In this contribution, we report a series of 1D MOF-in-2D COF hetero-structured composite membranes prepared by a two-phase interfacial polymerization (IP). The chain-like 1D MOFs can be orientally (in the (1 0 -1) plane) developed inside the 1D channels of the 2D COF membrane benefiting from the nano-confined effect, ligand trapping, and interlocking growth process, therefore achieving the strategic construction of heterogeneous ion transport channels with narrowed pore size and weakened K+-channel wall interactions (Fig. 1). The prepared composite membrane presents a molecular-level interlinked hybridization of covalent- and metal- organic frameworks, which is induced by the coordination between the -NH groups in COFs and the Cu centers from MOFs. Most importantly, the 1D MOF-in-2D COF composite membrane exhibits unprecedented single-species selective properties similar to KcsA channels with ultrahigh K+/Na+ selectivity of 82.52 and K+/Mg2+ selectivity of 1131.07.
As a proof of concept, 2D TAPA-TFP COF membranes with adjustable thickness (Supplementary Fig. 1a-d, Supplementary Data 1), rigid skeleton, superior tenacity and flexibility, and vertical 1D channels are chosen as the solid scaffolding and nano-confined template for composite membrane fabrication (Fig. 1). To achieve the molecular-level interlinked hybridization of covalent- and metal- organic frameworks, the chain-like 1D NH-CuBDC MOFs featuring accessible Cu sites and coordination-exchange characteristics are selected as a showcase (Fig. 1b, Supplementary Fig. 1e, Supplementary Data 2). Mulliken population analysis (Fig. 1c, Supplementary Table 1) indicates that electron-rich regions are localized over N atoms of TAPA-TFP COFs and the N atoms from -NH groups with a Mulliken atom charge of -0.63 are inclined to coordinate with the Cu centers from MOFs due to the lower steric hindrance (Fig. 1d). Simulation results indicates the short-range coordination interaction between MOFs and COFs (Supplementary Fig. 1f).
Typically, after the successful preparation of the self-standing TAPA-TFP COF membrane at the interface, NH-BDC is slowly injected into the underlying oil phase by an injection syringe to allow the -NH of NH-BDC to fully react with the -CHO at the COF defects (Note that research and simulation have confirmed that regardless of how perfect the crystal growth and stack, topological defects of COFs are inevitable) (Fig. 2a). Furthermore, NH-BDC can be captured and in-situ immobilized into the 1D COF channels waiting for the coordination with Cu atoms due to van der Waals and hydrogen bonding interactions (Fig. 2b). Subsequently, Cu(NO) is gently injected into the aqueous phase. Due to its small hydrated diameter (0.8 nm) and capillary effect, Cu can easily drill into the channels and coordinate with NH-BDC. The obtained composite membrane is denominated as TAPA-TFP-x-NH-CuBDC CMOF membrane, with x being the concentration of NH-BDC (Supplementary Fig. 2). The TAPA-TFP-x-NH-CuBDC CMOF membrane with superior mechanical properties shows a Janus morphology and its bottom surface is uniformly distributed with broken thin COF vesicles that nearly occupy the surface, in contrast to the pristine COF membrane (almost identical top and bottom surface morphologies) (Fig. 2c, Supplementary Fig. 2, Supplementary Movie 1). With the hierarchical increase of MOF ligand concentration, the thickness of the CMOF membrane progressively rises (Supplementary Fig. 2). The multifold increase in thickness suggests that the MOFs can grow into the interlayer or surface of the pristine COF membrane, or the insertion of MOFs can promote the secondary growth of the pristine COF membrane. Surprisingly, we failed to find definite MOFs on the top and bottom surfaces of CMOF membranes (Supplementary Fig. 3). The uniform Cu signals from energy-dispersive X-ray spectroscopy (EDXS) suggest a good dispersion of MOFs in the CMOF membranes (Fig. 2c, Supplementary Fig. 4). Noteworthy, individual copper ions from Cu(NO) cannot coordinate with COFs (Supplementary Fig. 5). HRTEM (high-resolution transmission electron microscope) images reveal the interlinked covalent- and metal- organic hetero-frameworks, and the lattice fringes from both MOFs (0.27 nm for the (5 0 -5) plane) and COFs (0.39 nm for the (0 0 1) plane) are observed (Fig. 2d, Supplementary Fig. 6). The TAPA-TFP-0.25-NH-CuBDC CMOF composite membrane features a main pore diameter of 0.68 nm (basically consistent with simulation results), which is smaller than that of COFs (1.22 nm) or MOFs (1.42 nm), also demonstrating the molecular-level interlinked hybridization frameworks (Fig. 2e, Supplementary Fig. 7a-e). It is noteworthy that the CMOF composite membrane exhibits another weaker peak at 1.20 nm, which is similar to the pore size of pristine COF membrane. This suggests that the chain-like MOFs may not fully occupy the COF pore channels in a complete top-down filling manner (Supplementary Fig. 7f). Computation results indicate that in the complex confined channels with a diameter of 0.68 nm, although the -COOH groups carry significant charges in an aqueous solution environment, they are almost unable to combine water molecules to form stable hydrated layers. Therefore, the pore size obtained from BET can represent the effective aperture for ion transport (Supplementary Fig. 7g). Moreover, the CMOF composite membrane displays a smaller surface area (88.24 m g) in comparison with the pristine COF membrane (307.41 m g) due to the presence of MOFs in the 1D channels. The actual MOF weight loadings in CMOF membranes are confirmed by a thermogravimetric analyzer (Supplementary Figs. 8, 9).
