × 10–8 mol∙m–2∙s–1∙Pa–1 accompanied by a selectivity of 58.7 for H2/CO2 via ligand vapor phase transformation of the Co-based gel layer (Fig. 4(b)) [57].
Fig.4 (a) Illustration of Zn2(bim)4 nanosheet membranes formation via GO guided self-conversion of ZnO nanoparticles. Reprinted with permission from Ref. [55], copyright 2018 The Royal Society of Chemistry. (b) Illustration of Co2(bim)4 nanosheet membranes formation via vapor phase transformation of Co-based gel. Reprinted with permission from Ref. [57], copyright 2018 Elsevier. |
4.1.3 COF membranes
In recent years, COF nanosheets have also been assembled for membranes. Many simulation results have proved that 2D COF is a type of promising material for membrane separation [58–60]. Interfacial synthesis and exfoliation methods are classical methods for the fabrication of high-performance 2D COF membranes. Recently, Yao et al. prepared ultrathin Tp-AD, Tp-Pa and Tp-BD nanosheets with a large area via azobenzene-assisted exfoliation method. The obtained Tp-AD membrane displayed a water flux of 596 L∙m–2∙h–1∙bar–1 and a RB rejection rate up to 98% [61]. In addition to the widely used exfoliation method, Dey et al. prepared COF membranes by an interfacial crystallization method, and the resulting Tp-Bpy film demonstrated a high solute rejection with an acetonitrile permeance of 339 L∙m–2∙h–1∙bar–1 [62]. Matsumoto et al. reported the formation of free-standing TAPB-PDA COFs films with a high rejection (up to 91%) of R-WT via interfacial polymerization catalyzed by Lewis acid [63]. Besides, Fan et al. reported tubular imine-linked COF-LZU1 membranes via in situ solvothermal growth. The resulting membrane possessed an outstanding water stability and a good water permeance about 760 L∙m–2∙h–1∙MPa–1 with a rejection rate of 98% for chrome black T [64].
At present, 2D COF membranes are limited in the separation of small molecules. One of the main problems is the relatively larger pore size of 2D COFs membranes than the size of most gas molecules. Hence, applications of 2D COF membranes for separation of small molecules are still a great challenge. Discussions regarding the control of the pore sizes of 2D COF membranes have dominated research in recent years [24,25,65,66]. Caro’s group designed a COF-COF bilayer membrane via a temperature-swing solvothermal approach. Owing to the formation of interlaced pore networks, the obtained COF-LZU1-ACOF-1 bilayer membrane with more appropriate pore size showed higher separation selectivity for H2/CO2, H2/N2 and H2/CH4 than the individual COF-LZU1 and ACOF-1 membranes [65]. Yang et al. reported a series of COF membranes assembled by 2D COF nanosheets and 1D cellulose nanofibers (CNFs) (Fig. 5(a)). The pore size of the 2D COF membrane reduced with the sheltering function of CNFs, endowing the resulting membrane with accurate molecular sieving ability to meet different separation demands [24]. More recently, Li et al. reduced the pore size to sub-nm scale (~0.6 nm) via adjusting the stacking mode of COF layers from AA to AB stacking, and the resulting AB stacking COF membranes with smaller pore showed high potentials in membrane separation (Fig. 5(b)). The idea in their study can also be applied to design other 2D membranes with suitable pore size for the accurate molecular sieving [66]. Recently, Ying et al. designed a 2D COF membrane using anionic TpPa-SO3Na and cationic TpEBr nanosheets with different pore sizes via layer-by-layer assembly. Compared to the COF-COF bilayer membrane (ca. 1 mm) reported by Caro’s group, the obtained ultrathin hybrid TpEBr@TpPa-SO3Na membrane (41 nm) with small pore and compact dense nanosheet membrane structure displayed a higher H2 permeance of 2566 GPU accompanied by a selectivity of 22.6 for H2/CO2 [25].
4.1.4 Graphitic carbon nitride membranes
Because of the ordered triangular 0.311 nm-sized nanopores, g-C3N4 have received more and more attentions for membrane separation. Many molecular simulations provide a bright prospect of g-C3N4 membranes in separation [67,68]. In 2017, our group prepared the 2D g-C3N4 membranes with artificial nanopores and self-supporting spacers (Fig. 6). The obtained 160 nm-thick g-C3N4 membrane displayed a water flux of 29 L∙m–2∙h–1∙bar–1 and a rejection rate of 87% for Evans blue [26]. To enhance the water permeance of g-C3N4 membranes further, our group prepared polyacrylic acid-modified g-C3N4 nanosheet membranes with the polyacrylic acid tuning nanochannels. Owing to a slight increase in interlayer spacing, the obtained g-C3N4-PAA hybrid membrane demonstrated a higher water flux of 117 L∙m–2∙h–1∙bar–1 accompanied by a similar rejection rate of 83% for Evans blue [69]. Recently, in order to improve the separation performance of the pure g-C3N4 membrane, Ran et al. enlarged the interlayer spacing of g-C3N4 adjacent nanosheets by embedding molecules containing SO3H. They successfully prepared the SPPO/g-C3N4 membranes, exhibiting a superior selective rejection of 100% to methyl blue and with a high water permeability of 8867 L∙m–2∙h–1∙bar–1 [27]. Besides, Wang et al. designed g-C3N4 based membranes by incorporating sulfate anion or camphorsulfonic acid anion into protonated layers. The obtained membrane showed highly enantioselective permeation, endowing graphite phase carbon nitride as a promising material for chiral membrane [70]. Moreover, several reviews about g-C3N4-based MMMs have emerged [71–77]. For instance, Gao et al. prepared acidified g-C3N4-PA MMMs by incorporating acidified g-C3N4 into a polyamide selective layer by interfacial polymerization, and the as-obtained g-C3N4-PA MMMs exhibited a higher water flux (45 L∙m–2∙h–1) than the pure PA membranes with an NaCl rejection of 98.6% [76].
