Two-dimensional lamellar nanosheet membranes with intrinsic size-sieving nanopores for ultrafast hydrogen separation

Abstract

Hydrogen separation and purification are essential for the large-scale application of green hydrogen energy. Two-dimensional (2D) material-based membrane technology provides an energy-efficient approach; however, the existing 2D lamellar membranes and nanoporous membranes struggle with low permeability and complex preparation procedures. Here, we constructed 2D lamellar nanosheet membranes with intrinsic size-sieving nanopores by assembling nanoporous polymeric carbon nitride (PCN) nanosheets into a lamellar membrane. These membranes exhibit high selectivity and permeance and ease of preparation. By applying MXene as the strong interaction patch for covering and repairing defects on PCN nanosheets, the resultant membranes demonstrated an ultrahigh H₂ permeance of 870 to 8046 GPU, considerable selectivity, and superior long-term stability, outperforming most current state-of-the-art membranes. Economic analysis revealed that MXene/PCN membranes achieved ultralow energy consumption and minimal membrane area demands for practical applications. This study inspires the construction of 2D lamellar membranes with intrinsic nanopores and provides a simple and scalable approach to preparing high-performance 2D nanosheet membranes for hydrogen separation.

A carbon nitride nanosheet membrane with intrinsic nanopores enables effective hydrogen separation with ultrahigh permeance.

INTRODUCTION

Hydrogen (H₂) energy is regarded as a sustainable alternative in the context of energy and environmental crises owing to its zero-carbon emission and high energy conversion efficiency (1, 2). Now, large-scale H₂ production industry, such as methane reforming process, usually requires further H₂ purification from other impurities (3, 4). Compared with conventional purification methods such as pressure swing adsorption (PSA) and cryogenic distillation, membrane-based H₂ separation technologies have gained widespread attention due to their high efficiency, easy operation, and low energy consumption (5). However, traditional polymer membranes suffer from the trade-off between permeability and selectivity, which limits their further development (6).

Recently, two-dimensional (2D) membranes with highly ordered structures constructed from nanosheets, such as graphene oxide (GO) (7, 8), transition-metal carbides/carbonitrides (MXene) (9–11), metal organic frameworks (MOFs) (12), covalent organic frameworks (COFs) (13), and nanoporous graphene (NPG) (14), have revolutionized membrane material design. In general, 2D membranes were classified as lamellar membranes and nanoporous membranes, depending on whether the assembled building blocks are nonporous or porous (15). Previous studies have shown that 2D lamellar membranes (e.g., GO and MXene) used the interlayer nanochannels between the adjacent nanosheets and thus achieved efficient gas sieving performance, outperforming the current Robeson upper bound (9). However, strict control of narrow interlayer spacing at a subnanometer level and tortuous gas transport channels leads to low gas permeation flux (11, 15, 16). In contrast, 2D nanoporous membranes (e.g., MOFs, COFs, and NPG) have intrinsic nanopores and short transport pathways, resulting in higher permeation flux compared with 2D lamellar membranes (17). Nevertheless, this type of membrane faces challenges for practical applications related to complex preparation procedures, low yields, and mismatched aperture sizes of membrane materials. Therefore, developing advanced construction strategies for 2D membranes is essential to overcome the above limitations for nanoporous membranes and lamellar membranes and thus broaden the practical application of 2D membranes for H₂ separation.

Here, we present a strategy for constructing 2D membrane by assembling nanoporous nanosheets into lamellar membranes (Fig. 1). This approach would have the advantages of both lamellar membranes and nanoporous membranes, including (i) sufficient nanopores shorten the transfer pathways, allowing for high permeance; (ii) suitable aperture size provides selective molecular sieving, showing high selectivity; and (iii) self-assembly of porous nanosheets by like simple vacuum filtration simplifies the fabrication steps, demonstrating great potential in large-scale preparation. The selection of 2D nanoporous material is crucial for the membrane construction and gas separation performance. However, the further development of MOFs (e.g., ZIF-8) and COFs might be restricted by the instability, high costs, and inappropriate aperture (18, 19). In contrast, polymeric carbon nitride (PCN) is an ideal building block for 2D lamellar nanosheet membrane due to its abundant intrinsic nanopores (16), appropriate aperture size (~3.11 Å) (20), and low preparation cost (21). Theoretical researches have revealed that ultrahigh H₂ separation selectivity can be achieved with PCN due to its lowest diffusion energy barriers for H₂ compared with other gases (22, 23). PCN can be assembled into a lamellar membrane through the ordered stacking of ultrathin nanosheets obtained via exfoliation (24).

Fig. 1. Construction strategy of lamellar membrane with intrinsic nanopores.

Nanoporous membranes achieve high gas permeance but face challenges related to complex preparation procedures and low mechanical strength. Lamellar membranes are convenient to fabricate but demonstrate low gas permeance. Lamellar membranes with intrinsic nanopores would simultaneously have high selectivity, permeance, and simple preparation procedures.

Although both inherent properties and theoretical research showed that PCN-based membranes enable H₂ sieving capability, few experimental results were unsatisfactory, exhibiting low selectivity approaching the Knudsen diffusion value (25). The main reason is that in-plane tri-s-triazine structures are prone to deterioration, and defects are formed in the exfoliation process (e.g., ultrasonication) (21, 26), which is commonly used for weakening the π-π interaction in bulk PCN to obtain the ultrathin nanosheets (27). Gas molecules would preferentially pass through the large defects rather than the intrinsic nanopores, severely degrading the membrane separation performance. The control and repair of defects on the 2D nanosheet membranes present considerable challenges, while effective and scalable strategies are highly desired for efficient gas separation.

In this study, PCN nanosheets were used as the main building blocks for constructing lamellar membranes, while MXene acted as the patch for the coverage and repair of the defects on the PCN nanosheets due to the strong interfacial interaction between MXene and PCN nanosheets (Fig. 2A). The resultant membranes demonstrated an ultrahigh H₂ permeance of 870 to 8046 gas permeation unit (GPU), considerable selectivity, and superior long-term stability, outperforming current state-of-the-art membranes. Economic analysis detailed the cost-effectiveness and feasibility of MXene/PCN nanosheet membranes for practical gas separation. Our approach provided insights into the development of 2D lamellar membranes with intrinsic nanopores and proposed a simple and scalable approach to prepare 2D nanosheet membranes for energy-efficient H₂ separation.

