Nanosheet-bridged metal-organic framework membranes for durable hydrocarbon separation under high pressure

Abstract

Metal-organic framework (MOF) membranes have garnered notable interest in molecular separation. However, their high-pressure application is severely hampered by the mechanical fragility stemming from weak interactions at grain boundaries. Inspired by the cuttlebone’s wall-cavity structure, we develop a class of nanosheets-bridged MOF membranes (NB-MOFs). These graphene oxide nanosheets act as a rigid wall, bridging soft MOF grains and dispersing accumulated stresses at grain boundaries under external loadings, thereby preventing structural cracking and enhancing the mechanical robustness of crystal membranes. By fine-tuning the morphology and content of nanosheets, the optimized NB–ZIF-8 membranes are endowed with unprecedented 50-bar pressure resistance and superior C₃H₆/C₃H₈ separation performance, with a separation factor >240 maintained above 300 hours at industry-relevant pressure. We also confirm our strategy’s versatility by fabricating pressure-resistant NB–ZIF-67 membrane and commercial polymer-supported NB-MOF membranes. We envision that our strategy will establish a platform for developing durable crystalline membranes and unlock their potential in real-world scenarios.

Bioinspired nanosheets-bridged MOF membranes are endowed with high-pressure resistance and durable separation performance.

INTRODUCTION

Metal-organic framework (MOF), an emerging kind of crystalline materials through coordination bonding between metal nodes and organic linkers, has established a transformative platform for chemical separations owing to their precisely tunable ultramicropores and modular designability (1–5). MOF polycrystalline membranes, consisted of highly interconnected MOF crystals, are considered as a disruptive molecular sieving membrane for gas separation (6), particularly transformative potential in olefin/paraffin separation—a critical yet challenging industrial process (7). Over the past decade, various MOF membranes have been fabricated by in situ crystallization (8–10), interfacial polymerization (11, 12), chemical vapor deposition (7), etc., which notably outperform the traditional polymeric materials. However, MOF membranes that are capable of continuous operation under extreme-pressure conditions remain quite rare, stemming from the inherently weak intercrystallite interactions (primarily van der Waals forces and hydrogen bonds) among intergrown MOF crystals (13, 14). Under external pressure loadings, the mechanical stresses propagate along grain boundaries and accumulate in these interlocked boundaries (15), ultimately triggering catastrophic structural collapse through cracking or plastic deformation in MOF membranes. The olefin and paraffin molecules (0.2-Å difference in kinetic diameter) are sensitive to these structural defects, and even angstrom-scale defects can lead to the complete loss of the sieving selectivity (16, 17).

So far, only a handful of studies have reported to improve the pressure resistance of MOF membranes by focusing on MOF cage rigidification through fortified metal node-linker coordination (18, 19) or filling guest molecules (20–22). Despite the notable advances, the inherent limitations arising from weak grain boundaries and consequent mechanical fragility remain largely unaddressed. Resolving this critical challenge will not only deepen the fundamental understanding of strengthening the structural resistance of MOF crystal membranes but also pave the way for their deployment in practical applications.

Deep-sea organisms usually have unique bone structure, as represented by cuttlebone, which can withstand high hydrostatic pressures exceeding 20 atm. Cuttlebone has a unique “wall-cavity” structure wherein the horizontally arranged walls bridge numerous cavities, forming a pressure-resistant structure (23). Under external loading, cavities experience deformation under compressing. These walls can redistribute and transfer stress, reducing stress concentration, thereby notably enhancing the structural resistance and preventing catastrophic structural collapse (24–26) (Fig. 1A and fig. S1). Hence, we envision that the pressure resistance of MOF membranes can be greatly enhanced if rigid nanosheets as artificial “walls” are incorporated to bridge the MOF grains.

Fig. 1. Structural design of cuttlebone inspired NB-MOF membrane and schematic illustration of membrane structural evolution mechanism under long-term high pressure.

(A) Typical rigid wall-cavity structure of natural cuttlebone with high hydrostatic pressure resistance. (B) Conventional single-phase MOF crystal membrane. (C) NB-MOF membrane. The red dash arrows represent the path of stress conduct; the blue dash arrows represent the path of stress dissipation.

