Skip to main content
NCBI home page
Chemical Science logoLink to Chemical Science
. 2026 Jul 13;17(34):16301–16310. doi: 10.1039/d6sc03370a

Scalable spray-coating of covalent organic framework membranes via microdroplet confined assembly for H2/CO2 separation

Feng Zhang a, Zihao Wang a, Zhiwen Fan a, Jiansong Liu a, Chunyue Pan a, Shuai Gu a, Juntao Tang a, Pengpeng Shao a, Jiayin Yuan b, Guipeng Yu a,✉
PMCID: PMC13360833  PMID: 42445522

Abstract

The scale-up and practical implementation of covalent organic frameworks (COFs) in gas separation remains constrained by intrinsic limitations, including insolubility, infusibility and mismatched pore sizes (>0.8 nm). Drawing inspiration from air-spray principles, herein we develop a facile bottom-up spray-coating strategy for the fabrication of lasagna-like COF membranes for the first time. Theoretical calculations and detailed experiments revealed the core role of spray-coating-induced microdroplet confined assembly, wherein atomized droplets generate spatially confined environments that accelerate molecular ordering and polycondensation. Within these microdroplet-confined spaces, oligo(ethylene glycol) (OEG) side-chains suppress molecular diffusion to enhance crystallization. Such synergy enables dynamic spray cycles to drive COF growth into continuous membranes within 30 minutes. The resulting COF-OEG-4 membrane exhibits outstanding gas separation performance, with an ultrahigh H2 permeance of 3650 GPU and an H2/CO2 selectivity of 41.6, ranking it among the highest-performing H2-selective membranes. Moreover, the strategy demonstrates excellent scalability, enabling the fabrication of large-area COF composite membranes exceeding 1750 cm2 directly on commercial substrates. This work critically advances both the COF processing methods and gas separation technologies.

.A bottom-up spray-coating strategy via microdroplet confined assembly enables rapid fabrication of large-area COF membranes for H2/CO2 separation.graphic file with name d6sc03370a-ga.webp

Introduction

Covalent Organic Frameworks (COFs) represent a class of crystalline porous solids constructed from light elements through strong covalent bonds, featuring highly ordered open pore structure, tunable pore functionality, and excellent chemical and thermal stability.1–6 These characteristics enable COFs to overcome the trade-off between permeability and selectivity inherent in traditional polymer membranes, making them ideal candidates for next-generation gas separation membranes.7–11 However, despite substantial research demonstrating the feasibility of COF membranes in gas separation,12–16 their practical application still faces significant challenges. Unlike conventional polymers, COFs generally exist as insoluble and infusible crystalline powders, which hinders their processing into continuous, defect-free films via general solution-casting or extrusion techniques.17–19 Therefore, developing new membrane fabrication methods is crucial for overcoming these challenges and advancing the practical application of COF materials.

To address the above challenges, a variety of fabrication strategies, including interfacial polymerization (IP),20–23 layer-by-layer assembly (LBL),24–28 and in situ growth (ISG),29–31 have been recently developed for COF membranes enabling important advances in crystallinity and structural controllability. For example, Banerjee disclosed the preparation of highly crystalline COF membranes via water/dichloromethane interfacial polymerization,20,21 whereas Jiang achieved the controlled fabrication of oriented COF membranes through liquid-phase epitaxial LBL assembly.27,28 Nevertheless, most methods still rely on static interfaces or prolonged crystallization steps, often involving tedious procedures and demanding substrate pretreatments, which severely limit their efficiency and controllability for large-area, continuous fabrication.32 Furthermore, achieving high-performance gas separation with COF membranes typically requires precise control over pore size and nanochannel environments, common strategies to this end—such as constructing COF/COF33,34 or COF/metal organic framework (MOF) heterostructures,35,36 pore functionalization,37,38 or developing COF-polymer composites39—often introduce complex post-treatment steps. While these additional procedures can enhance membrane performance, they significantly increase process complexity and hinder scalable fabrication. Consequently, the development of a simple, efficient, and scalable COF membrane fabrication method has become an urgent requirement to promote their practical application.

Inspired by air-spray technology, we developed a convenient bottom-up spray-coating strategy for the rapid fabrication of COF membranes (Scheme 1). This method capitalizes on the confined assembly of atomized microdroplets and the diffusion-inhibiting effect of oligo(ethylene glycol) (OEG) side chains, wherein alternating spray cycles effectively drive COF growth. In contrast to conventional COF membrane fabrication methods, this approach exhibits superior processing characteristics, enabling not only the straightforward formation of uniform coatings on diverse substrates, but also facile patterning and the continuous fabrication of membranes. The resulting COF-OEG-4 membrane exhibits H2/CO2 separation performance that far surpasses the 2008 Robeson upper bound, ranking it among the highest-performing H2-selective membranes. Moreover, this strategy demonstrates excellent scalability, enabling the continuous fabrication of uniform, large-area (over 1750 cm2) COF composite membranes on commercial substrates. Together, these advances offer a promising technical pathway to address the long-standing processing and integration challenges of COF materials for practical applications in separation, energy conversion, and flexible devices.

Scheme 1. Schematic of the bottom-up spray-coating strategy for fabrication of COF-OEG membranes.

Scheme 1

Open in a new tab

Results and discussion

Bottom-up spray-coating strategy for fabrication of COF membranes

The low processability of COFs powder at room temperature hinder their use in top-down coating processes. To tackle this issue, we developed a simple spray-coating processing strategy utilizing the confined assembly of atomized microdroplets, enabling bottom-up preparation of COF membranes from molecular building blocks (Fig. 1a). As shown in Fig. 1b, 1,3,5-triformylbenzene (TB) and terephthalic dihydrazide monomers bearing oligo (ethylene glycol) methyl ether side chains of varying lengths (TPH-OEG-X, X = 2, 3, 4 ethylene oxide units) were first dissolved in acetonitrile (ACN) and water, respectively. Subsequently, an air-spraying device alternately sprayed the solution as atomized droplets, inducing confined assembly of TPH-OEG and TB monomers within mixed microdroplets to accelerate COF nucleation and growth.40–42 OEG side chains grafted to hydrazide monomers enhanced their reactivity and water solubility (Fig. S4). Concurrently, hydrogen bonding between OEG side chains and solvent molecules suppressed diffusion of hydrazide units in mixed microdroplets (from 0.282 to 0.034 Å2 ps; Fig. S5 and Table S1), promoting the reversible formation of hydrazone bonds and dynamic structural reorganization. This synergy enabled the ultrafast formation of COF on the substrate (Video 1), thereby making the subsequent spray-coating of continuous COF coatings feasible.

