Graphene quantum dot membranes with tailorable pores for efficient gas separation

Abstract

Membrane materials and fabrication methods often need to be co-developed to meet the specific permeance and selectivity requirements of different gas separation application, which limits the adaptability of membrane technology. Here we propose a post-regulation strategy—adjusting the pore structures of a standard “primitive” membrane to meet various separation needs—that greatly simplifies membrane preparation and accelerates technological advancement. Graphene quantum dots (GQDs) are introduced as innovative building blocks: a continuous GQD membrane is constructed by tightly stacking GQDs, and then its pore structure is tuned via heat treatment and in-situ cross-linking with small amine molecules. Combining adjustable angstrom-scale pores with preferential CO₂ adsorption, the resulting GQD membranes exhibit widely tunable CO₂/N₂ and CO₂/CH₄ separation performances. The CO₂ permeance and separation factors surpass most reported membranes and can be tuned to exceed industrial targets for CO₂ capture. By varying the heat treatment temperature, the separation scope of the membrane is further extended to challenging gas pairs such as C₃H₆/C₃H₈, demonstrating the high potential of this customizable post-regulation pore structure strategy.

By post-tuning the GQD primitive membrane through heat treatment and in-situ crosslinking of small molecule amines, the as-prepared membrane achieved excellent CO₂ separation performance and was further extended to the C₃H₆/C₃H₈ separation system.

Introduction

Membrane-based gas separation technology is highly efficient and energy-saving, making it a promising alternative to traditional separation methods, such as distillation and adsorption^1–3. As the core of this technology, membrane materials have been extensively studied to achieve ultra-high permeance and selectivity. Advanced microporous materials, such as zeolites and metal-organic frameworks (MOFs) exhibit significant advantages in molecular sieving due to well-defined angstrom-scale pores^4–8. For example, poly[Zn₂(benzimidazole)₄]^9–11 and zeolitic imidazolate framework-8 (ZIF-8)^12–16 membranes have shown excellent H₂/CO₂ and C₃H₆/C₃H₈ separation performance, respectively, by virtue pore sizes that ideally match the target molecules.

However, the molecular sieving capability of these porous materials, which is primary advantage, is also their fundamental limitation, as the separation efficacy is critically dependent on the intrinsic pore size. This necessitates searching for materials with pore sizes lying between the kinetic diameters of specific gas pairs and developing corresponding membrane preparation technologies, limiting the materials available for separation. Consequently, developing standardized primitive membranes with adjustable pore structures to meet specific permeance and selectivity requirements would greatly simplify membrane manufacturing process and facilitate commercial developments. This strategy circumvents the need to rebuild membranes from the ground up for each separation scenario and performance target. To make gas transport independent of the intrinsic pore structure of building blocks, the building block particles should be extremely small so that the bulk structure has minimal impact on the assembled membrane. In this context, zero-dimensional quantum dots are highly appealing for membrane construction.

Among them, emerging graphene quantum dots (GQDs, size <10 nm) possess stable graphitized carbon cores and abundant surface functional groups, enabling them to tightly self-assemble into three-dimensional networks through cross-linking chemical reactions^17–20. Thus, GQDs are promising building blocks for assembling dense, defect-free membranes for gas separation. GQDs self-assembly is driven by their surface functionality, forming networks through covalent bonding, hydrogen bonding and electrostatic interactions among GQDs or between GQDs and cross-linkers^21–23. For example, amino-functionalized GQDs are typically cross-linked with trimesoyl chloride (TMC) to engineer controlled nanochannels, successfully enhancing separation performance in nanofiltration membranes^24,25. In addition, ultramicropores have been observed in microporous carbons derived from thermal annealing of GQDs with KOH activation, where the annealing induces “chemical welding” via C-O-C bonds formation²⁶. Therefore, regulating the crosslinking process is a key process for achieving GQD membranes with tunable pore sizes. However, in existing studies, GQDs were mainly used as nanofillers to regulate the structure of the polymer continuous phase, and gas separation membranes utilizing GQDs as the main skeleton have rarely been reported.

Herein, we present a strategy to prepare a continuous GQD primitive membrane via thermal spraying, followed by pore-size regulation through molecular cross-linking using pyrolyzed polyethyleneimine (PEI). The thermally sprayed GQD primitive membrane is calcined under argon while simultaneously being cross-linked with PEI decomposition products to obtain a post-treated GQD (P-GQD) membrane. Benefiting from the pore size sieving effect of the regulated pores and the strong CO₂ affinity of introduced surface groups, the P-GQD membrane exhibits CO₂-selective permeation with permeance and selectivity tunable only by adjusting the cross-linking temperature. Moreover, at higher cross-linking temperatures, the P-GQD membrane can selectively separate C₃H₆/C₃H₈ mixtures. This small-molecule cross-linking approach provides a feasible path to prepare GQD membranes with flexible, post-fabrication regulation of pore structure, enabling customization for a variety of separation targets.

