Precise modulation of MOF pore structures via functional group dimensions and spatial configuration for membrane separation

Abstract

Traditional petroleum distillation faces high energy demands, necessitating innovative alternatives like membrane separation. This study presents a breakthrough in dual-range and precise pore size modulation of metal-organic frameworks (MOFs) through a ligand functionalization strategy. By tailoring steric configurations and spatial orientations of light-responsive azobenzene groups, we achieved broad-range pore tuning (0.41 to 0.68 nanometers) via functional group length variation, coupled with subnanometer precision through reversible trans-to-cis photoisomerization. Four representative branched alkanes were selected to validate the MOF’s high selectivity. Results showed its capacity to generate a constant carbon-atom-count–dependent permeation gradient, realizing a four-step sequential separation that increased C₆H₁₄ purity from 25 to 92.2%. This synergistic approach uniquely combines large-scale pore adjustment with dynamic fine-tuning, decoupling separation efficiency from energy-intensive processes. The membranes’ structural stability and reversible light responsiveness further highlight their potential for sustainable hydrocarbon processing. By integrating molecular design with stimuli-responsive control, this work advances MOF-based membranes as a transformative solution for energy-efficient petroleum fractionation and precise molecular sieving.

MOF membranes can achieve pore structure and aperture regulation via functional group-light modulation for alkane separation.

INTRODUCTION

Separation techniques play a crucial role in petrochemical production (1–4). Crude oil is a complex multiphase mixture that requires fractionation via distillation techniques, whether for research purposes or for processing, to obtain petroleum fractions with relatively narrow boiling point ranges (5–7). Global refineries process more than 100 million barrels of crude oil daily, with energy consumption for petroleum fractionation exceeding 1100 TWh annually, accounting for 1% of global energy usage (2, 8, 9). Apart from petroleum distillation, the cracking process is also one of the most energy-intensive individual processes in the petrochemical industry (10). As environmental issues escalate, the reallocation of energy structures has garnered notable attention, making the development of low-energy alternative technologies, such as membrane separation, particularly vital (11–13). Membrane separation technology has the potential to reduce energy consumption by 23 TWh per year, equivalent to the energy produced by burning 3.21 million tons of coal (8, 9). The use of membrane separation technology for hydrocarbon separation not only reduces energy costs but also offers higher efficiency (14).

In crude oil fractional distillation, the oil is separated into various fractions, such as light, medium, and heavy fractions (14), with molecular weights ranging from tens to thousands (15–17). This process places stringent requirements on the pore size of the separation membrane to achieve effective fractionation. In addition, to further reduce energy consumption, post-cracking petroleum purification necessitates membranes capable of precise molecular sieving, enabling the effective separation of different molecules within the same fraction (10, 18). Therefore, future research should focus on enhancing the tunability of membrane pore sizes to achieve both broad adjustability and precise control, ultimately improving separation accuracy while minimizing energy consumption.

Among various membrane materials, metal-organic frameworks (MOFs) stand out as a class of porous materials, owing to their tunable structures and functionalization potential, particularly for their distinct advantages and great potential in fractional distillation of crude oil (19–28). The ordered porous structure of MOFs, combined with their inherent design flexibility, enables the precise modulation of pore sizes through the functionalization of surface groups on its framework (29–34). By adjusting the length and composition of these functional groups, a wide range of pore sizes can be tuned, enabling the effective separation of different oil fractions (5, 35–37). Moreover, the inherent flexibility of MOFs allows the integration of stimulus-responsive molecules, enhancing both the precision of pore size modulation and the ability to dynamically respond to external stimuli (26, 38–42). This characteristic notably enhances the flexibility and efficiency of the separation process, making it possible to separate different molecules within the same fraction (43). Therefore, as membrane materials, MOFs show great promise in achieving effective separation between different fractions by tuning the size of functional groups, while also enabling precise sieving of distinct molecules within the same fraction after petroleum cracking through variations in spatial configuration.

Here, we present an innovative MOF membrane system featuring a dual-mode pore control strategy that combines wide-range adjustability with subnanometer precision (Fig. 1A). The Azo-UiO-66 series MOFs were precisely designed with azobenzene side chains of systematically varied lengths (L1–L4), enabling macroscale pore size adjustment from 0.41 to 0.68 nm—a range perfectly spanning critical hydrocarbon dimensions (C₆-C₁₆). Through innovative integration with polyvinylidene difluoride (PVDF) substrates via vacuum filtration, these MOFs form robust separation membranes where the static pore structure can be further dynamically fine-tuned at the subnanometer scale (0.01- to 0.04-nm precision) via light-induced azobenzene isomerization. This unique combination of chemical modification and photoresponsive control allows for (i) initial coarse separation of petroleum fractions between light and medium fractions by carbon number through size-selective pores, followed by (ii) on-demand precision separation of similarly sized molecules (e.g., distinguishing C₆H₁₄ from C₉H₂₀) through light-activated aperture size modulation. The membranes’ exceptional performance is evidenced by their ability to efficiently separate complex mixtures including C₆H₁₄, C₉H₂₀, C₁₁H₂₄, and C₁₆H₃₄ after four consecutive separation cycles, achieving unprecedented liquid-phase purity of C₆H₁₄ enhancement from 25 to 92.2%. This work establishes a paradigm in membrane technology where static and dynamic pore control mechanisms operate synergistically to address the full spectrum of hydrocarbon separation challenges.

Fig. 1. Schematic diagram of MOF membrane separation.

(A) Schematic diagram of the overall alkane separation process: First, the pore size of the MOF is regulated through functional group control, enabling wide-range pore size adjustment. The dotted box in the upper right of the diagram displays four organic ligands with different functional groups and the structures of the MOFs synthesized from them. Subsequent stages use UV irradiation for fine-tuned, small-range aperture size modulation within the MOF, enabling efficient separation of small-sized molecules. The lower right dashed box depicts the isomerization process of the responsive organic ligands under UV irradiation and heating conditions. SEM images of MOF materials: Corresponding to (B) Azo-UiO(L1), (C) Azo-UiO-CH₃(L2), (D) Azo-UiO-C₂H₅(L3), and (E) Azo-UiO-C₄H₉(L4), respectively. The inset regions within each image depict individual MOF particles of the corresponding material, clearly revealing their microstructural features.