Fig. 2. Preparation and microstructure of nanosheets and membranes.

(A) Schematic illustration of the construction of MXene/PCN nanosheet membrane. MXene is expected to act as the patch for repairing the defects on the PCN nanosheets during the vacuum filtration process. (B and C) TEM images of PCN nanosheets, showing the uniform distribution of intrinsic size-sieving nanopores. (D) TEM image evidence for the surface defects formation on the PCN nanosheet and the defects range from 10 to 50 nm. Cross-sectional SEM image of (E) PCN and (F) MXene/PCN nanosheet membrane, indicating an obvious lamellar structure. In addition, increased thickness could be found after the incorporation of MXene. (G) Surface SEM image of MXene/PCN nanosheet membrane, showing a continuous surface without defects.

RESULTS

Preparation of PCN nanosheets

Ultrathin PCN nanosheets were prepared through liquid-assisted ultrasonic exfoliation with a mixed solvents of isopropanol (IPA) and H₂O. IPA molecules can intercalate into the interlayer region of nanosheets, weakening the π-π interaction between the adjacent layers, which leads to a high yield of ultrathin nanosheet (28). Tyndall scattering was observed in the as-prepared PCN suspension (fig. S1), indicating the successful exfoliation and good dispersion of the nanosheets. The micromorphology and structure of PCN nanosheets were further examined using high-resolution transmission electron microscopy (HRTEM) and atomic force microscopy (AFM). HRTEM images show a high density of nanopores (Fig. 2, B and C, and fig. S2), which allows a spherical guest with a diameter below ~0.311 nm to pass through (fig. S3). Moreover, it was clearly seen that surface defects ranged from 10 to 50 nm on the PCN nanosheets (Fig. 2D), likely caused by the structural damage and defect formation during the ultrasonic exfoliation process. A similar phenomenon has been reported in previous studies (25, 29). On the basis of the large-scale AFM results (figs. S4 and S5), the lateral size and thickness of the PCN nanosheets were ~320 to 730 and 0.43 to 1.12 nm, respectively. Given that the theoretical thickness of a monolayered PCN was 0.3 nm, the as-prepared PCN nanosheets in this work contained two to three monolayers (30). In addition, MXene demonstrated an intact structure without surface defects (figs. S6 and S7) and the corresponding lateral size and thickness of ~63 to 111 and 1.06 to 1.40 nm, respectively. The scanning electron microscopy (SEM) images of PCN nanosheets before and after MXene incorporation were provided, as displayed in fig. S8. PCN nanosheet demonstrated rough surface with a large number of defects, while MXene nanosheet exhibited a smooth surface without any defects. It was clearly seen that MXene could be attached to PCN nanosheet, and thus, the defects within the PCN nanosheets were expectedly covered and repaired after the incorporation of MXene, as confirmed by the remarkable decrease of defects.

Preparation of MXene/PCN nanosheet membrane

The MXene/PCN nanosheet membranes were prepared by ordered stacking of nanosheets on porous α-Al₂O₃ substrate (figs. S9 and S10) using the vacuum filtration method. Note that the self-assembly of MXene nanosheets was influenced by the water flow during the vacuum filtration process. In general, water molecules would tend to rapidly flow through the large defects rather the intrinsic nanopores on the PCN nanosheet, due to the difference of flowing resistance. The membranes assembled by PCN and MXene/PCN nanosheets demonstrated a distinct lamellar structure, as shown in the cross-sectional SEM images (Fig. 2, E and F). Energy dispersive spectrometer (EDS) mapping indicated the uniform distribution of Ti, C, and O elements in both surface and cross-sectional SEM images, verifying the high dispersion of MXene on the MXene/PCN nanosheets membrane (figs. S11 and S12). The thickness of the PCN nanosheet membranes increased from 1 to 1.3 μm after the incorporation of MXene. Surface SEM images showed that the porous α-Al₂O₃ substrate was completely covered by PCN nanosheets and MXene, with no obvious pinhole observed on the membrane surface (Fig. 2G and figs. S13 and S14). In addition, the obtained PCN nanosheets in this study contained two to three monolayers, and the membrane structure constructed by these nanosheets might be different from the membranes by monolayer nanosheet (fig. S15).

The crystalline structures of the prepared nanosheet membranes were investigated by x-ray diffraction (XRD) patterns, as shown in Fig. 3A. Two diffraction peaks at 12.8° and 27.4° corresponded to the (100) in-plane tri-s-triazine rings units and (002) interlayer graphitic stacking of PCN, respectively (31). The low relative intensity of the (100) plane for PCN was observed, which is attributed to the damage of in-plane tri-s-triazine units (29). For the MXene/PCN nanosheet membrane, an additional characteristic diffraction peak appeared at 5.98° corresponded to the interlayer stacking (002) of MXene, indicating the successful fabrication of the MXene/PCN nanosheet membrane. However, the relative intensity of this stacking plane decreased compared with pristine MXene, which was primarily due to the interfacial interaction between the two types of nanosheets affecting the original assembly of MXene. Moreover, the stacking structure of the nanosheets on the MXene/PCN nanosheet membranes was comprehensively analyzed (fig. S16), and the stacking types occurred on (i) adjacent PCN nanosheets (d₁), (ii) MXene and PCN nanosheets (d₂), and (iii) adjacent MXene nanosheets (d₃). Moreover, these three stacking modes were strongly related to the MXene mass fraction in the MXene/PCN nanosheet membranes. From XRD spectra (fig. S17A), no remarkable diffraction peak was observed as the MXene mass fraction was relatively low (<10%), which was attributed to the fact that (i) uniform mixing of MXene/PCN nanosheets and (ii) the stacking frequency of MXene/PCN nanosheets were much lower than that of adjacent PCN nanosheets (fig. S16B). As the MXene mass fraction gradually increased (>10%), the stacking frequency of MXene/PCN nanosheets remarkably increased as well as the adjacent MXene/MXene nanosheets, resulting in a new diffraction peak at 6.15°. However, it was noted that the location of this diffraction peak almost coincided with the diffraction peak of MXene. Such similar location of these two diffraction peaks made it difficult to accurately identify the MXene/PCN and MXene nanosheets, and we speculated that the characteristic peak of MXene/PCN might be covered by the MXene. For further clarifying the speculation, we calculated the relative potential energy of MXene/PCN nanosheet with different distances (centroid distance between the N atoms on the PCN and surface H atoms on the MXene), as shown in fig. S17B. It was found that the relative potential energy demonstrated the trends of decreasing and then increasing as the distance increased from 0.2 to 1.0 nm. The most stable state of MXene/PCN nanosheets was achieved at the lowest relative potential energy with a distance of 0.25 nm and thus the free spacing (subtracting the effective radius of N and H atoms) of MXene/PCN nanosheets was estimated to be ~0.05 nm, indicating strong adhesion between the two types of nanosheets. On the basis of the monolayer thickness of MXene (~0.98 nm) and PCN (~0.3 nm) and free spacing, the d-spacing of MXene/PCN was ~1.33 nm. Therefore, the corresponding diffraction peak would appear at ~6.7° (calculated on the basis of Bragg’s law), likely obscured by the broad peak of MXene nanosheets, and the phenomenon of obscuration was consistent with the XRD spectra.

