Construction of molecular fans in metal-organic framework to overcome diffusion resistance for trace benzene capture

Abstract

Molecular size-matched pores of adsorbents provide strong adsorptive forces for benzene, but high diffusion resistance. Overcoming the diffusion resistance within molecular size-matched pores is essential to efficient capture of trace benzene from indoor air or industrial exhaust gas, but challenging. Herein, we report a stable Fe-based metal-organic framework (ZJU-701) featuring hierarchical micropores (5.8 ± 0.9, 10.2 ± 2.3, and 21.2 ± 1.4 Å) and molecular fans with 1,4-naphthylene as rotors between micropores. ZJU-701 shows adsorption rate constant of 5.84 × 10⁻³ min⁻¹ and adsorption capacity of 2.05 mmol g⁻¹ for benzene at P/P₀ = 0.001 and 298 K, surpassing other benchmark adsorbents. Strikingly, ZJU-701 exhibits only 10.1% loss of dynamic adsorption capacity for trace benzene as the gas velocity increases from 0.03 to 0.12 m·s⁻¹, much lower than other benchmark adsorbents (30.6% ~ 76.9%). The rapid adsorption of benzene is attributed to the rotational motion of molecular fans in ZJU-701, verified by in-situ Infrared/Raman Spectroscopy and solid-state NMR. This work highlights a dynamic pore-engineering strategy with molecular fans to alleviate diffusion resistance within molecular size-matched pores and balance adsorption rate and capacity for trace gas capture.

Molecular size-matched pores provide strong adsorptive force but high diffusion resistance for benzene. This study installs molecular fans in size-matched pores of metal-organic framework to promote diffusion and dynamic capture of trace benzene.

Introduction

Benzene is classified as a Group 1 carcinogen with no recommended safe level of exposure by the World Health Organization^1–3, and commonly exists at trace concentrations in indoor air and industrial exhaust gas^4–7. Long-term exposure to trace benzene was reported to cause irreversible damage to human nervous and reproductive systems, and contribute to millions of premature deaths globally each year^8,9. The low energy-consuming and byproduct-free physisorption using porous adsorbents is a promising method for removing benzene^10–12. Meanwhile, designing the pore size of adsorbent to match the molecular size of benzene (i.e., molecule-pore size matching effect) is widely regarded as a crucial strategy for enhancing adsorption of trace benzene, by providing adsorption sites with multiple interactions and strong binding forces^13–15. For example, a number of metal-organic frameworks (MOFs) with pore size (6–10 Å) close to the kinetic diameter of benzene (5.80 Å), such as ZJU-520, ZJU-620, BUT-55, PCN-250, MIL-125-Zn, and MFM-300, were reported successively to refresh the record of equilibrium adsorption capacity for trace benzene^13,16–20. Recent representative adsorbents reported for trace benzene capture are summarized in Supplementary Table 1, including their uptake, dominant pore size, and design strategy. However, the molecule-pore size matching effect inevitably slows down the diffusion of benzene within the molecular size-matched pores of adsorbents, due to the steric hindrance effects and the congestion with large amount of coexisted gases (e.g., air, N₂), and subsequently lowers the adsorption rate^13,21,22. It could cause nonequilibrated dynamic adsorption of benzene under the fast gas velocity in the practical applications of adsorbents for air purification or exhaust gas treatment^23–25. Furthermore, the dynamic adsorption capacity would decrease with the increase of gas velocity, and finally be much lower than the equilibrium adsorption capacity, largely inhibiting the practical applications of adsorbents with molecular size-matched pores for removing trace benzene^23,26,27. Therefore, designing porous adsorbents not only with high equilibrium adsorption capacity but also with rapid adsorption rate for trace benzene is essential to achieve high dynamic adsorption capacity for removing trace benzene under fast gas velocity in practical applications, but challenging.

Inspired by the development of molecular machines^28–32, organic linkers containing rotatable functional groups (i.e., rotors) could be used to construct molecular fans in MOFs, which may be driven by the carrier gas (e.g., air, N₂) to accelerate benzene diffusion and adsorption rate in the molecule size-matched pores of MOFs. Hence, constructing molecular fans in the molecular size-matched pores of MOFs may simultaneously achieve high equilibrium adsorption capacity and rapid adsorption rate for trace benzene (Fig. 1), but has never been reported. Herein, we report a stable Fe-based MOF material (i.e., ZJU-701, ZJU = Zhejiang University) featuring hierarchical micropores (i.e., 5.8 ± 0.9 Å, 10.2 ± 2.3 Å, and 21.2 ± 1.4 Å) and molecular fans with rotatable 1,4-naphthylene group as rotor. The 1,4-naphthylene rotor integrated in the framework not only can accelerate the diffusion of benzene molecules, but also can enhance the adsorption of trace benzene by providing strong π-π interaction^33,34. ZJU-701 shows adsorption capacity of 2.05 mmol g⁻¹ and adsorption rate constant of 5.84 × 10⁻³ min⁻¹ for benzene at P/P₀ = 0.001 and 298 K, both of which surpassing other benchmark adsorbents (e.g., MOFs, activated carbons, and zeolites). Strikingly, ZJU-701 exhibits only 10.1% loss of dynamic adsorption capacity for trace benzene as the gas velocity increases from 0.03 to 0.12 m·s⁻¹, much lower than other benchmark adsorbents (30.6–76.9%). Please see Supplementary Table 2 for the trace benzene uptake, the adsorption rate constant k₁, the dynamic capacity loss under enhanced gas velocity, the dominant pore size, and the strategy of several benchmark adsorbents in Supplementary Table 2. This work highlights the potential of ZJU-701 as a versatile adsorbent for capturing trace harmful gases in practical applications.

Fig. 1. Adsorption capacity and rate for trace benzene by different adsorbents.