Fig. 1. Bottom-up COF coating fabrication. (a) Schematic diagram of the spray formation process of COF-OEG membranes. (b) Schematic illustration of the spray-coating apparatus and process. (c) Representation of COF spray-coating on diverse substrates. (d) Customized patterning via spatially controlled spray-coating. (e) COF composite film attached to non-woven fabric and PAN. (f) Optical photograph of a COF membrane supported on an AAO substrate.

Fig. 1

Open in a new tab

The proposed bottom-up spray-coating strategy demonstrates excellent substrate versatility, yielding robust COF coatings on glass, metallic, and non-metallic surfaces (Fig. 1c, S6 and S7). Beyond material compatibility, this approach also accommodates complex geometries, allowing the fabrication of COF membranes with customizable patterns (Fig. 1d). Highly flexible COF composites can be readily produced on non-woven fabric and polyacrylonitrile (PAN) substrates (Fig. 1e, S8 and S9). Moreover, continuous and dense COF membranes are successfully formed on anodic aluminum oxide (AAO) substrates within 30 minutes (Fig. 1f and S10), confirming the integrity of the coating and the broad processability of this method.

As illustrated in Fig. 2a and b, our proposed bottom-up spray-coating model demonstrates the formation process of COF-OEG membranes via microdroplet confined assembly. Initially, the hydrazide monomers preferentially anchor onto the substrate surface due to hydrogen bonding between their OEG side chains and the substrate, creating favorable sites for subsequent COF nucleation (Fig. S11).43 Subsequently, microdroplets carrying TB collide and mix with TPH-OEG microdroplets. Within the spatially confined droplets, diffusion limitation and rapid solvent evaporation synergistically drive the confined assembly of monomers, inducing rapid nucleation of COF oligomers (Fig. 2c). As spraying continues, the continuous supply of fresh monomers promotes the extension and interconnection of COF oligomers through epitaxial growth and layer-by-layer stacking, eventually transforming the COF structure from discrete oligomers into a continuous and dense membrane (Fig. 2d and S12). In situ Fourier-transform infrared (FT-IR) spectroscopy shows that during the spraying cycles, the intensity of the characteristic imine (C Created by potrace 1.16, written by Peter Selinger 2001-2019 N) stretching vibration band at 1620 cm−1 increases continuously, directly confirming the gradual formation of the covalent network (Fig. 2e and S13). Powder X-ray diffraction (PXRD) patterns reveal that the appearance of the (100) diffraction peak and the decrease in its full width at half maximum indicate enhanced crystallinity and long-range order during the spraying process (Fig. 2f and S14). Furthermore, white light interferometry (WLI) over an area of 870 µm × 670 µm reveals a surface roughness of Ra = 788 nm for COF-OEG membrane (Fig. 2g and S15), the pronounce micron-scale topological undulation suggests the possible presence of a rich porous infrastructure beneath the dense surface layer.44

Fig. 2. Dynamic evolution mechanism in spray-coating process of COF-OEG membranes. (a) The bottom-up spray-coating theoretical model. (b) Schematic of COF formation by microdroplet confined assembly. (c) Optical microscopy images of the formation process of monomer microdroplet confined assembly (400x microscope). From left to right: hydrazide microdroplets, mixed microdroplet confined assembly, and COF oligomers, respectively. (d) SEM image of the morphological evolution process of COF membranes. (e) In situ FT-IR spectra collected during COF-OEG membranes growth at different spraying cycles. (f) PXRD patterns of the COF-OEG membranes prepared with different spraying cycles. (g) 3D profile (830 µm × 630 µm) WLI spectrum of COF-OEG membranes. (h) Cross-sectional SEM image of the COF-OEG membranes prepared by the present strategy and a schematic illustration of its lasagna-like structure. (i) SEM image of COF-OEG grown in large droplets. (j) SEM images of COF-OEG membranes with two-phase liquid surface polymerization.

Fig. 2

Open in a new tab

This inference was directly confirmed by cross-sectional scanning electron microscopy (SEM) imaging, which clearly reveals a unique lasagna-like architecture: an internally connected porous network capped by a fully dense surface layer (Fig. 2h). This morphology is closely tied to the intrinsic features of the spray-coating process. During cyclic spray coating, the newly deposited microdroplets can hardly penetrate into the underlying layer and merge completely with the preformed COF membrane due to the rapid evaporation of ACN. As a result, monomers and oligomers rapidly concentrate, nucleate, and crosslink predominantly in the surface or near-surface region. Consequently, each spraying cycle tends to generate a new COF layer atop the previous one rather than fully fusing into a uniformly dense membrane. After multiple cycles, interfacial traces and partially interconnected voids are retained between the stacked layers, ultimately giving rise to the distinctive lasagna-like structure (Fig. S16). In contrast, conventional methods (e.g., in situ growth with large droplets or interfacial polymerization45 at two-phase liquids) yield structurally monotonous COFs (e.g., large particles or densely packed membranes) due to the absence of such processes (Fig. 2i, j and S17). Moreover, the strategy exhibits excellent in situ reparability, where additional spray cycles can remedy pre-existing microdefects on the substrate, restoring surface homogeneity without compromising solvent resistance post-repair (Fig. S18).