4.1.5 Nanoporous graphene (NPG) membranes
Graphene, as a well-known 2D material, was first exfoliated from graphite by Geim and Novosolov in 2004 [2]. The defect-free single-layer graphene is impermeable to molecules (even helium gas molecules), making it unsuitable for any separation application [78]. However, many molecular simulations theoretically demonstrated that NPG held great potential in separation, especially water desalination [79], ion separation [80], and gas separation [81–88]. At first, great attention was paid to various perforation techniques such as focused electron beam drilling [89,90], focused ion beam irradiation [91], ion bombardment and chemical oxidative etching [92,93], ultraviolet-induced oxidative etching [94], and oxygen plasma etching [95] to perforate selective pores in graphene sheets to create NPG membranes. Several reviews dealing with this topic have compared the advantages and disadvantages of various perforation techniques [96,97]. Later, NPG membranes were also prepared by some other methods such as direct synthesis [98–100].
One of the major challenges in high-performance NPG membrane is the control of pore size. Huang et al. proposed an ozone functionalization-based etching and pore-modification chemistry to prepare membranes, and the resulting graphene membrane with smaller pore size and higher pore density displayed an increased H2 permeance and an increased H2/CH4 selectivity [101]. Wang et al. used molecular dynamics simulations to prove that the bilayer NPG membrane with the lateral shift of the two graphene layers can continuously tune the effective pore size [102]. Zhao et al. reported a partially decoupled defect nucleation and pore expansion to prepare the graphene membrane for highly efficient gas separations (Fig. 7(a)). The obtained membrane exhibited a good H2/CH4 selectivity (15.6 to 25.1), a good H2/C3H8 selectivity (38.0 to 57.8), and a high H2 permeance (1340 to 6045 GPU) [103]. Recently, Sun et al. theoretically demonstrated that the nonselective large pores can be tuned to the selective pores via adding charges on the graphene surfaces. Because of electrostatic interactions between CO2 and the surface charges, the graphene nanopore with surface charges displayed increased CO2/N2 selectivity [104]. Moreover, larger graphene nanosheets almost inevitably have more defects. In order to utilize the polymer coatings to eliminate defects of membranes, Choi et al. fabricated the 20-nm-thick graphene-polymer hybrid membrane by a vapor-liquid interfacial polymerization process [105]. Yang et al. fabricated an ultrathin graphene-nanomesh (GNM)/single-walled carbon nanotube (SWNT) hybrid membrane with excellent mechanical strength (Fig. 7(b)). In this structure, the SWNT network with high mechanical strength served as a microscopic framework to support the GNM, preventing the hybrid membrane from tearing and solute leakage during the efficient water purification over large areas [106].
Fig.7 (a) Illustration of the partially decoupled defect nucleation and pore expansion to prepare NPG membranes. Reprinted with permission from Ref. [103], copyright 2019 American Association for the Advancement of Science; (b) Schematic illustration of GNM/SWNT membranes fabrication. Reprinted with permission from Ref. [106], copyright 2019 American Association for the Advancement of Science. |
4.2 2D membranes based on non-porous nanosheets
4.2.1 GO and reduced GO membranes
Among the graphene-based material family, besides NPG, more and more attentions have been paid to GO and reduced GO (rGO) for membrane separation. As the most widely studied graphene-based material for membrane separation, GO with oxygen-containing groups can be easily dispersed in water and enhance interactions with transport components. rGO can be obtained when oxygen atoms are removed from GO by reduction such as thermal reduction, chemical reduction, and solvothermal reduction. During the reduction process, defects could be formed in rGO, serving as nanopores to provide selective separations. To date, several articles about molecular simulations of GO- and rGO-based membranes have provided useful guidelines and directions for the preparation of high-performance separation membranes [107–110]. The following part will introduce recent advances in GO and rGO as 2D membrane materials for molecular and ionic separations.
In 2012, Nair et al. reported that the GO membranes were impermeable to liquids, vapors and gases, but allowed water to permeate unimpeded, making them promising for selective removal of water [111]. However, pure GO membranes are unstable in water environment. Hence, improving their performance stability for long-term membrane operation is very important. To solve this problem, much attention has been paid to the modification of GO by cross-linking, which can improve the water stability of GO membranes and control the charges, functionality, and spacing of the GO nanosheets. Different types of crosslinkers such as metal ions [112–114], polymers [5,115–117], and multifunctional small molecules [118–121] were used in the cross-linking process for membrane separation. Recently, Liang et al. achieved interlayer cross-linking by introducing sodium 1,4-phenylenediamine-2-sulfonate as the water transport promoters between GO nanosheets. The obtained membrane with a more stable structure and water transport promoters possessed a flux of 5.94 kg∙m–2∙h–1 accompanied by a separation factor of 3965 for water/butanol [122]. More recently, Pan et al. designed the GO membranes with dual crosslinkers, where polyvinylamine (PVAm) was used as the first crosslinker and silica generated from the biomineralization of PVAm served as the second crosslinker. The resulting GO-PVAm-silica membrane with a fixed interlayer spacing demonstrated a flux of 12.9±0.3 kg∙m–2∙h–1 and a separation factor of 1188±39 for water/butanol [123]. Nie et al. designed small-flake GO (SFGO) membranes with La3+ serving as a cross-linker and spacer. The resulting SFGO-La3+ membrane with shorter transport pathways of small-flake GO and a larger interlayer spacing could keep methanol permeance>100 L∙m–2∙h–1∙bar–1 for 24 h under 3-bar pressure [13]. Besides, reducing GO nanosheets is effective in protecting membranes from swelling, which can improve the stability of GO membranes [124]. Huang et al. reported solvent solvated rGO membranes with high performance for the organic solvent nanofiltration, which was stable in organic solvents, and strong acidic, alkaline, or oxidative media [125].