An adsorbent with molecular size-matched pores (blue color) shows a high adsorption capacity but a low adsorption rate due to strong interactions but slow diffusion. An adsorbent with large pores (green color) shows a high adsorption rate but a low adsorption capacity due to fast diffusion but weak interactions. An adsorbent with size-matched pores (magenta color) and molecular fans show both a high adsorption capacity and a high adsorption rate due to strong interactions and fast diffusion. Magenta and blue arrows indicate increases and decreases, respectively. Check marks and crosses indicate favorable and unfavorable characteristics, respectively. The molecular and fan symbols represent benzene and a molecular fan, respectively.

Results and discussion

Synthesis and structure

In previous study, MFM-300(Fe) synthesized by iron salts and biphenyl-3,3’,5,5’-tetracarboxylic acid (H₄BPTC) was reported to have a uniform pore size of 6.8 Å close to the kinetic diameter of benzene (5.80 Å), and show high equilibrium adsorption of trace benzene²⁰. In this study, organic linker combined H₄BPTC with a rotatable 1,4-naphthylene group as rotor, i.e., 5,5’-(naphthalene-1,4-diyl)diisophthalic acid (H₄NPDI), was used to construct Fe-based MOF (i.e., ZJU-701) with molecular fans (Fig. 1). Specifically, ZJU-701 was synthesized through a solvothermal reaction between Fe(NO₃)₃·9H₂O and H₄NPDI linker in N,N-dimethylformamide (DMF) with acetic acid as modulator at 423 K for 24 h. The scanning electron microscopy (SEM) images of ZJU-701 exhibits the formation of rod-shaped crystals (Supplementary Fig. 1). The powder X-ray diffraction (PXRD) results of ZJU-701 with sharp diffraction peaks indicates their good crystallinity (Supplementary Fig. 2). The crystallographic data of ZJU-701 determined by single-crystal X-ray diffraction (SCXRD) analysis are listed in Supplementary Table 3. It reveals that ZJU-701 adopts the hexagonal crystal system with space group of P6₁ (a = 46.53 Å, b = 46.53 Å, c = 27.46 Å), has a 3D framework consists of Fe₃ clusters Fe₃(μ₃-O)(-CO₂)_x(CH₃COO)_y and tetracoordinated NPDI^4- linkers (Fig. 2a-a’). In this structure, each two Fe₃ clusters are connected by two phthalic groups from two NPDI⁴⁻ linkers, and the alternating arrangement of Fe₃ clusters and phthalic groups in a line form one-dimensional rod building blocks (RBB) (Fig. 2b-b’). Each six RBBs are connected by NPDI⁴⁻ linkers and form a hexagonal prism (Fig. 2c-c’). In each lateral face of the hexagonal prism, NPDI⁴⁻ linkers bridges two parallel RBBs with angles of 15° and 165°, alternatively (Fig. 2c), and the 15° inclined NPDI⁴⁻ linkers in six lateral faces of the hexagonal prism build a right-handed helical network (Fig. 2d-d’). Inside the hexagonal prism and along with its right-handed helical network, two parallelly helicoidal networks consist of alternating Fe₃ clusters and NPDI⁴⁻ linkers are anchored vertically on the lateral faces (Fig. 2e), forming two helical channels with layer distances of 6.2 Å and 12.6 Å, respectively (Fig. 2e’). Furthermore, each two hexagonal prisms containing two helical channels are connected by sharing one lateral face, and each six hexagonal prisms form a hexagonal channel with diameter of 20.6 Å (Fig. 2f-f’). The narrow helical channels, wide helical channels and hexagonal channels of ZJU-701 are interconnected with each other, generating a hierarchically microporous structure (i.e., 6.2 Å, 12.6 Å and 20.6 Å). As the layer distance of narrow helical channel (6.2 Å) is close to the kinetic diameter of benzene (5.8 Å), molecule-pore size matching effect and high equilibrium adsorption capacity of trace benzene are expected. The windows between the narrow helical channel and other two channels are two NPDI^4- linkers with a distance of 6.2 Å coordinated by two Fe₃ clusters (Fig. 2e’). Hence, molecular fans with the rotatable 1,4-naphthylene groups of NPDI^4- linkers as rotors are successfully constructed in the molecular size-matched pores of ZJU-701 (Fig. 2f-f’).

Fig. 2. Structure of ZJU-701.

a, a′ The trinuclear iron cluster and organic linker in ZJU-701. b, b′ Schematic diagram of one-dimensional rod building blocks (RBB). c, c′ The hexagonal prism formed by six RBBs connected via NPDI4- linkers. d, d′ The two parallel helicoidal networks inside the hexagonal prism. e, e′ Two helical channels formed by two parallel helicoidal networks anchored inside the hexagonal prism. f, f ′ Structural diagram of ZJU-701 formed by six hexagonal prisms surrounding a hexagonal channel, determined by the SCXRD method. Color code: C, gray; O, red; Fe, green. Hydrogen atoms are omitted for clarity.

N₂ adsorption isotherm of ZJU-701 at 77 K is shown in Fig. 3a. The specific surface area of ZJU-701 calculated by the Brunauer-Emmett-Teller (BET) method is 1497 m²·g⁻¹ (Supplementary Fig. 3). The pore size distribution of ZJU-701 calculated by the Nonlocal Density Functional Theory (NLDFT) method shows hierarchical micropores, i.e., 5.8 ± 0.9 Å, 10.2 ± 2.3 Å, and 21.2 ± 1.4 Å (Fig. 3b), in good agreement with the hierarchical framework with narrow helical channel (6.2 Å), wide helical channel (12.6 Å) and hexagonal channel (20.6 Å) determined by the SCXRD analysis of ZJU-701 (Fig. 2f-f’). N₂ adsorption-desorption isotherms as well as PXRD patterns of ZJU-701 before and after treatments with aqueous solutions at pH from 2 to 12 for one week are almost unvaried (Fig. 3c, d), demonstrating that the framework structure and the pore size distribution of ZJU-701 can be well maintained in humid, acidic or basic environment. Thermalgravimetric analysis (TGA) indicates the thermal stability of ZJU-701 up to 350 °C, despite the slight mass loss due to solvent removal (Supplementary Fig. 4). Hence, the stable ZJU-701 with excellent chemical and thermal stability exhibit high potential for applications under various conditions such as high humidity, high temperature, and high acidity or basicity.

