Phase-transformable metal-organic polyhedra for membrane processing and switchable gas separation

Abstract

The capability of materials to interconvert between different phases provides more possibilities for controlling materials’ properties without additional chemical modification. The study of state-changing microporous materials just emerged and mainly involves the liquefication or amorphization of solid adsorbents into liquid or glass phases by adding non-porous components or sacrificing their porosity. The material featuring reversible phases with maintained porosity is, however, still challenging. Here, we synthesize metal-organic polyhedra (MOPs) that interconvert between the liquid-glass-crystal phases. The modular synthetic approach is applied to integrate the core MOP cavity that provides permanent microporosity with tethered polymers that dictate the phase transition. We showcase the processability of this material by fabricating a gas separation membrane featuring tunable permeability and selectivity by switching the state. Compared to most conventional porous membranes, the liquid MOP membrane particularly shows the selectivity for CO₂ over H₂ with enhanced permeability.

Introduction

Microporous materials that contain well-defined and controllable pores for selective sorption of guest molecules are deemed promising in gas separation¹, water purification², sensing³, catalysis⁴ and so on. Most of them are rigid, crystalline solids that are advantageous for the ease of structural engineering but adverse in terms of material processing. The requirement to harness good processability for practical use has led to the study of microporous adsorbents that can change phases between liquid, glass, and crystal^5,6. The state-of-the-art researches mainly focus on molecularly networked liquids and glasses from crystalline metal-organic frameworks (MOFs) that have porous three-dimensional structures made of metal nodes and organic ligands^7,8,9,10,11. The transition of MOFs from crystal to amorphous liquid or glass phase was achieved by the heat- or mechanical force-induced cleavage and rearrangement of their coordination bonds^12,13,14, similar to the phase transition of inorganic materials like silica glasses^15,16. For the sake of processability, the synthesis of MOF liquids or glasses is beneficial for their integration with various industrial applications. From the perspective of porosity, the amorphization of MOFs with bond breakages often leads to the collapse of their porous networks^10,17 and the uncontrollable densification of materials^18,19. Consequently, most of the MOF-derived liquids or glasses are less porous or even non-porous, compared to their initial crystalline structures.

Inspiration can be taken from polymer science. Organic polymers feature a reversible liquid-glass-crystal phase interconversion facilitated by molecular motion, conformational flexibility and noncovalent interactions between the polymer chains²⁰. To achieve preserved porosity and phase interconvertibility, one can design composite materials by integrating microporous adsorbents with processable polymers. Indeed, a similar idea has been adopted in the fabrication of porous liquids (PLs), which are synthetic liquid systems containing permanent pores capable of adsorbing guest molecules²¹. PLs are mainly classified into three categories according to their composition; type II and type III PLs are solution and suspension of microporous adsorbent, respectively. Therefore, they are inherently composite materials with massive solvents^22,23,24. Type I PLs are neat liquids with permanent pores and are mainly synthesized by chemically tethering polymeric chains on the surface of adsorbents like MOFs or molecular porous hosts such as porous organic cages (POCs) and metal-organic polyhedra (MOPs)^25,26,27. The existence of tethered polymers diminishes the intermolecular packing of the core adsorbents and imbues liquidity to the resulting products without penetrating or blocking their pores²⁸. Despite the achievements in PL synthesis, the related phase transition behavior and its impact on porosity are rarely studied due to the lack of structural understanding underneath the transition process. This lack of knowledge about the relationship among the structures, the phases and the properties has inhibited the development of phase-transformable porous materials.

Here we show microporous soft materials featuring interconvertible liquid-glass-crystal phase transitions and well-defined persistent microporosities. The key to success relies on the modular synthetic approach, in which molecular MOPs are used as inherent porous hosts and phase-transformable polymeric chains are tethered on the surface of MOPs via coordination bonds, leading to the resulting star polymer-like systems. This modular approach allows for the systematic alternation of hosts and polymer chains to produce six different MOP derivatives with tunable phase transitions. Upon different heating treatments, the intermolecular packing of the tethered polymer chains from adjacent MOPs is changed to afford a reversible transition between liquid, glass and crystal phases (Fig. 1a). Taking advantage of its processibility and porosity, a supported membrane composed of the phase-transformable MOP is fabricated for use in permeation-based gas separation. The gas permeability highly depends on the material phase, either liquid or crystal, even measured at the same temperature. In particular, the membrane in the liquid phase demonstrates an unconventional CO₂ selectivity over H₂ with high permeability. These results provide a deep insight into the processing and property manipulation of microporous materials by controlling the phase transition behavior.

Fig. 1: Liquid-glass-crystal interconvertible MOP and its synthetic route.

a Illustration of MOPs undergoing liquid-glass-crystal phase transition with permanent porosity. b Synthesis of rhodium-based cuboctahedral MOPs (RhMOP) from the coordinative assembly of rhodium acetate ([Rh₂(OAc)₄]) and isophthalic acid derivatives (R-H₂bdc). Color key: green, rhodium; red, oxygen; gray, carbon. c Modification of C₁₂RhMOP via the coordination of predesigned polymeric ligands, TX405-imi, to the axial site of dirhodium paddlewheel (highlighted). The polymer-tethered MOP, denoted as MOP-1A, exists as a purple liquid under ambient conditions.

Results

Design and synthesis of phase-transformable MOPs

The first challenge is to synthesize a MOP material with a lower melting point than its decomposition temperature. We select rhodium-based cuboctahedral MOPs (RhMOP; [Rh₂(R-bdc)₂]₁₂), formed by bridging twelve dirhodium paddlewheel units with twenty-four isophthalate derivatives (R-bdc)²⁹, as the porous host for material synthesis (Fig. 1b). Due to the strong coordination of equatorial carboxylate groups to dirhodium units, RhMOPs feature high structural integrity and maintain the internal microporosity even after guest removal (thermal activation). By taking advantage of twelve available open metal sites at the axial position of the dirhodium moieties³⁰, twelve polymeric chains can be tethered on the MOP periphery through coordination bonds. The isotropic shape of cuboctahedron allows the globular surface modification like star-shaped polymers to restrict their effective molecular packing for crystallization³¹.