In order to achieve better membrane separation, extensive effort has been focused on controlling the interlayer spacing in GO membrane. Creating more space and channels between layers always lead to a considerably enhanced membrane permeance. A widely adopted method to expand interlayer spacing is to introduce nano-sized spacers such as copper hydroxide nanostrands [126], nanoparticles [127–129], carbon nanodots [130], nanorods [131] and nanotubes [132–134] into the interlayers. For instance, Han et al. prepared a GO-based membrane intercalated with TiO2 nanoparticles generated from the oxidation of lamellar MXene. Compared to the traditional intercalation method, MXene served as a sacrificial template to generate continuous water nanochannels because of the consecutive distribution of TiO2 nanoparticles. The resulting membrane demonstrated a water permeability of 89.6 L∙m–2∙h–1∙bar–1 with a rejection rate over 97% for four different organic dyes, Rhodamine B, methylene blue, crystal violet and neutral red [135]. Recently, Yu et al. designed a 2D-2D rGO-TiO2 nanochannel membrane with TiO2 nanosheets on an rGO template via a solvothermal-induced assembly method, and the resulting rGO-TiO2 membrane exhibited a permeance of 9.82 L∙m–2∙h–1∙bar–1 accompanied by a rejection rate of 98.5% for RhB [136]. Moreover, the typical reduction of GO to rGO is a useful way to control the interlayer spacing. Han et al. designed ultrathin rGO membranes with well packed layer structure, which possessed a water permeance of 21.8 L∙m–2∙h–1∙bar–1 with a rejection rate of 99.2% for methyl blue [137].
Some 2D materials including MoS2 [138], WS2 [139], MXene [140,141] and niobate nanosheets [142] have also been combined with GO nanosheets to construct hybrid 2D laminar membranes. For example, Endo et al. fabricated GO/few-layer graphene hybrid 2D membranes followed by thermal treatment and Ca2+ crosslinking. The obtained GO/FLG membrane showed a steady performance (NaCl rejection of 85% and acid blue 9 rejection of 96%) for up to 120 h under intense cross-flow [143]. As for gas separation of GO membranes, Kim et al. reported that gas transport behavior of layered GO membranes can be controlled via different stacking methods in 2013 [144]. Simultaneously, Li et al. prepared 1.8 nm-thick GO membranes with an H2/CO2 selectivity of 3400 accompanied by an H2/N2 selectivity of 900 via a facile filtration process [14]. Encouraged by this success, GO membranes have also been considered as a promising candidate for gas separation. However, controlling the nanochannels/subnanochannels in GO membrane is also a challenge for their application in gas separation. For example, Shen et al. reported a facile external force driven assembly strategy to prepare GO membrane with an interlayer height of ~0.4 nm via the synergetic manipulation of “outer” and “inner” external forces (Fig. 8) [5].
In biological membranes, precise molecule separation not only can be realized by the molecular sieving relying on the size of nanochannels, but also can be realized by facilitated target molecule transport relying on carriers fixed in the nanochannel surfaces. Recently, inspired by dual transport mechanisms of biological membranes, Dou et al. designed Ag/IL-GO membranes with silver ions served as ethylene transport carrier and ionic liquid (IL) tuning the size of nanochannels. Owing to the cooperative effect of the in-plane molecular sieving and plane-to-plane carrier-facilitated transport, the obtained membrane showed a good C2H4 permeance of 72.5 GPU accompanied by a selectivity of 215 for C2H4/C2H6 (Fig. 9(a)) [145].
2D membranes have good chemical and thermal stability, and their nanoscale-thickness makes the water permeation rate very high, which can be used in the treatment of wastewater. In order to realize high separation performance for nanofiltration process, attention should be focused on the design and strategies employed to control interlayer spacing to enhance permeability and selectivity. Besides, the high-quality nanosheets with high yield, and assembling them to 2D membranes with refined structure as designed, as well as the nanofiltration device also have effect on the separation performance. Inspired by the charge interaction in biological membranes, Zhang et al. created surface charges on GO membrane to control ion transport. Owing to the high repulsive interaction, the obtained highly charged GO membrane repelled doubly charged ions, possessing the same charge of the membrane surface. Meanwhile, in order to balance the overall solution charge, the obtained highly charged GO membrane also restrained the permeation of singly charged ions, possessing the opposite charge of the membrane surface (Fig. 9(b)). Therefore, the resulting membrane could separate AB2- or A2B-type salts from water, while simultaneously maintaining a high water flux [15]. More recently, Wen et al. reported a planar heterogeneous interface desalination of a planar heterogeneous graphene oxide membrane (PHGOM) with a NaCl rejection rate of 97.0%, but a water flux of 1529 L∙m–2∙h–1∙bar–1. The PHGOM was prepared via the dual-flow filtration with a clapboard to separate negatively charged GO nanosheets and positively charged GO nanosheets (Fig. 9(c)). In the presence of an electric field, the ion transport behavior of the sodium and chloride resulted in a salt concentration depletion state transition zone, and thus deionized water can be extracted from the transition zone via an inverted T-shaped water extraction mode (Fig. 9(d)) [146]. PHGOM membrane exhibited good salt rejection with ultrahigh water flux, almost two or three orders of magnitude higher than other 2D membranes. Compared with the GO membrane [15], the composite GO-PVAm-silica membrane [123] present higher water flux and higher ion rejection due to the specially designed structure. In addition to the manipulation of the interlayer spacing and improving membrane performance stability, special attention should also be focused on the design of the desalination device to enhance permeation and selectivity for desalination processes.