Fig. 3. Porosity and stability of ZJU-701.

a N₂ adsorption-desorption isotherms of ZJU-701 at 77 K. b Pore size distribution profile of ZJU-701. c N₂ adsorption-desorption isotherms and d PXRD patterns after exposure in air or treatments with aqueous solution of ZJU-701 at pH 2–12 for 7 days.

Additionally, ZJU-702 was synthesized by replacing the NPDI^4- linker of ZJU-701 with BZDI^4- (5,5’-(benzene-1,4-diyl) diisophthalic acid) containing a smaller rotor (i.e., 1,4-phenylene), while ZJU-703 was synthesized by replacing the NPDI^4- linker of ZJU-701 with ATDI^4- (5,5’-(anthracene-9,10-diyl) diisophthalic acid) containing a bigger rotor (i.e., 9,10-anthrylene). The SEM images of ZJU-702 and ZJU-703 are similar to that of ZJU-701 (Supplementary Figs. 5, 6). The PXRD data with Rietveld refinement indicate that ZJU-702 and ZJU-703 share a similar framework structure with ZJU-701 (Supplementary Fig. 7 and Supplementary Tables 4, 5). The BET specific surface areas of ZJU-702 and ZJU-703 are 1641 and 1170 m²·g⁻¹ (Supplementary Figs. 8, 9), respectively, and they have similar pore size distributions and thermal stability to ZJU-701 (Supplementary Figs. 10, 11). Hence, the rotors with size difference in organic linker did not change the hierarchical framework structure of those MOFs. However, they alter the size of window between different channels and their rotational motions (Fig. 4), and subsequently may affect the diffusion and adsorption of benzene, which will be explored later.

Fig. 4. Molecular fans in the helical channels of ZJU-701, ZJU-702 and ZJU-703.

a Schematic diagram of the narrow helical channel and wide helical channel. The rotation of 1,4-naphthylene, 1,4-phenylene and 9,10-anthrylene groups within the helical channels of b ZJU-701, c ZJU-702, and d ZJU-703, respectively.

Benzene adsorption and capture

Adsorption isotherms of benzene at 298 K by ZJU-701, ZJU-702, ZJU-703 as well as MFM-300(Fe) (with the BET analysis shown in Supplementary Fig. 12) are displayed in Fig. 5a. Benzene uptake at a trace concentration (P/P₀ = 0.001) by the four MOFs follows the order of MFM-300(Fe) (0.90 mmol·g⁻¹) <ZJU-702 (1.00 mmol·g⁻¹) ≈ ZJU-703 (1.00 mmol·g⁻¹) <ZJU-701 (2.05 mmol·g⁻¹). Hence, the introduced 1,4-naphthylene group in ZJU-701 can significantly increase the trace benzene uptake, while the introduced 1,4-phenylene or 9,10-anthrylene can slightly increase it. Furthermore, after ten adsorption-desorption cycles, ZJU-701 shows unvaried adsorption capacity for trace benzene (Fig. 5b and Supplementary Table 6), as well as unvaried crystallinity and pore structure indicated by the N₂ adsorption isotherms (Supplementary Fig. 13a) and consistent PXRD patterns (Supplementary Fig. 14b), demonstrating excellent regeneration performance and potential for efficient applications. Adsorption kinetic curves of trace benzene (P/P₀ = 0.001) at 298 K by the four MOFs are displayed in Fig. 5c. The time required for trace benzene to approach adsorption equilibrium on the four MOFs follows the order of ZJU-701 (500 min) <ZJU-702 (800 min) <ZJU-703 (1400 min) <MFM-300(Fe) (>1400 min), while the Pseudo-first-order rate constant k₁ and the calculated diffusion coefficients D/r² of trace benzene increase in the order of MFM-300(Fe) (k₁ = 1.21 × 10⁻³ min⁻¹, D/r² = 1.471 × 10⁻⁶ s⁻¹) < ZJU-703 (k₁ = 2.10 × 10⁻³ min⁻¹, D/r² = 2.175 × 10⁻⁶ s⁻¹) < ZJU-702 (k₁ = 2.65 × 10⁻³ min⁻¹, D/r² = 3.739 × 10⁻⁶ s⁻¹) < ZJU-701 (k₁ = 5.84 × 10⁻³ min⁻¹, D/r² = 6.546 × 10⁻⁶ s⁻¹) (Supplementary Fig. 15). Hence, the introduction of rotatable functional group can largely accelerate the diffusion and adsorption of trace benzene in the molecular size-matched pore structure, and 1,4-naphthylene shows better acceleration performance than 1,4-phenylene or 9,10-anthrylene. Thus, ZJU-701 exhibits significantly higher adsorption capacity and faster adsorption rate of trace benzene than MFM-300(Fe), as well as ZJU-702 and ZJU-703.

Fig. 5. Benzene vapor adsorption data.

a Benzene adsorption isotherms at 298 K of ZJU-701, ZJU-702, ZJU-703 and MFM-300(Fe). b Ten benzene vapor adsorption-desorption cycles for ZJU-701. c Benzene adsorption kinetic curves at P/P₀ = 0.001 and 298 K. d Comparison of adsorption capacity (500 min) and pseudo-first-order rate constant k₁ of benzene at P/P₀ = 0.001 on various adsorbents. All adsorption capacities and pseudo-first-order rate constants were experimentally measured in this study under identical conditions. e Benzene breakthrough curves of the adsorbents at gas velocity from 0.03 m·s⁻¹ to 0.12 m·s⁻¹ at 298 K. f Relationships of dynamic adsorption capacity with gas velocity for various adsorbents. Error bars represent the standard deviations of three independent measurements.