The polymer chains are designed based on three principles. (1) Polymer chains should be flexible enough to attain mobility and conformational deformability. (2) One of the terminals of polymer chains should contain a Lewis basic coordination moiety to bind to the open metal sites of MOP. (3) The other terminal should contain a bulky functional group to avoid the self-penetration³² into the MOP apertures and maintain the internal porosity. Following these principles, we synthesized a suitable polymeric compound, TX405-imi, featuring a flexible polyethylene glycol (PEG) chain with an imidazole at one end and a bulky tert-octylphenyl group at the other end (Fig. 1c and Supplementary Figs. 1–5).

One archetypal RhMOP, C₁₂RhMOP ([Rh₂(C₁₂-bdc)₂]₁₂; C₁₂-bdc = 5-dodecoxybenzene-1,3-dicarboxylate), was first employed to coordinate with TX405-imi due to their good solubilities in non-coordinative dichloromethane (DCM), which does not interfere the coordination reaction between the MOP and the polymer. After adding 12 molar eq. of TX405-imi into C₁₂RhMOP under stirring, the color of the solution turned from green to purple, which is attributed to the formation of coordination bonds between the dirhodium paddlewheel of MOP and the imidazole moiety of TX405-imi³³. Further removal of DCM yielded a purple, neat liquid phase product, which is expected to be a 12-armed star polymer with a MOP core ([C₁₂RhMOP(TX405-imi)₁₂]; hereafter named MOP-1A) (Fig. 1c).

UV-vis spectroscopy confirmed the coordination reaction of C₁₂RhMOP with TX405-imi in DCM. Upon titration, the maximum absorption band (λ_max) assigned to the π*-σ* transition of dirhodium paddlewheels³³ shifted from 656 to 570 nm after the addition of 12 molar eq. of TX405-imi (Fig. 2a), resembling the control experiment using the imidazole derivative of 1-dodecyl-1H-imidazole (diz) (656 to 561 nm λ_max shift), which has been also demonstrated in our previous study³⁴. The consistent blue shift suggests the formation of [C₁₂RhMOP(TX405-imi)₁₂] rather than the physical mixture of C₁₂RhMOP with TX405-imi. ¹H NMR analysis further supports the formation of coordination bonds between C₁₂RhMOP and TX405-imi (Supplementary Figs. 6 and 7).

Fig. 2: Structural characterization of MOP-1A.

a The UV-vis spectra of the DCM solution of C₁₂RhMOP coordinated by 1-dodecyl-1H-imidazole (diz) or TX405-imi with different equivalent ratios show the shift of maximum absorbance, confirming the average number of coordinating ligands per MOP. DOSY NMR (500 Hz) spectra of (b) TX405-imi, (c) C₁₂RhMOP and (d) MOP-1A in chloroform-d (CDCl₃) with the diffusion coefficient and the size of molecule estimated by the Stokes-Einstein equation.

Diffusion-ordered NMR spectroscopy (DOSY) was conducted to investigate the change in molecular size after the formation of MOP-1A (Fig. 2b–d). A set of signals in the spectra revealed the diffusion coefficient (D) values of 2.42 × 10^–10 m²/s for TX405-imi, 1.88 × 10^–10 m²/s for C₁₂RhMOP and 9.80 × 10^–11 m²/s for MOP-1A, respectively. The hydrodynamic diameter of MOP-1A in chloroform was estimated to be 8.3 nm by the Stokes-Einstein equation, significantly surpassing that of C₁₂RhMOP (4.3 nm). Dynamic light scattering (DLS) experiments (Supplementary Fig. 8 and Table 1) support this size difference: 6.2 nm for MOP-1A and 3.8 nm for C₁₂RhMOP, implying the formation of a larger MOP-1A molecule.

Controllable liquid-glass-crystal interconvertibility of MOP-1A

The phase transition behavior was first examined by differential scanning calorimetry (DSC). Inheriting the polymeric property from the tethered TX405-imi, MOP-1A showed a melting point of 25 °C, which is lower than that of neat TX405-imi (37 °C) (Supplementary Fig. 9a–c). The control DSC experiments of the physical mixture revealed the in situ formation of MOP-1A during the heating, confirming that the change in melting point stems from the coordination of TX405-imi with C₁₂RhMOP (Supplementary Fig. 9d). From the viscoelastic measurements, the loss modulus (G”) of MOP-1A also gave clear indications of its linear frequency dependence as a liquid phase in the range of 25–60 °C above the melting point (Supplementary Fig. 10).

As a polymeric substance, the state of MOP-1A is expected to vary upon different thermal treatment. For instance, a typical way to obtain a polymer glass is through quenching melted polymers, while the slow cooling of liquid affords its crystal phase³⁵ (Fig. 3a). By screening the DSC experimental conditions, we found that different from the neat TX405-imi, MOP-1A formed a glass phase with the glass transition temperature at −56 °C under a faster cooling rate (30 °C/min) (Fig. 3b–d and Supplementary Fig. 11). The subsequent heating of the glass MOP-1A induced the recrystallization at −29 °C, followed by the melting transition at 30 °C (Fig. 3b). In contrast, the slower cooling rate of 2 °C/min resulted in the crystal phase with a distinctive crystallization temperature at −24 °C and a melting point at 25 °C (Fig. 3d). At medium heating/cooling rate of 10 °C/min, both the crystallization and glass transition of MOP-1A can be observed (Fig. 3c), showing the dependence of phase transition on the thermal treatment.

Fig. 3: Liquid-glass-crystal interconvertibility of MOP-1A.

a Illustration of tunable phase of MOP-1A through designed thermal treatments. The DSC curves of MOP-1A scanned under different ramp rates of (b) 30 °C/min, (c) 10 °C/min and (d) 2 °C/min, each showing specific phase transition such as melting, crystallization, recrystallization and glass transition within the temperature range from -75 °C to 100 °C. Variable-temperature synchrotron XRD to monitor the structural changes in MOP-1A upon the thermal treatments including (e) the quenching process to acquire glass phase and (f) the slow cooling process to acquire crystal phase. (g) Photos of MOP-1A captured by variable-temperature polarized-light microscope through the same thermal treatment to the corresponding synchrotron XRD patterns in (f): The thermal treatment starts at 40 °C and decreases to −60 °C, then increases to 40 °C at a constant ramp of 6 °C/min.