Structural characterization of COF-OEG membranes

Three COF-OEG membranes (COF-OEG-2, COF-OEG-3 and COF-OEG-4) were then prepared via the bottom-up spray-coating strategy, employing TB with different OEG functionalized hydrazide monomers (TPH-OEG-2, TPH-OEG-3 and TPH-OEG-4, respectively) (Fig. 3a–c and S19). FT-IR spectroscopy confirmed the formation of hydrazone linkages, with characteristic absorption bands observed at 1675 (C Created by potrace 1.16, written by Peter Selinger 2001-2019 O) and 1610 cm−1 (C Created by potrace 1.16, written by Peter Selinger 2001-2019 N) across all COF-OEG variants (Fig. S20).38,46 Further elemental analysis by X-ray photoelectron spectroscopy (XPS) verified the presence of C, N, and O elements, with N 1s XPS spectra showing characteristic peaks at 399.5 (C–N) and 400.4 eV (C Created by potrace 1.16, written by Peter Selinger 2001-2019 N), and O 1s XPS spectra displaying signals at 530.8 (C Created by potrace 1.16, written by Peter Selinger 2001-2019 O) and 532.7 eV (C–O) (Fig. S21–S23).47 SEM images revealed that all three membranes possessed dense and textured surfaces, indicative of robust film formation, whereas their cross-sectional views display a lasagna-like structure with an interconnected porous network (Fig. S24). Atomic force microscopy (AFM) measurements revealed a surface roughness of Ra = 60–100 nm for the COF-OEG membrane, further confirming the homogeneity of the membranes (Fig. S25).

Fig. 3. Characterization of the structure and composition of COF-OEG membranes. Schematic structure of (a) COF-OEG-2, (b) COF-OEG-3 and (c) COF-OEG-4 in antiparallel stacking. GIWAXS patterns of (d) COF-OEG-2, (e) COF-OEG-3, and (f) COF-OEG-4. PXRD patterns of (g) COF-OEG-2, (h) COF-OEG-3 and (i) COF-OEG-4. HR-TEM images of (j) COF-OEG-2, (k) COF-OEG-3, and (l) COF-OEG-4.

Fig. 3

Open in a new tab

The crystalline properties of the COF-OEG membranes were characterized using Grazing-incidence wide-angle X-ray scattering (GIWAXS) and PXRD. The GIWAXS patterns revealed characteristic diffraction signals (Fig. 3d–f). Correspondingly, the in-plane diffraction profiles exhibited peaks at qxy = 0.27, 0.53, and 1.72 Å−1, which are assignable to the (100), (200), and (001) crystallographic planes, respectively (Fig. S26). Meanwhile, the PXRD peaks showed a corresponding trend and were consistent with the antiparallel AA stacking simulation results (Fig. 3g–i and S27–S29). Pawley refinement further corroborates the structural assignment, with excellent agreement between experimental and calculated PXRD data and minimal residuals (COF-OEG-2: Rp, 1.34%, Rwp, 1.69%; COF-OEG-3: Rp, 1.48%, Rwp, 1.92%; COF-OEG-4: Rp, 3.74%, Rwp, 4.79%). High-resolution transmission electron microscopy (HR-TEM) images further support this finding, revealing lattice fringes with a spacing of 0.37–0.38 nm, consistent with the interlayer spacing expected for this stacking geometry (Fig. 3j–l).48,49 Collectively, these results unequivocally demonstrated the successful formation of a crystalline COF framework with the expected topology.

The porosity characteristics of the three COF-OEG membranes were evaluated by gas adsorption measurements. Brunauer–Emmett–Teller (BET) surface areas derived from CO2 adsorption–desorption isotherms at 195 K were determined to be 262.2, 243.2, and 102.1 m2 g−1 for COF-OEG-2, COF-OEG-3, and COF-OEG-4, respectively (Fig. S30a). Pore size distributions calculated from the same isotherms using nonlocal density functional theory (NLDFT) reveal a systematic decrease in pore size with increasing OEG side-chain length, shifting from the microporous regime for COF-OEG-2 (0.75 nm) to the ultra-microporous regime for COF-OEG-4 (0.54 nm) (Fig. S30b). Additionally, all COF-OEG membranes exhibit excellent thermal stability, retaining structural integrity up to 320 °C without significant mass loss (Fig. S31).

Gas separation performance of COF-OEG membranes

To evaluate their separation performance, the COF-OEG membranes were tested for single-component and mixed-gas permeation using the Wicke–Kallenbach method (Fig. S32). In all single-gas measurements, H2 consistently exhibited the highest permeance across the COF-OEG series, with values of 5450 GPU for COF-OEG-2, 4635 GPU for COF-OEG-3, and 3924 GPU for COF-OEG-4 (Fig. 4a–c). The gradual decrease in H2 permeance with increasing OEG side-chain length is attributed to progressive occupation of the pore space, which increases transport resistance. Nevertheless, owing to the well-developed lasagna-like porous architecture that facilitates molecular transport, all COF-OEG membranes still exhibit substantially higher H2 permeance than most reported COF, MOF, and polymer membranes. For all membranes, the permeance followed the order H2 > N2 > CH4 > CO2, which deviates from the trend expected solely based on kinetic diameters (H2: 0.289 nm, CO2: 0.33 nm, N2: 0.36 nm, CH4: 0.38 nm). Notably, despite its smaller kinetic diameter, CO2 exhibits lower permeance than N2 and CH4 due to stronger interactions with the ether-rich pore environment, which restrict its effective diffusivity.50 Density functional theory (DFT) calculations further show that the CO2 binding energy becomes more negative with increasing OEG chain length, indicating progressively stronger CO2 affinity (Fig. S33–S36 and Table S2). Further calculations revealed that the ideal selectivity of COF-OEG-2, COF-OEG-3, and COF-OEG-4 for H2/CO2 were 8.5, 22.1, and 33.5, respectively, all exceeding the Knudsen selectivity (Table S3).