Herein, inspired by this cuttlebone structure, we propose a nanosheets-bridging strategy for fabricating pressure-resistant MOF membranes (denoted as NB-MOF). Specifically, graphene oxide nanosheets with high porosity (PGO), acting as rigid wall, are incorporated into the MOF crystal membranes via an electrochemical coassembly method (Fig. 1C). The grains of ZIF-8 are bridged by PGO nanosheets through fortified coordination interactions between metal ions in the MOF and oxygen-containing functional groups on PGO nanosheets. Such bridging effect not only enhances the stiffness and hardness of MOF membrane but also establishes additional energy dissipation pathways to avoid stress accumulation, thus preventing structural cracking under high pressure. By fine-tuning the morphology and contents of bridging nanosheets, the mechanically fortified NB–ZIF-8 with excellent pressure resistance (up to 50 bar) and outstanding C₃H₆/C₃H₈ separation performance under industry-favorite pressure for 300 hours (7 bar) can be achieved, outperforming the reported MOF membranes. Without nanosheets bridging, the ZIF-8 membrane completely loses its separation performance within 10 hours of continuous high-pressure test. Moreover, the pressure-resistant NB–ZIF-67 membrane can be also fabricated by this strategy. These results indicate that nanosheets-bridging strategy can establish a platform to fabricate durable MOF membranes for the applications under high-pressure separation conditions.

RESULTS

Fabrication of NB-MOF membranes

The NB–ZIF-8 membranes were fabricated through a one-step electrochemical coassembly method in a single chamber (fig. S2). Briefly, the PGO nanosheets-bridging modules were homogeneously dispersed in the ZIF-8 precursor solution and acted as negatively charged templates that facilitated the assembly of Zn²⁺ ions onto their surfaces through Zn-O coordination. This preorganization of high density of metal ions on the two-dimensional surface can reduce the nucleation energy barrier of ZIF-8 and facilitate the rapid assembly of composite membrane (27, 28). Then, these PGO@Zn²⁺ composite nanosheets were driven to the surface of the conducting substrate via electrophoresis. During this process, the nucleation and growth of ZIF-8 crystals and the deposition of PGO nanosheets were simultaneously conducted. Last, the NB–ZIF-8 membranes were fabricated.

We synthesized PGO nanosheets by the oxidative etching method (29, 30). Their morphology and structure remain intact, as confirmed by scanning electron microscope (SEM) and x-ray diffraction (XRD) analysis (figs. S3 to S5). The PGO nanosheets display a high aspect ratio (>1000), with lateral dimensions ranging from 5 to 10 μm and a thickness less than 2 nm (Fig. 2A and fig. S4A). Moreover, the high-resolution atomic force microscope (AFM) images reveal pores ranging from 2 to 20 nm, with a pore density of ~600 pores per μm² (fig. S4, B and C). The increased nanopores are also confirmed through nitrogen (N₂) adsorption and the results of pore size distribution analysis (fig. S5), the total pore volume of PGO nanosheets is increased from 0.04 to 0.29 cm³ g⁻¹, which is expected to promote mass transfer. Fourier transform infrared (FTIR) spectroscopy and x-ray photoelectron spectroscopy (XPS) spectra (figs. S6B and S7) confirm the presence of various oxygen-containing groups, such as epoxy, hydroxyl, and carboxylic groups. The oxygen content of PGO nanosheets increases from 22.72 to 27.06% (table S1), which results in a higher surface charge density and greater hydrophilicity, as evidenced by zeta potential and water contact angle measurements (Fig. 2B and fig. S6, C and D). Quartz crystal microbalance (QCM) is used to further confirm the rapid assembly of Zn²⁺ by PGO nanosheets (Fig. 2C). The QCM sensors were coated with PGO nanosheets, and then the mass loading was recorded after immersion in an aqueous solution containing either Zn²⁺ or 2-methylimidazole (2-mIm) ligands. The result shows that negatively charged PGO nanosheets can adsorb 57.0 μg cm⁻² Zn²⁺ in 50 s. Notably, the Zn²⁺ adsorption capacity exceeds that of other reported porous nanosheets such as ionic covalent organic framework (fig. S8) (31). The rapid adsorption process demonstrates that PGO nanosheets can quickly capture Zn²⁺ ions through strong coordination interactions. This leads to the formation of positively charged PGO@Zn²⁺ composites, which are then driven to the cathode substrate through electrophoresis and participated in the membrane crystallization process.

Fig. 2. Microstructure of PGO nanosheets and NB-MOF membranes.

(A) AFM image of bridging modules: PGO nanosheets (inset: thickness measurement via Nanoscope software). (B) Zeta potential of precursors and products in the MOF-PGO electrochemical coassembly system. (C) PGO nanosheets mass evolution in Zn²⁺ and 2mIm aqueous solutions, respectively, by QCM monitoring. (D) Time-dependent nanosheets-bridging process characterized via SEM; scale bar, 1 μm. (E) Focused ion beam SEM image of NB–ZIF-8-2 membrane cross section. (F) SEM image of ZIF–8-PGO powders collected after the electrochemical coassembly process; the dashed lines indicate the bridged grains. (G) XRD patterns of ZIF-8 and NB–ZIF-8 membranes with different PGO loadings. (H) Chemical compositions of ZIF-8 and NB–ZIF-8 membranes, calculated by the high-solution XPS results. a.u., arbitrary unit.