RESULTS AND DISCUSSION

Separation membranes with a wide range of tunable pore size refinement are achieved by modulating the side-chain functional groups size and spatial configuration of MOF ligands. Therefore, we firstly designed and synthesized four different lengths of azo phenyl groups as light-responsive organic ligands: L1: 2-[(1E)-phenyldiazenyl]benzene-1,4-dicarboxylic acid; L2: 2-[(1E)-(4-methylphenyl)diazenyl]benzene-1,4-dicarboxylic acid; L3: 2-[(1E)-(4-ethylphenyl)diazenyl]benzene-1,4-dicarboxylic acid; and L4: 2-[(1E)-(4-butylphenyl)diazenyl]benzene-1,4-dicarboxylic acid (as shown in Fig. 1, see figs. S1 to S4 for the synthesis details). Nuclear magnetic resonance (NMR) spectroscopy, mass spectrometry (MS), and Fourier transform infrared (FTIR) spectroscopy confirmed the successful synthesis of the four ligands (figs. S5 to S14). The ultraviolet-visible (UV-Vis) absorption spectrum (figs. S15 to S16) and NMR spectra (fig. S17) demonstrate that all four light-responsive ligands can undergo a reversible transformation from trans-isomer to cis-isomer under UV irradiation.

The MOF of UiO-66, which are renowned for their stability, including water and chemical stability (44–46), are then synthesized. Typically, the reaction mixture consisting of L1, terephthalic acid, ZrCl₄, acetic acid, and dimethylformamide (DMF) was heated at 120°C for 24 hours to obtain Azo-UiO(L1) crystals. Using the same procedure, Azo-UiO-CH₃(L2), Azo-UiO-C₂H₅(L3), and Azo-UiO-C₄H₉(L4) were also obtained. The scanning electron microscopy (SEM) images of the four Azo-UiO-type MOFs are presented in Fig. 1 (B to E). All four MOFs exhibit similar octahedral shapes, with particle sizes approximately 400 nm (the fine particles on the MOFs were sprayed platinum to enhance conductivity).

The formation of four MOFs is further corroborated by powder x-ray diffraction (PXRD) patterns and FTIR characterization (Fig. 2, A and B). The PXRD results (Fig. 2A) demonstrate that the diffraction peaks of Azo-UiO(L1) to Azo-UiO-C₄H₉ (L4) are in accordance with those of UiO-66, thereby confirming the UiO-type structure. Three distinct characteristic peaks at 7.42°, 8.53°, and 26° correspond to the (111), (200), and (600) crystal planes, respectively (20). The alignment of the diffraction peaks indicates that the incorporation of azobenzene side chains does not disrupt the crystallinity of UiO-66. The FTIR spectrum of the MOFs (Fig. 2B) displays a distinctive band at around 1600 cm⁻¹, which can be attributed to the asymmetric stretching vibration of the O─C=O group. The peaks observed in the 1420- to 1390 cm⁻¹ range are ascribed to the symmetric stretching of COO- groups. An emerging absorption peak at 1650 cm⁻¹ has been identified as the characteristic C─N bond of the Schiff base (47). The bands observed near 780 and 660 cm⁻¹ are indicative of the symmetric and asymmetric stretching vibrations of O─Zr─O bonds. This evidence corroborates the formation of coordination bonds between Zr⁴⁺ ions and the terephthalic acid or its azobenzene derivatives (45, 48). The absorption peaks at 2960 and 830 cm⁻¹ are indicative of the stretching and bending vibrations of the C─H bonds, respectively. This evidence corroborates the presence of methyl groups and serves to confirm the successful synthesis of Azo-UiO-CH₃(L2). Furthermore, the absorption peaks at 2960 and 830 cm⁻¹, in conjunction with the absorption peak in the C─C stretching vibration region near 950 cm⁻¹, indicate the presence of ethyl groups, thereby providing additional confirmation of synthesis of Azo-UiO-C₂H₅(L3). Ultimately, the observed absorption peaks at 2960 cm⁻¹, 950 cm⁻¹, and the absorption peak in the C─H bending vibration region at 720 cm⁻¹ substantiate the presence of butyl groups, thereby corroborating the successful synthesis of Azo-UiO-C₄H₉(L4).

Fig. 2. Characterization of the Azo-UiO series of MOF.

(A) Powder x-ray diffraction (PXRD) patterns of UiO-66, dark blue curve; Azo-UiO(L1), sky blue curve; Azo-UiO-CH₃(L2), bright blue curve; Azo-UiO-C₂H₅(L3), dark red curve, and Azo-UiO-C₄H₉(L4), red curve. (B) FTIR spectra of Azo-UiO(L1), dark blue curve; Azo-UiO-CH₃(L2), sky blue curve; Azo-UiO-C₂H₅(L3), dark red curve, and Azo-UiO-C₄H₉(L4), red curve. (C) N2 sorption isotherms of cis-Azo-UiO(L1), black curve; cis-Azo-UiO-CH3(L2), dark blue curve; cis-Azo-UiO-C2H5(L3), red curve; cis-Azo-UiO-C4H9(L4), pink curve. (D) N2 sorption isotherms of trans-Azo-UiO(L1), black curve; trans-Azo-UiO-CH3(L2), dark blue curve; trans-Azo-UiO-C2H5(L3), red curve; trans-Azo-UiO-C4H9(L4), pink curve. Pore size profiles of (E) trans-Azo-UiO(L1), (F) cis-Azo-UiO(L1), (G) trans-Azo-UiO-CH₃(L2), (H) cis-Azo-UiO-CH₃(L2), (I) trans-Azo-UiO-C₂H₅(L3), (J) cis-Azo-UiO-C₂H₅(L3), (K) trans-Azo-UiO-C₄H₉(L4), and (L) cis-Azo-UiO-C₄H₉(L4). a.u., arbitrary units.