Fig. 4. Gas separation performance of COF-OEG membranes. Single gas (H2, CO2, N2, CH4) permeance and ideal H2/gas selectivity of (a) COF-OEG-2, (b) COF-OEG-3 and (c) COF-OEG-4. (d) Separation performance of equimolar H2/CO2 binary gas mixtures with COF-OEG membranes. (e) Simulation system with snapshots at 0 ps and 1000 ps for the permeation of an equimolar H2/CO2 mixture (100 H2 and 100 CO2 molecules in the feed chamber) through COF-OEG-4 membrane. (f) Calculated mean square displacement (MSD) values simulation time for H2 and CO2. (g) Long-term operational stability of gas separation with COF-OEG-4 membrane. (h) Comparison of H2/CO2 separation performance with the upper bound and other literature reports.

Fig. 4

Open in a new tab

The separation performance of three membranes was evaluated using equimolar H2/CO2 binary gas mixtures. As shown in Fig. 4d, the H2/CO2 selectivity of COF-OEG membranes increased significantly from 13.2 for COF-OEG-2 to 41.6 for COF-OEG-4 (Table S4). This enhancement correlates directly with the extension of the OEG side chains. As the side chains lengthen, the pore size progressively narrows to 0.54 nm for COF-OEG-4, increasing steric hindrance and further restricting CO2 transport. 77 K N2 adsorption measurements confirm this, showing that N2 can barely enter the pores in the low-pressure region (P/P0 < 0.1) and indicating that the steric hindrance caused by the OEG side chains has substantially reduced pore accessibility (Fig. S37).51 Notably, while the elongated OEG side chains introduce more polar adsorption sites, CO2 uptake does not increase but instead exhibits a monotonic decline (Fig. S38). Grand Canonical Monte Carlo (GCMC) simulations reveal the underlying reason at the molecular level, demonstrating that as the side chains lengthen, the number of CO2-accessible adsorption sites decreases, with the average loading dropping from 96.3 to 73.5 (Fig. S39, S40 and Table S5). This indicates that the reduction in effective adsorption space caused by pore narrowing and side-chain occupation outweighs the benefit of additional adsorption sites. Therefore, the elongated OEG side chains simultaneously increase steric hindrance and compress the effective pore transport space, thereby suppressing CO2 transmembrane transport and ultimately driving the continuous enhancement of H2/CO2 separation performance.

To further investigate the distinct transport behaviors of H2 and CO2 under mixed-gas conditions, GCMC competitive adsorption simulations and molecular dynamics (MD) permeation simulations were performed. GCMC results for equimolar H2/CO2 mixtures show that CO2 exhibits both higher adsorption energy and greater average molecular loading than H2 within the COF-OEG pore channels, and preferentially distributes near the ether oxygen atoms of the OEG side chains, whereas H2 is mainly located in the relatively open pore regions (Fig. S41, S42 and Table S6). This suggests that during mixed-gas separation, the strongly adsorbing CO2 tends to be preferentially adsorbed and retained, thereby suppressing its own effective transport.52 In contrast, H2 is less affected due to its smaller kinetic diameter and weaker interaction with the COF-OEG framework. MD permeation simulations directly verify this transport disparity. With increasing OEG side-chain length, the number of permeated CO2 molecules within 1000 ps drops from 2 to 0, and the amount retained within the membrane decreases from 60 to 42, indicating severely restricted transmembrane transport, while the number of permeated H2 molecules decreases only from 58 to 47, still passing through the membrane rapidly (Fig. S43–S46). Clearly, CO2 is readily trapped by adsorption within the confined pore channels and can hardly penetrate through, whereas H2 passes almost unimpeded. Furthermore, in COF-OEG-4, the H2 diffusion coefficient (6.1 × 10−4 cm2 s−1) is an order of magnitude higher than that of CO2 (1.53 × 10−5 cm2 s−1) (Fig. 4e, f and Table S7), yielding a diffusivity selectivity (DH2/DCO2 ≈ 40) that close to the experimental separation selectivity (41.6). Thus, under mixed-gas conditions CO2 transport is subjected to dual suppression—steric hindrance and preferential adsorptive retention—while H2 remains largely unaffected. This competitive adsorption effect amplifies the H2/CO2 transport disparity, leading to an actual separation selectivity that exceeds the ideal selectivity.53

Notably, owing to its well-integrated lasagna-like porous architecture, COF-OEG-4 membrane achieved the highest selectivity (41.6) while maintaining an excellent H2 permeance of 3650 GPU. The membrane also shows excellent operational stability, with no discernible performance decay over 240 h of continuous testing (Fig. 4g). As demonstrated in Fig. 4h, COF-OEG-4 membranes exhibits exceptional H2/CO2 separation performance, far surpassing the 2008 Robeson upper bound. Compared to polymer membranes, mixed matrix membranes (MMMs), MOF membranes, and other COF membranes (Table S8), it achieves a significant advantage in selectivity and permeance, ranking it among the highest-performing H2-selective membranes.

Scale-up processing and application potential

To demonstrate the scalability and practical potential of the bottom-up spray-coating strategy, we successfully extended it to a flexible commercial polysulfone (PSF) substrate. Despite the constraints of laboratory-scale equipment, a complete and continuous COF-OEG-4 separation layer with an area of over 1750 cm2 was constructed on the PSF substrate in only 3 hours, demonstrating the inherently efficient processing potential of this method beyond current practical limitations (Fig. 5a). SEM images collected from different regions showed that even at this large scale, the COF layer remained continuous and compact across the membrane surface, while the corresponding cross-sections revealed a regular lasagna-like structure with a relatively consistent thickness of 2.9 ± 0.1 µm (Fig. 5b, S47 and S48). Gas permeation tests further revealed that the resultant COF-OEG-4/PSF composite membrane delivered excellent and uniform separation performance across different regions, with average values of 3390 GPU and 38.1 for H2 permeance and H2/CO2 selectivity, respectively (Fig. 5c), comparable to the performance achieved on the rigid AAO substrate.