The time-dependent nanosheets-bridging process of NB-MOF membranes is characterized by SEM and XRD (Fig. 2D and figs. S10 and S11). Compared to ZIF-8 membrane (fig. S12), PGO@Zn²⁺ composite nanosheets are firstly uniformly driven to the anodic aluminum oxide (AAO) substrate surface within the first 10 min, and we note that the nucleation rate of ZIF-8 on the AAO substrate is suppressed. We assume that this phenomenon originates from the enrichment effect of PGO on Zn²⁺, leading to a decrease in the concentration of Zn²⁺ in the bulk solution, which in turn promotes the heterogenous nucleation growth of ZIF-8 along the PGO surface. In the subsequent 20 min, the substrate surface is densely covered with ZIF-8 crystals, forming a continuous crystal membrane, while the deposition of the ZIF-8@PGO composite structure is also observed. After the following 30-min deposition, a continuous NB–ZIF-8 membrane with distinct crystalline features is formed, with no PGO nanosheets observed on the membrane surface, possibly because of the decrease of PGO nanosheets concentration. Regarding the membrane thickness, during the 20 to 40 min, the membrane thickness sharply increases from 120 to 420 nm, which coincides with a dramatic change in peak intensity observed in the XRD spectra (fig. S13). Over the next 20 min, there is little change in membrane thickness due to the self-limiting effect of electrodeposition. The distribution and bridging microstructure of PGO nanosheets in the NB–ZIF-8 membrane are confirmed by focused ion beam SEM (FIB-SEM) cross-sectional analysis (Fig. 2E and fig. S14). Furthermore, quantitative analysis via XPS depth profiling determines that the relative content of PGO increases from 1.07 to 4.21 mol % from the top to the bottom of the membrane (fig. S15 and table S2). This combined evidence indicates that the PGO nanosheets are aligned parallel to the substrate, forming a uniform in-plane distribution but a gradient through the membrane’s thickness, with density increasing from the top surface downward.

The findings presented above provide additional evidence for our hypothesis; through the preassembly of metal ions, PGO nanosheets initiate a rapid heterogeneous nucleation process, thereby substantially improving the assembly of the membrane. Compared to the GO nanosheets, PGO nanosheets exhibit the enhanced adsorption capacity for Zn²⁺ (figs. S8, S9, and S16), which facilitate their deposition onto the AAO substrate and provide more bridging modules within ZIF-8 membranes. After the bridging process, the ZIF-8–PGO powder was collected. The heteroepitaxial growth of ZIF-8 crystals on the surface of PGO nanosheets is clearly observed. The crystalline properties of the bridged ZIF-8 exhibit no obvious changes (figs. S17 to S19), confirming that the PGO nanosheets have a negligible influence on its crystal-oriented growth. The grain density bridged by each PGO nanosheet is estimated to be 25 per square micrometer by marking the ZIF-8 grains in the magnified SEM images (Fig. 2F). Theoretically, a single PGO nanosheet larger than 10 μm can bridge more than 2500 ZIF-8 grains. We find that ZIF-8–PGO composite nanosheets exhibit spontaneous bending due to the impact of the electron beam in SEM observations. In this case, no intercrystallite cracking is observed (fig. S20). On the contrary, ZIF-8 membrane samples are easily penetrated along the ZIF-8 grain boundaries during routine SEM scans on the same conditions, suggesting that the bridged ZIF-8 grain boundaries are more stable.

A series of NB–ZIF-8 membranes were fabricated by incorporating different PGO nanosheets. The thickness and XRD spectra of the as-prepared NB–ZIF-8 and ZIF-8 membranes are similar (Fig. 2G and fig. S21), indicating that the intrinsic structure of ZIF-8 is well preserved. The PGO molar concentration of each NB–ZIF-8 membrane has been calculated to be 1.18, 3.18, and 5.91 mol %, respectively, as evidenced by XPS spectra (Fig. 2H and table S1). Notably, an increase in PGO content corresponds with a slight reduction in membrane thickness, which decreases from 520 to 450 nm (fig. S21). This observation suggests that a higher concentration of nonconductive PGO nanosheets leads to a more rapid coverage of the conducting substrate, thereby resulting in an earlier cessation of membrane growth.