Results

Preparation and characterization of GQDs

Hydrophilic GQDs were synthesized by direct pyrolysis of citric acid (CA)²⁷ at 210 °C for 0.5 h, as shown in Fig. 1a. Freeze-drying was used to avoid destroying the surface groups of GQDs during drying (Fig. 1b). A distinct Tyndall effect was observed in the aqueous GQDs sol (Fig. 1b right), indicating that the as-prepared GQDs are homogeneously dispersed in water. GQDs exhibit two key structural characteristics: (1) a stable graphitized carbon core, and (2) an abundance of oxygen-containing functional groups (-OH, -COOH, etc.) on the surface. In Fig. 1c, the high-resolution transmission electron microscopy (HR-TEM) image shows that the GQDs have irregular fragment morphologies with lateral dimensions less than 10 nm. Clear lattice fringes are observed with a lattice spacing of 0.32 nm, corresponding to the graphite (002) plane²⁸. The image of atomic force microscopy (AFM) in Fig. 1d further indicates GQD thicknesses on the order of 3–5 nm, consistent with the HR-TEM results. Dynamic light scattering (DLS) spectrum of the aqueous GQDs sol displays a single dispersed peak of ~ 5 nm, confirming the uniformity of GQD nanoparticles (Supplementary Fig. 1).

Fig. 1. Preparation and characterization of GQDs.

a Schematic illustration of GQD synthesis from citric acid. b Digital photos of GQD powder (left) and aqueous sol (right); the sol exhibits a clear Tyndall effect. c HR-TEM image of GQDs showing irregular <10 nm fragments with lattice fringes (0.32 nm spacing). d AFM image of GQDs with height profile along the white dotted line, indicating a thickness of ~3–5 nm. e Photoluminescence spectrum of an aqueous GQD sol (excitation at 350 nm). The a. u. represents arbitrary units. f ATR-FTIR spectrum of GQDs showing various surface functional groups. g, h High-resolution XPS C1s (g) and O1s (h) spectra of GQDs. Source data are provided as a Source Data file.

The structure-induced absorption and emission spectral characteristics of GQDs were examined by UV-visible and photoluminescence (PL) spectroscopy. The UV-vis absorption spectrum (Supplementary Fig. 2) of GQDs displays broad bands in the near-UV region, which is the characteristic of surface functional groups and carbon core of GQDs²⁹. Under 350 nm light excitation, the GQD solution shows a distinct fluorescence emission peak in the PL spectrum (Fig. 1e), attributed to the influence of surface functional groups^30,31. The surface functional groups of GQDs were further identified by attenuated total reflectance-Fourier-transform infrared spectroscopy (ATR-FTIR), as shown in Fig. 1f. Peaks at 3035 cm⁻¹ and 1406 cm⁻¹ correspond to -OH stretching and bending vibration³². Peaks at 1703 cm⁻¹ and 1177 cm⁻¹ are attributed to C = O and C-O stretching vibrations of -COOH, respectively³³. In addition, peaks at 2645 cm⁻¹ and 905 cm⁻¹ correspond to C-H and C = C vibration³⁴. X-ray photoelectron spectroscopy (XPS) analysis of C 1 s and O 1 s spectra (Fig. 1g, h and Supplementary Fig. 3) further confirms the abundance of oxygen-containing groups on the GQD surfaces. The proportions of C = O, C-O and C = C groups are calculated to be 23.62 %, 12.57 % and 63.81 %, respectively (Supplementary Table 1). These functional groups facilitate cross-linking between GQDs, making them attractive building blocks for separation membranes.

Preparation and characterization of P-GQD membranes cross-linked by PEI pyrolysis

As shown in Fig. 2a, a “primitive” GQD membrane was obtained by spraying the GQD sol onto a porous α-Al₂O₃ support. To prevent the small GQD nanoparticles from infiltrating into the macroporous support, which would impede gas transport and prevent formation of a surface layer, a thermal spraying method was used: the Al₂O₃ support was heated to 180 °C so that the solvent evaporated rapidly upon contact, thereby minimizing the penetration of GQDs and maintaining the substrate’s original permeability. At this optimized support temperature (180 °C, see Supplementary Fig. 4), GQDs remain on the surface. For the as-sprayed primitive GQD membrane, the proportion of C = O is almost constant, and the proportion of C-O increases, indicating that the carboxyl and hydroxyl groups are cross-linked (Supplementary Table 1). After spraying a total volume of 2 mL GQD sol, the particles stack into a thin continuous membrane, which can more effectively reduce gas transport resistance compared to thicker membranes (Supplementary Fig. 5). Moreover, the employed spray-coating is a scalable industrial process, which suggests its potential for the deposition of primitive GQD membranes onto large-area substrates with high reproducibility.