Subsequently, the light-responsive properties of MOFs are then investigated. Following 60 min exposure to UV radiation, the diffraction peaks of the four MOFs remained unaltered, as illustrated in fig. S18. This observation indicates that the original crystalline structures were preserved after UV radiation. The framework itself is inert under UV light, thereby confirming the excellent stability of the four MOFs. The isomerization of azobenzene on the MOF side chains was investigated using UV-Vis spectroscopy (fig. S19). Following UV irradiation (λ = 365 nm), the absorption band corresponding to the trans-azobenzene at 324 nm exhibited a gradual decrease, while the absorption band corresponding to the cis-azobenzene at 435 nm demonstrated a notable increase. This suggests that the azobenzene present in the MOF side chains has undergone a transformation from the trans to cis form. Conversely, following irradiation with visible light, the absorption peak intensity at 430 nm decreased, while the absorption peak intensity at 325 nm increased, thereby achieving a cis-to-trans transformation.

Furthermore, the four MOFs demonstrated the capacity for persistent photoisomerization following multiple cycle of “on-off” experimentation, thus establishing a foundation for the reuse of MOFs as membrane materials for separation processes. Furthermore, NMR spectroscopy was used to examine the decomposition of the MOF under UV irradiation for varying durations (fig. S20). Initially, the black spectrum underwent a transformation into a red spectrum, which subsequently transformed into a blue spectrum upon prolonged UV irradiation. It can be observed that as the irradiation time increased, the area of the characteristic peak cis-isomer in the NMR spectrum gradually increased. This suggests that the quantity of cis-azobenzene present within the MOF has increased, thereby demonstrating that the Azo-MOF is capable of undergoing photoisomerization. The above results demonstrate that Azo-MOF displays substantial photoisomerization in the presence of UV irradiation.

The porosity of the materials was studied using nitrogen adsorption methods characterized. All four MOF materials exhibited type I adsorption isotherms, with a rapid increase in adsorption at low pressures until saturation was reached in the micropores (Fig. 2C). While the isotherm shapes of all four MOFs are similar, the adsorption capacity decreases with the increasing length of the azobenzene side chains. The introduction of the azobenzene groups occupies part of the pore volume, reducing the effective pore space for adsorption. Following exposure to UV radiation, the adsorption capacities of the four cis-MOFs exhibited minimal variation (Fig. 2D). Following UV irradiation, the pore size of Azo-UiO(L1) was observed to increase from 0.64 nm (Fig. 2E) to 0.68 nm (Fig. 2F). This alteration is attributable to the reduction in molecular dimensions that occurs during the trans-to-cis isomerization of the azobenzene groups, which consequently results in an increase in pore size. A comparable phenomenon was observed in the remaining three MOFs. In particular, the pore size of trans-Azo-UiO-CH₃(L2) is 0.58 nm (Fig. 2G), whereas it increases to 0.60 nm in the cis form (Fig. 2H). With regard to trans-Azo-UiO-C₂H₅(L3), the pore size is 0.57 nm (Fig. 2I), increasing to 0.58 nm in the cis form (Fig. 2J). Similarly, trans-Azo-UiO-C₄H₉(L4) has a pore size of 0.41 nm (Fig. 2K), which increases to 0.42 nm in the cis form (Fig. 2L), shown as table S1.

As the side chain length of the MOFs increases, there is a corresponding decrease in pore size, with a more pronounced reduction observed as the length of the functional groups increases. Moreover, upon exposure to UV radiation, the pore size of the MOF exhibits a slight increase, which can be attributed to the trans-to-cis isomerization of the azobenzene groups present in the side chains. The azobenzene molecules undergo a contraction, which results in an expansion of the aperture size. Furthermore, it was observed that as the length of the side chains increased, the extent of the pore size increase upon UV irradiation gradually diminished. It is speculated that the reason lies in the fact that when azobenzene units are branched onto the rigid MOF framework, the longer the side chain length, the more pronounced the steric hindrance effect becomes. Longer side chains impede the isomerization reaction of azobenzene groups, leading to a stronger restriction on azobenzene isomerization. This, in turn, weakens its ability to regulate pore aperture, thereby limiting the extent of pore structure (49–51). When UV light is shone upon the MOF, the conformational change in azobenzene alters the orientation of the side chains within the cage (fig. S21), thereby exerting a limited effect on the pore structure. Consequently, UV irradiation enables fine-tuning of the pore structure. It is evident that the pore size variation resulting from functional group modification is notably greater than that induced by isomerization, as illustrated in Fig. 2 (E to L). Consequently, alterations to functional groups permit a more extensive spectrum of pore size modulation in MOFs, whereas UV light is only capable of precise adjustment within a more limited range. This characteristic offers notable potential for the separation of petroleum fractions and the purification of products derived from petroleum cracking.

The separation membrane was prepared using a vacuum filtration method, with a detailed account of the steps provided in the Supplementary Materials. Photographs of the membranes are presented in Fig. 3 (A to D). All four membranes exhibit a yellow hue, a flat surface, and no discernible deformation, cracks, or defects. Figure 3 (E to H) illustrates the SEM images of the cross-linked MOF membrane surfaces. The images demonstrate a relatively flat surface, with MOF particles exhibiting a uniform distribution and a high degree of packing density densely packed. The energy-dispersive x-ray spectroscopy (EDS) mapping shows a uniform distribution of the Zr element over the whole membrane, further confirming that the MOF is uniformly distributed. The cross-sectional morphology of the MOF membranes was observed using SEM (Fig. 3, I to L). The SEM images clearly show an MOF layer with a thickness of 5 to 6 μm without obvious defects or pores. The EDS elemental mapping further corroborates the uniform distribution of the Zr element within the membrane, thereby substantiating the uniformity of the MOF layer. Figure 3 (M to P) illustrates the three-dimensional (3D) atomic force microscopy surface morphology of the MOF membrane. The images demonstrate fluctuations and irregularities in the membrane’s microscopic structure, with the surface exhibiting distinct waviness. Table S2 presents the results of the surface roughness analysis, which indicates that the MOF membrane exhibits relatively high surface roughness. This feature is of paramount importance for the separation process as it provides a larger filtration area for solvent molecules (52, 53). The enhanced surface area increases the membrane’s permeability, thereby optimizing its separation performance. Consequently, this surface characteristic exerts a beneficial influence on the MOF membrane’s efficacy in molecular separation applications.