Fig. 5. Fabrication and separation performance characterization of large-area COF-OEG-4/PSF composite membranes. (a) Photograph of the COF-OEG-4/PSF composite membrane with an area of over 1750 cm2. (b) SEM images of different regions on the composite membrane surface. (c) H2/CO2 separation performance measured at different regions of the membrane. (d) Area comparison between the COF-OEG-4/PSF composite membrane and other advanced COF membranes. (e) Key parameter comparison among different fabrication methods for representative COF composite membranes used in H2/CO2 separation, including processing time, membrane area, H2 permeance, H2/CO2 selectivity, and membrane thickness.

Fig. 5

Open in a new tab

Building on the above scale-up demonstrations, this strategy achieves an order-of-magnitude improvement in controllable fabrication area compared with most reported COF membrane fabrication methods (Fig. 5d and Table S9). Furthermore, when benchmarked against current mainstream techniques for H2-selective COF membranes, such as IP, ISG, LBL and knife-coating (KC), our approach demonstrates significant advantages in processing efficiency, scalability, and overall H2/CO2 separation performance (Fig. 5e and Table S10). This work not only proves the substantial potential and extensibility of the bottom-up spray-coating strategy but also provides a crucial technical pathway and valuable insights for advancing the scalable manufacturing and application of COF separation membranes toward practical industrial gas separation.

Conclusions

In summary, we have developed a bottom-up spray-coating strategy that effectively addresses the long-standing challenges in the scalable and controllable fabrication of COF membranes. This spray-based approach relies on the confined environment of atomized microdroplets and the diffusion inhibiting effect of OEG side chains, which synergistically accelerate molecular ordering and enhance crystallization to yield a continuous COF membrane with a lasagna-like structure within 30 min. The resulting COF-OEG-4 membrane exhibits outstanding gas separation performance, with an ultrahigh H2 permeance of 3650 GPU and an H2/CO2 selectivity of 41.6, which positions it among the highest-performing H2-selective membranes. Moreover, the strategy demonstrates remarkable scalability and process robustness, producing uniform, defect-free membrane over an area of 1750 cm2 within 3 h. This work not only provides a practical and scalable platform for COF membrane manufacturing, but also establishes microdroplet confined assembly as a compelling mechanism for the rapid and controllable fabrication of advanced separation membranes.

Experimental parts

Materials and chemicals

All materials, methods, and instruments used in this study are described in detail in Section 1 (Materials and methods) of the SI. The same SI provides the synthetic procedures and structural characterization data of the three COF precursors (TPH-OEG-2, TPH-OEG-3, and TPH-OEG-4) in Section 2 (Synthesis of Monomers), along with the NMR data shown in Fig. S1–S3. A schematic illustration of the synthesis of the three COF-OEG membranes and the detailed preparation methods are presented in Section 3 (Spray-coating preparation of COF membranes) of the SI.

Preparation of COF-OEG-2 membrane

Using COF-OEG-2 as an example, the spray-coating preparation process of the COF membrane is described below. 16 mg of 1,3,5-Triformylbenzene (TB, 0.1 mmol) was dissolved in 4 mL of acetonitrile solution and 63.8 mg of TPH-OEG-2 (0.15 mmol) was dissolved in 4 mL of 3 M acetic acid solution. Using the air spray-coating method (spray rate: 4 mL min−1), the TPH-OEG-2 solution was first sprayed onto the substrate, and then the acetonitrile solution of TB was sprayed, and the COF oligomers were rapidly formed. To achieve a uniform and dense coating, the membrane was cycled back and forth 10 times. The oligomers in the COF-OEG-2 membrane were then removed with water, dichloromethane and ethanol.

Gas separation performance test of COF-OEG membranes

The gas separation performance of the membrane was tested using the Wicke–Kallenbach technique in a homemade gas separation unit. As shown in SI Fig. S32, the working principle of the Wicke–Kallenbach technique and the schematic diagram of the gas separation setup are illustrated. In single-component gas tests (H2, CO2, N2, CH4), the flow rate of the test gas was set to 50 mL min−1. For the two-component gas mixture (H2/CO2 = 50/50, v/v), the total flow rate was maintained at 100 mL min−1. During testing, the flow rates of both single-component gases and gas mixtures were calibrated using a soap film flow meter, while the purge gas (Ar) was maintained at 30 mL min−1. The feed pressure was maintained at 1.1 bar, resulting in a transmembrane pressure difference of 0.1 bar. The temperature was kept constant at 25 °C throughout all experiments. Tests were initiated only after all system parameters stabilized. The effective test area of the membrane is about 2.0 cm2, and the permeance of all membranes needs to be measured at least 5 times under stable conditions to ensure the reliability and reproducibility of the data. The gas permeance (P) is defined by the following eqn (1):

graphic file with name d6sc03370a-t1.webp 1

where Pi is the permeance of component gas (GPU), Ni is gas permeation rate (mol s−1), ΔPi was the transmembrane pressure difference of component i (Pa), and A is the effective membrane area (m2). The ideal gas selectivity is calculated as the ratio of their permeance (i/j) according to eqn (2),

graphic file with name d6sc03370a-t2.webp 2

The gas permeance was calculated as above and the separation factor (α) for H2/CO2 mixture was calculated using the following eqn (3),

graphic file with name d6sc03370a-t3.webp 3

In this equation, yH2 and yCO2 represent the molar concentrations of H2 and CO2 on the permeate side, respectively. Similarly, xH2 and xCO2 denote the corresponding molar concentrations on the feed side. All permeation data presented are averaged from measurements conducted on three individual samples.

Density functional theory (DFT) calculations

During the calculations, the fragments of COF-OEG was used to simplify the calculations. The weak interaction of COF-OEG with H2/CO2 was achieved by the Interaction Region Indicator (IRI). The geometry optimization and frequency calculations for COF-OEG, H2/CO2 and COF-OEG@H2/CO2 performed on Gaussian 16 at the chosen B3LYP/6-311G (d, p). Consequently, the resulting structures were analyzed and imaged using the Multiwfn and VMD software. The energy difference between reactants and products can be calculated from Edifference = Eproducts − Ereactants. Since Eproducts and Ereactants are estimated at the same basis set level, the effect of the basis set superposition error (BSSE) can be excluded. The energy difference during the H2/CO2 adsorption process on COF-OEG was calculated using the above equation. All structures were geometrically optimized by B3LYP/6-311G (d, p), and the energies were also computed using B3LYP/6-311G (d, p).