Fig. 2. Preparation and characterization of the P-GQD membrane.

a Schematic of P-GQD membrane fabrication process. b Thermogravimetric analysis (TGA) curves of GQDs and PEI under N₂. c,d Top-view and cross-sectional SEM images of the P-GQD membrane with corresponding EDS elemental mapping. e XPS of C 1s spectra of the sprayed GQD, S-GQD, and P-GQD membrane. f Raman spectra of the sprayed GQD, S-GQD and P-GQD membrane. g Water contact angles of sprayed GQD, S-GQD and P-GQD membrane. Source data are provided as a Source Data file.

To enable selective gas transport across the membrane, either the pore size at angstrom level and/or preferential adsorption sites for specific gases need to be rationally regulated. The as-prepared GQD primitive membrane is initially gastight after thermal spraying at 180 °C, so a calcination step is required to create pores. We employed the pyrolysis of PEI to introduce abundant small amine molecules (e.g., diethylenetriamine and ethylenediamine)³⁵ that can cross-link with the carboxyl groups on the GQD surface, thereby regulating the pore size of the membrane. Additionally, the introduction of amino groups, which have high CO₂ affinity, should enhance CO₂ selectivity^36–38. Thermogravimetric analysis (TGA, Fig. 2b) shows that PEI begins to pyrolyze at 280 °C under inert atmosphere, releasing small amines that can cross-link with the GQDs framework. Above 330 °C, the weight loss of GQDs slows markedly, indicating a transition from rapid volatilization of unstable groups to slower removal of more stable groups. Based on these results, 350 °C was chosen as an initial trial temperature to ensure efficient cross-linking between PEI pyrolysis products and the GQD primitive membrane. The GQD membrane obtained via PEI pyrolysis cross-linking is hereafter designated as P-GQD.

The evolution of P-GQD membrane surface groups with calcination temperature was probed by quasi-in-situ infrared spectroscopy (Supplementary Fig. 6). Below 230 °C, neither GQDs nor PEI pyrolyze, and thus there is no significant change in functional groups. At 280 °C, the carboxyl and hydroxyl group of GQDs react with each other to form esters. After PEI starts to pyrolyze, the amino groups of small molecule amines react with the carboxyl groups of GQDs to form amide. Simultaneously, generation of -CN and enhanced adsorption of CO₂ further indicate that small amines from PEI have chemically cross-linked onto the GQDs. As calcination temperature further increases, the content of O, N-containing functional groups has decreased, which is due to the continuous thermal decomposition of the cross-linked P-GQD membrane, causing the loss of partial elements.

The surface and cross-sectional morphologies of the P-GQD membrane were observed by a high-resolution SEM with energy-dispersive X-ray spectroscopy (EDS). As shown in Fig. 2c, a continuous, dense P-GQD membrane is formed on the alumina support, accompanied by a uniform distribution of N (Supplementary Fig. 7). This confirms that the GQD membrane was uniformly cross-linked by PEI pyrolysis products. To illustrate the role of PEI, we also prepared a control membrane (S-GQD) by calcining the single GQD membrane under Ar in the absence of PEI. In contrast to P-GQD membrane, the S-GQD membrane has a rougher surface showing only C and O from GQDs and Al from the Al₂O₃ support in EDS, with no N detected (Supplementary Fig. 8). Cross-sectional EDS mapping (Fig. 2d) clearly shows enrichment of C and N in the P-GQD membrane layer on the Al₂O₃ support surface and no significant penetration of GQDs into the support, indicating that the P-GQD membrane remains on top of the support. The uniformity of the prepared P-GQD membrane was further demonstrated by the enrichment of C and N elements on the support. The outline of alumina particles of the support is still visible (Fig. 2c), and Al signals appear in the XPS survey (Supplementary Fig. 9), suggesting the thickness of P-GQD membrane is ultrathin, which likely contributes to its high gas permeance.

The cross-linking by PEI pyrolysis was further verified by XPS analysis of nitrogen functionalities. The N 1 s spectrum of P-GQD membrane (Supplementary Fig. 10) shows a peak attributable to C-N bonds in aromatic rings and secondary amine (C-NH-C) bonds, as shown in Supplementary Fig. 11³⁹. Changes in the C 1 s spectra (Fig. 2e and Supplementary Table 1) also reflect the introduction of PEI pyrolysis products during the preparation of the P-GQD membrane. The C 1 s peaks at 284.8, 286.0, and 288.8 eV correspond to C = C, C-O/C-N, and C = O bonds, respectively⁴⁰. After calcination, the intensity of the C = O peak decreases in both S-GQD and P-GQD membrane, indicating consumption of carboxyl groups via cross-linking. In the S-GQD (no PEI) membrane, a decrease in the C-O peak is also observed, due to partial volatilization of -COOH/-OH groups upon heating. In the P-GQD membrane, however, the 286.0 eV peak (C-O/C-N) is enhanced, reflecting the formation of new C-N bonds from cross-linking (in addition to residual C-O).