To further confirm distinct phases, temperature-dependent synchrotron X-ray diffraction (XRD) measurements were performed on MOP-1A under different heating treatments. Glass MOP-1A formed by quenching the liquid MOP-1A at −60 °C ( > 80 °C/min cooling rate) showed a less-defined XRD pattern (Fig. 3e). By contrast, the crystalline MOP-1A obtained by gradual cooling (6 °C/min) to -60 °C showed characteristic diffraction peaks at 2θ = 9.6° and 11.8° (λ = 0.77489 Å), corresponding to the lamellar structure of PEG chains³⁶ (Fig. 3f). This crystallization process was observed to occur at −20 °C, consistent with the above DSC results. The variable-temperature polarized-light microscope images of the sample in the same cooling process also revealed the appearance of birefringent Maltese cross patterns at -20 °C and the continuous growth of spherulites up to −60 °C (Fig. 3g). After heating the sample back to 30 °C, both XRD and polarized-light microscope exhibited the disappearance of the diffraction pattern and spherulites, respectively. This reversible feature represents the phase transition of PEG moieties³⁷, supporting that the liquid-glass-crystal interconversion process of MOP-1A is dominated by the tethered polymer chains and can be controlled through tailored thermal treatments.

Generalizability of modular synthetic approach to synthesize phase-transformable MOPs

The replicable formation of the coordination bond between the dirhodium paddlewheels and polymer chains suggests that this modular approach can serve as a versatile tool to create a large library of MOPs with tunable phase transition behaviors. As a proof of concept, the other five types of phase-transformable MOPs were synthesized by the combination of three types of MOPs and two types of polymer chains (Fig. 4a). The three MOPs are differentiated by the functional groups substituted at the 5-position of benzene-1,3-dicarboxylate, where dodecoxy (−OC₁₂H₂₅), hydrogen (−H), and hydroxy (−OH) groups provide variant polarity on the MOP surface, denoted as C₁₂RhMOP, HRhMOP, and OHRhMOP, respectively. The two polymer chains are TX405-imi and TX305-imi with different polymerization degrees of PEG at 40 and 30, respectively. By mixing one of each species at the MOP/polymer molar ratio of 1:12, we acquired the six corresponding MOPs liquids at ambient conditions, named MOP-1A, MOP-2A, MOP-3A, MOP-1B, MOP-2B and MOP-3B (Fig. 4a). UV-vis spectra, DOSY NMR and DLS measurements were comprehensively performed to confirm the synthesis of all the MOP samples (Supplementary Figs. 12–15 and Table 1).

Fig. 4: Versatility of modular synthetic approach.

a Illustration of MOPs and polymers used for the fabrication of polymer-tethered MOPs with diverse phase transition behaviors. DSC curves (10 °C/min) of polymer-tethered MOPs, including (b) MOP-1A, (c) MOP-2A, (d) MOP-3A, (e) MOP-1B, (f) MOP-2B and (g) MOP-3B. The melting point (T_m) was labeled along with the estimated heat of fusion (ΔH_f).

The above results showed that the melting or crystallization of MOPs correlates to the degree of ordering of the PEG chain moiety. Therefore, the longer PEG chains are supposed to feature higher crystallinity and impart MOPs with higher melting points³⁸. From the DSC experiments (Fig. 4b–g), the MOPs tethered with the longer PEG chains of TX405-imi (MOP-1A, -2A and -3A) showed similar melting points at 25, 25 and 21 °C, respectively. With the shorter polymer of TX305-imi, MOP-1B melted at a lower temperature of 15 °C, while MOP-2B and MOP-3B did not show an obvious peak of crystallization with a much lower melting point at around 17 and 8 °C, respectively. These results suggest that the melting point is mainly affected by the designed PEG chain length.

When comparing the heat of fusion (ΔH_f) of each compound, we see the influence of the MOP core; MOP-1A, MOP-2A and MOP-3A showed ΔH_f = 37.2, 34.4 and 18.1 J/g, respectively (Fig. 4b–d). This result implies that the crystalline domains of PEG are strongly affected by the surface functional groups of the MOP, even though all MOP samples have the same tethered TX405-imi. A similar trend was observed for the series of MOPs with the shorter TX305-imi; MOP-1B, MOP-2B and MOP-3B showed ΔH_f = 13.3, 0.9 and 1.4 J/g, respectively (Fig. 4e–g). This difference in ΔH_f can be most likely attributed to the different intramolecular interactions between the substituent groups on MOPs and the polymer chains that facilitate or prohibit the formation of the crystalline domain.

To investigate the impact of substituent groups of MOP on the conformation and packing of polymer chains, we employed synchrotron small and wide angle X-ray scattering (SWAXS) experiments at 40 °C for the liquid MOP samples (Supplementary Fig. 16a). The SWAXS curves of all samples exhibited two characteristic peaks at q = ~ 0.8 and 0.4 Å^-1, reflecting the distances between the inner Rh atoms of the rhodium paddlewheel moieties at the adjacent ( ~ 0.8 nm) and opposing ( ~ 1.6 nm) positions, respectively (Supplementary Fig. 16b). The presence and positions of these two peaks further confirm the intact structural integrity of the MOP core, consistent with previous studies^39,40. MOP-1A, -2A and -3A showed humps at different q-values (0.183, 0.109 and 0.072 Å^–1), corresponding to d-spacings of 3.44, 5.77 and 8.68 nm, respectively. Similar trends were observed for MOP-1B, -2B and -3B with decreased d-spacings (3.27, 5.38 and 7.62 nm, respectively), consistent with the shorter chain length of TX305-imi. Considering the size of the MOP core determined by single-crystal X-ray crystallography (ca. 2.5 nm)²⁹, this d-spacing can be assigned to the distance between MOPs in the liquid phase, which should be dictated by the intramolecular compacting and/or intermolecular interaction of the tethered polymer chains^41,42.

We anticipate that the tethered polymers are compacted in the liquid MOP-1A and -1B due to their preferential interaction with the dodecoxy groups, which results in the shorter distance between the MOPs. On the other hand, the polymers are stretched out from the OHRhMOP cores in MOP-3A and -3B, leading to a longer inter-MOP distance. These SWAXS results, together with DSC and synchrotron XRD experiments, illustrate the corresponding crystallization behavior, further elucidating the impact of MOP cores on molecular-level interactions (Supplementary Figs. 17–22, Table 2). In detail, the compacting of PEG chains found in the liquid MOP-1A induces a higher crystallinity in the solid state with a higher ΔH_f value, while the stretching of PEG chains found in the liquid MOP-3A disturbs the ordered packing of MOPs, resulting in the decrease of ΔH_f.