Nanosheet-bridging effect of MOF membrane mechanical reinforcement

It can be inferred that there exists a relatively strong interaction between ZIF-8 and PGO, which not only induces the fast nucleation and growth but also constructs bridging connection among ZIF-8 crystals. The nanosheets-bridging structure was investigated by FTIR and high-solution XPS spectra. Two characteristic peaks at 468 and 566 cm⁻¹ are observed in the NB–ZIF-8 membrane (fig. S22), corresponding to the formation of Zn─O bonds between the Zn²⁺ ions and the oxygen-containing groups (hydroxyl, carbonyl, and carboxyl) in both monodentate and bidentate coordination forms (32). The Zn 2p spectra show two characteristic peaks at 1022.3 and 1045.3 eV for both ZIF-8 and NB–ZIF-8-2 membranes (Fig. 3A), corresponding to Zn 2p_3/2 and Zn 2p_1/2, respectively, indicating that Zn²⁺ is in a tetrahedral coordination state. With the increasing PGO content, the two peaks gradually shift to lower binding energy, which is mainly attributed to the variation in Zn local coordination environment via Zn-O coordination, well consistent with the FTIR results. Compared to the pure PGO, the peak areas of the C─O and C═O in PGO@Zn²⁺ and NB–ZIF-8 membrane decrease, indicating that the carboxyl and hydroxyl groups are set as bridge sites and intimately involved in the internal assembly of ZIF-8 membrane (Fig. 3B). Furthermore, the presence of the O─C═O peak at 288.4 eV and the C─O peak around 287.0 eV in the NB–ZIF-8 provides evidence for the successful integration of PGO. In addition, the C─O and C─N peaks of NB–ZIF-8-1 exhibit a shift from 286.7 and 285.9 eV to 287.1 and 286.2 eV, respectively, corresponding to an increase content of PGO. The N─Zn/N─C peak shifts to a lower energy level, while the characteristic peak of the imidazole ring (C═N) still remained at 397.8 eV according to the N 1s spectra (fig. S23). Combined with the trend of the C─N peak, which shifts to a higher energy level, it can be inferred that the binding energy of the N─Zn bond is decreased. The decreased binding energy of the N─Zn bond corresponds to a more stable and rigid structure. In addition, molecular dynamics simulations were performed to elucidate the intensity interactions. We calculated the interactions between ZIF-8 and PGO. The interaction energy between ZIF-8 and PGO exhibits a notable reduction from −1.09 to −3.77 eV. This observation suggests that the introduction of PGO-mediated coordination interactions, as opposed to the weaker interactions such as hydrogen bonds or van der Waals forces associated with the original terminal ligands, leads to a substantial enhancement in interaction (Fig. 3C).

Fig. 3. Mechanism of nanosheets bridging–induced mechanical enhancement of NB-MOF membrane.

High-resolution XPS spectra of MOF-PGO membranes with peak fitting: (A) Zn 2p, (B) C 1s. (C) The calculated interaction energy between ZIF-8 and PGO by molecular dynamics simulation. (D) The morphology, Young’s modulus, and energy dissipation of ZIF-8 and NB–ZIF-8-2 membrane, measured by AFM; scale bar, 1 μm. (E) Load-displacement curves of ZIF-8 membrane and NB–ZIF-8 membranes, measured by nanoindentation. (F) Young’s modulus and hardness of ZIF-8 and series of NB–ZIF-8 membranes. (G) Top-view SEM morphology of membranes after bending test; all the test membrane samples were fabricated on the flexible PAN substrate. (H) The C₃H₆/C₃H₈ separation of NB–ZIF-8-2 membrane on PAN substrate during the different curvature bending, equimolar C₃H₆/C₃H₈ mixture, 25°C, feed pressure at 1.1 bar.