Fig.9 (a) Synergistic effects of molecular sieving and carrier-facilitated transport by imitating biological protein nanochannels. Reprinted with permission from Ref. [145], copyright 2019 John Wiley and Sons; (b) Schematic of the design of surface-charged GO membranes. Reprinted with permission from Ref. [15], copyright 2019 Nature Publishing Group; (c) Scheme illustration of the preparation of the PHGOM. (d) Scheme of the working mechanism of a PHGOM-based device for water desalination. Reprinted with permission from Ref. [146], copyright 2020 Wiley VCH. |
4.2.2 LDH membranes
LDHs, as a representative of anionic clays, are formed by positively charged layers and interlayer galleries containing charge-compensating anions [147]. In 2008, Kim et al. prepared the LDH thin films via electrophoretic deposition method, exhibiting slight permselective property toward CO2 [148]. Then, they prepared MgAl-CO3 LDH membranes by a vacuum-suction method. In order to reduce voids and pinholes in the membranes, the obtained LDH membrane was repaired by silicone coating, exhibiting a CO2/N2 selectivity of 34.4 [149]. In 2014, Caro’s group fabricated NiAl-CO3 LDH membranes via the in situ hydrothermal growth, realizing an H2/CH4 separation factor of 80 [150]. Recently, Liu et al. reported 2D LDH membranes with inherent CO2-selective transport channels. Owing to the intrinsic breathing effect of LDH toward CO2, the intercalated ions transformed from CO2 regulated the channel from 0.7 nm to 0.3 nm by electrostatic interaction with LDH layers, endowing the resulting membrane with a good separation performance for CO2/CH4 [151]. Furthermore, Caro’s group designed a ZIF-8@MgAl-CO3 LDH membrane, introducing a new concept of “LDH buffer layer”. First, vertically aligned MgAl-CO3 LDH layers were fixed on substrates. Then, ZIF-8 seeds were deposited on MgAl-CO3 LDH layers, which could prevent ZIF-8 seeds from falling off (Fig. 10(a)). Finally, the ZIF-8@MgAl-CO3 LDH membrane with an H2/CH4 separation factor of 12.9 accompanied by an H2 permeance of 1.4 × 10–7 mol∙m–2∙s–1∙Pa−1 was achieved via secondary growth [152]. Furthermore, Caro’s group further prepared vertically aligned ZnAl-CO3 LDH buffer layers on substrates via in situ solvothermal growth by the urea hydrolysis method (Fig. 10(b)). Contrary to the physical interaction between ZIF-8 and MgAl-CO3 LDH network, this work reported the high-affinity metal-imidazole interaction between ZnAl-CO3 LDHs and ZIF-8 crystals, where ZIF-8 membranes can be directly synthesized on the ZnAl-CO3 LDHs. The as-prepared composite membrane demonstrated an H2/CH4 separation factor of 12.5 with an H2 permeance of 1.4 × 10–7 mol∙m–2∙s–1∙Pa–1 [153]. In order to further improve the gas selectivity of ZIF-8@ZnAl-NO3 LDH composite membrane, Caro’s group used the controlled calcination of a ZnAl-NO3 LDH membrane to create a ZnO buffer layer, which was subsequently turned into ZIF-8 membrane through a solvothermal treatment with the mixed H2O/DMF solution containing ligand of 2-methylimidazole (Fig. 10(c)). The obtained ultra-thin membrane exhibited an excellent H2/CH4 separation factor of 83.1 accompanied by an H2 permeance of 1.9 × 10–8 mol∙m–2∙s–1∙Pa–1 [154]. Recently, Caro’s group designed COF-in-LDH membranes via in situ growth of 2D COF (COF-LZU1 or TFB-BD) on vertically aligned CoAl-LDH nanosheets (Fig. 10(d)). In contrast to 2D COF membranes based on intrinsic pore sieving, the separation of the as-obtained membrane was mainly based on the COF interlayer space. The obtained COF-LZU1 membrane demonstrated an H2 permeance of 3655 GPU accompanied by an H2/CO2 selectivity of 31.6 [155].
Besides gas separation, LDH-based membranes have received increasing attention for liquid separation [156–159]. For example, Wang et al. prepared an intercalated MgAl-LDH membrane intercalated with amino acids on porous tubes via an in situ hydrothermal process. The glycine-intercalated LDH composite membrane possessed a water permeance of 566 L∙m–2∙h–1∙MPa–1 with a rejection rate of 98.5% for Eriochrome Black T molecules [160]. Recently, as revealed by Ang et al., the solvent permeation of LDH membranes improved by altering the divalent cation in the LDH crystal structure, providing preliminary guidelines for their application as organic solvent nanofiltration membranes [161].
Fig.10 (a) The formation of ZIF-8 membranes on vertically aligned MgAl-CO3 LDH layers. Reprimted with permission from Ref. [152], copyright 2014 The Royal Society of Chemistry; (b) The formation of ZIF-8 membranes on ZnAl-CO3 LDH buffer layers. Reprinted with permission from Ref. [153], copyright 2014 American Chemical Society; (c) The formation of ZIF-8 membranes via partial conversion of a ZnO buffer layer from a ZnAl-NO3 LDH layer. Reprinted with permission from Ref. [154], copyright 2015 Wiley VCH; (d) Illustration of the vertically aligned COF membrane formation. Reprinted with permission from Ref. [155], copyright 2020 American Chemical Society. |
4.2.3 MXene membranes
Transition metal carbides and/or nitrides, named MXenes, were first reported by Gogotsi’s and Barsoum’s groups [162,163]. As a family of emerging 2D materials, MXene nanosheets are similar to GO nanosheets and can be used into 2D membranes for size-selective separations. Ti3C2Txis one of the representative MXenes, usually prepared by a HF or HCl/LiF etching process. Gogotsi’s group reported the very first example of 2D Ti3C2Tx MXene membranes via a vacuum-assisted filtration with exfoliated 2D Ti3C2Tx sheets. Because of the presence of H2O between the hydrophilic Ti3C2Tx layers, the obtained 2D Ti3C2Tx membrane exhibited a water flux of 37.4 L∙m–2∙h–1∙bar–1 [164]. Then, our group designed a 2D Ti3C2Tx membrane with positively charged Fe(OH)3 colloidal nanocrystals serving as the distance holder to create expanded nanochannels followed by HCl dissolution (Fig. 11(a)). The obtained MXene membrane exhibited an excellent water permeance of 1084 L∙m–2∙h–1∙bar–1 accompanied by an Evans blue rejection rate of 90% [16]. Besides liquid separation, our group also applied the MXene membranes to gas separation, exhibiting aligned and regular subnanochannels. The obtained membrane exhibited a good H2/CO2 selectivity>160 with an H2 permeability>2200 Barrer [17]. Subsequently, Li et al. studied the gas transportation mechanism of different gas molecules in MXene membranes by MD simulations. The simulation results proved that the structure and