Phase-dependent gas sorption and separation capability of MOP-1A

The high structural integrity of RhMOP under thermal activation allows us to explore the porosities of polymer-tethered MOPs by conducting CO₂ adsorption measurements. At 30 °C, the liquid MOP-1A and C₁₂RhMOP showed CO₂ adsorption of 0.98 and 2.77 cm³/g at around 100 kPa, respectively, while TX405-imi showed negligible uptake (Fig. 5a). The higher CO₂ uptake of MOP-1A than TX405-imi implies the contribution of C₁₂RhMOP intrinsic porosity. The normalized CO₂ uptake explained the origin of adsorption; MOP-1A and C₁₂RhMOP showed similar uptake of 1.5 and 1.3 moles of CO₂ per mole of MOP, respectively (Fig. 5b). This matching indicates the preservation of CO₂-accessible cavities of C₁₂RhMOP in MOP-1A without being blocked by the tethered polymer chains.

Fig. 5: The CO₂ sorption- and permeation-based separation of MOP-1A.

a CO₂ sorption isotherms of C₁₂RhMOP, MOP-1A, and TX405-imi measured at 30 °C, where C₁₂RhMOP exhibits as solid, MOP-1A exhibits as liquid, and TX405-imi as supercooled liquid. b CO₂ sorption isotherms of C₁₂RhMOP and MOP-1A measured at 30 °C with normalized unit. For each sample, three different sample batches were measured to obtain the corresponding CO₂ isotherms with error bars of each data point. c CO₂ sorption isotherms of MOP-1A at different states and temperatures. d Photos of processing MOP-1A into a membrane on an α-Al₂O₃ disc and the cross-sectional SEM image of the supported MOP-1A membrane. e CO₂ permeability of MOP-1A at different states and temperatures in comparison with TX405-imi. f CO₂/N₂ separation performance of MOP-1A and TX405-imi along with the performance upper bound for polymers reported in 2019⁴⁸. g CO₂/H₂ separation performance of MOP-1A and TX405-imi with the performance upper bound for polymers reported in 1999^46,47, compared to other reported porous materials including polymers, 2D materials, and MOFs (Supplementary Table 5). All the sorption, gas permeability and selectivity data of MOP-1A were plotted in purple to be distinguishable from the reference results.

The phase interconvertibility of MOP-1A allows us to prepare both liquid and crystal phases of MOP-1A at the same temperature (0 °C) and investigate the influence of physical states on their corresponding CO₂ accessibility. The liquid phase of MOP-1A is maintained by gradually cooling the liquid sample from 30 to 0 °C to afford a supercooled liquid without any recrystallization. In contrast, the quenching of liquid MOP-1A by liquid nitrogen (-196 °C) into glass state, followed by raising the temperature back to 0 °C, turns MOP-1A into a crystal phase by inducing the recrystallization process above its glass transition temperature (Fig. 3b). At around 100 kPa and 0 °C, the liquid MOP-1A adsorbed 0.32 cm³/g CO₂, while its crystal phase adsorbed 0.11 cm³/g (Fig. 5c). This three times lowered gas capacity in the crystalline MOP-1A can be attributed to both the generation of crystalline domains and the largely decreased molecular chain mobility, which suppress the diffusion of CO₂ molecules and thus gas accessibility to the MOP core⁴³. Note that raising the temperature from 0 to 30 °C increased the CO₂ capacity of liquid MOP-1A from 0.32 to 0.98 cm³/g (Fig. 5c), which is opposite to the gas adsorption tendency observed in most conventional microporous solids. Considering the increasing thermal motions of polymer chain segments at higher temperatures, the generation of more transient voids or free volume is anticipated to enhance both the gas diffusion and capacity in the liquid MOP-1A⁴⁴. All these results suggest that the CO₂ uptake within MOP-1A follows the solution-diffusion process, in which the accessibility of CO₂ to the MOP core is largely influenced by the molecular packing and mobility of tethered polymer chains³². The final CO₂ capacity of MOP-1A is determined by both the MOP porosity and the CO₂ diffusivity/permeability within the polymers tethered on MOP surfaces.

The tunable CO₂ permeability of MOP-1A, as well as its processibility imbued by phase convertibility, prompted us to fabricate a supported membrane for CO₂ separation application. A facile drop-coating method was applied by first adding a drop of liquid MOP-1A at the center of a mesoporous α-Al₂O₃ disc, followed by covering it with a polyvinylidene difluoride (PVDF) membrane to assist the liquid to spread and wet the surface of α-Al₂O₃, leading to the formation of a supported MOP-1A membrane with a thickness of approximately 100 μm (Fig. 5d). Notably, the liquid MOP-1A is viscous enough (236 Pa·s at 25 °C; Supplementary Fig. 10) to be localized as the fixed filler in α-Al₂O₃.

Gas permeation measurements were conducted to investigate the CO₂ permeability of MOP-1A at different states. The liquid MOP-1A showed a three-times higher CO₂ permeability (21.6 Barrer at 10 °C; 51.2 Barrer at 15 °C) than its crystal phase (7.1 Barrer at 10 °C; 19.3 Barrer at 15 °C) (Fig. 5e). Moreover, the temperature increase over the melting point of MOP-1A revealed continuously increasing permeability (21.6, 51.2, 107 and 152 Barrer at 10, 15, 30 and 40 °C, respectively), which is attributed to the higher free volume generated in the polymer moiety of MOP-1A^42,45. Consistent with the CO₂ adsorption results, the gas permeability of MOP-1A membrane can be controlled via switching the state and temperature.