Fig. 3. Characterization of the Azo-UiO series of MOF membranes.

Photograph of (A) Azo-UiO(L1) membrane, (B) Azo-UiO-CH₃(L2) membrane, (C) Azo-UiO-C₂H₅(L3) membrane, and (D) Azo-UiO-C₄H₉(L4) membrane. SEM surface images of (E) Azo-UiO(L1) membrane, (F) Azo-UiO-CH₃(L2) membrane, (G) Azo-UiO-C₂H₅(L3) membrane, and (H) Azo-UiO-C₄H₉(L4) membrane, and their EDS mapping of Zr (inset). SEM cross-sectional photograph of (I) Azo-UiO(L1) membrane, (J) Azo-UiO-CH₃(L2) membrane, (K) Azo-UiO-C₂H₅(L3) membrane, (L) Azo-UiO-C₄H₉(L4) membrane, and their EDS mapping of Zr (inset). Atomic force microscopy 3D image of (M) Azo-UiO(L1) membrane, (N) Azo-UiO-CH₃(L2) membrane, (O) Azo-UiO-C₂H₅(L3) membrane, and (P) Azo-UiO-C₄H₉(L4) membrane. (Q to T) UV-Vis spectral changes shown as absorbance vs. wavelength of trans-to-cis isomerization of (Q) Azo-UiO(L1) membrane, (R) Azo-UiO-CH3(L2) membrane, (S) Azo-UiO-C2H5(L3) membrane, and (T) Azo-UiO-C4H9(L4) membrane, where absorbance is defined as a dimensionless logarithmic ratio, i.e., the base-10 logarithm of the ratio of incident light intensity before passing through the analyte to the transmitted light intensity after passage through the same. h, hours.

The cis-trans isomerization of the azobenzene MOF membranes has been confirmed through UV-Vis spectroscopy. The light-responsive characteristics of the MOF membranes are illustrated in Fig. 3 (Q to T). Initially, the MOF membranes display two distinct absorption peaks at 325 and 450 nm, respectively. These peaks correspond to the π-π* transition of the trans isomer and the n-π* transition of the cis isomer, respectively. Following a period of 60 min under UV irradiation, the absorption band at 325 nm exhibits a decrease, indicating a transition from the trans to the cis isomer. Concurrently, the peak at 435 nm displays an increase, confirming the formation of the cis isomer (26, 38, 54, 55). Upon exposure to visible light, the azobenzene molecule undergoes a transition from the cis to trans form, which results in a decrease in UV absorption at 435 nm and an increase at 325 nm. This photoisomerization phenomenon persists and remains evident even with an increased MOF content in the composite membrane (fig. S22). Furthermore, the light responsiveness of the MOF composite membrane remains intact after five consecutive UV-Vis cyclic tests, indicating that the photoisomerization of the MOF is reversible and exhibits good repeatability (see figs. S19, S22, and S23). Obviously, the photoisomerization of the azobenzene side chains in the MOF is predominantly reversible, with some incomplete reversibility potentially attributable to site-blocking effects. Furthermore, the reversible variation in the absorption band intensity of Azo-MOF can be repeatedly observed (figs. S19, S22, and S23), indicating that the photoisomerization of azobenzene is driven by alternating UV and visible light irradiation during the switching cycle. This characteristic enables the reversible modulation of the MOF’s pore size in response to light. The photoisomerization characteristics also provide a theoretical basis for the separation membrane to change the membrane aperture size, improve the separation efficiency and effectively reduce energy consumption (29). The permeance of various solvents through four types of MOF membranes was evaluated by using a self-made permeation test device (fig. S24). As shown in fig. S25 and Fig. 4 (A and B), the polarity of the solvents had a substantial impact on the permeation flux. Specifically, as shown in fig. S25A, there is no notable difference in the permeance of the water through different MOF membranes and the permeance is very low. However, the water molecule is notably smaller than the pore sizes of the MOF membranes, the molecular sieve effect does not substantially hinder their permeance. The contact angles of the MOF membranes range from 50° to 60° (fig. S26), indicating similar surface polarity. Therefore, the polarity of the membrane materials plays a more prominent role in determining water permeance. This is attributed to the strong interactions between the polar solvent (water) and the MOF membrane, which slow down the permeance of water. Similarly, the same trend was observed for methanol. As shown in fig. S25B, the permeance of different MOF membranes to methanol is not notably affected, but its permeation rate is clearly higher than that of water. This is due to methanol having weaker polarity compared to water, and its molecular size being smaller than the pore size of the MOF. Furthermore, the molecular sieving effect of the MOF membrane has a more pronounced impact on the permeation properties of nonpolar solvents, such as hexane and nonane shown in Fig. 4 (A and B). From Fig. 4A, it can be seen that the first three MOF membranes have less effect on the permeance of hexane, which is due to the fact that the pore sizes of these three membranes are larger than the molecular size of hexane. In contrast, the Azo-UiO-C₄H₉(L4) membrane has the longest side-chain length and the smallest pore size, resulting in the lowest permeation rate. In contrast, the pore size of Azo-UiO-C₄H₉(L4) is smaller than that of hexane, and although UV light slightly enlarges the aperture via photoisomerization, the permeance only increases marginally. When the solvent molecule size is further increased, the influence of MOF side-chain functional groups on separation performance becomes even more notable. As shown in Fig. 4B, the permeance of nonane is lower than that of hexane. For nonane, as the length of the side chains increases, the permeation rate of the MOF membranes notably decreases. Under UV irradiation, the permeation of MOF membranes was slightly increased. The above data suggest that by modifying the size and spatial arrangement of functional groups, the separation performance of the membrane can be optimized under specific conditions, thereby improving the separation effect of the membranes. This can be explained by the fact that the pore sizes of the first three MOF membranes are considerably larger than the diameter of hexane molecules, resulting in higher permeance. When the solvent molecule size is further increased, the influence of MOF side-chain functional groups on separation performance becomes even more notable. As shown in Fig. 4B, the permeance of nonane is lower than that of hexane. Furthermore, nonane permeance decreases as the length of the MOF side chains increases. After UV irradiation, a slight increase in nonane permeance is observed, especially for the Azo-UiO-C₄H₉(L4) membrane, although the overall change remains minor. This behavior can be attributed to the combined effects of molecular sieving and photoisomerization. In addition, control experiments confirm that solvent viscosity has little influence on permeance (fig. S27). Instead, the permeation behavior is predominantly governed by the molecular sieve effect and the polarity of the MOF membrane.