Grand Canonical Monte Carlo (GCMC) simulations

GCMC simulations were performed using the adsorption module of Materials Studio (MS) to simulate both single-component CO2 adsorption and equimolar H2/CO2 competitive adsorption at constant pressure. Antiparallel stacked 2 × 2 × 2 supercells of COF-OEG-2, COF-OEG-3, and COF-OEG-4 were used as adsorbents. For single-component CO2 adsorption, the CO2 partial pressure was set to 101.3 kPa at 298 K; for mixed-gas competitive adsorption, the total pressure was maintained at 101.3 kPa (partial pressure of each component: 50.65 kPa) at 298 K. The COMPASS II forcefield was employed in the simulations. The van der Waals interactions were calculated by the atom-based method with a fine cutoff distance of 12.5 Å, whereas the Ewald and group method summed the coulombic interactions.

Molecular dynamics (MD) simulations

The MD simulations were carried out using the COMPASS II force field in the Materials Studio software. Three layers of antiparallel stacked COF-OEG were placed in a simulation chamber with dimensions of 63 Å × 63 Å × 109.3 Å. A baffle consisting of a (100) crystalline surface composed of He atoms was added to the left side of the simulation chamber. An equimolar mixture of H2 and CO2 (100 molecules of each component) was added to the left side of the chamber, and vacuum was applied to the right side. Simulations were performed using the NVT ensemble (constant particle number, volume, and temperature), and the system was optimized prior to the diffusion simulation. The initial velocities were randomized, and an Andersen thermostat was used to maintain a constant temperature of 298.0 K throughout the simulation. MD simulations were performed for 1 ns with a time step of 1 fs using the Forcite module. The mean square displacement (MSD) was analyzed as a function of simulation time. The diffusion coefficients for H2 and CO2 were calculated from the slope of a linear fit to the MSD versus time plot (where the diffusion coefficient equals the slope divided by 6).

Author contributions

Guipeng Yu, Chunyue Pan, and Juntao Tang proposed the idea and designed the experiments. Feng Zhang performed the experiments and carried out the structural characterizations. Zihao Wang, Zhiwen Fan, and Jiansong Liu helped to analyse the results of gas separation performance. Shuai Gu helped with the structural characterization analysis. Pengpeng Shao helped to synthesize the materials. Feng Zhang, Zihao Wang, and Zhiwen Fan performed the DFT, GCMC, and MD simulations. Feng Zhang, Jiayin Yuan, and Guipeng Yu wrote the manuscript with contributions from all authors.

Conflicts of interest

There are no conflicts to declare.

Supplementary Material

SC-017-D6SC03370A-s001
SC-017-D6SC03370A-s002
Download video file (23.8MB, mp4)

Acknowledgments

This work was supported by the National Natural Science Foundation of China (No. 22409216 and 22301014), Hunan Provincial Key Research and Development Program (No.2024JK2066, 2025JK2085 and 2025JK2086), the Natural Science Foundation of Hunan Province (2025JJ40010), and the State Key Laboratory of Fine Chemicals, Dalian University of Technology (KF2415). We are grateful to the High-Performance Computing Center of Central South University for assistance with the computations.

Data availability

The authors declare that the remaining data supporting the findings of this study are available within the article and its supplementary information (SI), and the full image data set is available from the corresponding author upon request. Supplementary information is available. See DOI: https://doi.org/10.1039/d6sc03370a.