The capability of MOP-1A to separate CO₂ from N₂ or H₂ was evaluated by performing the single-gas permeation tests of N₂ and H₂ under the same conditions (Supplementary Fig. 23 and Table 4). Based on the gas permeability, the corresponding ideal selectivities of CO₂/N₂ and CO₂/H₂ of MOP-1A at different phases and temperatures were estimated and plotted in comparison with TX405-imi and the Robeson upper bound^46,47,48 (Fig. 5f, g). At 10 °C, the crystalline MOP-1A showed an enhanced CO₂ permeability of 7.1 Barrer and CO₂/N₂ selectivity of 30.4. At such conditions, TX405-imi exists in a densely packed crystal phase that only allows Knudsen selectivity⁴⁹ (0.21) with a much lower permeability of CO₂ (0.03 Barrer) through defects at the grain boundaries among adjacent crystalline domains (Supplementary Table 3 and Fig. 24). Note that such low CO₂ permeability was even close to the detection limitation of the permeation setup and therefore cannot be accurately measured (Supplementary Fig. 25). In contrast, CO₂ mainly diffuses through the CO₂-philic PEG chains⁵⁰ in the pin-hole free MOP-1A membrane, affording high CO₂ selectively. Simultaneously, the high permeability of MOP-1A membrane can be attributed to the existence of accessible MOP cavities that facilitate faster gas diffusion. Furthermore, the CO₂/N₂ separation performance of MOP-1A can be tuned by changing the physical state of MOP-1A even at the same temperature; the liquid MOP-1A at 10 °C showed a higher CO₂ permeability of 21.6 Barrer than the crystalline membrane while still maintaining a CO₂/N₂ selectivity of 14.1 (Fig. 5f), which can be attributed to the simultaneous increase of both CO₂ and N₂ permeabilities in the liquid sample. Note that the previous studies of the ultrathin solid films of both C₁₂RhMOP and HRhMOP(diz)₁₂ showed a similar CO₂/N₂ selectivity (13 and 10) but a much higher CO₂ permeance of up to 195 GPU, which can be attributed to both the ultrathin thickness and high surface density of MOPs in these pure MOP films in the absence of tethered polymer chains^40,51.

For CO₂/H₂ separation, the MOP-1A membranes also show temperature- and phase-dependent permeability and selectivity (Fig. 5g). At 10 °C, the crystalline TX405-imi membrane is H₂-selective with an H₂/CO₂ selectivity of 15 (CO₂/H₂ selectivity = 0.07), while MOP-1A presented an opposite selectivity with CO₂/H₂ selectivity of 1.36 in the crystal phase and 2.29 in the liquid phase. Further increasing the temperature in the liquid phase provided a higher selectivity of 3.70 at 40 °C. This CO₂ selectivity over H₂ is distinctive from most of porous membranes made of polymers, 2D materials, and MOFs (Fig. 5g and Supplementary Table 6), which are usually H₂-selective due to the size exclusion mechanism^52,53. The unique selectivity of the MOP-1A membrane is attributed to the CO₂-philic PEG moieties that facilitate the dissolution and diffusion of CO₂ over H₂ and can be highly advantageous for hydrogen purification. Currently, the most prevalent method for generating hydrogen is through the steam methane reforming reaction, which yields a CO₂/H₂ molar ratio of 1:4⁵⁴. In contrast to H₂-selective membranes, the CO₂-selective MOP-1A membrane, which we propose here, excels at eliminating the low-concentration CO₂ from such gas mixture products and exhibits promising CO₂/H₂ selectivity for enhancing the overall separation efficiency.

To confirm the reproducibility and reliability of the liquid MOP-1A membrane in gas separation, different sample batches of liquid MOP-1A were tested in the permeation measurements of both single gas and gas mixtures at 30 °C (see Supplementary Tables 6-10). Compared to the pure TX405-imi membrane, MOP-1A gave a much more reproducible gas permeability and ideal selectivity with acceptable value variation among different batches (Supplementary Tables 7-9). With a mixed-gas feed consisting of CO₂ and N₂ in a 1:1 molar ratio, the permeabilities of CO₂ and N₂ in the mixture were 197 ± 18 and 7.18 ± 2.80 Barrer, respectively. The CO₂/N₂ selectivity was maintained at 30.4 ± 11.7 (Supplementary Table 10). For the mixed CO₂/H₂ (1:1 in mol) system, the liquid MOP-1A also revealed the reproducible permeabilities of CO₂ (180 ± 13 Barrer) and H₂ (24.6 ± 5.7 Barrer) with a high selectivity of CO₂/H₂ (7.51 ± 1.56) (Supplementary Table 11). These results demonstrate the reproducibility and the intact performance of MOP-1A in CO₂/H₂ and CO₂/N₂ separation even in the mixed-gas conditions.

Discussion

In summary, we showed the modular synthetic approach to produce phase-transformable RhMOPs. Tethering 12 predesigned polymer chains to the RhMOP surface by coordination bonds afforded a star polymer-like system featuring interconvertible liquid-glass-crystal phases by controlled heating protocols. The molecular entity of MOPs was preserved during the phase transition, leading to significantly increased CO₂ adsorption capacity compared to the neat polymer counterparts. The gas uptake and diffusion through the material were confirmed to be phase-dependent. By taking advantage of the processability given by phase interconvertibility, we fabricated a defect-free RhMOP-based membrane and demonstrated gas separation performance that can be controlled by phase change. Particularly, the liquid MOP-1A membrane showed CO₂ selectivity over H₂ with high permeability, providing an alternative type of processable porous membrane for challenging separation targets. Fundamentally, the modular synthetic strategy allows for the creation of a large library of phase-transformable porous materials by the combination of available MOPs with different metals and topologies and polymers with different structures and functions. The resulting versatile porous soft matters with designed phases pave a feasible way to realize their practical synthesis, fabrication and applications.

Methods

Materials

Triton X-405 (TX405) and Triton X-305 (TX305) were supplied as a 70% aqueous solution purchased from Sigma-Aldrich. Rhodium (III) chloride trihydrate was supplied by Tanaka Holdings Co. Isophthalic acid, 5-hydroxyisophthalic acid, triphenylphosphine, carbon tetrabromide, and imidazole were purchased from Tokyo Chemical Industry Co. Sodium acetate trihydrate and sodium carbonate were provided by Nacalai Tesque, Inc. Sodium hydride was supplied by Kanto Chemical Co., Inc. Solvents were purchased from Wako Pure Chemical Industries.

The porous α-Al₂O₃ discs, measuring 40 mm in diameter and 2 mm in thickness, were supplied by YOMO TECH. These discs comprise α-Al₂O₃ particles with an average particle size of approximately 100 nm and a porosity of 34%.

Synthesis of MOPs

Synthesis of rhodium acetate

RhCl₃·3H₂O (4.00 g, 15.2 mmol) and NaOAc·3H₂O (8.00 g, 58.8 mmol) were dissolved in 70 mL AcOH and 70 mL EtOH. The mixture was stirred and fluxed for 4 h under N₂, followed by naturally cooling down to precipitate the solid products. The precipitates were separated from the reaction mixture by filtration and then recrystallized by MeOH to obtain the pure solids of rhodium acetate ([Rh₂(AcO)₄·(MeOH)₂], ~ 3.30 g) (Supplementary Fig. 26).