Raman spectroscopy (Fig. 2f) was used to analyze the carbon microstructure of the P-GQD membrane before and after calcination. Two characteristic peaks appear at ~1370 cm⁻¹ (D-band, disordered carbon) and ~1605 cm⁻¹ (G-band, graphitic carbon)^41,42. The as-sprayed primitive GQD membrane exhibits a strong fluorescence background, making its D/G bands difficult to detect⁴³; a faint G-band is visible, consistent with the high graphitization of GQDs. In both S-GQD and P-GQD membranes, pronounced D and G bands are observed. Notably, the P-GQD membrane shows a relatively stronger D-band (weaker G-band) compared to S-GQD membrane, indicating that the incorporation of PEI-derived alkyl chains increases the disorder in the carbon framework. XRD and HRTEM were also used to characterize the structural features of the P-GQD membrane, showing that the graphitic structure of GQD particles is preserved within the P-GQD membrane, and that interparticle stacking builds transport pathways (Supplementary Figs. 12 and 13).

Water contact angle measurements further confirmed the effect of PEI cross-linking on P-GQD membrane (Fig. 2g). The as-sprayed GQD primitive membrane is highly hydrophilic (low contact angle) due to abundant surface carboxyl and hydroxyl groups. After heat treatment without PEI (S-GQD), some of these groups are lost, resulting in reduced hydrophilicity of S-GQD membrane. In contrast, the P-GQD membrane surface becomes hydrophobic, owing to the hydrophobic alkyl chains introduced by cross-linking and the substantial reduction of carboxyl and hydroxyl groups. This enhanced hydrophobicity should help prevent moisture from disrupting the cross-linked network, thereby improving P-GQD membrane stability. Consistently, zeta potential measurements (Supplementary Fig. 14) show that pristine GQDs carry a significant negative charge (from deprotonated –COOH), while the zeta potential of P-GQD membrane is much less negative due to the neutralization of carboxylates by the cross-linked amines⁴⁴.

Gas separation performance of P-GQD membranes

The gas separation performance of P-GQD membranes was evaluated by the Wicke-Kallenbach method for equimolar binary gas mixtures. The control S-GQD membrane shows low CO₂/N₂ separation factor (Supplementary Fig. 15). In contrast, P-GQD membranes exhibit greatly enhanced CO₂ separation due to the PEI-derived cross-linked structure. As shown in Supplementary Fig. 16, for a P-GQD membrane prepared using 2 mL GQD sol and 6 mg PEI, the CO₂ permeance increases with higher calcination temperature, while the mixed-gas CO₂/N₂ and CO₂/CH₄ separation factor decrease. This trend is attributed to more C, N, O being volatilized at higher temperatures, enlarging pore aperture and reducing steric hindrance to gas transport. Overall, the P-GQD membrane calcined at 350 °C exhibits the best performance, with a CO₂ permeance of ~1504 GPU and a CO₂/N₂ separation factor of ~43, along with a CO₂/CH₄ separation factor of ~51.

To further probe permeation behavior, we varied the feed pressure in separation tests. As shown in Fig. 3a, b, both the CO₂ permeance and separation factors of the P-GQD membrane decline as feed pressure increases, while the permeance of CH₄ increases. This phenomenon is likely because the elevated pressure increases the CH₄ loading in the membrane, thereby promoting CH₄ permeation and hindering CO₂ diffusion⁴⁵. Even at 0.6 MPa feed pressure, the CO₂/N₂ and CO₂/CH₄ separation factors remain above 20, demonstrating robust high-pressure separation performance. Moreover, when the feed pressure is returned to 0.1 MPa, the CO₂ permeance and separation factors recover to their initial high values, indicating that the membrane structure is not harmed by the pressure cycling.

Fig. 3. CO₂ separation performance of P-GQD membranes.

a, b CO₂/N₂ (a) and CO₂/CH₄ (b) separation performance of P-GQD membrane as a function of feed pressure. The colored regions represent the feed pressure returning to 0.1 MPa. c 25 h operational stability test for CO₂/N₂ separation (equimolar feed). d Long-term stability of P-GQD membrane for CO₂/CH₄ separation under cyclic feed pressure swings. e, f Comparison of P-GQD performance (red stars) with other reported membranes for CO₂/N₂ (e) and CO₂/CH₄ (f) separations (data from Supplementary Tables 3 and 4). The colored regions represent the target areas for industrial CO₂ separation. Source data are provided as a Source Data file.

We also evaluated the P-GQD membrane under different feed compositions. As shown in Supplementary Fig. 17, altering the CO₂ content in the CO₂/N₂ or CO₂/CH₄ feed has minimal effect on CO₂ permeance and separation factors, demonstrating that the P-GQD membrane offers high separation efficiency in separation scenarios with different CO₂ contents. Increasing the test temperature accelerates gas diffusion⁴⁶ and thus raises the CO₂ permeance (Supplementary Fig. 18). However, the CO₂/N₂ and CO₂/CH₄ separation factors decrease at higher temperature because the adsorption affinity for CO₂ is reduced. The performance of the P-GQD membrane was also evaluated under humid conditions. When exposed to humid feed gas at 60 °C, the P-GQD membrane maintains moderate permeance and selectivity, with performance recovery observed after returning to dry gas at room temperature (Supplementary Fig. 19).