Fig. 3. Interfacial interaction between PCN and MXene.

(A) XRD spectra of PCN, MXene, and MXene/PCN, revealing the stacked structure between PCN and MXene nanosheets, and detailed structures were also shown in fig. S16. (B) N 1s XPS spectra of PCN, MXene, and MXene/PCN. (C) FTIR spectra of PCN, MXene, and MXene/PCN, showing the hydrogen bonding interaction between PCN and MXene. (D) MD simulations of the calculated nonbonded interaction energy between PCN nanosheet and MXene, revealing that van der Waals, electrostatic, and hydrogen bonding interaction occurred. Structural model of MXene/PCN nanosheet in MD simulations. (E) Top view and (F) side view. a.u., arbitrary units.

Furthermore, the elemental compositions, chemical structure, and interfacial interaction between MXene and PCN were characterized by x-ray photoelectron spectroscopy (XPS) analysis, Fourier transform infrared (FTIR) spectra, and molecular dynamics (MD) simulations. Four characteristic peaks were identified in the XPS spectra of the MXene/PCN (Fig. 3B and fig. S18), corresponding to N 1s, C 1s, Ti 2p, and O 1s signals, respectively, which confirmed the presence of MXene in MXene/PCN nanosheet membrane. In particular, the binding energy peaks at 398.87, 399.85, 401.2, and 404.52 in the N 1s XPS spectra were ascribed to the C─N═C, N─(C)₃, N─H, and π-excitations, respectively (32). The characteristic peaks of N─H shifted from 401.2 to 401.4 eV after the incorporation of MXene (table S1), likely caused by the strong interfacial interaction between MXene and PCN (33). As displayed in Fig. 3C, the bands attributed to the vibration of ─OH redshifted from 3450.1 cm⁻¹ for MXene to 3420.2 cm⁻¹ for MXene/PCN, indicating a hydrogen bonding interaction between MXene and PCN (34). This hydrogen bonding interaction might occur between the ─NH₂/NH of PCN nanosheets and ─OH of MXene. To further clarify the interaction, the nonbonded interaction energies between MXene and PCN nanosheets were calculated by MD simulations (Fig. 3, D to F). The results indicated that van der Waals, electrostatic, and hydrogen bonding interaction were involved in the interaction between MXene and PCN nanosheets (35). This strong interfacial interaction was conducive to inducing the ordered stacking of the nanosheets, resulting in defect repair of PCN nanosheets and enhanced mechanical strength. Compared with pristine PCN nanosheet membrane, the MXene/PCN nanosheet membrane demonstrated an enhanced mechanical strength (fig. S19 and table S2). The strain, tensile stress, and Young’s modulus of MXene/PCN nanosheet membrane were 2.76%, 4.57 MPa, and 112.72 MPa, respectively. The obtained MXene/PCN nanosheet membranes also demonstrated superior stability and maintained structural integrity in water for 48 hours (fig. S20). In addition, the efficient repair of defects by MXene was demonstrated by the pore diameter distribution (fig. S21). The large nanopores (~305.16 Å) resulting from the exfoliation process disappeared after MXene modification, indicating that MXene could effectively repair the defects on the PCN nanosheets. Overall, the characterizations and simulation results confirmed that strong interfacial interactions (i.e., van der Waals, electrostatic, and hydrogen bonding interaction) occurred between MXene and PCN nanosheets, which was conducive to facilitating the repair of defects on the PCN nanosheets.

Role of MXene in defect repair of PCN nanosheet membranes for hydrogen separation