interlayer spacing of MXene membrane were important for the gas diffusion [165]. Shen et al. achieved precise manipulation of stacking behaviors and interlayer spacing of MXene nanosheets by borate and polyethylenimine molecules [6]. Wang et al. designed 2D lamellar MXene membranes with regular and straight interlayer channels via the direct self-stacking of rigid MXene nanosheets. The resulting membrane with ordered 2 nm channels showed a precise molecular (larger than 2 nm) rejection and a water permeance of 2300 L∙m–2∙h–1∙bar–1 [7]. Wu et al. applied hydrophilic (–NH2) and hydrophobic (–C6H5, –C12H25) groups to tune the size and wettability of nanochannels. Compared to hydrophobic nanochannels with disordered molecular configuration, hydrophilic nanochannels with highly ordered alignment aggregates displayed better separation permeance. However, the permeance of the nonpolar molecule was comparable in both hydrophilic and hydrophobic nanochannels [166]. Recently, Xing et al. adopted a freeze-drying method to directly synthesize the crumpled Ti3C2Tx MXene nanosheets from flat Ti3C2Tx MXene nanosheets for the fabrication of c-Ti3C2Tx MXene membranes via vacuum filtration (Fig. 11(b)). The obtained c-Ti3C2Tx MXene membrane with larger interlayer channels demonstrated ultrafast permeation for water and acetone [167]. Moreover, the swelling of MXene membranes in aqueous solution is still a challenge to overcome. Recently, our group fabricated non-swelling Al3+-intercalated MXene membranes. Affinity interactions between Al3+ and the abundant oxygen functional groups terminating at the MXene surface made the non-swelling MXene membrane exhibit a long-term stability in liquid water up to 400 h with a water permeance of 2.8 L∙m−2∙h−1 and a rejection rate of 96.5% for NaCl [168]. Moreover, our group reported a self-crosslinked MXene membrane in the presence of Ti–O–Ti bonds for a stable interlayer spacing, which could improve the performance of ion exclusion and anti-swelling property [169].
Besides pure MXene membranes, MXene-based separation membranes have also received increasing attention. For instance, Wu et al. prepared Ti3C2Tx/PEI and Ti3C2Tx/PDMS MMMs. Compared to the pure polymer membranes, the alcohol fluxes of the as-obtained MMMs increased by 30% and 162%, respectively [170]. Then, Hao et al. further incorporated functionalized Ti3C2Tx nanosheets with four types of groups into polymers. As a consequence, the obtained membrane with multi-functional Ti3C2Tx nanosheets enhanced the fluxes [171]. More recently, Shamsabadi et al. introduced 2D Ti3C2Tx MXene nanosheets into the Pebax-1657 to improve the separation performance [172]. Furthermore, Gao et al. reviewed a great deal of work about MXene/polymer membranes and provided perspectives and useful guidelines for the development of MXene/polymer membranes [173].
Fig.11 (a) The preparation of MXene membranes with Fe(OH)3 served as distance holder. Reprimted with permission from Ref. [16], copyright 2017 Wiley VCH. (b) The fabrication of Ti3C2Tx membranes and c-Ti3C2Tx membranes. Reprinted with permission from Ref. [167], copyright 2020 American Chemical Society. |
4.2.4 TMD membranes
2D TMDs are a new type of 2D layered materials, which have received intense attention after graphene. The general chemical formula of TMDs is MX2, where M is the transition metal element and X represents chalcogen. Many molecular simulations showed a bright prospect of TMD materials for membrane-based separation, especially MoS2 nanosheets and WS2 nanosheets [174–176]. Sun et al. reported the first example of laminar MoS2 membranes with a water permeance of 245 L∙m–2∙h–1∙bar–1 accompanied by an Evans blue rejection rate of 89% by the assembly of atom-thick MoS2 sheets [177]. Then, they also prepared 2D nanostrands-channelled WS2 membranes via filtering the nanostrands/WS2 nanosheet suspension on the porous AAO substrates. Notably, they found that the nanochannels created by the ultrathin nanostrands between the WS2 sheets cracked under high pressure. Owing to the new fluidic nanochannels created by the crack, the obtained membrane showed an increase in the flux without a markedly reduced rejection for Evans blue [178]. Motivated by their work, this topic received increasing attention. In order to improve the separation performance, Hirunpinyopas et al. proposed a chemical dye functionalization strategy to protect membranes from swelling when exposed to the water environment. The main procedure is to filter exfoliated MoS2 dispersion on a PVDF membrane and subsequently immerse in the dye solution. The resulting membrane showed a good stability without swelling in water for periods exceeding 6 months [179]. Ang et al. used cationic LDH nanosheets to control the interlayer spacing of anionic TMD nanosheets, and the obtained membrane displayed almost 100% rejection of methyl orange, dissolved in acetone [180]. In addition, Ries et al. reported the covalently functionalized molybdenum disulfide (MoS2) nanosheets to prepare the 2D MoS2 membranes with different functional groups. Modification with different functional groups could not only control the interlayer spacing, but also control the surface chemistry of 2D MoS2 membranes. The acetamide-functionalized 2D MoS2 membrane demonstrated a water flux of 33.7 L∙m–2∙h–1∙bar–1 accompanied by a NaCl rejection of 82% [181].
In addition to the widely used solvent-exfoliation method to prepare nanosheets, Li et al. successfully prepared 7 nm-thick 2D MoS2 membranes via the integration of chemical vapor deposition-grown centimeter-scale nanosheets onto porous polymeric substrates (Fig. 12). The obtained ultrathin membrane displayed a water permeance of 322 L∙m–2∙h–1∙bar–1 accompanied by the ion rejection rates of 99% for various seawater salts including Na+, K+, Ca2+ and Mg2+ [18]. Owing to the high-quality nanosheets and assembling them to 2D membranes with refined structure as designed, such atom-thick membrane made from chemical vapor deposition showed a ultrahigh water permeance with excellent ionic sieving capability compared to other 2D MoS2 membranes made from exfoliated nanosheets reported by Ries et al. [181]. Recently, Hu et al. designed tannic acid-modified MoS2 nanosheets via a tannic acid assisted exfoliation method. The obtained hybrid membrane with 1 wt% TAMoS2 in MoS2 membrane displayed a water flux of 10000 L∙m–2∙h–1∙bar–1 accompanied by a rejection of 98% for methylene blue [182].