We confirmed the mechanical response through quantitative nanomechanical mapping in the AFM. The as-prepared membranes exhibit alternating regions of bright and dark contrast (Fig. 3D and figs. S24 and S25). These contrast variations correspond to areas exhibiting high and low Young’s modulus distribution. Notably, the NB–ZIF-8-2 membrane demonstrates an average Young’s modulus increase from 2.5 to 3.3 GPa, indicating a 32% enhancement in mechanical stiffness. The NB–ZIF-8-2 membrane displays numerous bright regions along the grain boundaries, suggesting that PGO nanosheets effectively bridge these grains and improve structural strength. Furthermore, a systematic statistical analysis of the distribution of Young’s modulus reveals that 61.72% of the ZIF-8 membrane’s area exhibits a Young’s modulus of lower than 2.5 GPa (fig. S25E). This finding suggests the existence of numerous relatively soft regions within the ZIF-8 membrane, which may experience considerable deformation and cracking under high pressure. Conversely, after the PGO bridging, it is observed that 78.97% of the NB–ZIF-8-2 membrane’s area has a Young’s modulus exceeding 2.5 GPa (fig. S21F). Furthermore, the increased energy dissipation values are observed, ranging from 3.79 to 6.03 keV (Fig. 3D), indicating that NB–ZIF-8 has the capacity to dissipate mechanical energy through inelastic deformation. The highest elastic recovery of NB–ZIF-8-2 membrane is also confirmed in the load-displacement measurement by applying a controlled 100-nm indentation depth on the membranes (Fig. 3E). The hardness calculated from the nanoindentation test exhibits the same trend of Young’s modulus, and the hardness of NB–ZIF-8-2 membrane reaches 0.367 GPa, which is 2.36 times greater than that of the ZIF-8 membrane (Fig. 3F). We speculate that the external stress can be transferred from the inherently brittle grain boundaries to the more rigid PGO nanosheets, thereby providing extra energy dissipation pathways. Consequently, the NB–ZIF-8 membranes demonstrate the formation of supplementary energy dissipation pathways, which effectively improves their mechanical properties and capacity to withstand applied mechanical loads.

The results of the bending tests conducted on the ZIF-8 and NB–ZIF-8 membranes substantiate our findings. We fabricated the ZIF-8 and NB–ZIF-8 membranes on a polyacrylonitrile (PAN) substrate (figs. S26 to S28). The ZIF-8 membrane demonstrates characteristic cracking along its grain boundaries even under minimal bending test (Fig. 3G). In contrast, the NB–ZIF-8 membrane maintains its integrity and exhibits stable separation performance (fig. S28), achieving a C₃H₆ permeance of 23.5 GPU and a C₃H₆/C₃H₈ selectivity of 267 at a bending curvature of 60 m⁻¹ (Fig. 3H). At a bending curvature of 80 m⁻¹, cracking patterns become evident, resulting in a deterioration of its selective separation performance (fig. S28D). Collectively, these findings provide strong evidence that incorporating PGO effectively bridges ZIF-8 grains and improves the mechanical properties of crystal membrane, thereby guaranteeing stable performance under external mechanical stresses.

High-pressure resistance and long-term stability of membranes

The gas separation performance of the as-prepared membranes was evaluated using the Wicke-Kallenbach method (fig. S29) (33). The ZIF-8 membrane exhibits a C₃H₆/C₃H₈ separation factor of 111 and a C₃H₆ permeance of 109.7 GPU, well consistent with previous reports (34). The enhanced separation selectivity is observed after PGO bridging. The NB–ZIF-8-2 membrane with 3.18 mol % PGO demonstrates a superior C₃H₆/C₃H₈ selectivity of 273, accompanied by an excellent C₃H₆ permeance of 130.5 GPU (Fig. 4A). Compared with GO nanosheets, PGO nanosheets with high-density pores promote mass transfer, thereby improving selectivity without compromising permeance (fig. S30). Moreover, the PGO nanosheets spread parallel to the substrate surface can be regarded as gutter layers, which may avoid the pore penetration of ZIF-8 and lead to a slight permeance increase of NB–ZIF-8 membranes. The decrease in separation performance under high PGO content may be attributed to the high loading of PGO nanosheets, resulting in increased transport resistance and reduced permeance. The molecular sieving properties of NB–ZIF-8-2 are assessed using various gas molecular probes, including H₂, CO₂, N₂, CH₄, C₃H₆, and C₃H₈ (fig. S31). A sharper cutoff between C₃H₆ and C₃H₈ is observed, with the ideal C₃H₆/C₃H₈ selectivity exceeding 308. In addition, the slight adsorption selectivity (C₃H₆/C₃H₈ = 1.02) tested by pure C₃H₆ and C₃H₈ adsorption indicates that the enhanced selectivity originates from the precise molecular sieving (fig. S32). The separation performance at different assembly times indicates the formation of a selective continuous membrane between 20 and 30 min (fig. S33), which is consistent with the results of SEM and XRD analyses. These results indicate that PGO bridging promotes the assembly of the membrane and eliminates membrane defects, thereby enhancing the kinetic separation of C₃H₆/C₃H₈.

Fig. 4. Hydrocarbon separation performance of NB–ZIF-8 membrane and the high-pressure stability of NB–ZIF-8 membrane.