XPS (X-ray photoelectron spectroscopy), UV-Vis-NIR, XANES (synchrotron-based X-ray absorption near edge structure) and ATR-FTIR (attenuated total reflection-Fourier transform infrared microscope) confirm that different from the reported COF-MOF hybrids where definite COFs or MOFs can be observed and core-shell or coating structures can be achieved by a simple combination of COFs and MOFs, the prepared TAPA-TFP-x-NH-CuBDC CMOF composite membrane presents a molecular-level interlinked hybridization of covalent- and metal- organic frameworks, which is induced by the coordination between the -NH groups in COFs and the Cu centers from MOFs (Supplementary Figs. 10-14, Supplementary Table 2). The slight difference in surface wettability between the pristine COF membrane and CMOF composite membranes results from the introduction of MOFs (Supplementary Fig. 15). Close examinations of the XRD peak position and intensity and GIWAXS (grazing incidence wide angle X-ray scattering) confirm that the TAPA-TFP-5-NH-CuBDC CMOF membrane is preferentially oriented and the diffraction peaks at 6.1° and 12.2° are respectively assigned to the peaks from the (1 0 -1) and (2 0 -2) crystallographic planes of MOFs (Figs. 1b and 2f). When reducing the concentration of NH-BDC or increasing the thickness of the pristine COF membrane, the XRD peaks of MOFs almost disappear, suggesting that the XRD peaks of MOFs are easily obscured after growing into COFs (Supplementary Fig. 16). It is worth mentioning that the sharp XRD peaks of the pristine COF membrane that should have been detected (The ordered and definite structure of the COF membrane had already been established before forming the CMOF membrane as displayed in Fig. 2a.) are not distinctly observed in the CMOF composite membrane, which may result from: 1) the introduction of MOFs into the pore structure or interlayer of COFs based on coordination interactions masking the XRD signals from COFs, and 2) the intense XRD signals from MOFs overwhelmingly suppressing the COF signals, making the XRD peaks from COFs extremely difficult to observe when MOFs and COFs coexist. Still, the (110) crystallographic plane with weak signals of pristine COF membrane can be observed in CMOF composite membrane (Fig. 2f), and its right-shift compared with pristine COF membrane confirms the effective growth of MOFs within the COF channels and the relatively strong interaction between MOFs and COFs. When increasing the thickness of the pristine COF membrane or changing the distribution of MOF monomers in two phases during the IP process, the obtained CMOF membranes exhibit comparable characteristics and definite MOFs are still not observed (Supplementary Figs. 17-23, Supplementary Table 3). After the introduction of MOFs, the COF vesicles/nanotubes distributed in the bottom surfaces of membranes change from plump to shriveled, which results from the interfacial COF membrane hindering heat (generated from the MOF growth process) dissipation across the water/oil interface and the heat accumulation leading to local fracture and shriveling of pristine COF vesicles/nanotubes.