Although TMD membranes are often used for liquid separation, a growing number of articles have been published on TMDs membranes for gas separation. In 2015, Wang et al. prepared the first example of ultrathin 2D MoS2 membranes via a simple filtration for the gas separation. Owing to the presence of the larger stacking space, the gas transportation in 17 nm-thick MoS2 membranes present a Knudsen diffusion behavior, and the obtained membrane demonstrated a high H2 permeance of 27440 GPU but a poor H2/CO2 selectivity (~3) [183]. Then, Achari et al. prepared 500 nm-thick MoS2 membranes with an H2 permeability of 1175 Barrer and an H2/CO2 selectivity of 8.29. They also discovered a phase transition of MoS2 via heating at 160 °C, inducing the decrease of membrane interlayer spacing from 11.32 to 6.12 Å and consequent enhancement (30%) in H2 permeability of the membranes without reducing the selectivity [184]. Shen et al. designed a composite MMM via a simple drop-coating and evaporation method. The resulting MoS2-Pebax/PDMS/PSf membrane possessed a CO2 permeability of 64 Barrer accompanied by a CO2/N2 selectivity of 93 [185]. Moreover, Chen et al. reported that 1-butyl-3-methylimidazolium tetrafluoroborate IL was confined into the channels of MoS2 membranes and WS2 membranes via infiltration. Owing to the high CO2 solubility of the IL, membranes with the nanoconfined IL exhibited a good CO2 permeance and excellent CO2/gas selectivities [19,186].
Tables 1 and 2 summarize detailed fabrication approaches of various 2D-material membranes mentioned in this review.
Tab.1 Summary of gas separation for 2D-material membranes |
| Feed condition | Membrane | Fabrication approach | Permeability/permeance (of faster species) | Selectivity | Ref. |
|---|---|---|---|---|---|
| p-Xylene/o-xylene | MFI | Secondary growth | 4 × 10–7 mol∙m–2∙s–1∙Pa–1 | 25–45 | [32] |
| p-Xylene/o-xylene | Silicalite-1 | Gel-free secondary growth | 1.3 × 10–7 mol∙m–2∙s–1∙Pa–1 | 1050 | [33] |
| p-Xylene/o-xylene | MFI | Gel-free secondary growth | 2.4 × 10–7 mol·m–2·s–1·Pa–1 | 500 | [34] |
| p-Xylene/o-xylene | MFI | Gel-free secondary growth | 5.6 × 10–7 mol·m–2·s–1·Pa–1 | 2000 | [35] |
| p-Xylene/o-xylene | MFI | Gel-free secondary growth | 2.9 × 10–7 mol·m–2·s–1·Pa–1 | >10000 | [20] |
| H2/CO2 | AMH-3/PBI | Casting | 1 Barrer | 35 | [39] |
| H2/CO2 | Zn2(bim)4 | Hot-drop coating | 2700 GPU | 291 | [4] |
| H2/CO2 | Zn2(bim)3 | Hot-drop coating | 8 × 10–7 mol·m–2·s–1·Pa–1 | 166 | [22] |
| H2/CO2 | CuBDC-GO | Vacuum filtration | 9.6 × 10–7 mol·m–2·s–1·Pa–1 | 95.1 | [52] |
| H2/CO2 | Zn2(bim)4/GO | Direct growth | 1.4 × 10–7 mol·m–2·s–1·Pa–1 | 106 | [55] |
| H2/CO2 | MAMS-1 | Hot-drop coating | 553±228 GPU | 235±14 | [23] |
| H2/CO2 | [Cu2(ndc)2(dabco)]n/PBI | Casting | 6.13±0.03 Barrer | 26.7 | [49] |
| H2/CO2 H2/N2 H2/CH4 | Zn2(bim)4 | Direct growth | 2.04 × 10–7 mol·m–2·s–1·Pa–1 | 53 67 90 | [56] |
| H2/CO2 | Co2(bim)4 | Vapor phase transformation | 1.72 × 10–7 mol·m–2·s–1·Pa–1 | 58.7 | [57] |
| H2/CO2 | c-Oriented ZIF-L | Secondary growth | 1.95 × 10–7 mol·m–2·s–1·Pa–1 | 24.3 | [53] |
| H2/CO2 | MXene | Vacuum filtration | 2226.6 Barrer | 167 | [17] |
| H2/CO2 | Polyimide-PGM | Vapor-liquid interfacial polymerization | 6.85 × 10–6 mol·m–2·s–1·Pa–1 | 6.41 | [105] |
| H2/CO2 | TpEBr@TpPa-SO3Na | Layer-by-layer assembly | 2566 GPU | 22.6 | [25] |
| H2/CO2 | COF-LZU1 | In situ growth | 3654.8 GPU | 31.6 | [155] |
| H2/CO2 H2/N2 H2/CH4 | COF-LZU1-ACOF-1 | Temperature-swing solvothermal approach | 2.24 × 10–7 mol·m–2·s–1·Pa–1 2.38 × 10–7 mol·m–2·s–1·Pa–1 1.82 × 10–7 mol·m–2·s–1·Pa–1 | 24.2 83.9 100.2 | [65] |
| H2/CO2 H2/N2 | GO | Vacuum filtration | 10–7 mol·m–2·s–1·Pa–1 | 3400 900 | [14] |
| H2/CO2 H2/C3H8 | EFDA-GO | External force driven assembly | 840–1200 Barrer 3.9 × 10−7 mol·m–2·s–1·Pa–1 | 29–33 260 | [5] |
| H2/N2 | MCM-22/Silica | Layer-by-layer deposition | 2.09 × 10–8 mol·m–2·s–1·Pa–1 | 7.5 | [37] |
| H2/N2 | MCM-22/Silica | Deposition cycles | 10–8 mol·m–2·s–1·Pa–1 | >100 | [38] |
| H2/CH4 | JDF-L1/6FDA-4MPD+ 6FDA-DABA | Casting | 137±14 Barrer | 35.6±1.4 | [41] |
| H2/CH4 | JDF-L1/polysulfone | Casting | 12.5 Barrer | 128±13 | [42] |
| H2/CH4 | MCM-41+ JDF-L1/ 6FDA-4MPD+ 6FDA-DABA | Casting | 440 Barrer | 32 | [43] |
| H2/CH4 | NiAl-CO3 LDH | In situ growth | 4.5 × 10–8 mol·m–2·s–1·Pa–1 | 78 | [150] |
| H2/CH4 | ZIF-8@MgAl-CO3 LDH | In situ growth Secondary growth | 1.4 × 10–7 mol·m–2·s–1·Pa–1 | 12.9 | [152] |
| H2/CH4 | ZIF-8@ZnAl-CO3 LDH | In situ growth | 1.4 × 10–7 molm–2·s–1·Pa–1 | 12.5 | [153] |