Fig. 4. The penetration properties of different solvents.

(A) Nonane permeance for membranes made from Azo-UiO series MOF before (blue) and after (red) UV irradiation. (B) Hexane permeance for membranes made from Azo-UiO series MOF before (blue) and after (red) UV irradiation. (C) Petroleum ether permeance for membranes made from Azo-UiO series MOF before (blue) and after (red) UV irradiation. (D) Solvent naphtha permeance for membranes made from Azo-UiO series MOF before (blue) and after (red) UV irradiation. (E) Kerosene permeance for membranes made from Azo-UiO series MOF before (blue) and after (red) UV irradiation. (F) Paraffin permeance for membranes made from Azo-UiO series MOF before (blue) and after (red) UV irradiation. (G) The separation factors of petroleum ether and solvent naphtha before (blue asterisk) and after (pink triangle) UV irradiation. (H) The separation factors of solvent naphtha and kerosene before (blue triangle) and after (red square) UV irradiation. (I) The separation factors of kerosene and paraffin before (blue triangle) and after (red hexagon) UV irradiation.

To further examine the impact of MOF membrane pore size on the permeance of nonpolar solvents, four nonpolar solvents with varying carbon chain lengths (petroleum ether, solvent oil, kerosene, and paraffin) were subjected to testing at room temperature (Fig. 4, C to F). The results of the test for the permeance of petroleum ether are presented in Fig. 4C. The results demonstrate minimal variation in petroleum ether permeance across the Azo-UiO(L1) to Azo-UiO-C₂H₅(L3) membranes, whereas a notable decline is evident in the Azo-UiO-C₄H₉(L4) membrane. Following exposure to UV radiation, the permeance of the Azo-UiO(L1) to Azo-UiO-C₂H₅(L3) membranes exhibited a slight increase compared to before UV light irradiation, whereas the Azo-UiO-C₄H₉(L4) membrane demonstrated notable fluctuations before and after UV light irradiation. This phenomenon can be attributed to the introduction of functional groups on the MOF side chains, which occupy a portion of the pore volume (51). As the size of the functional groups increases, the occupied volume expands, which results in a reduction in pore size. Following exposure to UV radiation, the azobenzene groups present on the side chains undergo a trans-to-cis isomerization, resulting in a contraction of the azobenzene molecules and an expansion of the MOF pore size (56). The primary constituents of petroleum ether are C₅H₁₂ and C₆H₁₄ hydrocarbons, with molecular sizes ranging from 0.38 to 0.43 nm, shown as table S3. The Azo-UiO-C₄H₉(L4) membrane pore size is within this range. Consequently, the permeance of petroleum ether through the Azo-UiO-C₄H₉(L4) membrane is less than that observed for the first three membranes. Following exposure to UV light, the aperture size of the Azo-UiO-C₄H₉(L4) membrane increases, as shown in Fig. 2L. Consequently, the permeance is observed to increase in comparison to the flux observed before light exposure. A comparable pattern is evident in the solvent oil test (Fig. 4D). As the length of the MOF side chain increases, the permeance of solvent oil gradually decreases. However, following exposure to UV light, an increase in permeance was observed. The Azo-UiO-C₄H₉(L4) membrane demonstrates minimal alteration in permeance before and following light exposure. This phenomenon can be attributed to the solvent oil’s primary component, C₉H₂₀, which has a molecular diameter of approximately 0.56 nm. The pore size of the Azo-UiO-C₄H₉(L4) membrane is smaller than that of the solvent oil molecules, both before and after light irradiation, which restricts its permeance. Consequently, there is no notable alteration in the permeance before and following UV light exposure. In the kerosene test, a notable decrease in permeance was observed between the Azo-UiO(L1) and Azo-UiO-CH₃(L2) membranes, whereas the reduction between Azo-UiO-CH₃(L2) and Azo-UiO-C₄H₉(L4) was relatively minimal. Furthermore, the Azo-UiO(L1) membrane demonstrates a notable alteration in permeance before and following UV light exposure. Nevertheless, for solvents with larger molecular sizes, such as liquid paraffin, the modulation effect of the MOF membrane is not notable (Fig. 4F). The pore size range of the MOF membranes is exceeded by the hydrocarbons comprising liquid paraffin, which consists of molecules larger than C₂₀H₄₂. Consequently, the permeance is exceedingly low across all four MOF membranes, remaining below 1 liter m³ hour⁻¹ bar⁻¹. Furthermore, no notable enhancement in permeance was discerned following UV irradiation. In conclusion, the pore size modulation of MOF membranes exerts a considerable influence on the permeance of nonpolar solvents with varying molecular sizes. Furthermore, through control experiments such as varying the concentration of cross-linking agents, it was confirmed that the high selectivity of the membrane stems from the differences in MOFs (see fig. S28 and note S1). This trend of change persists in practical applications (fig. S29), and the MOF exhibits excellent stability (figs. S30 to S38 and note S2), and the experiment was repeated six times and the permeances remained constant, which indicates no collapse of membrane under 1-bar applied pressure (fig. S39).