Synthesis of 5-(dodecyloxy)isophthalic acid

Dimethyl-5-hydroxyisophthalate (4.00 g, 19 mmol) and K₂CO₂ (7.90 g, 57 mmol) were added in 50 mL dimethylformamide (DMF) and stirred at 70 °C for 20 min. 1-Bromododecane (5.28 g, 21 mmol) was then added into the solution, followed by reacting at 70 °C for 24 h. After cooling down, 200 mL water was added to the reaction mixture and the obtained precipitates were collected by filtration. Then the solids were washed with water and diethyl ether, then dried at 120 °C in vacuum overnight to obtain dimethyl-5-(dodecyloxy)isophthalate ( ~ 7.20 g) (Supplementary Fig. 27a). The obtained product (7 g, 19 mmol) was dissolved in 50 mL THF and 10 mL MeOH, followed by the addition of NaOH (3.00 g, 75 mmol) in 60 mL water. The mixture was heated at 65 °C overnight under stirring for reaction. The solvents of THF and MeOH in the reaction mixture were removed by evaporation and the remaining solution was neutralized by 37 wt% HCl. The precipitates were collected by filtration, washed by diethyl ether and water, and dried in vacuum to obtain the products ( ~ 6.70 g) (Supplementary Fig. 27b).

Synthesis of C₁₂RhMOPs

C₁₂RhMOP was synthesized following our previously published protocol³³. Rhodium acetate Rh₂(AcO)₄·(MeOH)₂ (100 mg, 0.2 mmol) was reacted with 5-dodecoxybenzene-1,3-dicarboxylic acid (174 mg, 0.5 mmol) in 10 ml of dimethylacetamide (DMA) in the presence of Na₂CO₃ (52 mg, 0.5 mmol) at 100 °C for 48 h. The resulting green solution was centrifuged, followed by precipitating the supernatant with MeOH to obtain C₁₂RhMOP solids. The solid was redissolved in dichloromethane, filtered, and evaporated. Finally, the solids were collected, redispersed and washed in EtOH, and dried overnight at 120 °C under vacuum to obtain the pure C₁₂RhMOP (~ 140 mg).

Synthesis of HRhMOPs

HRhMOP was synthesized according to our previous reports³⁴. Rhodium acetate Rh₂(AcO)₄·(MeOH)₂ (200 mg, 0.4 mmol) was reacted with benzene-1,3-dicarboxylic acid (328 mg, 1.97 mM) in 14 ml of DMA in the presence of Na₂CO₃ (210 mg, 2 mmol) at 100 °C for 48 h. The resulting precipitating solid was separated from the reaction mixture, followed by successive washing with DMA, EtOH and MeOH. Finally, the solids were collected and dried overnight at 120 °C under vacuum to obtain the MOP products ( ~ 160 mg).

Synthesis of OHRhMOPs

OHRhMOP was synthesized according to our previous reports³⁰. Rhodium acetate Rh₂(AcO)₄·(MeOH)₂ (200 mg, 0.4 mmol) was reacted with benzene-1,3-dicarboxylic acid (360 mg, 1.92 mM) in 14 ml of DMA in the presence of Na₂CO₃ (210 mg, 2 mmol) at 100 °C for 48 h. The resulting precipitating solids were washed with fresh DMA for three times, followed by dissolving in 35 mL water. HCl solution (37% HCl:H₂O = 3:97 v/v) was added into the green aqueous solution until a pH of ~3 to precipitate the OHRhMOP solids. The solids were collected, washed by water and redissolved in MeOH. The solution was dried by vacuum, and the residue was washed by acetone and diethyl ether. Finally, the powders were dried at room temperature under vacuum to obtain the MOP products ( ~ 210 mg).

Synthesis of 1-dodecyl-1H-imidazole (diz)

diz was synthesized following our previously published protocol³³. Imidazole (1.63 g, 24 mmol) and of KOH (1.4 g, 25 mmol) were mixed in acetonitrile (60 ml) and stirred at room temperature for 2 h. Then 1-bromododecane (4.98 g, 20 mmol) was added and the solution was stirred for 24 h at room temperature. Finally, the solvent was evaporated and water (35 ml) was added to the resulting solid. The product was extracted in 40 ml DCM for three times, treated with anhydrous MgSO₄, filtered, and dried under vacuum to obtain the products (~5.50 g) (Supplementary Fig. 27c).

Synthesis of polymeric ligands

Pretreatment of TX405 and TX305

Both TX405 and TX305 are commercially available as 70% solutions, which need to be dried prior to use to eliminate the water content. The drying procedure began by employing a spiral plug evaporator at a temperature of 90 °C for 8 h to eliminate the majority of the water. Following that, DCM was added to dissolve the resulting transparent and viscous liquids while still hot, preventing them from solidifying into a white solid. Furthermore, anhydrous Na₂SO₄ was introduced to remove the trace amount of water and filtered by vacuum filtration. The resulting solution was dried using a rotary evaporator and subjected to an 8 h drying period in a vacuum oven at 120 °C prior to use.

Synthesis of TX405-Br

In a round-bottom flask connected to the balloon filled with N₂, 4.00 g of TX405 (2.01 mmol), 1.05 g of triphenylphosphine (4.00 mmol), and 8 mL of a THF:DCM mixture (5:1 v/v) were combined and placed in an ice bath while vigorously stirring. Subsequently, a solution containing 1.33 g of carbon tetrabromide (4.01 mmol) and 8.5 mL of THF:DCM (5:1 v/v) was added to the mixture. The solution was left in the ice bath for 8 h, during which the temperature naturally increased to ambient conditions. The resulting light-yellow solution was collected using centrifugation before being dried with a spiral plug evaporator. The resulting viscous liquid was then subjected to purification through column chromatography on silica gel. The purification process involved using an initial eluent mixture of ethyl acetate:methanol (90:10 v/v), followed by a second eluent of DCM:methanol (80:20 v/v). Yield: 92%. ¹³C NMR (126 MHz, CDCl₃): δ 156.37, 142.34, 127.00, 113.76, 71.21, 70.57, 67.28, 50.74, 56.98, 37.92, 32.31, 31.76, 31.67, 30.33.