(A) C₃H₆/C₃H₈ separation performance of NB–ZIF-8 membrane with various PGO nanosheets loading at 1.1 bar and ambient temperature. (B) C₃H₆/C₃H₈ separation performance of NB–ZIF-8-2 membrane with elevated pressure up to 7 bar. (C) Comparison of C₃H₆/C₃H₈ selectivity change from 1.1 to 7 bar with different size GO in ZIF-8 membrane. (D) Comparison of the C₃H₆/C₃H₈ separation performance after high-pressure impact test; all membrane were pressed under 50 bar for 6 hours (h); the high pressure N₂ was chosen to as the impact source. (E) Comparison of mechanical properties and high-pressure stability of ZIF-8 and NB–ZIF-8 membrane. (F) Long-term stability test at 1.1 bar and five cycles of pressurization-depressurization test. (G) Long-term stability test at 7 bar of ZIF-8 and NB–ZIF-8 membrane. (H) Comparison of the separation performance of NB–ZIF-8 membrane with the state-of-art works under high feed pressure (≥7 bar) (table S3).

Stable separation performance is the cornerstone for practical separation, and several researches reported that the hybrid distillation-membrane technology can save 20 to 50% energy consumption (35–37). On the basis of this scenario, the membrane structure needs to withstand the high-pressure loading (≥7 bar). However, the C₃H₆/C₃H₈ selectivity of ZIF-8 membrane declines by 69% as the pressure increased to 7 bar (fig. S34). In contrast, the NB–ZIF-8-2 only shows an 11% decrease; the C₃H₆/C₃H₈ selectivity still remained at 243, 8 times higher than ZIF-8 membrane (Fig. 4B). We hypothesize that this could be attributed to the effective bridging of the ZIF-8 grains, which not only enhances the mechanical properties of the membrane but also provides additional stress dissipation pathways, thereby ensuring its pressure resistance. To verify our hypothesis, we fabricated unbridged ZIF-8 membrane by incorporating the same content GO quantum dots (GOQDs) into the ZIF-8 membrane. The GOQD’s size is smaller than 50 nm, notably smaller than the size of ZIF-8 crystals (100 to 300 nm) (fig. S35). Consequently, the GOQDs mainly exist as discrete points within the ZIF-8 crystals (38) and are unable to bridge these grains effectively. The ZIF-8–GOQD membranes (fig. S36), prepared under identical conditions, exhibit a selectivity decline of more than 70% under high pressure (fig. S37). These findings suggest that the enhancement of the membrane’s pressure resistance can be only achieved by establishing effective bridging structures between ZIF-8 grains (Fig. 4C). Without PGO bridging, the grain boundaries continuously accumulate external stresses, leading to structural instability under high pressure; both the unbridged ZIF-8 membrane and the ZIF-8–GOQD membrane fail to meet the stability requirements under high pressure. Notably, we directly introduce high-pressure gas into the feed side of the membrane, thereby exposing the membrane to an immediate pressure impact of 6 bar, which corresponds to an instantaneous load of 6 kgf cm⁻². Despite this, the NB–ZIF-8 membrane remains stable, highlighting its potential for practical applications.

To investigate the upper bound of pressure resistance, high-pressure nitrogen (50 bar, which is equivalent to a pressure load of 50 kgf cm⁻²) was used to evaluate the structural stability of the membrane (Fig. 4D). Notably, we found that ZIF-8 membrane exhibits a permeation behavior reminiscent of the aging effects observed in polymers under high pressure (fig. S38), characterized by a gradual decrease in N₂ permeance. In contrast, the NB–ZIF-8-2 membrane maintains a consistent permeance throughout the testing. Following the high-pressure impact tests, the separation performance of NB–ZIF-8 remains largely unaffected, while the selectivity of ZIF-8 experiences a reduction of approximately 65%. This observation suggests that minor defects within the ZIF-8 membrane are exacerbated under high-pressure conditions, resulting in irreversible structural damage.