To further evaluate the separation performance of different solvents, we define the separation factor as the ratio of the permeance of two different solvents through the same MOF membrane (S = P₁/P₂) (26). The separation factor of these nonpolar solvents was calculated, and the results are shown in Fig. 4 (G to I) and table S4. The data demonstrate that the Azo-UiO-C₄H₉(L4) membrane exhibits the most effective separation of petroleum ether and solvent oil under UV light irradiation (see Fig. 4G). With regard to the separation of solvent oil and kerosene, the Azo-UiO-C₂H₅(L3) membrane is observed to exhibit the most favorable performance (see Fig. 4H). Furthermore, the Azo-UiO-CH₃(L2) membrane demonstrates the most pronounced impact on the separation of kerosene and liquid paraffin under UV light irradiation (see Fig. 4I). The above data suggest that the selection of appropriate MOF membranes and the adjustment of UV light irradiation conditions facilitate the effective achievement of separation.

On the basis of the aforementioned results, we further tested the separation performance of the membrane for a flow-phase separation method (note S3). Two light petroleum fractions (C₆H₁₄ and C₉H₂₀) were combined with two medium petroleum fractions (C₁₁H₂₄ and C₁₆H₃₄) in an equal ratio, and separation experiments were conducted at room temperature and 1-bar pressure. The schematic experimental design is presented in Fig. 5A. For the detailed experimental procedures, please refer to the note S1 and figs. S40 and S41. Through flow-phase separation experiments, the mixed alkanes exhibited a remarkable interception rate of 82.6% for C₁₆H₃₄ after passing through the first layer of Azo-UiO(L1) membrane. This separation efficiency is attributed to the differences in the size of the functional groups within the MOF membrane, which facilitated the separation of different fractions. Similarly, after passing through the Azo-UiO-CH₃(L2) membrane and Azo-UiO-C₂H₅(L3) membrane, C₁₁H₂₄ and C₉H₂₀ were successfully separated, with separation efficiencies of 79.4 and 68.3%, respectively. In the fourth layer of the flow phase, by combining functional group modification and UV irradiation as a joint control strategy, the aperture size was precisely regulated within the range of 0.41 to 0.42 nm, resulting in the successful isolation of C₆H₁₄ with a purity of 92.2%. These results demonstrate that by integrating functional group engineering with remote-control strategies, we achieved broad-range and precise pore size modulation, enabling both coarse separation of light and medium fractions (L1 to L3: C₁₆H₃₄, C₁₁H₂₄, and C₉H₂₀) and fine separation of components with closely related carbon chain lengths (L4: C₆H₁₄ and C₉H₂₀). Overall, within the Azo-UiO series of MOF membranes, the pore structure and aperture size are controlled through the synergistic regulation of MOF pore size by functional groups and light, thereby governing hydrocarbon molecule transport. On the one hand, the crystalline order of the MOF forms a continuous pore network. The pore structure and aperture are modulated through the hierarchical adjustment of azobenzene side chain length. As the side chain length increases from alkyl-free (L1) to butyl (L4), size-selective transport pathways are formed, accommodating hydrocarbons with carbon numbers ranging from C₆H₁₄ to C₁₆H₃₄. For instance, the large-pore L1 membrane exhibits low-resistance transport for a broad range of hydrocarbons, while the small-pore L4 membrane permits only small molecules such as C₆H₁₄ to pass through, consistent with molecular sieving mechanisms (57). On the other hand, UV-induced azobenzene isomerization causes contraction of the side-chain molecular dimensions and slight expansion of aperture diameter, enabling lossless fine-tuning of transport resistance and selectivity (58). Simultaneously, the ordered pore topology of the MOF crystal provides a structural foundation for low tortuosity, thereby reducing latent transmission resistance (57). The enhanced hydrophobicity from alkylation of the side chains, coupled with affinity for nonpolar hydrocarbons, further minimizes friction resistance between molecules and pore walls (58). Ultimately, this synergistic mechanism constructs a carbon number–dependent permeation gradient, elevating C₆H₁₄ purity from 25 to 92.2% and enabling precise molecular separation. These findings highlight the notable potential of MOF membranes for high-efficiency and precise separation of petroleum fractions and broaden their prospects for advanced applications in petrochemical separations.

Fig. 5. Continuous separation of mixed alkanes.

(A) Schematic diagram of mixed alkane mobile phase separation sequentially represents the staged separation of mixed alkanes under pressure: A(i) passing successively through the Azo-UiO(L1) membrane (screening out C₁₆H₃₄, dark green spheres), A(ii) the Azo-UiO-CH₃(L2) membrane (screening out C₁₁H₂₄, green spheres), A(iii) Azo-UiO-C₂H₅(L3) membrane (screening out C₉H₂₀, purple spheres), and A(iv) and UV-irradiated Azo-UiO-C₄H₉(L4) membrane (screening out C₆H₁₄, red spheres). Shown is the percentage composition of intercepted liquid components for each MOF membrane layer in the mobile phase separation experiments: These correspond to the interstitial liquid composition ratios for the Azo-UiO(L1) membrane (B), Azo-UiO-CH₃(L2) membrane (C), Azo-UiO-C₂H₅(L3) membrane (D), and Azo-UiO-C₄H₉(L4) membrane (E) before and after UV irradiation.