Generally, after the successful preparation of x-NH-CuBDC MOFs at the oil-water interface, TFP dissolved in dichloromethane was slowly injected into the underlying oil phase by an injection syringe to allow the -NH of MOFs to fully react with the -CHO of TFP (this movement should be as slow and gentle as possible to avoid interface disturbance). One day later, TAPA dissolved in dichloromethane was slowly injected into the underlying oil phase to initiate the Schiff base reaction. After three days of reaction, x-NH-CuBDC-TAPA-TFP MCOF membranes can be observed at the oil-water interface, with x being the concentration of NH-BDC (Fig. 3a). On the noticeably smoother top surfaces, unpenetrated crater-like structures appear, which may result from the MOF lamellae retarding heat dissipation, leading to the local formation of gas nanobubbles (Fig. 3b, Supplementary Figs. 24, 25). Different from the CMOF membranes, with the hierarchical increase of MOF ligand concentration, the thickness of the MCOF membrane progressively decreases, which is attributed to the inhibition effect of MOFs at the interface on the out-of-plane stacking of COF membrane (Supplementary Fig. 24). Expectedly, MOF lamellae are observed on both the top and bottom surfaces of MCOF membranes and the Cu signals from EDXS confirm a good distribution of MOFs in the MCOF composite membranes, which suggests that the TAPA-TFP COFs can capture and suck the MOF lamellae at the interface to grow integrally (Fig. 3b, Supplementary Figs. 26, 27). Regarding the pore size distribution, the pristine COF membrane exhibits sharp peaks centered at 1.22 and 2.52 nm, while 0.25-NH-CuBDC-TAPA-TFP MCOF composite membrane shows only one peak at 1.22 nm (Fig. 3c). The disappearance of larger pores (2.52 nm) from inevitable topological defects is attributed to the insertion and suturing function of MOFs. Non-significant pore size changes compared to pristine COF membranes imply that the molecular-level linkage between MOFs and COFs in the MCOF membrane is very weak and the post-MOF synthesis strategy is more favorable to the formation of cross-linked heterogeneous frameworks (Figs. 2e and 3c). The MCOF composite membrane exhibits a significantly larger surface area (528.10 m g) than the pristine COF membrane (307.41 m g), due to the MOFs at the interface inhibiting heat loss and the higher temperature increasing the porosity of COF membrane. Interestingly, 0.5-NH-CuBDC-TAPA-TFP MCOF composite membrane exhibits the sharpest XRD peaks (Fig. 3d), and the HRTEM detected the lattice fringes from both MOFs (0.27 nm for the (5 0 -5) plane) and COFs (0.39 nm for the (0 0 1) plane) (Fig. 3e). The actual weight loading of MOFs in the MCOFs is generally higher than that of CMOFs (Supplementary Fig. 28). ATR-FTIR, XPS, UV-Vis-NIR and XANES demonstrate the successful insertion of MOFs but the very weak and even negligible coordination interactions between MOFs and COFs (Supplementary Fig. 14b, Supplementary Figs. 29-32, Supplementary Table 4). The top surface of the Synchronous composite membrane also presents a crater-like structure and definite MOFs are not observed, indicating that the 1D MOFs easily creep into the COF skeleton and disappear unless the MOFs are synthesized beforehand (Supplementary Fig. 33).
The cation transport properties in as-prepared composite membranes are investigated using a concentration-driven configuration (Fig. 4, Supplementary Fig. 34). For the CMOF composite membrane, the ion conductivity of the permeate side increases linearly with time, indicating a concentration-driven diffusion process (Fig. 4a, d, Supplementary Fig. 35). Unprecedented K-selective transport characteristics are displayed by the TAPA-TFP-0.25-NH-CuBDC CMOF composite membrane (COFs: 1 mM TAPA + 1 mM TFP), and the ion permeation rate strictly depends on the hydrated ionic diameter, presenting a sharp size cutoff of 6.62 Å based on the pore diameter of 6.8 Å (Figs. 2e and 4b, Supplementary Table 5). Definitely, when the hydrated ionic diameter (7.16 Å for Na, 7.64 Å for Li, and 8.56 Å for Mg) is higher than 6.8 Å, the ion permeation rate is less than 0.60 mmol m h, which is at least 80-time lower than that of K with a hydrated ionic diameter of 6.62 Å, thus, featuring extremely high K/Na (82.52) and K/Mg (1131.07) ideal selectivities (Fig. 4c), significantly higher than those of previously reported membranes with various channel configurations (Fig. 4g). After a long period of testing, the membrane still maintains excellent integrity, toughness, and K-selective transport property (Supplementary Fig. 36). It is noteworthy that the growth and coordination of the 1D MOFs in the 1D channels of 2D COF membranes is intricate and dependent not only on the MOF ligand concentration but also on the thickness of the pristine membrane, resulting in significant variations in ion selectivity between different CMOF membranes (Fig. 4a-f). The ion permeation rate of the TAPA-TFP-5-NH-CuBDC CMOF composite membrane (COFs: 1 mM TAPA + 1 mM TFP) is higher than that of the pristine COF membrane, suggesting that the high ligand concentration may cause defects in the composite membrane, leading to nonselective