For better evaluating the adsorption performance of ZJU-701, other six benchmark adsorbents for trace benzene are used for comparison. The PCN-250(Fe₂Co), MIL-125-Zn, HKUST-1 and BUT-55 were synthesized following the methods reported in the literature^13,18,19,35. The well-known commercial activated carbon (AC) F400d and zeolite ZSM-5^36,37 were purchased from the market. The detected PXRD patterns, BET specific surface area and pore size distribution of the six benchmark adsorbents (Supplementary Figs. 16–22 and Supplementary Table 7) are consistent with the results in literature^13,18,19,38, indicating the successful synthesis or appropriate purchase of those adsorbents. All the adsorbents used in the adsorption kinetic experiment and the following breakthrough experiment are with consistent particle size distributions (i.e., 4.08 ± 0.07 ~ 4.62 ± 0.20 μm, Supplementary Fig. 23), activation protocols (i.e., degassing under vacuum at 150 °C for 3 h, Supplementary Figs. 24, 25), sample loading densities for kinetic uptake measurements (i.e., 10.93–11.08 kg·m⁻³, see the Methods), and bed packing densities for the following breakthrough experiments (i.e., 46.94–47.45 kg·m⁻³, see the Methods), to exclude the interferences of these factors for the comparison of kinetic and dynamic performances of the adsorbents. The adsorption kinetic curves of benzene on the six benchmark adsorbents are displayed in Supplementary Fig. 26. As displayed in Fig. 5d, Supplementary Figs. 27, 28 and Supplementary Table 8, ZJU-701 has the highest adsorption capacity and fastest adsorption rate of trace benzene (P/P₀ = 0.001) among the benchmark adsorbents.

To simulate the trace benzene capture in practical applications, the breakthrough curves of trace benzene (P/P₀ = 0.005) carried with N₂ by ZJU-701 and other nine adsorbents at different gas velocities were obtained (Fig. 5e, Supplementary Figs. 29, 30 and Supplementary Table 9). The gas velocities in the breakthrough experiments are from 0.03 to 0.12 m·s⁻¹, which are realistic for volatile organic compounds (VOC) air purification when the powder adsorbents are used³⁹. The calculated benzene uptakes at different gas velocities are shown in Fig. 5f. Notably, the benzene uptake of MFM-300(Fe) sharply decreased by 63.5% as the gas velocity increases from 0.03 to 0.05 m·s⁻¹ and decreased by 76.9% as the gas velocity approaches 0.12 m·s⁻¹. The decreases of benzene uptake by the increase of gas velocity are also observed for other adsorbents, and follows the order of MFM-300(Fe) (76.9%) > ZJU-703 (67.2%) > ZSM-5 (63.5%) > HKUST-1 (51.7%) > ZJU-702(51.7%) > BUT-55 (48.1%) > MIL-125-Zn (44.2%) > AC (37.9%) > PCN-250(Fe₂Co) (30.6%) > ZJU-701 (10.1%). The decrease of benzene uptake could be attributed to the slow diffusion of benzene within the narrow pore structure and the resulted nonequilibrium adsorption at elevated gas velocity^13,23,26,40. Fortunately, ZJU-701 exhibits the lowest and negligible decrease percentage (10.1%) of benzene uptake as the gas velocity increases from 0.03 to 0.12 m·s⁻¹, which should be attributed to the fastest adsorption rate of benzene as displayed in Fig. 5d. During the dynamic breakthrough adsorption, higher gas velocity shortens the residence time (i.e., adsorption time) of benzene with porous adsorbents, and thus often reduces the dynamic adsorption capacity, especially for the adsorbent with low adsorption rate because of nonequilibrium adsorption^13,23. ZJU-701 with faster adsorption rate (i.e., faster diffusion) than other benchmark adsorbents for benzene can reach the equilibrium adsorption within a shorter adsorption time, and thus could suffer from a less loss of dynamic adsorption capacity for trace benzene as the gas velocity increases. All the breakthrough curves are fitted well by the Yoon-Nelson model, indicated by the high R² values (Supplementary Tables 10, 11). ZJU-701 has higher values of breakthrough kinetic coefficient (i.e., mass-transfer coefficients) k_YN than ZJU-702, ZJU-703 and other seven benchmark adsorbents (Supplementary Table 12), demonstrating the better dynamic adsorption performance of ZJU-701 for trace benzene. The quantitative comparison of mass-transfer coefficients together with the capacity loss under increasing velocity can convincingly exhibit the better dynamic adsorption performance of ZJU-701 for trace benzene.

The relative humidity (RH) of real air streams can be up to 80%^41,42. Thus, the breakthrough curves of trace benzene by ZJU-701 under 0%, 20%, 40%, 60% and 80% RH were obtained. The calculated benzene uptake of ZJU-701 decreases by 16.4% as the RH increases from 0% to 80% (Supplementary Fig. 31a), indicating that the influence of humidity on the adsorption performance of ZJU-701 for trace benzene is insignificant. Other organic vapors (e.g., toluene, o-xylene, m-xylene, p-xylene, and ethylbenzene) as interferents may also be present in the real air streams⁴³. Thus, the multicomponent breakthrough curves of trace benzene (600 ppm) with another organic vapor (600 ppm) by ZJU-701 are conducted. The calculated benzene uptake of ZJU-701 decreases by 9.8%, 12.3%, 15.8%, 19.7% and 17.6% in the presence of toluene, o-xylene, m-xylene, p-xylene, and ethylbenzene, respectively (Supplementary Fig. 31b–f), demonstrating that the influences of other organic vapors as interferents on the adsorption performance of ZJU-701 for trace benzene are also insignificant. Therefore, ZJU-701 shows superior potential for removing trace benzene under fast gas velocity in practical applications, due to its high adsorption capacity, fast adsorption rate, low decrease of adsorption capacity at high gas velocities, readily achievable regeneration, and good performance in the real air streams with high RH and the presence of interferents.