The P-GQD membrane shows excellent stability over time. Over a 25 h continuous test with an equimolar CO₂/N₂ mixture, the CO₂ permeance remains nearly constant and the CO₂/N₂ separation factor stays around 50 (Fig. 3c). Over longer-term tests totaling 100 h for CO₂/CH₄ and CO₂/N₂ separation with periodic feed pressure cycling, the CO₂ permeance shows only a slight decline while the separation factor slightly increases (Fig. 3d and Supplementary Fig. 20). Under actual industrial conditions involving fouling and various gas impurities, the durability of P-GQD membrane needs to be further investigated. Notably, CO₂/N₂ separation performance of P-GQD membrane meets the industrial requirements for post-combustion CO₂ capture (CO₂ permeance ≥ 1000 GPU and CO₂/N₂ selectivity ≥ 20)⁴⁷, and its CO₂/CH₄ separation performance far exceeds that of commercial natural gas separation membranes (typically CO₂ permeance > 100 GPU, CO₂/CH₄ selectivity > 30)³⁷, as illustrated in Fig. 3e, f. Importantly, by simply adjusting the calcination temperature, the permeance and selectivity of P-GQD membrane can be tuned over a wide range for both CO₂/N₂ and CO₂/CH₄ separation. For example, increasing the calcination temperature from 320 °C to 360 °C adjust the CO₂ permeance from 120 to 1975 GPU, while the corresponding CO₂/CH₄ separation factor decreases from 138 to 28. This flexibility enables convenient customization of membrane performance for different separation scenarios.

Gas separation mechanism of P-GQD membrane

The P-GQD membrane has high CO₂ permeability and selectivity, mainly originating from two key effects: preferential adsorption of CO₂ and size-sieving by the regulated ultra-micropores (Fig. 4a). Since PEI contains abundant amino groups, the cross-linking of PEI pyrolysis products effectively enhances the CO₂ affinity of the membrane, leading to selective CO₂ adsorption and transport. Accordingly, compared with the GQD primitive membrane and the S-GQD membrane, the P-GQD membrane shows a distinct FTIR absorption band for adsorbed CO₂ at 2340 and 2360 cm⁻¹ spectrum (Fig. 4b).

Fig. 4. Mechanistic analysis of CO₂ separation by P-GQD membrane.

a Schematic illustration of the two key separation mechanisms in P-GQD membrane: preferential CO₂ adsorption and molecular sieving. b FTIR spectra of the sprayed GQD, S-GQD and P-GQD membranes, highlighting the emergence of CO₂ adsorption bands in P-GQD. The shaded area shows the characteristic absorption peaks of CO₂. c Single-gas adsorption isotherms of CO₂, CH₄, N₂, and H₂ on P-GQD powders at 298 K. d Effect of PEI dosage on CO₂/N₂ and CO₂/CH₄ separation performance of P-GQD membranes. The error bar is the standard deviation from at least three samples. e Pore size distribution of P-GQD derived from NLDFT analysis of CO₂ adsorption at 273 K. The shaded area shows a pore size of 0.35 nm. f Single-gas permeation of P-GQD membrane at 298 K. The shaded area shows the cut-off between CO₂ and N₂. g Ideal selectivities of CO₂ over H₂, N₂, and CH₄ calculated from single-gas permeation data. Source data are provided as a Source Data file.

To further elucidate the separation mechanism, gas adsorption experiments were conducted for single-component H₂, CO₂, N₂ and CH₄. As shown in Fig. 4c, the equilibrium adsorption capacity on P-GQD powder follows the order CO₂ >> CH₄ > N₂ > H₂, reflecting a strong preferential adsorption of CO₂ by groups containing N. Temperature-dependent adsorption measurements (Supplementary Fig. 21) further verify the high CO₂ affinity of P-GQDs: the zero-coverage adsorption enthalpy for CO₂ is −28.14 kJ mol⁻¹, much higher than that for CH₄ (−16.66 kJ mol⁻¹) or N₂ (−11.86 kJ mol⁻¹). In comparison, for S-GQDs calcined at 350 °C alone, the zero-coverage adsorption enthalpy of CO₂ was measured to be −25.97 kJ mol⁻¹, which is lower than that of P-GQDs (Supplementary Fig. 22). By fitting the 298 K isotherms (Supplementary Fig. 23) using ideal adsorbed solution theory (IAST), we estimated the adsorption selectivities to be ~14.6 and ~4.7 for equimolar binary CO₂/N₂ and CO₂/CH₄, respectively.