Furthermore, the long-term stability of ZIF-8 and NB–ZIF-8 membranes was tested at 1.1 and 7 bar, respectively. After 350 hours of long-term stability testing at 1.1 bar and five cycles of pressurization-depressurization from 1.1 to 7 bar, the separation performance of NB–ZIF-8-2 membrane has no obvious degradation (Fig. 4F). We further tested the high-pressure long-term stability of the ZIF-8 and NB–ZIF-8 membranes. The ZIF-8 membrane exhibits a sharp decline in C₃H₆ flux within 10 hours, while the C₃H₈ flux increases, resulting in a 90% loss of C₃H₆/C₃H₈ selectivity (Fig. 4G). Under the long-term high-pressure loading, the unbridged grain boundaries are prone to accumulating external stress and leading to the formation of defects. In contrast, because of enhanced stiffness, hardness, and energy dissipation, our NB–ZIF-8-2 membrane exhibits the record-breaking pressure-resistance property, with the selectivity maintained at nearly 240 with the C₃H₆ flux of 1.3 × 10⁻² mol s⁻¹ m⁻² (55.8 GPU, 1 GPU = 3.348 × 10⁻¹⁰ mol m⁻² s⁻¹ Pa⁻¹) at 7 bar under high-pressure conditions for 300 hours and maintained stable performance after 60 days of storage testing (Fig. 4G). Furthermore, the NB–ZIF-8-2 membrane exhibits excellent stability under industrially relevant conditions, including a simulated five-component feed gas for propane dehydrogenation as well as an elevated operating temperature of 50°C (figs. S39 and S40) (39, 40).

We have prepared larger NB–ZIF-8 with 14-fold increased area and confirmed the uniformity of the PGO bridging strategy through the comparison of optical interference, membrane morphology, crystallinity, and separation performance (figs. S41 to S44). All the test samples exhibit the enhanced C₃H₆/C₃H₈ separation performance (fig. S41) with a C₃H₆ permeance surpassing 100 GPU and a C₃H₆/C₃H₈ selectivity surpassing 250. Meanwhile, the PGO bridging strategy can be extended to fabricate other pressure-resistant MOF membrane, especially those MOFs that are generally considered prone to grain boundary defects (41). As an illustration, we successfully fabricated PGO nanosheets–bridged ZIF-67 membrane (fig. S45). The NB–ZIF-67 membrane also exhibits superior separation performance, with a C₃H₆/C₃H₈ selectivity of 296 and a C₃H₆ permeance of 55.8 GPU. At 7 bar, its separation performance remains stable, with an 8.2% decrease in C₃H₆/C₃H₈ selectivity (fig. S45E).

Last, we compared the performance of our NB-MOF membranes with the state-of-art membranes operating under 7 bar (Fig. 4H and table S3). The ZIF-8 membrane shows a sharp decrease in selectivity and near total loss of its separation performance after 10 hours of continuous operation. In contrast, our NB–ZIF-8-2 membrane remains stable and exhibits selectivity that is 24 times higher. Currently, only a few studies have reported the performance of MOF membranes at 7 bar. Compared to these MOF membranes, despite operating for over 300 hours at 7 bar, our membrane remained positioned in the upper-right corner of the images, representing both the highest separation performance and ever-recorded longest stability for MOF membranes.

DISCUSSION

Inspired by the natural cuttlebone structure, we propose a facile nanosheet-bridging strategy to fabricate durable NB-MOF membranes for practical hydrocarbon separation. The PGO nanosheets serve as the rigid wall and bridge the MOF grains through the fortified coordination interaction between PGO nanosheets and MOF. Such bridging effect enhances mechanical properties of crystal membranes and provides additional energy dissipation pathways to avoid stress accumulation, thus preventing structural cracking under high pressure. By tuning the morphology and content of nanosheets, the optimized NB–ZIF-8-2 membrane exhibits excellent pressure resistance (up to 50 bar) and superior long-term stability at industry-favorite C₃H₆/C₃H₈ separation pressure (7 bar, 300 hours), with a selectivity of 240 and a steady C₃H₆ flux of 1.3 × 10⁻² mol s⁻¹ m⁻², outperforming the state-of-art reports. This strategy also allows fabricating defect-free NB–ZIF-67 membranes with pressure resistance, as well as NB-MOF membranes on commercial polymer substrate. The facile and scalable nanosheet-bridging strategy affords a platform for fabricating pressure-resistant MOF membranes and many other crystalline membranes toward a broader range of applications.

MATERIALS AND METHODS

Materials

2-mIm (98%, Macklin), anhydrous zinc acetate [Zn (CH₃COO)₂, 99%, Aladdin], anhydrous cobalt acetate [Co (CH₃COO)₂, 98%, Macklin], methanol (MeOH, 99%, Aladdin Co. Ltd.), hydrogen peroxide (H₂O₂, 30 wt % aqueous solution, Tianjin JiangTian Chemical Co. Ltd.), deionized water (DI water, laboratory-made), single-layer GO nanosheets (0.5 mg/ml), and GOQDs (0.5 mg/ml) were supplied by XFNANO Co. Ltd. (Jiangsu, China). All chemicals were used directly without further purification. The AAO (with pore diameter of 110 to 150 nm, Pu-Yuan Nanotechnology Co. Ltd.) and PAN (with average pore diameter <20 nm, Shandong Megavision Membrane Engineering and Technology Co.) were used as the porous substrate. Before electrochemical coassembly, all the substrates need to coat platinum (Pt) layer by ion sputtering deposition to increase electrical conductivity.