Benzene binding sites

The structures of benzene-loaded ZJU-701, ZJU-702 and ZJU-703 were analyzed by Rietveld refinements of the experimental PXRD patterns using the structural model obtained from the DFT optimizations, following previous studies (Supplementary Figs. 32–34 and Supplementary Tables 13–15)^44,45. The results indicate that there are three distinct binding sites within the framework of ZJU-701 for benzene. Site I and Site II are located in the narrow helical channel and the wide helical channel with layer distances of 6.2 Å and 12.6 Å, respectively (Fig. 6b), while site III is located in the hexagonal channel with diameter of 20.6 Å (Fig. 6c). For binding site I (Fig.6d), two −CH from two benzene rings of NPDI^4- linkers interact with the π electrons of benzene molecule, while the π electrons from three benzene rings of NPDI^4- linkers interact with two −CH and π electrons of benzene molecule, forming two CH_framework···π_benzene interactions (3.356–3.815 Å), two CH_benzene···π_framework interactions (3.629–3.702 Å), and one π···π interaction (3.829 Å). For binding site II (Fig. 6e), the π electrons from one benzene molecule interact with three −CH from three benzene rings of NPDI^4- linkers, forming three weaker CH_framework···π_benzene interactions (3.229–3.984 Å) and one CH_benzene···π_framework (3.883 Å). Similarly, there are one CH_framework···π_benzene interaction (3.079 Å), and one CH_benzene···π_framework interaction (3.746 Å) at binding site Ⅲ (Fig. 6f). The interaction patterns of benzene molecules in ZJU-702 and ZJU-703 are essentially consistent with that in ZJU-701 (Supplementary Figs. 35, 36). Notably, in binding site Ⅰ of ZJU-703, steric hindrance from the 9,10-anthrylene group results in the confinement of benzene molecules between two parallel ligands (Supplementary Fig. 36a), with only one CH_benzene···π_framework interaction (3.871 Å) and two π‑π stacking interactions (3.552–3.723 Å). Attributed to the molecule-pore size matching effect, the calculated adsorption energies at site I of ZJU-701, ZJU-702 and ZJU-703 for benzene (i.e., −60.99 kJ·mol⁻¹, −52.38 kJ·mol⁻¹, and −65.42 kJ·mol⁻¹, respectively) are all higher than the calculated adsorption energies at site II and site III (Supplementary Fig. 37). These values are consistent with the isosteric heats of adsorption (Q_st) calculated from the adsorption isotherms of benzene at different temperatures (Supplementary Figs. 38, 39). The Grand Canonical Monte Carlo (GCMC) simulation results show that benzene molecules quickly occupy sites I and II at low relative pressure, and then adsorb at site III as the relative pressure increases (Supplementary Figs. 40–42). Although the calculations above indicate that ZJU‑703 has a higher adsorption energy for benzene than ZJU‑701, experiments show that ZJU-703 has a lower adsorption capacity for trace benzene than ZJU-701 (Fig. 5a, c). This could be attributed to the higher steric hindrance for benzene adsorption within ZJU-703 framework (Fig. 4d). The occupancies of Site I in ZJU-701 and ZJU-703 by benzene at different pressures, and the accessible volumes of Site I are calculated using the GCMC simulations (Supplementary Figs. 43–45 and Supplementary Table 16). The Site I in ZJU-703 is not fully accessible at low pressure (P/P₀ = 0.001). The occupancy of Site I in ZJU-703 by benzene at low pressure (P/P₀ = 0.001) is 10.48%, and then increases to 29.95% by benzene at higher pressure (P/P₀ = 1.0). The calculated accessible volume of Site I in ZJU-703 is 0.270 cm³·g⁻¹. While the occupancies of Site I in ZJU-701 by benzene at P/P₀ = 0.001 and P/P₀ = 1.0 are 50.37% and 58.10%, respectively. The calculated accessible volume of Site I in ZJU-701 is 0.574 cm³·g⁻¹. The much lower benzene occupancies and accessible volume of Site I in ZJU-703 than that in ZJU-701 indicates the higher steric hindrance of Site I in ZJU-703 for benzene (i.e., unfavorable kinetics), and can reasonably explain the lower experimental benzene uptake by ZJU-703 than that by ZJU-701, despite the higher binding energy of Site I in ZJU-703 for benzene (i.e., favorable thermodynamics). It demonstrates that the bigger rotor 9,10-anthrylene causes more steric hindrance for benzene than 1,4-naphthylene within the MOFs, supporting the lower adsorption capacity and slower adsorption rate of trace benzene by ZJU-703 than ZJU-701.

Fig. 6. Benzene binding sites in ZJU-701.

a Distribution of benzene binding sites in ZJU‑701 determined by the Rietveld refinements of the experimental PXRD patterns. b Site I located within the narrow helical channel and Site II within the wide helical channel. c Site III located within the hexagonal channel. Views of d Site I, e Site II and f Site Ⅲ in ZJU-701. Color code: H, white; C, gray; O, magenta; Fe, green; benzene, blue.