To further clarify the mechanism of separation, we performed molecular dynamics (MD) calculations to investigate the interaction between different MOFs and various alkanes (fig. S42 and table S5). The results show that the adsorption enthalpy of the same alkane varies minimally across different MOFs, indicating that even with differences in the side-chain length of the MOFs, the interaction strength with the same alkane remains largely consistent. This suggests that the adsorption ability of MOFs alone is not sufficient to achieve effective separation of different alkanes. To investigate the mechanism of alkane separation in greater depth, we conducted MD simulations of the permeation process through MOF membranes for various alkanes (as shown in Fig. 6, A to J). Results indicate that within 500 ps, longer MOF side chains permit passage of shorter alkane molecules; for instance, C₆H₁₄ permeation through the Azo-UiO(L1) membrane is 2.5 times that through the Azo-UiO-C₄H₉(L4) membrane. This trend was validated in additional alkane separation experiments (Fig. 6E) and aligns with prior permeation results. These findings indicate that pore size plays a decisive role in permeation separation of alkanes with varying carbon chain lengths by MOFs bearing side chains of differing functional group lengths. This sieving effect constitutes a critical factor for successful alkane separation.

Fig. 6. Theoretical simulations of alkane molecules permeating through MOF membranes.

The simulation system includes cross-sectional views at 0 and 500 ps, demonstrating equal quantities of C₆H₁₄/C₉H₂₀/C₁₁H₂₄/C₁₆H₃₄ molecules (40 C₆H₁₄ molecules, 40 C₉H₂₀ molecules, 40 C₁₁H₂₄ molecules, and 40 C₁₆H₃₄ molecules) permeating through Azo-UiO(L1) (M1), Azo-UiO-CH₃(L2) (M2), Azo-UiO-C₂H₅(L3) (M3), and Azo-UiO-C₄H₉(L4) (M4) membranes, respectively: (A) C₆H₁₄ permeation through the M1 membrane, (B) C₉H₂₀ permeation through the M1 membrane, (C) C₁₁H₂₄ permeation through the M1 membrane, (D) C₁₆H₃₄ permeation through the M1 membrane; (E) C₆H₁₄ permeation through M2 membrane, (F) C₉H₂₀ permeation through M2 membrane, (G) C₁₁H₂₄ permeation through M2 membrane; (H) C₆H₁₄ passing through M3 membrane, (I) C₉H₂₀ passing through M3 membrane, (J) C₆H₁₄ passing through M4 membrane; (K) summary diagram of alkane permeation through various membranes in the simulation system.

In summary, we achieved effective petroleum fractionation using a series of MOF membranes with precisely tunable aperture sizes. These membranes were synthesized from ligands featuring azobenzene side chains of varying lengths. When applied to nonpolar organic liquids, the membranes demonstrated distinct permeabilities and selectivities, along with excellent structural stability. Moreover, we showcased the potential for petroleum fractionation by modulating the size of functional groups within the MOFs and incorporating light-responsive moieties. This study integrates light-isomerizable molecules of different lengths into MOF architectures, paving the way for next-generation separation membrane technologies and advancing the feasibility of petroleum fractionation. Furthermore, the current strategy for precisely tuning MOF aperture sizes through variations in functional group size and spatial framework configuration can also be used to modulate the stability of 2D materials, such as black phosphorus (figs. S43 to S45 and note S4) (59, 60). This approach provides valuable insights into membrane separation mechanisms and offers a strategy for tailoring the properties of MOF-based functional materials.

MATERIALS AND METHODS

Materials

Dimethyl 2-aminoterephthalate, nitrobenzene, tetrahydrofuran, p-toluidine, potassium bisulfate, p-ethylaniline, p-butylaniline, zirconium tetrachloride, terephthalic acid, n-hexane, and nonane were purchased from Aladdin. Acetic acid, sodium bicarbonate, methylene chloride, sodium sulfate, ethyl acetate, methanol, sodium hydroxide, hydrochloric acid, DMF, and glacial acetic acid were purchased from National Pharmaceuticals (Shanghai Test). Petroleum ether, 1,3,5-benzenetricarboxylic acid chloride (TMC) and piperazine anhydrous (PIP) were purchased from Macklin. Solvent naphtha was purchased from Shanghai Rhawn Co. Ltd. Kerosene and paraffin were purchased from Shanghai Boer Chemical Reagents Co. Ltd. Undecane and hexadecane were purchased from Shanghai TCL Chemical Industry Development Co. Ltd. PVDF membranes (0.22 μm) (13.25 cm by 15 cm) were purchased from Solarbio and cut into 25-mm diameter discs by ourselves. Chemicals that can be purchased are used directly without further purification.

Characterization

¹H and ¹³C NMR spectra were measured on a Bruker Avance II 400 M spectrometer. High-resolution mass spectra were collected on BRUKER autoflex maX matrix-assisted laser desorption/ionization–time-of flight MS. The images of MOF particles were characterized using a Hitachi S-4800 scanning electron microscope. The surface and cross section of the MOF membrane were characterized using the ZEISS Sigma scanning electron microscope from Carl Zeiss AG. The sputter-coating parameters were set to 40 mA for 40 s, and the working voltage was maintained at 15 kV. Powder x-ray diffractograms data were obtained by testing with a Bruker’s x-ray diffractometer equipped with a Cu Kα target to generate x-rays, with a test voltage of 40 kV; a test current of 40 mA; and a scanning range of 5° to 40°, with a step size of 0.0167°, and a time of 0.1 s per step. The FTIR spectra data were recorded on a Thermo Fisher Scientific (China) Ltd. spectral Nicolet is10 spectrometer with a wave number of 4000 to 500 cm⁻¹. The data of the adsorption-desorption curves, specific surface area, and pore size distribution of the MOF were obtained using the Fully Automatic Specific Surface Area and Micropore Analyzer (3H-2000 PM2) from Beijing Best Instrument Technology Co. Ltd. Before testing, the MOF sample was degassed at 120°C for 300 min, with a saturated vapor pressure of 1.0305 bar. Testing was conducted at liquid nitrogen temperature (77.3 K), with the pore size distribution calculated using the Horvath-Kawazoe (H-K original) model. The 3D surface morphology of the MOF membrane was obtained using AFM (Bruker Dimension Icon). Liquid UV was obtained using Shimadzu UV-1780 UV-Vis absorption spectroscopy, and cis-trans isomerization of solid MOF was detected using Shimadzu UV-2550 UV-Vis absorption spectroscopy equipped with an integrating sphere attachment. The absorption spectra of the solutions were tested using a Shimadzu UV-1780 UV absorption spectrometer. The contact angle of the membrane was tested by using the JCY-4 contact angle measuring instrument of Shanghai Fangrui Instrument Company. Data for the separation of mixed alkanes were obtained using Shimadzu’s Gas Chromatography Mass Spectrometer GCMS-QP2010 Plus. Pore size tuning via the UV (λ = 365 nm) lamp with 125 W irradiation. The distance between the sample and the lamp (bulb) was controlled to be 10 cm for each irradiation, and the irradiation was vertical. The visible lamp (λ = 450 nm) used was a 10-W light-emitting diode lamp, and the irradiation distance and position were kept the same as that of the UV lamp.