Notes and references

  1. Cote A. P. Benin A. I. Ockwig N. W. O'Keeffe M. Matzger A. J. Yaghi O. M. Science. 2005;310:1166–1170. doi: 10.1126/science.1120411. [DOI] [PubMed] [Google Scholar]
  2. Huang N. Wang P. Jiang D. Nat. Rev. Mater. 2016;1:16068. doi: 10.1038/natrevmats.2016.68. [DOI] [Google Scholar]
  3. Ge S. Wei K. Peng W. Huang R. Akinlabi E. Xia H. Shahzad M. W. Zhang X. Xu B. B. Jiang J. Chem. Soc. Rev. 2024;53:11259–11302. doi: 10.1039/D4CS00521J. [DOI] [PubMed] [Google Scholar]
  4. Khalil I. E. Das P. Thomas A. Acc. Chem. Res. 2024;57:3138–3150. doi: 10.1021/acs.accounts.4c00491. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Mahato A. K. Paul S. Banerjee R. Chem. Soc. Rev. 2025;54:3578–3598. doi: 10.1039/D4CS01222D. [DOI] [PubMed] [Google Scholar]
  6. Song Y. Wang Q. Li E. Wang T. Wang W. Li J. Zhang F.-Y. Li B. Jiang D. Wang Y. Tong X. Yu X. Mahurin S. M. Yang Z. Dai S. Nat. Commun. 2025;16:5712. doi: 10.1038/s41467-025-61325-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Robeson L. M. J. Membr. Sci. 2008;320:390–400. doi: 10.1016/j.memsci.2008.04.030. [DOI] [Google Scholar]
  8. Chen B. Xie H. Shen L. Xu Y. Zhang M. Zhou M. Li B. Li R. Lin H. Small. 2023;19:2207313. doi: 10.1002/smll.202207313. [DOI] [PubMed] [Google Scholar]
  9. Asif M. B. Kim S. Nguyen T. S. Mahmood J. Yavuz C. T. J. Am. Chem. Soc. 2024;146:3567–3584. doi: 10.1021/jacs.3c10832. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Jin Y. Li M. Yang Y. Adv. Sci. 2024:2412600. [Google Scholar]
  11. Guo S. He J. Han H. Han R. Shou J. Ai C. Tang R. Gu S. Tang J. Yu G. Chem. Commun. 2025;61:8212–8215. doi: 10.1039/D5CC01328C. [DOI] [PubMed] [Google Scholar]
  12. Wang P. Peng Y. Zhu C. Yao R. Song H. Kun L. Yang W. Angew. Chem., Int. Ed. 2021;60:19047–19052. doi: 10.1002/anie.202106346. [DOI] [PubMed] [Google Scholar]
  13. Wang S. Yang Y. Liang X. Ren Y. Ma H. Zhu Z. Wang J. Zeng S. Song S. Wang X. Han Y. He G. Jiang Z. Adv. Funct. Mater. 2023;33:2300386. doi: 10.1002/adfm.202300386. [DOI] [Google Scholar]
  14. Fan H. Wang H. Peng M. Meng H. Mundstock A. Knebel A. Caro J. ACS Nano. 2023;17:7584–7594. doi: 10.1021/acsnano.2c12774. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Zhang H. Shao T. Cheng Z. Dong J. Wang Z. Jiang H. Zhao X. Xiaoteng Liu T. Zhu G. Zou X. Angew. Chem., Int. Ed. 2024;63:e202411724. doi: 10.1002/anie.202411724. [DOI] [PubMed] [Google Scholar]
  16. Deng L. Zhu S. Zou Q. Xie Q. Song G. Pan C. Wei B. Huang Z. Liu T. Tang J. Yuan J. Yu G. Angew. Chem., Int. Ed. 2024;63:e202408453. doi: 10.1002/anie.202408453. [DOI] [PubMed] [Google Scholar]
  17. Li C. Yang J. Pachfule P. Li S. Ye M.-Y. Schmidt J. Thomas A. Nat. Commun. 2020;11:4712. doi: 10.1038/s41467-020-18427-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Sasmal H. S. Kumar Mahato A. Majumder P. Banerjee R. J. Am. Chem. Soc. 2022;144:11482–11498. doi: 10.1021/jacs.2c02301. [DOI] [PubMed] [Google Scholar]
  19. Wang K. Qiao X. Ren H. Chen Y. Zhang Z. J. Am. Chem. Soc. 2025;147:8063–8082. doi: 10.1021/jacs.4c16485. [DOI] [PubMed] [Google Scholar]
  20. Dey K. Pal M. Rout K. C. Kunjattu H S. Das A. Mukherjee R. Kharul U. K. Banerjee R. J. Am. Chem. Soc. 2017;139:13083–13091. doi: 10.1021/jacs.7b06640. [DOI] [PubMed] [Google Scholar]
  21. Sasmal H. S. Halder A. Kunjattu H S. Dey K. Nadol A. Ajithkumar T. G. Ravindra Bedadur P. Banerjee R. J. Am. Chem. Soc. 2019;141:20371–20379. doi: 10.1021/jacs.9b10788. [DOI] [PubMed] [Google Scholar]
  22. Yang R. Liu S. Sun Q. Liao Q. Xi K. Su B. J. Am. Chem. Soc. 2022;144:11778–11787. doi: 10.1021/jacs.2c03864. [DOI] [PubMed] [Google Scholar]
  23. Fang L. Xu H. Qiu S. Ye T. Wang T. Shang J. Gu C. Kitagawa S. Li L. Angew. Chem., Int. Ed. 2025;64:e202423220. doi: 10.1002/anie.202423220. [DOI] [PubMed] [Google Scholar]
  24. Ying Y. Tong M. Ning S. Ravi S. K. Peh S. B. Tan S. C. Pennycook S. J. Zhao D. J. Am. Chem. Soc. 2020;142:4472–4480. doi: 10.1021/jacs.9b13825. [DOI] [PubMed] [Google Scholar]
  25. Ying Y. Peh S. B. Yang H. Yang Z. Zhao D. Adv. Mater. 2022;34:2104946. doi: 10.1002/adma.202104946. [DOI] [PubMed] [Google Scholar]
  26. Guo Z. Wu H. Chen Y. Zhu S. Jiang H. Song S. Ren Y. Wang Y. Liang X. He G. Li Y. Jiang Z. Angew. Chem., Int. Ed. 2022;61:e202210466. doi: 10.1002/anie.202210466. [DOI] [PubMed] [Google Scholar]
  27. Wang M. Zhang P. Liang X. Zhao J. Liu Y. Cao Y. Wang H. Chen Y. Zhang Z. Pan F. Zhang Z. Jiang Z. Nat. Sustain. 2022;5:518–526. doi: 10.1038/s41893-022-00870-3. [DOI] [Google Scholar]
  28. Zhao J. Wang Y. Zhu Z. Sun Q. Lyu B. Bie M. Hou J. Zhang S. Pan F. Jiang Z. Nat. Water. 2025;3:1459–1467. doi: 10.1038/s44221-025-00538-0. [DOI] [Google Scholar]
  29. Fan H. Mundstock A. Feldhoff A. Knebel A. Gu J. Meng H. Caro J. J. Am. Chem. Soc. 2018;140:10094–10098. doi: 10.1021/jacs.8b05136. [DOI] [PubMed] [Google Scholar]