Spectral characterization for dynamics of molecular fans within MOFs

The apparent rotational barriers of rotors within ZJU-701, ZJU-702, and ZJU-703, obtained by the low-field time-domain ¹H NMR spin-lattice relaxation (T₁) measurements and Arrhenius-type analysis^46,47, are 4.45, 8.17, and 16.57 kJ·mol⁻¹, respectively (Fig. 7a and Supplementary Table 17). The measured rotational energy barriers of rotors within ZJU-701, ZJU-702, and ZJU-703 are negatively correlated with the benzene diffusion coefficients and uptake rate constants (Fig. 7b), indicating that the enhancements of benzene diffusion coefficient and uptake rate are resulted from the low rotational barrier of rotor (i.e., high rotor motion), and thus can directly support the proposed rotor-assisted transport mechanism. The rotation motion of molecular rotor within MOFs is indeed often a hindered rotation, with a certain apparent rotational energy barriers related to the steric environment^48,49. Under the N₂ gas flow, the molecular rotor can obtain higher rotational kinetic energy and momentum transferred from N₂ gas by molecular collision⁵⁰. Thus, the rotation motion of rotor within MOFs under the N₂ gas flow mainly depends on two factors, i.e., the apparent rotational energy barriers and the collision frequency with N₂ molecules. The 1,4-naphthylene rotor in ZJU-701 has lower rotational energy barrier (4.45 kJ·mol⁻¹) than 1,4-phenylene in ZJU-702 (8.17 kJ·mol⁻¹). Meanwhile, the collision frequency of 1,4-naphthylene in ZJU-701 with N₂ molecules should be twice of 1,4-phenylene in ZJU-702, because the cross section of 1,4-naphthylene is twice of 1,4-phenylene. These results well support the higher rotational motion of 1,4-naphthylene in ZJU-701 than 1,4-phenylene in ZJU-702, and the better dynamic adsorption performance of ZJU-701 than ZJU-702. ZJU-703, with largest rotor 9,10-anthrylene, has highest apparent rotational energy barrier (16.57 kJ·mol⁻¹), and thus suffers from stronger steric hindrance from the steric environment. According to the framework structure of ZJU-703 (Supplementary Fig. 46c), the steric hindrance of rotor 9,10-anthrylene mainly arises from the confined framework environment, which physically restricts the free volume required for the larger 9,10-anthrylene rotor to rotate. Moreover, the 9,10-anthrylene rotor could also clash with neighboring rotors within ZJU-703, due to its relatively large size (Supplementary Fig. 47). Thus, it is reasonable that ZJU-703 has the poorest dynamic adsorption performance and the smallest spectral shift, due to the highest rotational energy barrier.

Fig. 7. Rotational energy barriers and their relationships with benzene diffusion and uptake rate.

a Apparent rotational energy barriers (E_a) derived from variable-temperature low-field T₁ relaxation measurements and Arrhenius-type analysis for ZJU-701, ZJU-702, and ZJU-703. b Correlations between rotational energy barriers and diffusion coefficients (D/r²) and uptake rate constants (k₁). Error bars represent the standard deviations of three independent measurements.

The in-situ low-frequency Raman Spectroscopy studies of ZJU-701, ZJU-702 and ZJU-703 at N₂ gas flow of 0 and 100 mL·min⁻¹ are displayed in Fig. 8a–c and Supplementary Fig. 48. The Raman peak at 71 cm⁻¹ for ZJU-701 is assigned to the out-of-plane rotation of aromatic rings (i.e., 1,4-naphthylene)^51,52, and it exhibits a blue-shift with Δ = 10 cm⁻¹ by the increase of N₂ gas flow from 0 to 100 mL·min⁻¹, confirming the accelerated rotation of 1,4-naphthylene by N₂ gas flow^53,54. Similarly, the Raman peaks at 70 cm⁻¹ and 75 cm⁻¹ for ZJU-702 and ZJU-703 are assigned to the out-of-plane rotation of 1,4-phenylene, and 9,10-anthrylene, respectively. However, the peak blue shifts of ZJU-702 and ZJU-703 are only Δ=4 cm⁻¹ and 1 cm⁻¹, respectively. This blue-shift can be interpreted as the enhancement of effective rotational/torsional potential of the rotor within MOFs under the gas flow⁵³. Moreover, there is a significantly negative linear relationship between the Raman peak shift (Δ = 10, 4, and 1 cm⁻¹ for ZJU-701, ZJU-702, and ZJU-703, respectively) and the rotational barrier (4.45, 8.17, and 16.57 kJ·mol⁻¹ for ZJU-701, ZJU-702, and ZJU-703, respectively) (Supplementary Fig. 49). Generally, the rotational motion of rotor is low when the rotational barrier is high. Hence, this negatively linear relationship indicates that a larger Raman peak shift means a lower rotational barrier and subsequently a higher rotational motion for a molecular rotor. The higher Raman peak blue-shift (10 cm⁻¹) of ZJU-701 than that of ZJU-702 (4 cm⁻¹) and ZJU-703 (1 cm⁻¹), indicates the higher molecular rotation motion of 1,4-naphthylene within ZJU-701 than that of 1,4-phenylene within ZJU-702 and 9,10-anthrylene within ZJU-703. Meanwhile, the in-situ infrared spectroscopy studies of ZJU-701, ZJU-702 and ZJU-703 at N₂ gas flow of 0 and 100 mL·min⁻¹ are shown in Fig. 8d–f. The infrared spectroscopy (IR) peak at 1033 cm⁻¹ for ZJU-701 is assigned to the ring breathing vibration⁵⁵ and C-H stretching vibration on the aromatic ring⁵⁶ (e.g., 1,4-naphthylene), respectively, and its intensity is enhanced by 10.4% as the N₂ gas flow increases from 0 to 100 mL·min⁻¹, attributed to the accelerated rotation of 1,4-naphthylene⁵⁷. While the enhancements of the IR peaks at 1017 cm⁻¹ and 1029 cm⁻¹ for ZJU-702 and ZJU-703 are relatively lower (i.e., 3.9% and 2.0%, respectively), again indicating the higher molecular rotation motion of 1,4-naphthylene within ZJU-701, as compared with 1,4-phenylene within ZJU-702 and 9,10-anthrylene within ZJU-703. The same in-situ Raman and IR measurements were also performed in the presence of 125 ppm benzene/N₂ gas flow. The obtained results show the negligible effects of trace benzene (125 ppm) on the Raman peak shift and the IR peak intensity enhancement of ZJU-701, ZJU-702 and ZJU-703 (Fig. 8 and Supplementary Fig. 50). Hence, rotational dynamics of ZJU-701, ZJU-702 and ZJU-703 are not affected by the trace benzene. This could be attributed to the very low relative pressure (i.e., 0.000125) of benzene in the 125 ppm benzene/N₂ gas flow, and the rotational dynamics of ZJU-701, ZJU-702 and ZJU-703 should be mainly driven by the dominant N₂ rather than the trace benzene in the gas flow. These results confirm that the rotation of molecular rotors within the adsorbents persists with trace benzene present, and can help to prove the causality between gas-driven rotation and accelerated benzene diffusion.