Gas permeation through the prepared PCN nanosheet membranes was measured using a homemade Wicke-Kallenbach module (Fig. 4A and fig. S22). As presented in Fig. 4B, pristine PCN nanosheet membranes demonstrated an ultrahigh H₂ permeance (10,874 GPU) but low gas separation selectivity (i.e., H₂/N₂ of ~4.23 and H₂/CO₂ of ~5.14), mainly due to the presence of defects on the nanosheets, as confirmed by the previous characterization results. With the incorporation of MXene, the H₂ permeance sharply decreased to 870 GPU, while the corresponding H₂/N₂ and H₂/CO₂ selectivity increased to 17.37 and 30.21, respectively. For comparison, other typical 2D materials [e.g., hexagonal boron nitride (h-BN), molybdenum disulfide (MoS₂), reduced graphene oxide (rGO), and GO] have been also considered to use as the patch for repairing the defects on PCN nanosheet, and the corresponding gas separation performances of the composite nanosheet membranes were investigated (fig. S23). It was observed that the H₂ permeance of the obtained h-BN/PCN, MoS₂/PCN, GO/PCN, rGO/PCN, and MXene/PCN nanosheet membranes were 6836.2, 5263.1, 1068.0, 1957.9, and 869.4 GPU, respectively, with corresponding H₂/CO₂ selectivity of 7.1, 12.9, 22.4, 11.5, and 31.3, respectively. Compared with the pristine PCN nanosheet membrane, the remarkable decrease of gas permeance and increased of gas separation selectivity indicated that the incorporation of these 2D nanosheets could cover and repair the defects on PCN nanosheets. However, the notable difference of the gas permeance and separation selectivity was also shown, which was mainly attributed to the interfacial interaction between PCN and these nanosheets. Actually, the physicochemical properties of different 2D materials and the corresponding interaction strength with PCN directly determined the structure of composite membrane and the capability for repair defects on the PCN nanosheets. For h-BN and MoS₂ nanosheets, the lack of specific active sites and functional groups resulted in the incomplete adhere to PCN nanosheets. For example, the diffraction peak of the h-BN/PCN nanosheet membranes was located at ~7.2°, implying that the interlayer spacing were ~1.2 nm and thus the free spacing of h-BN/PCN was ~0.6 nm (36). Such a large free spacing of the composite membrane resulted in the nonselective nanochannels and weakened the molecular sieving capability of the intrinsic nanopores within the PCN nanosheets, leading to a low H₂/CO₂ selectivity. Compared with h-BN and MoS₂ nanosheets, the incorporation of GO nanosheets with PCN nanosheets exhibited enhanced gas separation selectivity, but the gas separation performance is still unsatisfactory. Theoretically, GO nanosheets have abundant oxygen-containing functional groups such as epoxy group (C─O─C), hydroxyl group (─OH), and carboxyl group (─COOH), which could realize the combination with PCN nanosheets. Nevertheless, the extremely flexible characteristic of the GO nanosheets tended to contort during the vacuum filtration process, resulting in incomplete alignment with PCN nanosheets and the formation of nonselective interlayer nanochannels locally. In addition, these nanochannels would allow both the H₂ and CO₂ molecules to pass through, reducing the separation performance. In addition, similar to h-BN and MoS₂, the decrease of active sites on the surface of the rGO nanosheets also caused weak interaction with PCN nanosheet, making it difficult to achieve high gas separation performance. Moreover, rGO obtained from chemical reduction (e.g., ascorbic acid and hydrazine) often accompanied with the defects on the surface and cannot achieve effective coverage and repair the defects on the PCN nanosheets. In contrast, a framework constructed by the strong chemical bonding (e.g., Ti─Ti, Ti─C, etc.) demonstrated the rigidity of MXene, which could maintain its original structure, and insignificant deformation occurred during the vacuum filtration process. Moreover, a large number of oxygen-containing functional groups and Lewis acid Ti sites (Ti─OH) were uniformly distributed on the MXene nanosheet surface (37), which could exhibit appropriate flexibility and provide abundant active sites for interaction with PCN. Such structure with rigid and flexibility could achieve complete alignment with PCN nanosheets and effective repair of defects. Therefore, compared with other 2D materials, MXene represented superior capability for repairing defects on the PCN nanosheets.

Fig. 4. Experimental results and MD simulations of the gas permeation performance for PCN and MXene/PCN nanosheet membranes.

(A) Homemade Wicke-Kallenbach gas permeation module and schematic illustration of gas transport through MXene/PCN nanosheet membrane. The purple O-rings around the membrane represent the seal rings for improving the leak tightness of the system and preventing the contact between membrane surface and the stainless steel module. (B) Comparison of H₂ permeance, H₂/N₂, and H₂/CO₂ selectivity through PCN and MXene/PCN nanosheet membranes at 25°C and 1 bar, and the inserts are photographs of prepared PCN and MXene/PCN nanosheet membranes. The number of gas molecules of H₂/CO₂ mixture through the nanosheet membranes as a function of simulation time and the corresponding simulation snapshots at different times in MD simulations were provided. (C) PCN nanosheet membrane. (D) MXene/PCN nanosheet membrane. The total permeation time is set as 10 ns, and detailed data could be found in fig. S28.

The aforementioned characterization and additional experimental results (fig. S24) indicated that the interlayer nanochannels formed by the adjacent nanosheets were not responsible for the gas separation process due to the inappropriate d-spacing. Therefore, the intrinsic size-sieving nanopores within the PCN nanosheets primarily contribute to gas molecular sieving, while MXene acted as a patch for repairing defects on the PCN nanosheets. To elucidate the role of MXene in defect repair of PCN nanosheet membranes, the gas transport mechanisms through pristine PCN and MXene/PCN nanosheet membranes were investigated using MD simulations with an ideal model (fig. S25). The detailed validation and parameter testing steps were shown in figs. S26 and S27. The simulation time was set as 10 ns, and the number of molecules through the membrane as a function of time was presented in Fig. 4C and fig. S28. It was observed that both H₂ and CO₂ molecules passed through the PCN membrane with defects rapidly, with a diffusivity ratio of 25:10 and a H₂/CO₂ selectivity of 2.5 at 0.2 ns. As the simulation time extended to 1 ns, the H₂/CO₂ selectivity decreased to 1.3, implying that it was difficult to achieve efficient molecular sieving capability for PCN membrane with the presence of surface defects. Gas molecules preferentially passed through the large defects rather than the intrinsic nanopores of the PCN. Meanwhile, almost no CO₂ molecule passed through the MXene/PCN nanosheet membranes, even when the simulation time was extended to 10 ns, while H₂ molecules easily passed through (Fig. 4D). The MXene/PCN nanosheet membranes demonstrated enhanced gas molecular sieving capability, indicating that MXene could effectively remedy the surface defects on the PCN nanosheets.

Gas separation performance of MXene/PCN nanosheet membranes

Gas permeation of different gas molecules through the MXene/PCN nanosheet membranes were investigated, as displayed in Fig. 5A. A sharp cutoff in gas permeance between H₂ and large gases (i.e., CO₂, N₂, and CH₄) was observed, indicating that the MXene/PCN nanosheet membranes achieved excellent H₂ molecule sieving capability (5). The MXene/PCN nanosheet membranes exhibited an H₂ permeance of 870 GPU and ideal selectivity of 30.21 for H₂/CO₂, 17.37 for H₂/N₂, and 14.26 for H₂/CH₄, outperforming the corresponding Knudsen selectivity (4.7, 3.7, and 2.8, respectively). Moreover, no obvious relationship was observed between gas permeance and molecule weight (fig. S29), suggesting that the gas transport mechanism through MXene/PCN nanosheet membranes is primarily governed by size exclusion rather the Knudsen diffusion (9). Note that although the CO₂ molecules demonstrated a lower molecular kinetic diameter compared with N₂ and CH₄ molecules, CO₂ permeance through the MXene/PCN nanosheet membrane was lower than that of N₂ and CH₄ molecules. The phenomenon was mainly attributed to the strong interaction between CO₂ molecules and active sites (e.g., ─NH/NH₂, C≡N in SPCN, and ─OH in MXene) on the MXene/PCN nanosheet membrane. The higher diffusion energy barriers for CO₂ molecule resulting from the strong interaction resulted in a higher H₂/CO₂ selectivity, and similar results have also reported in previous studies (38, 39).