  30. Fan H. Peng M. Strauss I. Mundstock A. Meng H. Caro J. J. Am. Chem. Soc. 2020;142:6872–6877. doi: 10.1021/jacs.0c00927. [DOI] [PubMed] [Google Scholar]
  31. Liu M. Xu Y. Liu Y. Shang S. Gao W. Wang X. Hong J. Xu H. Hua C. You Z. Zhou Z. Guo S. Liu Y. Chen J. Angew. Chem., Int. Ed. 2025;64:e202505491. doi: 10.1002/anie.202505491. [DOI] [PubMed] [Google Scholar]
  32. Zhang Y. Wang H. Wang W. Zhou Z. Huang J. Yang F. Bai Y. Sun P. Ma J. Peng L. E. Tang C. Y. Shao L. Matter. 2024;7:1406–1439. doi: 10.1016/j.matt.2024.01.028. [DOI] [Google Scholar]
  33. Fan H. Mundstock A. Feldhoff A. Knebel A. Gu J. Meng H. Caro J. J. Am. Chem. Soc. 2018;140:10094–10098. doi: 10.1021/jacs.8b05136. [DOI] [PubMed] [Google Scholar]
  34. Feng T. Shi X. Ju T. Long J. Liu Q. Liu M. Wang Y. J. Membr. Sci. 2024;709:123144. doi: 10.1016/j.memsci.2024.123144. [DOI] [Google Scholar]
  35. Fan H. Peng M. Strauss I. Mundstock A. Meng H. Caro J. Nat. Commun. 2021;12:38. doi: 10.1038/s41467-020-20298-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Pu Y. Zhao M. Liang X. Wang S. Wang H. Zhu Z. Ren Y. Zhang Z. He G. Zhao D. Jiang Z. Angew. Chem., Int. Ed. 2023;62:e202302355. doi: 10.1002/anie.202302355. [DOI] [PubMed] [Google Scholar]
  37. Jing X. Zhang M. Mu Z. Shao P. Zhu Y. Li J. Wang B. Feng X. J. Am. Chem. Soc. 2023;145:21077–21085. doi: 10.1021/jacs.3c07393. [DOI] [PubMed] [Google Scholar]
  38. Ren L. Chen J. Han J. Liang J. Wu H. Chem. Eng. J. 2023;462:142112. doi: 10.1016/j.cej.2023.142112. [DOI] [Google Scholar]
  39. Fan H. Wang H. Peng M. Meng H. Mundstock A. Knebel A. Caro J. ACS Nano. 2023;17:7584–7594. doi: 10.1021/acsnano.2c12774. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Zhang Y. Wang Y. Zhang X. Wu L. Wang H. Wei X. Wu W. D. Wang X. Li W. Wu Z. Nanoscale. 2021;13:13764–13775. doi: 10.1039/D1NR03535E. [DOI] [PubMed] [Google Scholar]
  41. Lin H. Zhou L. Mei F. Dou W. Hu L. Yang H. Xu L. Angew. Chem., Int. Ed. 2023;62:e202301900. doi: 10.1002/anie.202301900. [DOI] [PubMed] [Google Scholar]
  42. Chen H. Li X. Li B. Chen Y. Ouyang H. Li Y. Zhang X. J. Am. Chem. Soc. 2025;147:11399–11406. doi: 10.1021/jacs.5c01072. [DOI] [PubMed] [Google Scholar]
  43. Liu Q. Liu M. Zhang Z. Yin C. Long J. Wei M. Wang Y. Nat. Commun. 2024;15:9221. doi: 10.1038/s41467-024-53625-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Wyart Y. Georges G. Deumié C. Amra C. Moulin P. J. Membr. Sci. 2008;315:82–92. doi: 10.1016/j.memsci.2008.02.010. [DOI] [Google Scholar]
  45. Meng Q.-W. Zhu X. Xian W. Wang S. Zhang Z. Zheng L. Dai Z. Yin H. Ma S. Sun Q. Proc. Natl. Acad. Sci. U. S. A. 2024;121:e2316716121. doi: 10.1073/pnas.2316716121. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Senarathna M. C. Li H. Perera S. D. Torres-Correas J. Diwakara S. D. Boardman S. R. Al-Kharji N. M. Liu Y. Smaldone R. A. Angew. Chem., Int. Ed. 2023;62:e202312617. doi: 10.1002/anie.202312617. [DOI] [PubMed] [Google Scholar]
  47. Xie Y. Rong Q. Mao F. Wang S. Wu Y. Liu X. Hao M. Chen Z. Yang H. Waterhouse G. I. N. Ma S. Wang X. Nat. Commun. 2024;15:2671. doi: 10.1038/s41467-024-46942-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Li X. Qiao J. Chee S. W. Xu H.-S. Zhao X. Choi H. S. Yu W. Quek S. Y. Mirsaidov U. Loh K. P. J. Am. Chem. Soc. 2020;142:4932–4943. doi: 10.1021/jacs.0c00553. [DOI] [PubMed] [Google Scholar]
  49. Li X. Xu H.-S. Leng K. Chee S. W. Zhao X. Jain N. Xu H. Qiao J. Gao Q. Park I.-H. Quek S. Y. Mirsaidov U. Loh K. P. Nat. Chem. 2020;12:1115–1122. doi: 10.1038/s41557-020-00562-5. [DOI] [PubMed] [Google Scholar]
  50. Jiang X. He S. Li S. Bai Y. Shao L. J. Mater. Chem. A. 2019;7:16704–16711. doi: 10.1039/C9TA03416A. [DOI] [Google Scholar]
  51. Shao P. Li J. Chen F. Ma L. Li Q. Zhang M. Zhou J. Yin A. Feng X. Wang B. Angew. Chem., Int. Ed. 2018;57:16501–16505. doi: 10.1002/anie.201811250. [DOI] [PubMed] [Google Scholar]
  52. Tian X. Huan H. Zhang K. Zhang R. Liu L. Liu X. Zhang X. Yu Y. Gu T. Wang S. Jiang Z. Adv. Mater. 2025;37:2504622. doi: 10.1002/adma.202504622. [DOI] [PubMed] [Google Scholar]
  53. Tian X. Cao L. Zhang K. Zhang R. Li X. Yin C. Wang S. Angew. Chem., Int. Ed. 2024:e202416864. doi: 10.1002/anie.202416864. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

SC-017-D6SC03370A-s001
SC-017-D6SC03370A-s001.pdf (7.4MB, pdf)
SC-017-D6SC03370A-s002
Download video file (23.8MB, mp4)

Data Availability Statement

The authors declare that the remaining data supporting the findings of this study are available within the article and its supplementary information (SI), and the full image data set is available from the corresponding author upon request. Supplementary information is available. See DOI: https://doi.org/10.1039/d6sc03370a.

Articles from Chemical Science are provided here courtesy of Royal Society of Chemistry

RESOURCES