Fig. 8. In-situ low-frequency Raman and infrared spectra.

In-situ low-frequency Raman spectra of a ZJU-701, b ZJU-702 and c ZJU-703 at 0 and 100 mL·min⁻¹ flow rate, respectively. In-situ infrared spectra of d ZJU-701, e ZJU-702 and f ZJU-703 at flow rates of  0 and 100 mL·min⁻¹, respectively.

Therefore, the high rotation of molecular fans (i.e., 1,4-naphthylene) within ZJU-701, as confirmed by the results of in-situ low-frequency Raman spectroscopy, in-situ infrared spectroscopy and low-field NMR, could be responsible for the higher adsorption rate of trace benzene within ZJU-701. As shown in Fig. 4, the higher rotation of 1,4-naphthylene within ZJU-701 than 1,4-phenylene within ZJU-702 should be attributed to the larger size and lower apparent rotational barrier of 1,4-naphthylene than 1,4-phenylene (Fig. 4c). The higher rotation of 1,4-naphthylene within ZJU-701 than 9,10-anthrylene within ZJU-703 should be due to the less steric hindrance of 1,4-naphthylene than that of 9,10-anthrylene within framework (Fig. 4d).

In summary, we report a stable metal-organic framework ZJU-701 featuring hierarchical micropores and molecular fans with 1,4-naphthylene groups as rotors. The molecular size-matched pores (5.8 ± 0.9 Å) are mainly responsible for the high adsorption capacity of trace benzene by providing multiple adsorptive forces. The molecular fans help to overcome the diffusion resistance within the molecular size-matched pores, and contribute to the rapid adsorption rate of trace benzene. Thus, ZJU-701 shows efficient capture of trace benzene and negligible loss of dynamic adsorption capacity at high gas velocities of practical applications. ZJU-701 also shows high adsorption capacity and rapid adsorption rate for many other trace gases (i.e., toluene, m-xylene, NH₃, SO₂, and SF₆) (Supplementary Figs. 51, 52 and Supplementary Table 18), exhibiting as a versatile adsorbent for capturing trace harmful gases. This work proposes a dynamic pore-engineering strategy (i.e., constructing molecular fans in molecular size-matched pores of MOFs) to alleviate the high diffusion resistance within narrow porous structures, and balance adsorption rate and capacity for trace gas capture. It also highlights the application of molecular machine-integrated MOFs for protecting our environment.

Methods

Synthesis of ZJU-701 (powder)

5,5’-(Naphthalene-1,4-diyl)diisophthalic acid (100 mg, 0.22 mmol) and ferric nitrate nonahydrate (250 mg, 0.62 mmol) were added to a mixed solvent of N,N-dimethylformamide (DMF, 10 mL) and acetic acid (5 mL). The mixture was sonicated until complete dissolution, sealed in a reaction vessel, and heated at 150 °C for 24 h. After cooling to room temperature, the earthy-yellow product was collected by filtration, washed three times with ethanol and DMF, and dried in a vacuum oven at 80 °C for 24 h.

Synthesis of ZJU-701 (single crystal)

5,5’-(Naphthalene-1,4-diyl)diisophthalic acid (100 mg, 0.22 mmol) and ferric nitrate nonahydrate (25 mg, 0.062 mmol) were added to a mixed solvent of DMF (14 mL) and acetic acid (7 mL). The mixture was sonicated until complete dissolution, sealed in a reaction vessel, and heated at 135 °C for 72 h. After cooling to room temperature, small orange-yellow single crystals suitable for X-ray diffraction were obtained.

Synthesis of ZJU-702

5,5’-(benzene-1,4-diyl) diisophthalic acid (80 mg, 0.20 mmol) and ferric nitrate nonahydrate (250 mg, 0.62 mmol) were added to a mixed solvent of DMF (10 mL), acetic acid (5 mL), and pentanol (1.5 mL). The mixture was sonicated until complete dissolution, sealed in a reaction vessel, and heated at 135 °C for 24 h. After cooling to room temperature, the earthy-yellow product was collected by filtration, washed three times with ethanol and DMF, and dried in a vacuum oven at 80 °C for 24 h.

Synthesis of ZJU-703

5,5’-(Anthracene-9,10-diyl)diisophthalic acid (100 mg, 0.20 mmol) and ferric nitrate nonahydrate (200 mg, 0.50 mmol) were added to a mixed solvent of DMF (10 mL) and acetic acid (5 mL). The mixture was sonicated until complete dissolution, sealed in a reaction vessel, and heated at 155 °C for 36 h. After cooling to room temperature, the earthy-yellow product was collected by filtration, washed three times with ethanol and DMF, and dried in a vacuum oven at 80 °C for 24 h.

Characterizations

All reagents were purchased commercially and used without further purification. PXRD data were collected on a German Bruker D8 Advance X-ray diffractometer equipped with Cu targets and ceramic tubes. PXRD patterns were processed and simulated using crystal structure data with the Mercury program. TGA was performed on a Mettler Toledo TGA2 thermogravimetric analyzer under a nitrogen atmosphere at a heating rate of 5 °C min⁻¹. The surface morphologies of the MOFs were examined using a ZEISS Gemini 300 scanning electron microscope (SEM).