| H2/CH4 | ZIF-8@ZnAl-CO3 LDH | In situ growth | 1.9× 10–8 mol·m–2·s–1·Pa–1 | 83.1 | [154] |
| H2/CH4 | Porous graphene | LPCVD | 6045 GPU | 15.6 | [103] |
| CO2/CH4 | AMH-3/cellulose acetate | Casting | 10.36±0.25 Barrer | 30.03±0.34 | [40] |
| CO2/CH4 | CuBDC/PI | Casting | 2.78±0.02 Barrer | 88.2±1.3 | [48] |
| CO2/CH4 | CuBDC/PIM-1 | Casting | 407.3 GPU | 15.6 | [50] |
| CO2/CH4 | CuBDC/6FDA-DAM | Casting | 430±10 Barrer | 43±3 | [51] |
| CO2/CH4 | LDH () | Spin-casting | 150 GPU | 33 | [151] |
| CO2/N2 | MgAl-CO3 LDH | Vacuum-suction | 2.07 × 10–7 mol·m–2·s–1·Pa–1 | 35 | [149] |
| CO2/N2 | MoS2-Pebax/PDMS/PSf | Drop-coating | 64 Barrer | 93 | [185] |
| CO2/N2 CO2/CH4 | MoS2 SILM | Vacuum filtration | 47.88 GPU | 131.42 43.52 | [186] |
| CO2/N2 CO2/CH4 | WS2 SILM | Vacuum filtration | 47.3 GPU | 153.21 68.81 | [19] |
| C2H4/C2H6 | Ag/IL-GO | Vacuum filtration Spin-coating | 72.5 GPU | 215 | [145] |
| n-Butane/i-butane | MFI | Vacuum filtration | 1923 GPU | 58 | [36] |
Tab.2 Summary of nanofiltration for 2D-material membranes |
| Feed system | Membrane | Fabrication approach | Water flux | Rejection/% | Ref. |
|---|---|---|---|---|---|
| Evans blue | g-C3N4 | Vacuum filtration | 29 L·m–2·h–1·bar–1 | 87 | [26] |
| Evans blue | g-C3N4-PAA | Vacuum filtration | 117 L·m–2·h–1·bar–1 | 83 | [69] |
| Evans blue | MXene | Vacuum filtration | 1084 L·m–2·h–1·bar–1 | 90 | [16] |
| Evans blue | MoS2 | Vacuum filtration | 245 L·m–2·h–1·bar–1 | 89 | [177] |
| Evans blue | WS2 | Vacuum filtration | 1850 L·m–2·h–1·bar–1 | 82 | [178] |
| Evans blue | NSC-GO | Vacuum filtration | 695 L·m–2·h–1·bar–1 | 83.5 | [126] |
| Evans blue | NbN/GO | Vacuum filtration | 20 L·m–2·h–1·bar–1 | 98 | [142] |
| Rhodamine B | CDs–GO | Vacuum filtration | 439 L·m–2·h–1 | 96.9 | [130] |
| Rhodamine B | Fe3O4@rGO | Filtration-disposition | 296 L·m–2·h–1·bar–1 | 98.14 | [128] |
| Rhodamine B | rGO-TH | Vacuum filtration | 8526 · 30 L·m–2·h–1·bar–1 | 99±1 | [121] |
| Rhodamine B | Tp-AD | Vacuum filtration | 596 L·m–2·h–1·bar–1 | 98 | [61] |
| Rhodamine B | SWCNT/GO | Vacuum filtration | 710 ·50 L·m–2·h–1·bar–1 | 97.4±0.3 | [132] |
| Rhodamine B | g-C3N4 NT/rGO | Vacuum filtration | 4.87 L·m–2·h–1·bar–1 | 98 | [134] |
| Rhodamine B | CA/GO-TiO2 | Vacuum filtration | 33.2 L·m–2·h–1 | 99.4 | [131] |
| Rhodamine B | GO/TiO2 | Vacuum filtration | 89.6 L·m–2·h–1·bar–1 | 99.3 | [135] |
| Rhodamine B | rGO-TiO2 | Secondary growth | 9.82 L·m–2·h–1·bar–1 | 98.5 | [136] |
| Methylene blue | TAMoS2 | Vacuum filtration | 10000 L·m–2·h–1·bar–1 | 98.26 | [182] |
| Methylene blue | BPEI/GO | Vacuum filtration | 2.09 L·m–2·h–1·bar–1 | 96.4 | [117] |
| Methylene blue | WS2/GO | Vacuum filtration | 159.6 L·m–2·h–1·bar–1 | 96.3 | [139] |
| Methylene blue | GO/MXene | Vacuum filtration | 71.9 L·m–2·h–1·bar–1 | 99.5 | [141] |
| Methyl blue | rGO | Vacuum filtration | 21.8 L·m–2·h–1·bar–1 | 99.2 | [137] |
| Methyl blue | SPPO/g-C3N4 | Vacuum filtration | 8867 L·m–2·h–1·bar–1 | 100 | [27] |
| Methyl red | Zn-TCP(Fe)/PEI | Vacuum filtration | 4243 L·m–2·h–1·bar–1 | 98.2 | [54] |
| Methyl orange | MgAlLDH | Vacuum filtration | 298 L·m–2·h–1·bar–1 | 99.5 | [161] |
| Methyl orange | MWNTs/GO | Vacuum filtration | 8.69 L·m–2·h–1·bar–1 | 96.1 | [133] |
| Congo red | GO/NH2-Fe3O4 | Vacuum filtration | 15.6 L·m–2·h–1·bar–1 | 94 | [129] |
| Congo red | GO/MoS2 | Pressure-assisted filtration | ~10.2 L·m–2·h–1·bar–1 | 99.6 | [138] |
| Eriochrome black T | MgAl-LDH | In situ growth | 566 L·m–2·h–1·MPa–1 | 98.5 | [160] |
| Chrome black T | COF-LZU1 | In situ growth | 760 L·m–2·h–1·MPa–1 | 98 | [64] |
| Acid yellow 14 | c-Ti3C2Tx | Vacuum filtration | 344 L·m–2·h–1·bar–1 | 76.4 | [167] |
| NaCl | PHGOM | Dual-flow filtration | 1529 L·m–2·h–1·bar–1 | 97 | [146] |
| NaCl | GNM/SWNT | O2 plasma drilling | 22 L·m–2·h–1·bar–1 | 98.1 | [106] |
| NaCl | MXene | Vacuum filtration | 2.8 L·m–2·h–1 | 96.5 | [168] |
| NaCl | MoS2 | Chemical vapor deposition | >322 L·m–2·h–1·bar–1 | >99 | [18] |
| NaCl | MoS2 | Vacuum filtration | 33.7 L·m–2·h–1·bar–1 | 82 | [181] |
| NaCl | GO-PVAm-Silica | Pressure-assisted filtration | 80.2 · 0.8 kg·m–2·h–1 | 99.99 | [123] |
| NaCl | g-C3N4-PA | Interfacial polymerization | 45 g·m–2·h–1 | 98 | [76] |
| MgCl2 | GO | Pressure-assisted filtration Dip-coating | 51.2 L·m–2·h–1·bar–1 | 93.2 | [15] |
| CoCl2 | Al-MOF | Vacuum filtration | 2.22 mol·m–2·h–1·bar–1 | 100 | [47] |
| Na2SO4 | COFs@CNFs | Vacuum filtration | 42.8 L·m–2·h–1·bar–1 | 96.8 | [24] |
5 Challenges and perspectives