Membrane separation performance is also influenced by the degree of cross-linking. As shown in Fig. 4d, increasing the PEI dosage increases the CO₂/N₂ and CO₂/CH₄ separation factors but lowers the CO₂ permeance because the excessive cross-linking reduces pore volume. Conversely, a too-low PEI amount (3 mg) is insufficient to form an effective cross-linked network, resulting in low separation factors (Supplementary Table 2). Thus, both excessive and insufficient cross-linking are detrimental to the separation performance of the P-GQD membrane.

The pore size distribution of P-GQD was examined using CO₂ adsorption isotherm at 273 K (Supplementary Fig. 21a), indicating a bimodal microporous structure, with pore size around 0.35 nm lies between the kinetic diameters of CO₂ (0.33 nm) and N₂ (0.364 nm) (Fig. 4e). Single-gas permeation tests yield ideal selectivities of ~29 for CO₂/N₂ and ~44 for CO₂/CH₄ (Fig. 4f, g), far exceeding the corresponding Knudsen diffusion selectivities (0.8 for CO₂/N₂, 0.6 for CO₂/CH₄) and also much higher than the IAST adsorption selectivities. This indicates that molecular sieving (exclusion of larger molecules by the ~0.35 nm pores) plays an indispensable role in the CO₂ separation process.

The temperature dependence of permeance provides further insight. As test temperature rises, permeances of all gases increase due to accelerated diffusion. However, the increase in CO₂ permeance is significantly smaller than that of CH₄ or N₂, because higher temperature weakens CO₂ adsorption affinity (Supplementary Fig. 18). From Arrhenius plots of permeance vs 1/T, we calculate permeation activation energies of 7.65 kJ mol⁻¹ for CO₂, 22.36 kJ mol⁻¹ for N₂, and 23.61 kJ mol⁻¹ for CH₄ (Supplementary Fig. 24). The notably lower activation energy for CO₂ indicates that the membrane facilitates CO₂ transport relative to the other gases. Finally, we observed that using a larger amount of GQD sol (forming a thicker membrane) decreases CO₂ permeance but increases separation factors (Supplementary Fig. 25), consistent with increased diffusion path length and additional selective adsorption sites. All these results demonstrate that the P-GQD membrane achieves CO₂ separation through a combination of strong CO₂ adsorption and size-sieving effects. It is worth noting that GQD membranes cross-linked with other small amine (e.g., diethylenetriamine) also show moderate CO₂ separation performance (Supplementary Fig. 26), suggesting the generality of this small-molecule post-cross-linking strategy. However, their efficacy is somewhat lower than that of PEI-cross-linking due to the weight loss mismatch and the inability of a single small molecule to precisely match the subtle variations in the GQDs stacking pores.

C₃H₆/C₃H₈ separation performance of regulated P-GQD membrane

Based on the pore formation mechanism and the CO₂ separation results, varying the calcination temperature modifies the degree of crosslinking and volatilization of some light elements, thereby altering the pore size of the P-GQD membrane (Supplementary Fig. 27). P-GQDs exhibit a bimodal ultra-microporous structure with sizes of 0.33–0.40 nm and 0.43–0.70 nm (Supplementary Fig. 28). The total micropore volume increases with calcination temperature due to enhanced release of volatile species. A fundamental trade-off exists: >0.4 nm pores become more, whereas <0.4 nm pores required for size-sieving of CO₂ is maximized at intermediate temperature (for instance 350 °C) and diminishes at 380 °C due to pore widening. Therefore, we expected that increasing the calcination temperature (and thus pore size) would enable separation of larger gas molecules.

To test this, we carried out single-component permeation experiments on P-GQD membranes calcined at various temperatures. As shown in Fig. 5a, b, at a relatively low calcination temperature of 320 °C, the small pore size of the membrane significantly blocks the permeation of gases larger than CO₂. As the calcination temperature increases, the permeance of all gases rises and the CO₂/N₂ selectivity gradually drops, but a clear cut-off between C₃H₆ and C₃H₈ emerges. In an equimolar C₃H₆/C₃H₈ mixture test, the P-GQD membrane calcined at 380 °C achieves a C₃H₆ permeance of 117 GPU with a C₃H₆/C₃H₈ selectivity of 12. This selectivity is far higher than the IAST-predicted adsorption selectivity (1.98, Supplementary Fig. 29), underscoring the dominant role of molecular sieving due to pore widening. The C₃H₆/C₃H₈ selectivity remains above 10 even as the feed pressure increases to 0.35 MPa (Fig. 5c). These results demonstrate that the P-GQD membrane has potential application in C₃H₆/C₃H₈ separation.

Fig. 5. P-GQD membranes for C₃H₆/C₃H₈ separation.

a Single-gas permeances of various gases through P-GQD membranes calcined at different temperatures. b Ideal selectivities for several gas pairs calculated from single-gas permeation results at different calcination temperatures. c Equimolar binary C₃H₆/C₃H₈ separation performance of a P-GQD membrane as a function of feed pressure. Source data are provided as a Source Data file.