Fig. 5. Gas separation performance of the MXene/PCN nanosheet membranes.

(A) Single gas permeance as a function of kinetic diameter at 25°C and 1 bar, and the insert is the gas separation selectivity for H₂ over other gases. Gas permeance and H₂/CO₂ separation selectivity as function of (B) MXene mass fraction, (C) membrane thickness, and (D) feed pressure at 25°C and 1 bar. (E) Long-term gas separation performance of MXene/PCN nanosheet membrane at 25°C and 1 bar. (F) Gas permeance and separation factors of the MXene/PCN membrane in mixed gas permeation at 25°C and 1 bar. The volume ratio of two gases in the feed side is 1:1, and the total gas flux is 50 ml min⁻¹.

The effects of various influencing factors (i.e., MXene size, MXene mass fraction, membrane thickness, and feed pressure) on the gas permeance and separation selectivity performance of MXene/PCN nanosheet membranes were systemically studied. The H₂ permeances were 896.4, 577.4, and 315.4 GPU with H₂/CO₂ selectivity of 31.3, 33.7, and 35.5, respectively, when the lateral sizes of MXene were 100, 200, and 500 nm, respectively (fig. S30). It must be noted that the increase of lateral size of MXene nanosheets would make impact on gas permeance, while the H₂/CO₂ selectivity slightly increased from 31.3 to 35.5. Considering the defect size on PCN nanosheets ranged from 10 to 50 nm [as confirmed by transmission electron microscopy (TEM) and SEM results], the lateral size of 100 nm for MXene nanosheets might be the optimal choice for repair defects. A larger lateral size (>100 nm) of MXene nanosheets would inevitably block the intrinsic nanopores on PCN nanosheets, resulting in low gas permeance. Moreover, the increase of MXene mass fraction, the obvious decrease in the gas permeance, and the increase in the H₂/CO₂ selectivity were observed, indicating the effective repair of defects on the PCN nanosheets. As the MXene mass fraction increased from 0 to 33.3%, H₂ permeance decreased from 10,874 to 870 GPU, and H₂/CO₂ selectivity increased from 5.14 to 30.21 (Fig. 5B). It was noted that the H₂ permeance remarkably decreased to 496.3 GPU, while the H₂/CO₂ selectivity slightly increased to 31.4 as the MXene mass fraction increased to 42.9%. This result implied that the excess addition of MXene might block the inherent nanopores within the PCN nanosheets and inhibit the gas transport. Such a low increase in the separation selectivity and remarkable decrease in gas permeance indicated that the MXene mass fraction of 33.3% was the optimal condition for PCN modification. In addition, MXene/PCN nanosheets membranes with different thickness were prepared by regulating the suspension concentration and volumes in the vacuum filtration process (fig. S31). As shown in Fig. 5C, MXene/PCN nanosheet membranes with a thickness of 40 nm demonstrated a high H₂ permeance of 13,487 GPU, but the H₂/CO₂ separation performance was 3.78, which might be the incomplete coverage of porous α-Al₂O₃ substrate by MXene/PCN nanosheets (40). Theoretically, the increase of membrane thickness could provide more size-sieving unit for gas separation, resulting in the increase of selectivity and decrease of gas permeance. As the membrane thickness increased from 40 to 1320 nm, the H₂/CO₂ selectivity obviously increased from 3.8 to 30.21, and the H₂ permeance decreased from 13,487 to 870 GPU. Further increase of the membrane thickness would inevitably result in a lower permeance, which is uneconomic to practical applications. Moreover, a gradual decrease in H₂/CO₂ selectivity was observed as the feed pressure increased from 1.0 to 2.0 bar (Fig. 5D), which was mainly attributed to the fact that the membrane structure constructed by nanosheets would contort under a high-pressure condition, resulting in the formation of nonselective nanochannels (10, 40).

Stability is a crucial factor in assessing the applicability and durability for practical industrial application. Therefore, the long-term stability of MXene/PCN nanosheet membrane for gas separation was investigated, as shown in Fig. 5E. It was observed that MXene/PCN nanosheet membrane demonstrated excellent stability and durability during the ~130-hour testing. The corresponding H₂/CO₂ selectivity slightly decreased from 33 to 29.3, and H₂ permeance decreased from 881.7 to 860.1 GPU during the long-term testing. Moreover, the MXene/PCN nanosheet membrane also demonstrated a superior long-term stability under humid conditions and elevated temperature (fig. S32). The result indicated that MXene/PCN nanosheet membrane demonstrated superior stability and promising potential in practical H₂ separation. To assess its practical applications in H₂ separation and purification, we also measured the gas permeance of the MXene/PCN nanosheet membrane in the mixture gas system (Fig. 5F), and the separation factors of 23.77 for H₂/CO₂, 13.80 for H₂/N₂, and 10.52 for H₂/CH₄ were obtained. The separation factors in the mixture gas system were slightly lower than the ideal selectivity in the single-gas system. This special phenomenon was mainly attributed to the competitive adsorption mechanism rather the Knudsen mechanism. Theoretically, the difference of separation selectivity between single-gas and mixed-gas systems caused by Knudsen diffusion was mainly due to the collision of gas molecules in the nanochannels. While this special result in this study was attributed to the fact that the large molecules (i.e., CO₂, N₂, and CH₄) might be adsorbed, the membrane surface weakens the surface adsorption of H₂ molecule, according to the decreasing H₂ permeance. Moreover, the large molecules near the nanopores would have a blocking effect that further decreased the H₂ permeance. It could be observed that the H₂ permeance through the MXene/PCN nanosheet membrane in the mixed-gas system demonstrated remarkable decrease compared with that in the single-gas system, which further confirmed our results. Moreover, our previous simulation results have confirmed that such nonpermeating component would make impacts on the permeation of permeating component and thus influence the gas separation performance (41). Similar experimental results were also found in the previous studies (11, 42). Moreover, the scaled-up MXene/PCN nanosheet membranes with different diameters were fabricated (fig. S33), and the large-area membrane also exhibited superior gas separation performance (fig. S34). Although the H₂/CO₂ separation selectivity slightly decreased after increasing the membrane area, the separation performance still far exceeded the Robeson upper bound for polymer membranes. This decrease may be caused by the formation of nonselective nanochannels and nonuniformity of the membrane structure related to the disordered stacking, due to the differences in driving force (e.g., pressure) between regions of the membrane (e.g., the center and edges) during the large-scale fabrication process by vacuum filtration. The aforementioned results demonstrated the promising potential of the MXene/PCN nanosheet membrane for large-scale applications.