Synthesis of TX405-imi

In a two-necked round-bottom flask, 0.27 g of imidazole (3.97 mmol), 0.13 g of NaH (2.98 mmol; dispersed in 60% mineral oil), and 5.3 mL of DMF were combined. This flask was then connected to a condenser equipped with a balloon filled with N₂ and stirred at room temperature for 30 min. Afterward, a solution consisting of 4 g of TX405-Br (1.95 mmol) and 2 mL of DMF was injected into the mixture using a syringe while under vigorous stirring. Subsequently, the solution was further stirred and heated to 50 °C for 4 h. Once the reaction was completed, the mixture was transferred to a centrifuge tube, which was then connected to a spiral plug evaporator for the overnight removal of DMF. Finally, the resulting viscous liquid was purified using a flash column with a DCM:MeOH ratio of 6:4 (v/v) as the eluent. Yield: 69%. ¹H NMR (500 MHz, CDCl₃): δ 7.54 (s, 1H), 7.25 (dt, J = 8.9, 2.2 Hz, 2H), 7.03 (s, 1H), 7.00 (s, 1H), 6.83 (dt, J = 8.9, 2.2 Hz, 2H), 4.20–3.40 (m, 160H), 1.69 (s, 2H), 1.33 (s, 6H), 0.70 (s, 9H).

Synthesis of TX305-Br

TX305-Br was synthesized using the same procedure as TX405-Br. The starting materials were replaced by 4.00 g of TX-305 (2.58 mmol), 1.35 g of triphenylphosphine (5.15 mmol), and 1.71 g of carbon tetrabromide (5.16 mmol). Yield: 87%. ¹³C NMR (126 MHz, CDCl₃): δ 156.39, 142.36, 127.01, 113.77, 71.23, 70.58, 69.84, 67.30, 57.00, 37.93, 32.32, 31.76, 31.68, 30.33.

Synthesis of TX305-imi

TX305-imi was synthesized using the same procedure as TX405-imi. The starting materials were replaced by 0.45 g of imidazole (6.61 mmol), 0.23 g of NaH (5.27 mmol; dispersed in 60% mineral oil), and 4 g of TX305-Br (2.48 mmol). Yield: 70%. ¹H NMR (500 MHz, CDCl₃): δ 7.55 (s, 1H), 7.25 (dt, J = 8.9, 2.2 Hz, 2H), 7.04 (s, 1H), 7.01 (s, 1H), 6.82 (dt, J = 8.9, 2.2 Hz, 2H), 4.30–3.40 (m, 120H), 1.69 (s, 2H), 1.33 (s, 6H), 0.71 (s, 9H).

Synthesis of phase-transformable MOPs

For C₁₂RhMOP and HRhMOP-based samples, 1 eq. of MOP (1.85 mmol) was dissolved in 0.5 mL of DCM, while a solution containing 12 eq. of polymeric ligand (22.16 mmol) with 0.5 mL of DCM was prepared. These two solutions were mixed under continuous stirring at room temperature, resulting in the formation of a purple solution. The solution was subsequently dried using a spiral plug evaporator and a vacuum oven set at 80 °C overnight to yield neat MOP-1A, -2A, -1B and -2B. For OHRhMOP-based samples, the solvent was replaced by DCM:MeOH (3:1 v/v) to afford miscibility for both species, following the same procedure to obtain MOP-3A and MOP-3B.

Instrumentation and characterization

UV-visible spectroscopy of the solution samples was performed in a V-670 spectrophotometer (JASCO). 50 mg samples and 3 mL CHCl₃ were added to a quartz cuvette, forming a clear solution to be measured.

Diffusion-ordered NMR spectroscopy (DOSY), ¹H NMR spectra, and ¹³C NMR spectra were carried out on a Bruker Biospin Avance III 500 spectrometer. For analysis, 5 mg of sample was dissolved in 600 μL CDCl₃.

Differential scanning calorimetry (DSC) data were collected by using a DSC 3500 Sirius instrument (NETZSCH-Gerätebau GmbH, Germany).

Temperature-dependent XRD was performed at Taiwan Photon Source (TPS) 19 A station at the National Synchrotron Radiation Research Center (NSRRC). The X-ray at a wavelength of 0.77489 Å (16 keV) was generated from a cryogenic undulator under vacuum (CU15). Supplementary Fig. 28 shows the proprietary setup for the measurements by adopting our previous report⁵⁵. The MOP liquids were packed in a capillary tube with a diameter of 0.7 mm. The sample was heated/cooled at a ramp of 6 °C/min by the flowing nitrogen as the XRD patterns were continuously recorded at intervals of 10 °C. Each diffraction pattern was recorded with exposure duration of 10 s. The temperature was controlled using Oxford Cryostream liquid nitrogen cooling system 800.

Dynamic light scattering (DLS) were performed on a Zetasizer Nano ZS instrument (Malvern Instruments, Malvern, UK). The light source was a HeNe laser working at λ = 633 nm. The observations were made at the backscattering angle θ = 173°.

CO₂ gas sorption isotherms were recorded on a BELSORP-MINI X volumetric adsorption instrument from BEL Japan Inc. Prior to gas sorption measurement, the samples were activated at 120 °C for 12 h. At each condition, three different batches of MOP-1A samples were performed to obtain the corresponding CO₂ isotherms with error bars of each data point.

Polarized-light microscope photos of samples was captured by using an OLYMPUS BX51 optical microscopy equipped with polarizers. A Linkam FTIR 600 stage enabling temperature control over the samples was used during the observation.

The rheological measurements. Macroscopic rheological measurements of MOP-1A were carried out at 25 °C, 40 °C and 60 °C in a shearing mode with a stress-controlled MCR502 rheometer (Anton-Paar, Austria). A cone-plate geometry with a diameter of 25 mm and a cone angle 1° was used.