Unlike previous studies, small amounts of GQDs doping only modulate the properties of mixed matrix membranes, here GQDs were employed as building blocks to assemble into dense membranes. More significantly, the membrane preparation strategy we developed allows straightforward tuning of pore size to match different separation targets by simply adjusting the calcination temperature and the extent of cross-linking. This is a convenient “post-regulation” approach compared to synthesizing entirely new membranes for each application. By formulating appropriate post-treatment conditions for different separation scenarios, this strategy could greatly accelerate the development of high-performance membranes for a broad range of gas separation.

Discussion

We developed a strategy to customize GQD membranes for CO₂-selective permeation. GQDs were successfully assembled into a membrane on an Al₂O₃ support via thermal spraying, followed by heat treatment and cross-linking with small amine molecules to regulate the pore structure of the membrane. This approach combines the intrinsic nanoporous framework of closely packed GQDs with strong CO₂-philic surface functionality of the membrane. The two effects synergistically enable extremely high CO₂ permeance (>1000 GPU) together with high separation selectivity (separation factor > 40) for both CO₂/N₂ and CO₂/CH₄ mixtures. The small-molecule post-cross-linking strategy demonstrated here allows convenient membrane customization: by adjusting the heat-treatment temperature, choosing different cross-linker molecules, or modifying the GQD surface groups, the pore structure of the membrane and affinity can be tailored to specific application. Using this approach, we extended the separation capabilities of the membrane to the C₃H₆/C₃H₈ systems. Overall, this customizable pore-regulation strategy holds great promise for broad applications in membrane-based gas separations.

Methods

Chemicals and Materials

CA (CA, 99%) was purchased from Macklin. Polyethyleneimine (PEI, M.W. 10000, 99%) was obtained from Aladdin. α-Al₂O₃ supports (top-layer pore size ~ 70 nm) were provided by Foshan Yirun Jingtao New material Co., Ltd.

Synthesis of GQDs

GQDs were synthesized through pyrolysis of CA. Specifically, 1.5 g of CA powder was placed in a beaker and heated at 210 °C for 30 min in an oven. After pyrolysis, the resulting brown-red solid was dissolved in 20 mL deionized water with stirring, then filtered with a 0.22 μm membrane to obtain a GQD sol. The concentration of GQDs aqueous sol is approximately 10 mg mL⁻¹. The GQD sol was freeze-dried to yield GQD powder for characterization.

Preparation of GQD primitive membrane by thermal spraying

A porous α-Al₂O₃ support (top layer pore size of 70 nm, diameter of 18 mm and thickness of 1 mm) was preheated to 180 °C on a hot plate. A GQD aqueous sol was then spray-coated onto the hot support using a handheld spray gun (0.3 mm nozzle) with compressed air (flow of 7–7.5 mL/min). The spray nozzle was held vertically ~15 cm above the Al₂O₃ support, and the GQD sol was delivered at ~0.15 mL/min. Upon contact with the heated support, the solvent water droplets evaporate quickly, leaving GQD nanoparticles on the support surface to form a continuous GQD primitive membrane. A total of 2 mL of GQD sol was sprayed. The actual amount of GQDs loaded on the support was determined by weighing the substrate before and after spraying (Supplementary Table 5). After coating, the GQD primitive membrane was further dried in an oven at 80 °C for 10 h.

Preparation of P-GQD membrane by PEI pyrolysis cross-linking

The dried GQD primitive membrane was placed in a tubular furnace together with a specified amount of polyethyleneimine (PEI) for heat treatment under flowing Ar (100 mL min⁻¹). The PEI was positioned upstream of the GQD primitive membrane so that its decomposition vapors could contact and cross-link with the surface functional groups of GQD membrane. The system was heated to the target temperature at 1 °C min⁻¹ and maintained for 10 h. After cooling to room temperature under Ar, the cross-linked P-GQD membrane was obtained with the regulated pore structure. For comparison, an S-GQD membrane was prepared by heat treating the sprayed GQD primitive membrane under the same conditions but without any PEI present.

Gas permeation tests

Each membrane was mounted in a custom stainless-steel permeation cell and sealed with silicone O-rings. The diameter of the tested P-GQD membrane is 10 mm. Gas permeation was measured by the Wicke-Kallenbach method at room temperature. For single-gas tests, H₂, CO₂, N₂, CH₄, C₃H₆, or C₃H₈ was fed to the upstream side at a flow rate of 100 mL/min. For mixed-gas tests, equimolar binary CO₂/N₂ or CO₂/CH₄ feed was used, and helium (He) at 50 mL/min was used as the sweep gas on the permeate side. The compositions of permeate gases were analyzed online by a gas chromatograph (Agilent 7890B) equipped with a high-sensitivity thermal conductivity detector. Gas permeance was calculated as:

$$P_i=\frac{J_i}{\mathrm{S}\mathrm{\Delta }{p}_i}$$ 1

where P_i is the permeance of component i (mol m⁻² s⁻¹ Pa⁻¹), J_i is the molar flow rate of component i in the permeate (mol s⁻¹), S is the effective membrane area (m²), and Δp_i is the transmembrane pressure drop of component i. Permeance is often reported in gas permeation units, GPU, where 1 GPU = 3.349 × 10⁻¹⁰ mol m⁻² s⁻¹ Pa⁻¹.