ion transports. When selecting the thinner pristine COF membrane, the ion selectivity of the as-prepared CMOF membrane decreases to a certain extent, which may be attributed to the shorter ion-channel wall interaction path (Fig. 4e, f). To evaluate the effect of driving force for ion diffusion, different concentration gradients including 0.02, 0.1, and 0.2 M are employed for K and Na transport. The K permeation rate of TAPA-TFP-0.25-NH-CuBDC CMOF composite membrane (COFs: 1 mM TAPA + 1 mM TFP) presents a proportional rise to the concentration gradient, but K/Na selectivity may show a nearly 10-fold decrease if the concentration gradient is low (Fig. 4h). The CMOF composite membrane is also employed for separating the practical brine system from Sichuan deep ground (pH = 6.74, K: 41617 ppm, Na: 275339 ppm, Li: 1553 ppm, Mg: 9853 ppm, and other unknown components in unknown concentrations), and after 60 h of diffusion, the K mass content increases by ∼2 times, which demonstrates the preferential transport of K than Na (Supplementary Fig. 37). The not entirely satisfactory actual brine separation performance is attributed to the complex brine composition and intricate transport process due to numerous unknown factors affecting diffusion. Compared with CMOF composite membranes, MCOF composite membranes have larger pore size, surface area and MOF loadings, leading to higher K permeation rate but poor ion-sieving performance (Fig. 4i, Supplementary Fig. 38). In binary-ion systems, the ion selectivity decreases due to the competition for occupying effective mass transport channels between cations (Supplementary Fig. 39). Thin-film composite (TFC) membranes were fabricated on anodic alumina oxide (AAO) substrates via in-situ confinement and growth (Supplementary Fig. 40) to investigate the practical application prospects. Due to the different nano-confined growth environment, the TFC membranes exhibit different properties from self-standing membranes, but still have K-recognition channels (Supplementary Figs. 41-44).
The CMOF membrane exhibits two different-sized pores (Fig. 2e). The 0.68-nanometer pores are dedicated to effective ion screening, while the 1.2-nanometer pores facilitate rapid ion migration (Fig. 5a). Pore-entrance size-sieving effect plays a significant role in the ultra-selective K transport. The hydration energies of Na (-365 kJ mol), Li (-475 kJ mol), and Mg (-1830 kJ mol) are much larger than that of K (-295 kJ mol), resulting in a higher transport energy barrier to entry into the narrow hetero-channels, and therefore come with exponentially lower permeation rates. In addition to the pore-entrance sieving effect, in-pore transport dynamics also play vital roles in ion sieving. In this stage, specific interactions (hydrogen bonding interactions, ion-binding site interactions) are often considered as the energy barriers for ion transport. The 1D MOFs have a large number of free -COOH groups with lower affinity for K (Fig. 5b, c), which is conducive to the fast transportation of K in the heterogeneous channels. Molecular dynamics (MD) simulations also confirm the selective transport of K ions over other metal ions (especially Na ions) in the CMOF channels, which is supported by the snapshots showing the transmembrane migration behavior of Na (yellow) and K ions (pink) through the CMOF membranes (Fig. 5d), and the ion transmembrane energy barrier (Fig. 5e). As displayed in Fig. 5e, Potentials of mean force (PMF) profiles exhibit a gradual increase in energy as all ions approach the membrane pores, and K ions face the lowest energy barriers during transmembrane migration, thereby facilitating K ion transport and rapid transmembrane (The transmembrane energy barrier for Mg is the highest (34.51 kcal mol), which is much greater than that of Li (18.73 kcal mol), Na (15.56 kcal mol), and K (8.12 kcal mol)). Furthermore, during the same period, the number of K ions passing through the CMOF membrane is significantly higher than that of Na ions, indicating superior K/Na ideal selectivity (Fig. 5f), which is consistent with experimental results (confirming the efficacy of CMOF membranes in ion separation and lower actual selectivity). These findings suggest that the excellent permselectivity of the CMOF membrane is attributed to the synergistic effect of pore-entrance size-sieving and in-pore transport dynamics.
Additional MCOF and CMOF composite membranes have also been prepared according to the synthetic procedure. The pristine DABA-TFP COF membrane features a thickness of 70.21 nm. After the introduction of MOFs, its thickness increases to 116.33 nm (Supplementary Fig. 45). Although no explicit MOFs are observed, EDXS confirms a uniform distribution of MOFs in the CMOF composite membrane (Supplementary Fig. 45). Conversely, the pristine TAPA-DHA COF membrane features a thickness of 102.56 nm, but the NH-CuBDC-TAPA-DHA MCOF composite membrane presents a thickness of 61.54 nm (Supplementary Fig. 46). MOF lamellae are mainly present on the top surface of the MCOF composite membrane, and EDXS confirms a uniform distribution of MOFs, but exhibiting no obvious ion selectivity (Supplementary Figs. 46, 47).