Comparisons of the gas separation performance with other reported membranes

Figure 6 (A to C) presents the gas separation performance of MXene/PCN nanosheet membranes compared to other reported state-of-the-art membranes (tables S3 to S5). Notably, as a type of 2D material membrane, all of the as-prepared MXene/PCN nanosheet membranes in this study exhibit high H₂ permeance and considerable separation selectivity, exceeding the Robeson upper bound for polymer membrane. To reduce costs in industrial application of gas separation applications, increasing gas permeance is more crucial than achieving ultrahigh selectivity (see the “Economic analysis” section); meanwhile, higher selectivity can be compensated by using multistage membrane in series (42, 43). In contrast to other 2D material membranes (e.g., MXene, MOFs, COFs, and MoS₂), MXene/PCN nanosheet membranes demonstrated an ultrahigh H₂ permeance. Unlike traditional 2D lamellar membrane that used the interlayer nanochannels as the main sieving pathway, lamellar PCN nanosheet membranes with intrinsic nanopores on the nanosheet surface shorten the gas transport distance, thereby improving gas permeance. However, surface defects on these 2D porous membranes can compromise gas separation selectivity. Therefore, an easy and versatile approach for precise control of surface defects on the nanosheets is essential for advancing 2D nanoporous membranes in gas separation applications. The superior separation performance together with stability recommended the MXene/PCN nanosheet membranes as a high-efficiency alternative for precise H₂ separation and purification in hydrogen energy industry.

Fig. 6. Gas separation performance of MXene/PCN nanosheet membranes compared with other reported gas separation membranes.

(A) H₂/CO₂. (B) H₂/N₂. (C) H₂/CH₄. The membranes were classified on the basis of the type of the building blocks, and the detailed references and data were provided in tables S3 to S5. PBI, polybenzimidazole; PIM, polymers of intrinsic microporosity; PEI, polyetherimide; CMS, carbon molecular sieves.

Actually, several defect control and repair methods have been developed in the previous studies such as the chemical vapor deposition (CVD). In a CVD method for the preparation of PCN nanosheet membrane, the N- and C-rich precursors (e.g., melamine and cyanamide) underwent decomposition, migration, and surface chemical reaction under high-temperature conditions and directly formed a PCN membrane (44). However, a CVD method for fabrication was mainly restricted by the complex condition, poor uniformity, and large-scale fabrication. First, it must be mentioned that the selection of substrate would make important impacts on the growth of PCN membrane. The substrate with low surface energy would reduce the nucleation site and thus result in a discontinuous membrane and uneven membrane thickness, while a high surface energy might trigger the agglomeration and increase the nonselective nanochannels for gas transport. The current PCN membrane prepared by the CVD method was limited in size to the level of the laboratory (<10 cm²), making it difficult to meet the large-scale requirements for industry application. Moreover, Zhou et al. (42) developed a novel bottom-up strategy for the one-step fabrication of PCN nanosheets without visible defects. Although superior separation performance was achieved with this PCN nanosheet membrane, the yields of nanosheets were extremely low (2.2-g precursor yielding 6- to 8-mg products), remaining a large-scale preparation challenge. Moreover, polymers such as poly(dimethylsiloxane) and polyether block amide were also used as filler materials to spin-coat onto the prepared membrane surface for filling defects and cracks (7, 10, 45, 46). However, this posttreatment modification using polymer was mainly applied to repair defects on the substrate due to incomplete coverage by a small amount of 2D nanosheets, particularly in ultrathin nanosheet membrane. Furthermore, additional posttreatment procedures might inevitably increase preparation cost. In contrast, the defect repair of PCN nanosheet in this study was achieved by ordered stacking of MXene/PCN nanosheets in the vacuum filtration process. This approach can greatly simplify the preparation procedure without posttreatment modification. Actually, current researches on 2D nanosheet membranes for gas separation remain at the laboratory stage with a small size, and their further development for practical applications is mainly limited by the scaled-up fabrication. The most significant issue brought by the large-scale membrane is the remarkable decrease in the gas separation performance. Therefore, the fabrication of large-area 2D nanosheet membrane must overcome a series of engineering challenges such as the nonuniformity of membrane structure and the formation of defects. Developing feasible scaled-up fabrication technology and regulation method is a prerequisite for realizing the practical gas separation application of 2D nanosheet membranes (47, 48). In this study, although the gas separation selectivity of MXene/PCN nanosheet membranes slightly decreased after increasing the membrane area (effective diameter of 60 mm), the separation performance still far exceeded the Robeson upper bound for polymer membranes, demonstrating the promising potential for large-scale applications.