Synthesis of PGO nanosheets

A dispersion of single-layer GO nanosheets was subjected to ultrasonic treatment and centrifugation to enhance its dispersibility. Subsequently, 50 ml of the GO nanosheet dispersion was heated to 100°C using an oil bath, to which 1.25 ml of an aqueous hydrogen peroxide (H₂O₂) solution was rapidly added. The mixture was allowed to react for either 2 or 4 hours to yield a dispersion of PGO. After reaction, the PGO nanosheets dispersion was dialyzed with deionized water. The concentration of the resulting PGO dispersion was adjusted to 0.5 mg/ml. On the basis of the reaction time, the resultant PGO nanosheets were designated as PGO-2h and PGO-4h, with the term “PGO” in this research specifically referring to PGO-4h.

Fabricating NB-MOF membrane via electrochemical coassembly method

Zn (CH₃COO)₂ was used as the zinc source (0.83 mmol, 0.152 g) and dissolved in 10 ml of DI water. Concurrently, a ligand solution was prepared by dissolving 22-mIm (50 mmol, 4.105 g) in 50 ml of DI water. Subsequently, different contents of GO/PGO nanosheets (0, 0.3, 0.9, and 1.5 mg) were integrated into the 2-mIm solution. This mixture was then stirred with the zinc solution to form the precursor solution. The current density during the coassembly process was maintained at 0.13 mA cm⁻². The resultant membranes were designated as ZIF-8 and NB–ZIF-8-X (X = 1, 2, 3), where X corresponds to the quantity of GO/PGO nanosheets added. For NB–ZIF-67 membrane, the metal sources were substituted for Co (CH₃COO)₂ (0.83 mmol), and the other processes are the same as fabricating NB–ZIF-8 membranes.

Gas permeation measurement

Gas permeation of the membranes was measured by a Wicke-Kallenbach method by using a homemade setup (fig. S29). The equimolar C₃H₆/C₃H₈ mixtures were set as feed from 1.1 to 7.0 bar in feed side. The membrane sample was installed in the membrane cell, and the chamber was sealed using O rings. The cell was then clamped securely and connected to the testing system. A sweep gas (Argon) flow rate of 8 ml min⁻¹ was regulated downstream using a mass flow controller. The downstream pressure was maintained at atmospheric pressure. A predefined method file was selected on the basis of the gas system, and the gas chromatograph was initialized by heating the column (140°C), injector (160°C), and thermal conductivity detector (160°C). Sample injection commenced once the baseline stabilized, with chromatograms acquired at 8-min intervals. Separation selectivity and gas permeance were calculated from peak integration. Stable performance over three to five consecutive injections confirmed data reliability, and the average value was recorded. Subsequent membranes were tested following the same protocol. For high pressure test, the N₂ molecules were set as high-pressure source. For long-term storage test, the membrane was stored in a desiccator during the intervals of each test. For polymer-supported NB-MOF membrane test, a thin layer of polydimethylsiloxane (0.5 wt %) was coated on the surface after bending to preliminary assess the impact of membrane crystalline defects on separation performance.

The gas permeance of as-prepared membranes was calculated by Eq. 1

$${\left(\frac{P}{l}\right)}_{\mathrm{i}}=\frac{Q_{\mathrm{i}}}{\mathrm{\Delta }{P}_{\mathrm{i}}A}$$ (1)

where P/l is the gas permeance, GPU, and 1 GPU = 1 × 10⁻⁶ cm³ [standard temperature and pressure (STP)] cm⁻² s⁻¹ cm Hg⁻¹ = 3.348 × 10⁻¹⁰ mol m⁻² s⁻¹ Pa⁻¹; Q_i is the volumetric flow of component gas, i, cm³ (STP)/s; l is the thickness of membrane, cm; A is the effective testing area of membrane samples, cm²; and ΔP_i refers to the partial pressure difference between feed side and permeate side of component i, cmHg.

The C₃H₆/C₃H₈ selectivity α_i/j for mixed gas is defined by Eq. 2

$${\mathrm{\alpha }}_{\mathrm{i}/\mathrm{j}}={\left(\frac{P}{l}\right)}_{\mathrm{i}}/{\left(\frac{P}{l}\right)}_{\mathrm{j}}$$ (2)

More details can be found in the Supplementary Materials.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S45

Tables S1 to S3

References