2D membranes, as the most promising family of separation membranes, have received increasing attention in recent years. The nanostructures/subnanostructures and microenvironments of nanopores and/or nanochannels have been tuned by ion/molecule intercalation, interlayer cross-linking, and charge modification, resulting in promising 2D membranes with precise molecular sieving. Besides, the ultrathin 2D membranes would greatly shorten the molecule transport pathway, minimizing mass-transfer resistance and thus increasing membrane permeance, while simultaneously maintaining a high selectivity, surpassing the upper bound of the separation performance of traditional membranes.
As summarized in Tables 1 and 2, even with significant breakthrough in the development of 2D membranes, there are remaining problems for 2D membranes that should be addressed before their practical utilization. 1) The swelling problem of some 2D lamellar membranes, such as MXene and GO membranes is still one of the most important issues, hindering their applications in liquid systems, such as ion sieving. Facile, effective and universal strategy should be proposed to solve this problem, with potential of promoting the development of lamellar membranes. 2) The second one is the facile scale-up technology. The main reason that 2D membranes cannot beat polymeric membranes is the scale-up industrialization of 2D membranes. It is totally different to produce 2D membranes with huge effective membrane area coupled to membrane modules comparing with lab-scale preparation. The nonselective defects within 2D membranes should be avoided during scale-up production. Most 2D membranes are prepared by the traditional method of vacuum-assisted filtration, where several hours are usually needed to assemble a membrane, which is time-consuming. More importantly, it is difficult to prepare defect-free and homogenous 2D membranes with large area by vacuum filtration, because the driving force across the porous substrate (pressure difference) along the radial direction will be different (different, e.g., at the edge and centre position), resulting in inhomogeneous membrane structure and sacrificial separation performance. Therefore, the preparation method of 2D lamellar membranes is a key bottle neck that blocks their scale-up for industrialization and applications. Electrostatic deposition seems to be a good way to produce 2D membranes with large area, which has many advantages such as facile and rapid operation, homogenous stacking structure and low energy cost. Up to now, it is still difficult to prepare 2D membranes with exactly homogeneous microstructure, especially in a large scale of membrane area. Hence, more simple and cost-effective ways for fabricating 2D membranes need to be studied further. Furthermore, the issues such as the steady production of high-quality nanosheets with high yield, together with how to scale up such atom-thick membranes into applicable separation devices need to be further studied. Besides, development of robust 2D membranes with stable separation performance under realistic operational conditions is also important. 3) Moreover, the mass transport mechanisms in the confined nanochannels/ subnanochannels within 2D membranes also need to be discovered via more cases, including the simulation models, diffusion laws that are different from the classical fluid mechanics, the interaction between molecules/ions and wall of nanochannels. The understanding of mass transportation behavior in 2D membranes from the atomic level and theoretical breakthrough of the molecules/ions transport in confined nanochannels/ subnanochannels will also help to design novel structures of 2D lamellar membranes or pave new routes for next generation of membranes.
In the next few years, 2D membranes will be focused on the types basing on porous nanosheets, where the prolonged zigzag pathway in traditional lamellar membranes based on non-porous naonsheets can be avoided effectively, while sharply reducing the permeation resistance. In addition, the ultrathin 2D membranes with thickness of only several nanometers will be another research hotspot, where the mass transportation route can be decreased maximumly, resulting in ultrahigh permeance and promoting further applications of 2D membranes.
Nevertheless, 2D membranes are considered to be promising candidates for separation as the next-generation membranes, which have the possibility to surpass the upper bound of separation performance. It is undeniable that there is still a long way to go for the wide practical utilization of 2D membranes