Small and wide angle X-ray scattering (SWAXS) data was conducted at the TPS 13 A BioSAXS beamline of the NSRRC. Liquid samples were loaded into washer of 1 mm X-ray path length and thermostat at 40 °C for data collection. With an X-ray beam energy of 15.0 keV (wavelength λ = 0.8266 Å) and two sample-to-detector distances of 2847 and 332 mm with the Eiger X9M SAXS detector and the Eiger X1M WAXS detector, respectively, the merged data covered a wide scattering vector q-range of 0.005 to 1.6 Å^-1. For comparison, the SWAXS profiles of the solution-state MOP sample in the solvent of acetone were collected with an adjusted X-ray beam energy of 19.5 keV (wavelength λ = 0.635 Å). The obtained data was further analyzed by Fourier transformation to obtain the distance distribution (p(r)) function (Supplementary Fig. 17) using the following equation:

$${{\rm{I}}}\left({{\rm{q}}}\right)=4{{\rm{\pi }}}{\int }_{0}^{{{{\rm{D}}}}_{\max }}{{\rm{p}}}({{\rm{r}}})\frac{\sin ({{\rm{qr}}})}{{{\rm{qr}}}}{{\rm{dr}}}$$

(1)

where I(q) represents the SWAXS intensity and D_max is the maximum dimension. This analysis allowed for the determination of the diameter (D_max) and shape of the structure.

Field emission scanning electron microscope (FE-SEM) was performed by JEOL JSM-7600F to evaluate the thickness of the MOP liquid on α-Al₂O₃. Before imaging, the samples were coated with platinum via sputtering deposition at a current of 10 mA for 90 s. SEM was performed under an acceleration voltage of 10 kV during image acquisition.

Gas separation setup and analysis

The liquid MOP-1A was transformed into membrane form to conduct single-gas permeation tests. Initially, we deposited approximately 50 mg of MOP-1A onto a 2 mm thick porous α-Al₂O₃ substrate (with a diameter of 40 mm, average particle size of 0.1 μm, and porosity of 0.34), forming a droplet (Fig. 5d). This droplet was then delicately pressed with a piece of polyvinylidene difluoride (PVDF) filtration membrane (with a pore size of 0.22 μm), which assists the liquid to spread and wet the surface of α-Al₂O₃, to create a thin layer of MOP-1A on the substrate. The substrate, along with the pressed MOP-1A and the PVDF filter, was placed in an oven at 35 °C for a minimum of 6 hours to stabilize MOP-1A on the substrate. Following this, the PVDF filter was removed, and a supported MOP-1A membrane with an approximately 100 μm thickness was formed due to the viscous nature of MOP-1A (236 Pa·s at 25 °C; Supplementary Fig. 10), as seen in the cross-sectional SEM image.

MOP-1A was masked with a piece of aluminum tape and secured using epoxy (3M^TM Scotch-Weld^TM Epoxy Adhesive DP100FR) prior to measurement (Supplementary Fig. 29). Only a small area of the membrane in a square shape remained exposed and was accessible to gas transport during the permeation measurement. Therefore, the membrane area was defined by the opening of applied epoxy resin. The single-gas permeation tests for the membrane samples were conducted using a homemade system⁵³ based on the constant-volume method (Supplementary Fig. 30). A masked membrane sample was outgassed at approximately 50 mtorr at room temperature for a minimum of 2 hours. For gas permeation tests involving the liquid phase of the sample, the system’s temperature was regulated either by a convection oven or a chilling system using ethylene glycol coolant, depending on the desired test temperature. Note that the lowest available temperature for the gas permeation tests is 10 °C to avoid the gas leakage issue. For gas permeation tests of samples at the solid crystalline phase, a chilling jacket with dry ice was employed to lower the temperature to −70°C for 40 minutes, facilitating the phase transition from liquid to solid. Subsequently, the jacket was removed, and the chilling system with ethylene glycol was utilized to control the temperature back to 10 and 15 °C to obtain the crystalline MOP-1A membrane for the gas permeation measurements.

The supported membrane of TX405-imi was fabricated by dropping the polymer liquid on the surface of mesoporous α-Al₂O₃ above its melting point (40 °C), followed by quickly freezing the sample in the refrigerator to obtain the polymer membrane in a solid state. Considering the low viscosity of liquid TX405-imi that can result in uncontrollable sample penetration through α-Al₂O₃ to create defects, the permeation experiment of TX405-imi membrane was only measured at 10 °C at its solid state.

Single-gas permeation analysis

In the typical measurement setup, a target gas (H₂, CO₂, or N₂) at a pressure of 1.1 to 1.2 bar was introduced on the feed side. The pressure on the product side of the membrane gradually increased as the target gas permeated from the feed side, and this was monitored using an MKS AA09A Baratron transducer. The increase in downstream pressure over time was then used to calculate the gas permeability of the membrane using the following equation:

$${{\rm{Permeability}}}=\,\frac{{{\rm{V}}}}{{{\rm{RTA}}}}\left(\frac{{{\rm{dp}}}}{{{\rm{dt}}}}\right)\frac{{{\rm{l}}}}{\triangle {{\rm{p}}}}$$

(2)

where R represents the gas constant, T is the temperature, A denotes the membrane area permitting permeation, V represents the volume of the downstream reservoir, dp/dt is the rate of change of pressure on the permeate side with respect to time, l signifies the thickness of the membrane, and ∆p represents the transmembrane pressure difference. The ideal selectivity of one gas species over another was determined as the ratio of their permeabilities obtained from their respective single-gas permeation tests. The effective membrane area used in Eq. (2) was determined by analyzing a photographic image of the membrane sample using an open-source software package, Image J⁵⁶. The thickness of the membrane samples was determined from their cross-sectional SEM images after the gas permeation measurements (Supplementary Figs. 31 and 32).

Mixed-gas permeation analysis

The same setup was employed for the mixed-gas permeation tests, which were conducted in a manner closely resembling the single-gas permeation tests. The feed gas consisted of either a CO₂/N₂ (50/50 in mol) or CO₂/H₂ (50/50 in mol) mixture, with a total pressure ranging from 1.1 to 1.2 bar at 30 °C. The gas composition on the permeate side was analyzed using a gas chromatograph (Shimadzu GC-2014) equipped with a thermal conductivity detector (TCD) and a Shincarbon-ST column. The separation factor for species i over j was calculated using the following equation:

$${{\rm{Separation}}}\; {{\rm{factor}}}=\frac{({{{\rm{y}}}}_{{{\rm{i}}}}/{{{\rm{y}}}}_{{{\rm{j}}}})}{({{{\rm{x}}}}_{{{\rm{i}}}}/{{{\rm{x}}}}_{{{\rm{j}}}})}$$

(3)

where x_i and y_i are the molar fractions of species i in the feed and permeate, respectively, and x_j and y_j are the molar fractions of species j in the feed and permeate, respectively.