Adsorbent pretreatment and gas adsorption

After synthesis and washing, all adsorbents were stored in a vacuum oven at 80 °C prior to experiments. Before adsorption measurements, the samples were degassed under vacuum (10⁻¹⁰ bar) at 423 K for 6 h to remove residual solvent molecules from the pores. N₂ adsorption-desorption isotherms were recorded at 77 K using a BSD-660M instrument (Beishide Instrument Technology (Beijing) Co., Ltd., China). Adsorption isotherms of benzene vapor were recorded using a JW-ZQ100 analyzer (Beijing JWGB Instrument Co., Ltd., China). For adsorption kinetics of benzene vapor, a 125 ppm benzene/N₂ mixture (P/P₀ = 0.001) was continuously introduced into the sample chamber. Subsequently, uptake–time profiles were recorded at 298 K using a BSD-VVS analyzer. Pseudo-first-order rate constants and diffusion coefficients were calculated from the kinetic data to evaluate the adsorption rates of the adsorbents.

Breakthrough experiments

All adsorbents were mixed with 5 μm quartz wool and packed into a glass breakthrough column with an inner diameter of 6 mm. Prior to measurements, the column was degassed at 423 K under a nitrogen flow of 200 mL·min⁻¹ for 6 h to remove solvent molecules from the pores¹⁶. The column temperature was maintained at 298 K using a water bath in a BSD-MAB system. A 600 ppm benzene/N₂ stream (P/P₀ = 0.005) was generated by bubbling high-purity nitrogen through liquid benzene. By adjusting the dual nitrogen velocities, the linear gas velocities were controlled at 0.03, 0.05, 0.06, 0.08, 0.10, and 0.12 m·s⁻¹, corresponding to volumetric flow rates of 50, 80, 110, 140, 170, and 200 mL·min⁻¹, respectively. The outlet benzene concentration was monitored in real time using a gas chromatograph (GC-9790Plus, Full Instrument Co., Ltd., China). Breakthrough curves were obtained by recording concentration–time profiles, from which the dynamic adsorption capacities of benzene at different velocities were calculated.

In-situ spectroscopic experiment

Prior to spectroscopic measurements, ZJU-701, ZJU-702, and ZJU-703 were degassed under vacuum at 423 K for 6 h to remove solvent molecules from the pores. In-situ low-frequency Raman and infrared spectroscopic measurements were performed under increasing nitrogen flow, using a HORIBA LabRAM HR Evolution system (532 nm excitation) and a Thermo Nicolet iS50 spectrometer, respectively, to probe the dynamic responses of molecular fans within the frameworks. Spectra were collected after equilibration at each flow step.

Low field NMR measurements

Low-field time-domain ¹H NMR T₁ relaxation measurements were performed on a MicroMR12-025V NMR analyzer equipped with a 25 mm coil and operating at 22 MHz, and the T₁ relaxation data were processed using the NMR analysis application software ver. 1.0. Prior to low-field NMR measurements, ZJU-701, ZJU-702, and ZJU-703 were ground and pretreated using the same degassing procedure described above. A standard oil sample was first placed into the magnet coil to calibrate the radio-frequency pulse frequency and determine the center frequency. After calibration, a weighed solid powder sample was loaded into the magnet coil. The spin-lattice relaxation signal was acquired using the inversion recovery sequence at different temperatures. The T₁ distributions of ZJU-701, ZJU-702, and ZJU-703 were collected at 20, 40, 60, 80, and 100 °C. The obtained recovery data were processed by inversion analysis using the software to generate the spin-lattice relaxation time (T₁) distribution. The T₁ value corresponding to the main peak maximum in the T₁ distribution was extracted for further Arrhenius-type analysis.

Author contributions

W.W. and Z.H. conceptualized the project and designed the experiments; Z.H. synthesized the materials; Z.H. and X.Yang. performed material characterization and gas adsorption experiments; Z.H., W.W. and Y.L. analyzed the data; Z.H., W.W., Y.H., X.Yun., K.Y., and D.L. discussed, wrote, and revised the manuscript. All authors discussed the results and contributed to the final paper. W.W. provided supervision and acquired funding for this work.

Peer review

Peer review information

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

Funding

This work was supported partly by the National Natural Science Foundation of China (22476179), the National Key Research and Development Program of China (2022YFC3702101, 2024YFD1501801), the Fundamental and Interdisciplinary Disciplines Breakthrough Plan of the Ministry of Education of China (JYB2025XDXM909), and the Fundamental Research Funds for the Central Universities (226-2025-00148).

Data availability

The crystallographic data generated in this study have been deposited in the Cambridge Crystallographic Data Centre under deposition numbers CCDC 2525905 [https://www.ccdc.cam.ac.uk/structures/Search?Ccdcid=2525905&DatabaseToSearch=Published] (ZJU‑701), 2525906 [https://www.ccdc.cam.ac.uk/structures/Search?Ccdcid=2525906&DatabaseToSearch=Published] (ZJU‑702), 2525907 [https://www.ccdc.cam.ac.uk/structures/Search?Ccdcid=2525907&DatabaseToSearch=Published] (ZJU‑703), 2563520 [https://www.ccdc.cam.ac.uk/structures/Search?Ccdcid=2563520&DatabaseToSearch=Published] (benzene@ZJU-701), 2563998 [https://www.ccdc.cam.ac.uk/structures/Search?Ccdcid=2563998&DatabaseToSearch=Published] (benzene@ZJU-702) and 2563999 [https://www.ccdc.cam.ac.uk/structures/Search?Ccdcid=2563999&DatabaseToSearch=Published] (benzene@ZJU-703). These data can be obtained free of charge from the CCDC via www.ccdc.cam.ac.uk/data_request/cif. The source data generated in this study have been deposited in Figshare under accession code https://doi.org/10.6084/m9.figshare.33106952⁵⁸. Source data are also provided with this paper. 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 https://doi.org/10.1038/s41467-026-77361-y.