Economic analysis

Economic analysis is essential for assessing the cost-effectiveness and feasibility of membrane gas separation. It must be mentioned that the main purpose of the economic analysis is not to provide detailed engineering designs or an exact cost estimation for a specific plant but rather to conduct a preliminary and comparative assessment of the potential economic feasibility of different membranes. This implies that these performance data (permeance and selectivity) obtained from the experimental measurement could be used as the input parameters for further calculation of economic analysis although the testing apparatus in the experiments and cross-flow mode used in economic calculations were different. Such cross-flow mode was considered the cornerstone that could realistically predict the separation performance of membrane module (fig. S35) (49), which has been commonly used for the calculation of energy consumption for evaluating the applicability of the membrane materials in the previous studies (50–53). Moreover, the membrane in the cross-flow mode for economic calculation referred to a separation system rather than a single membrane, and therefore, there are no strict requirements for the type and number of membranes in this system. It was noted that multistage membranes were commonly adopted for the practical applications due to the limitation of a single membrane. On the basis of the economic analysis, some key parameters that made most significant impact on the economic cost of the gas separation process could be obtained, further guiding the development of high-performance membrane materials. In addition, $Q_{\mathrm{y}}$ is the flow rate in the permeate side and is related to the $Q_{\text{in}}$ (flow rate in feed side), $R_1$ (recovery rate), and $y$ (molar fraction of H₂ in the permeate side). Namely, $Q_{\mathrm{y}}$ represents the gas processing capacity of a gas separation system, and it was assumed to be 50 kg s⁻¹ in this study, as adopted in the literature to perform economic analysis of gas separation membranes (51). The economic consumption (EC) depends on the gas separation selectivity, and the membrane area is determined by both selectivity and permeance. Note that the EC value decreased sharply from 7.81 to ~2 GJ/t as the gas separation selectivity increased from 7.6 to ~17.7 (fig. S36A); however, further increase in the gas separation selectivity above 17.7 produced negligible EC reduction. Moreover, increasing selectivity with reduced permeance could increase the required membrane area and preparation costs (fig. S36B). From this perspective, membrane materials with ultrahigh selectivity but very low permeance are uneconomical in practical gas separation processes, and an optimal balance between selectivity and permeance maximizes overall efficiency. As shown in Fig. 7A, the MXene/PCN nanosheet membrane achieved EC values of 1.62 to 1.96 GJ/t. Compared with the traditional gas separation method (e.g., PSA, EC value of 4.4 to 5.6 GJ/t) (54), the gas separation based on the MXene/PCN nanosheet membrane could reduce the economic cost. Notably, it must be pointed out that MXene/PCN nanosheet membrane required the smallest membrane area as low as ~554,313.5 m² compared with the current state-of-the-art membranes (Fig. 7B), owing to its ultrahigh molecular permeance and moderate selectivity. These low EC values and membrane area requirements demonstrate the practical viability of MXene/PCN nanosheet membranes and their potential for H₂ separation.

Fig. 7. Economic analysis.

(A) Energy consumption and required membrane area of MXene/PCN nanosheet membranes compared with other reported gas separation membranes, and the insert was the schematic illustration of the cross-flow mode for economic analysis. (B) Enlarged image of the target area for energy consumption and required membrane area. Detailed references and data were provided in table S6.

DISCUSSION

This study presents the construction of 2D lamellar nanosheet membranes with inherent in-plane nanopores and proposes a defect repair strategy of nanosheet membrane for H₂ separation. Strong interfacial interaction between MXene and PCN promoted the ordered stacking of nanosheets and repaired defects on the PCN nanosheets. Experimental results and MD simulations highlighted the importance of defect control in 2D nanoporous nanosheet membranes and clarified the key role of MXene in repairing defects on PCN nanosheets during the H₂ separation process. The MXene/PCN nanosheet membrane achieved an ultrahigh H₂ permeance of 870 to 8046 GPU and considerable selectivity, thanks to the well-defined pore size of PCN nanosheets and efficient defect repair by MXene. This performance surpassed that of most existing state-of-the-art membranes. Gas transport through the MXene/PCN nanosheet membrane was dominated by size exclusion rather than Knudsen diffusion. Moreover, MXene/PCN nanosheet membranes maintained the excellent stability and separation performance during a continuous test lasting over 130 hours. Economic analysis indicated that MXene/PCN nanosheet membranes exhibited low compression energy consumption and minimal membrane area demands for practical H₂ separation. This study demonstrated the potential of MXene/PCN nanosheet membranes for practical H₂ separation applications.

MATERIALS AND METHODS

Materials and reagents

Melamine (99%) was purchased from Aladdin Chemical Reagent Co. Ltd. (Shanghai, China). IPA (≥99.5%) was supplied by Macklin Chemical Reagent Co. Ltd. (Shanghai, China). Ti₃C₂-MXene suspension (5 mg ml⁻¹) was obtained from Nanjing Muke Nanotechnology Co. Ltd. (Nanjing, China). Porous ceramic substrates (pore size of 100 nm and diameter of 18 mm) were provided by Puyuan Nanotechnology Co. Ltd. (Hefei, China). The poly(ether sulfone) (PES) substrates (pore size of 100 nm and diameter of 25 mm) were obtained from Longjin Membrane Technology Co. Ltd. (Nantong, China). All materials and reagents were used as received directly without additional purification.

Preparation of bulk PCN powder and PCN nanosheets

PCN powder was synthesized using a traditional thermal polymerization method, and the procedure is outlined below. One gram of melamine was placed in a ceramic crucible with a lid and heated in a tubular furnace at 520°C for 6.5 hours with a heating rate of 2°C min⁻¹ under an air atmosphere. The resulting bulk PCN powder was further ground for at least 15 min to obtain a fine powder (fig. S37). The PCN nanosheet suspension was prepared via a liquid ultrasonic exfoliation method, with mixed solvents used to accelerate the intercalation and exfoliation of bulk PCN (28). Briefly, 1 mg of bulk PCN powder was dispersed in 50 ml of a mixed IPA-H₂O (volume ratio of 1:1) solution and then subjected to ultrasonication for 2 hours to obtain a PCN nanosheet suspension (0.02 mg ml⁻¹).

Preparation of MXene/PCN nanosheet membranes

The MXene suspension (5 mg ml⁻¹) was diluted to 0.02 mg ml⁻¹ with deionized water and then subjected to ultrasonication at a temperature below 5°C for 10 min under a nitrogen atmosphere. Subsequently, different volumes of MXene suspension (0.02 mg ml⁻¹) were directly added to PCN nanosheet suspension (0.02 mg ml⁻¹), followed by ultrasonication at a temperature above 5°C for 10 min under a nitrogen atmosphere to obtain mixed MXene/PCN solutions with varying concentrations. MXene/PCN nanosheet membranes were then prepared using a vacuum filtration method by filtering a specified volume of the MXene/PCN mixed solution onto porous α-Al₂O₃/PES substrate. Similarly, pristine PCN nanosheet membranes were fabricated without the addition of MXene. Last, the obtained pristine PCN and MXene/PCN nanosheet membranes were dried in a vacuum oven at room temperature for at least 12 hours.