Synthetic Azo-UiO series MOF

ZrCl₄ (46.6 mg) was dissolved in a mixture of 7.5 ml of DMF and 0.5 ml of glacial acetic acid and sonicated for 10 min to obtain a clarified solution. Then, 16.2 mg of L1 and 6.64 mg of terephthalic acid were added and stirred for 20 min to obtain a clarified solution of orange color. The solution was then transferred to a hydrothermal reactor containing a Teflon liner and heated at 120°C for 24 hour. After cooling to room temperature, the solution was washed three times each with DMF and methanol and dried at 60°C. The solution was then mixed with a mixture of 16.2 mg of L1 and 6.64 mg of terephthalic acid and stirred for 20 min. The syntheses of Azo-UiO-CH₃(L2), Azo-UiO-C₂H₅(L3), and Azo-UiO-C₄H₉(L4) were carried out by simply replacing the 16.2 mg of L1 in the previous step by 17.04 mg of L2, 18 mg of L3, and 19.62 mg of L4, respectively, and the rest of the steps remained unchanged. The homemade MOFs were treated with vacuum degassing at 110°C for 6 hours before preparing the membranes and performing other characterizations such as N₂ physisorption.

Acid digestion of MOFs

One milligram of MOF powder was added to 20 μl of DCl (35 wt % in D₂O) for elimination, followed by 0.65 ml of dimethyl sulfoxide–d₆ dilution, and was used to detect the extent of cis-trans isomerization after UV illumination.

Preparation of MOF membranes

The 13.25 cm by 15 cm PVDF membrane was cut into circles with a diameter of 250 mm, rinsed three times with ethanol, rinsed once with deionized water, and subsequently immersed in deionized water for use. Then, 5 ml of MOF aqueous solution (100 μg/ml) was diluted to 50 ml, and sonicated thoroughly for 1 hour. Subsequently, vacuum filtration was used to deposit the MOF solution onto a PVDF membrane. Filtration was continued for an additional 2 min after observing that there was no notable liquid on the surface. Subsequently, the cross-linked layer was prepared. Two milliliters of a 0.1 wt % PIP aqueous solution was added onto the prepared MOF membrane. After 2 min, the vacuum pump was turned on to allow the solution to slowly drip down. Once there was no visible liquid remaining, vacuum was further applied for an additional 2 min. Then, 2 ml of a 0.2 wt % TMC solution in n-hexane was added and cross-linked for 20 s. Last, the membrane was washed with deionized water to remove unreacted substances, and the prepared MOF membrane was activated. The activation treatment method refers to the literature (3). The cross-linked film was soaked in acetone, taken out after 1 min, connected to a vacuum pump, and rinsed with acetone. Using the same method, an MOF membrane was prepared with a loading amount of 2000 μg, with all other operations remaining unchanged.

Membrane performance evaluation

Flux separation

The flux test device is constructed by the laboratory itself, relying on the permeability formula; the device needs to contain a water pump, water pressure gauge, filters, etc. during the test process to record the parameters such as pressure, time, and the corresponding water output, which will be used for the calculation of the permeability. The self-priming pressurized pump is a DC mini-pump model DP-521, which is connected to a power adapter with a male and female DC connector. The water pressure gauge is Leerda Y-60 general pressure gauge with pressure range 0.1 to 60 MPa. The tube has an inner diameter of 10 mm and an outer diameter of 13 mm. The inlet is connected to a filter, and the outlet is connected to a DFilter Technology 25 mm replaceable membrane filter. The permeability (P) is calculated as follows (3)

$$P=\frac{V}{A\times ∆t\times ∆p}$$

where V is the volume of permeate collected (L), A is the membrane area (square millimeter), ∆t is the time required to collect the desired volume of permeate (hours), and ∆p is the transmembrane pressure (bar). Permeability is expressed in liters per square meter per hour per bar (L/m²·hour·bar), which is the conventional standard.

Simulation details

The MD simulations in this work were carried out using the Forcite module in Materials Studio software (61). The COMPASSII force field was used in all simulations to describe the interatomic interactions during the separation of C₆H₁₄, C₉H₂₀, C₁₁H₂₄, and C₁₆H₃₄ across the four triple-layer MOF structures. Individual solvent models for C₆H₁₄, C₉H₂₀, C₁₁H₂₄, and C₁₆H₃₄ were created and optimized. Subsequently, each of these solvent molecules was introduced into 1D channels of the four triple-layer MOFs [Azo-UiO(L1), Azo-UiO-CH₃(L2), Azo-UiO-C₂H₅(L3), and Azo-UiO-C₄H₉(L4)], respectively. Before production MD simulations, all systems underwent energy minimization as a preequilibration step to obtain stable initial configurations. The convergence criteria were set to 1.0 × 10⁻⁴ kcal/mol for energy, 0.005 kcal/(mol·Å) for force, and 0.00005 Å for displacement. Last, a 500-ps canonical ensemble (NVT) simulation was conducted with a time step of 5.0 fs and the temperature maintained at 298.5 K using the Andersen thermostat.

Supplementary Materials

This PDF file includes:

Figs. S1 to S45

Tables S1 to S5

Supplementary Notes 1 to 4

References