The ideal selectivity for gases i and j was calculated as:

$${\alpha }_{ij}=\frac{P_i}{P_j}$$ 2

where P_i and P_j are single gas permeance of component i and j, respectively.

For gas mixtures, the separation factor was determined by:

$${SF}_{ij}=\frac{y_i/y_j}{{x_i/x}_j}$$ 3

where x and y are the molar fractions of components i and j in the feed and permeate, respectively.

The temperature dependence of permeance was analyzed using the Arrhenius equation:

$$\mathrm{l}nP=-\frac{E_p}{R}\frac{1}{T}+C$$ 4

where P is the gas permeance, E_p is the activation energy of gas permeation, R is the gas constant (8.314 J mol⁻¹ K⁻¹), T is the absolute temperature (K) and C is constant. By plotting ln P is versus 1/T, the activation energy E_p can be obtained from the slope.

Characterization of GQDs

The morphology and size of GQDs were examined by transmission electron microscopy (TEM, JEM-2100, JEOL) and AFM (AFM, Nanowizard, JPK). Surface functional groups were characterized by attenuated total reflectance Fourier-transform infrared spectroscopy (ATR-FTIR, Nicolet iS50) and X-ray photoelectron spectroscopy (XPS, Thermo Fisher Escalab 250 Xi+). Optical properties were measured by Ultraviolet-visible spectroscopy (UV-vis, UV-1900) and photoluminescence spectroscopy (PL, PTI QM400). Thermogravimetric analysis (TGA) of GQDs and PEI was performed using a Netzsch STA 449 F5 thermal analyzer and a PerkinElmer Diamond TG/DTA, respectively, at a heating rate of 10 °C min⁻¹ from 50 °C to 800 °C under N₂ flow rate of 20 mL min⁻¹.

Characterizations of GQD membranes

Surface and cross-sectional morphologies of the sprayed GQD primitive membrane and P-GQD membrane were observed by scanning electron microscopy (SEM, JSM-7900F, JEOL) with energy-dispersive X-ray spectrometer (EDS) for elemental distribution mapping. Fourier-transform infrared spectroscopy (FT-IR, TENSOR27, Bruker) was used to monitor the evolution of membrane surface groups under different heat-treatment temperatures. Elemental composition of the membranes was analyzed by XPS (Thermo Fisher Escalab 250 Xi + ). Carbonaceous microstructures of membranes were examined by Raman spectroscopy (HNG-IS Raman, 355 nm excitation). Water contact angles were measured using a KRÜSS DSA100 contact angle goniometer for evaluating hydrophilicity and hydrophobicity of membranes. Zeta potential measurements of membrane samples were carried out using a Zetasizer Nano instrument (Malvern Co.).

Gas adsorption experiments

P-GQD powder samples for adsorption measurements were prepared by drop-casting the GQD aqueous sol onto a quartz support, followed by the same PEI pyrolysis cross-linking treatment used for membranes. Before measurement, samples were degassed at 200 °C for 3 h to remove residual volatiles. Gas adsorption isotherms for H₂, CO₂, N₂, and CH₄ were collected at 298 K and 273 K using volumetric adsorption analyzers (Micromeritics ASAP 2020 Plus and Quantachrome Autosorb iQ). The 298 K adsorption data for CO₂, N₂, and CH₄ were fitted with a single-site Langmuir-Freundlich model:

$$q=\mathrm{Q}\frac{b{p}^{1/n}}{1+b{p}^{1/n}}$$ 5

where q is the amount of adsorbed (mmol g⁻¹) at pressure p, Q is the saturation adsorption capacity (mmol g⁻¹), b is the affinity constant (kPa⁻¹), and n is the heterogeneity index.

The ideal adsorption solution theory (IAST) was used to predict adsorption selectivity for binary mixtures at equilibrium. The IAST selectivity (S_ads) for component 1 over component 2 is given by:

$$S_{ads}=\frac{q_1/q_2}{p_1/p_2}$$ 6

where q₁ and q₂ are adsorbed amounts of component 1 and 2, and p₁ and p₂ are the corresponding mole fractions in the gas phase.

Author contributions

X.J.Z., X.F.Z. and W.S.Y. conceived and designed the experiments. X.J.Z. performed the membrane fabrication, characterization and performance tests. X.F.Z. and W.S.Y. supervised the research. X.J.Z., Q.X.F., L.M.Z. and X.F.Z. analyzed the data and discussed the results. Z.W.C. and H.B.L. provided some experimental components. All the authors contributed to the drafting and revision of the paper.

Peer review

Peer review information

Nature Communications thanks Hailiang Liu, Bishnupada Mandal and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data that support the findings of this study are available within the article and its Supplementary Information. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-69938-4.