Characterization of the MXene/PCN nanosheets and membranes

Field emission SEM (ZEISS GeminiSEM 500, Germany) coupled with EDS (Oxford UltimMax100, UK) was performed to investigate the micromorphology and elemental distribution of prepared membranes. Crystal structures were identified by XRD (Shimadzu XRD-6100, Japan) with Cu Kα radiation (λ = 1.54 Å), and the detailed parameters for test were as follows. The scan rate was 5° min⁻¹; the 2θ range was 5° to 40°; the voltage and current were 40 kV and 40 mA, respectively. TEM (Thermo Talos L120C G2, UK) and AFM (Shimadzu SPM-9700HT, Japan) were conducted to observe the microstructures and defects of the nanosheets. The elemental composition and functional groups were explored by XPS (Thermo ESCALAB Xi+, UK) analyses and FTIR spectra (Thermo Nicolet Iz10, UK), respectively. The XPS peaks were corrected on the basis of C 1s peak (binding energy at 284.8 eV). N₂ adsorption-desorption measurements were conducted at 77 K using Micromeritics instrument (ASAP 2020 Plus HD88, USA), and the corresponding pore distribution was calculated on the basis of the Barret-Joyner-Halenda method.

Measurements of gas permeation rates

Gas permeation measurements of the prepared membranes were conducted using a homemade Wicke-Kallenbach module. Two seal rings were used to improve the leak tightness of the system and prevent the contact between membrane surface and the stainless steel module. Various gases (i.e., H₂, N₂, CO₂, and CH₄) were used as the feed gas, with Ar chosen as the sweep gas. The gas flow rates and the pressures on both the feed and permeate sides were regulated by mass flow controllers (Sevenstar, China) and a back pressure valve, respectively. Besides, the gas flow rate on the permeate side was measured using a bubble flowmeter. In single-gas permeation experiments, both the feed and sweep gas flow rate were set to 50 ml min⁻¹. In mixed-gas permeation experiments, two different gases were used as the feed, with each gas flowing at 25 ml min⁻¹. The gas compositions on the permeate side were analyzed using gas chromatography (FuLi GC9790II, China). Gas permeation experiments were carried out at 25°C.

Gas permeance ( $P_{\mathrm{i}}$ ) through the membrane was calculated by the following equation

$$P_{\mathrm{i}}=\frac{N_{\mathrm{i}}}{A\mathrm{\Delta }{P}_{\mathrm{i}}}$$ (1)

where $P_{\mathrm{i}}$ (mol m⁻² s⁻¹ Pa⁻¹) is the gas permeance of component i through the membrane, $N_{\mathrm{i}}$ (mol s⁻¹) is the flow rate of component i, A (m²) is the effective membrane area for a gas permeation test, and $\mathrm{\Delta }{P}_{\mathrm{i}}$ (Pa) is the transmembrane pressure difference of component i. In this work, gas permeance of the membrane was compared using the unit of GPU (55).

$$1 \text{GPU}=3.3928\times {10}^{-10} \text{mol} {\mathrm{m}}^{-2} {\mathrm{s}}^{-1} {\text{Pa}}^{-1}$$ (2)

The ideal selectivity ( $S_{\mathrm{i}/\mathrm{j}}$ ) of two components in single-gas permeation experiment was calculated by the following equation

$$S_{\mathrm{i}/\mathrm{j}}=\frac{P_{\mathrm{i}}}{P_{\mathrm{j}}}$$ (3)

where $P_{\mathrm{i}}$ and $P_{\mathrm{j}}$ were the permeance for gas components i and j through the membrane, respectively.

The separation factor ( ${\mathrm{\alpha }}_{\mathrm{i}/\mathrm{j}}$ ) of two components in mixed-gas permeation experiment was calculated by the following equation

$${\mathrm{\alpha }}_{\mathrm{i}/\mathrm{j}}=\frac{y_{\mathrm{i}}/y_{\mathrm{j}}}{x_{\mathrm{i}}/x_{\mathrm{j}}}$$ (4)

where $x_{\mathrm{i}}$ and $y_{\mathrm{i}}$ are the molar fraction for gas component i on the feed side and permeate side, respectively, while $x_{\mathrm{j}}$ and $y_{\mathrm{j}}$ were the molar fraction for gas component j on the feed side and permeate side, respectively.

MD simulations

MD simulations were conducted to investigate the nonbonded interaction (i.e., van der Waals and electrostatic interactions) energies between MXene and PCN nanosheet, as well as the permeation of mixed gases (H₂ and CO₂) through pristine PCN and MXene/PCN nanosheet membranes. All the MD simulations were performed using LAMMPS software. Van der Waals interaction was modeled using the Lennard-Jones 12-6 potential and electrostatic interactions using the Ewald summation method. Interaction parameters between hybrid atoms were determined using the Lorentz-Berthelot mixing rules. In the gas permeation simulations, the membrane was placed at the center of the simulation box. Sixty H₂ molecules and sixty CO₂ molecules were added to the feed side of the membrane, while the permeate side was initially set to be a vacuum. A piston was placed above the feed gas to control the gas pressure by applying driving forces. Periodic boundary conditions were applied in the x and y directions, and reflective boundary conditions were applied in the z direction. The intramolecular bond and bond angle bending energies of the gas molecules were modeled using a harmonic potential. The PCN, CO₂, and H₂ molecules were modeled using the Chemistry at HARvard Macromolecular Mechanics (CHARMM) force field (56), while MXene was described by the universal force field (57) with charge equilibration (QEq) method (58). A cutoff radius of 1 nm was applied for the calculation of atomic interactions. The simulations were performed using the canonical (NVT) ensemble, maintaining the temperature at 300 K with a Nosé-Hoover thermostat. The time step for the simulations was 0.5 fs.

Supplementary Materials

This PDF file includes:

Figs. S1 to S37

Tables S1 to S6

References