Methane storage using metal–dipyrazolate frameworks Authors: Xiang-Jing Kong, Varvara I Nikolayenko, Alan C Eaby, Lenore Miller, Debobroto Sensharma, Daqiang Yuan, Hongliang Huang, Guang-Rui Si, Xia Li, Soumya Mukherjee, Leonard J Barbour, Brian Space, Xian-He Bu, Zuo-Ren Nie, Tao He, Jian-Rong Li, Michael J Zaworotko Published: 2026 ABSTRACT AbstractAdsorbed natural gas storage using porous materials at ambient temperature and relatively low pressure promises to address safety and cost concerns of conventional natural gas storage technologies (liquefaction and compression), but its utility is hindered by low deliverable capacity. Flexible porous materials such as metal–organic frameworks can exhibit isotherms with the potential to afford enhanced deliverable capacity. However, prototypal flexible adsorbed natural gas sorbents, exemplified by the metal–organic framework cobalt benzenedipyrazolate, Co(bdp), suffer from hydrolytic instability and are unsuitable for pelletization. Here we report a family of metal–bipyrazolate frameworks, including hydrophobic sorbents, Zn(dpt) and Co(dpt) (H2dpt = 2,5-di(1H-pyrazol-4-yl)thiophene), that exhibit methane-induced reversible transformations between narrow-pore and large-pore phases. Zn(dpt) shows exceptionally high methane deliverable capacities for 5–35 and 5–65 bar, that is, 173 cm3 (STP) cm−3 and 225 cm3 (STP) cm−3, respectively, but, unlike Co(bdp), is hydrolytically stable. In situ structural characterization, high-pressure gas sorption and modelling provide an insight into the narrow-pore–large-pore transformations, whereas testing with 250-ml tanks reveals that Zn(dpt) retains high deliverable capacity over multiple cycles. We demonstrate a practical alternative to pelletization through a formulation approach that is probably generally suitable for flexible sorbents.Subject terms: Metal-organic frameworks, Metal-organic frameworks Main Increasing energy demand coupled with ever-rising CO2 emissions makes the development of cleaner and greener energy sources an urgent global challenge1,2. Natural gas (NG), which is primarily composed of methane (CH4), is a transportation fuel that can alleviate our reliance on petroleum because of its abundance, high research octane number (RON = 107) and low/non-existent CO/CO2/NOx/SOx emissions3,4. However, safe high-capacity storage and on-demand delivery of NG remain a challenge that hinders its broader implementation. State-of-the-art NG storage technologies typically use cryogenic temperature (liquefied NG) or high pressure (compressed NG) to achieve high volumetric energy density, but both have latent hazards and high energy footprints for cooling or compression5. Adsorbed natural gas (ANG) storage at ambient temperature represents a promising alternative technology, both in terms of cost and safety, as much lower pressures can be used. The key to ANG is the development of better and cheaper physisorbents with high deliverable capacity. Various porous materials have been investigated and evaluated in this context, including metal–organic frameworks (MOFs), covalent organic frameworks (COFs), zeolites and activated carbons4–11. MOFs and COFs are of special interest as they can offer high surface area and tuneable structures. Current benchmark ANG sorbents include rigid materials featuring high uptake at high pressure, flexible materials that enable high deliverable working capacities, and sorbents that balance volumetric and gravimetric capacities10–13. Rigid sorbents have demonstrated high methane uptake at high pressure and low temperature, but for practical utility in vehicles, deliverable capacity is more relevant for evaluating performance6,7. In the case of ANG storage, deliverable capacity is often calculated as the difference in CH4 uptakes between 5 bar and 35 bar or 65 bar, which can be achieved through single- or two-stage compressors, respectively. Flexible MOFs that switch between narrow-pore (np) and large-pore (lp) phases offer promise for ANG as their stepped S-shaped isotherms can combine high lp uptake with low or no np uptake along with intrinsic heat management during adsorption–desorption cycles14. This situation is exemplified by cobalt benzenedipyrazolate (Co(bdp)), which revealed a deliverable capacity of 197 cm3 (STP) cm−3 (STP is defined here as 273.15 K and 101.325 kPa)13. Beyond adsorption performance, real-world deployment requires ANG sorbents to integrate hydrolytic stability, synthesis scalability, facile regeneration and good recyclability. In NG, water is an inherent impurity, as it naturally occurs and remains in the gas regardless of the extraction process or source. In North America, 68–120 mg N−1 m−3 (STP) of water vapour is present in typical NG15. Long-term exposure to such a level of water vapour will degrade moisture-sensitive MOFs. To render a suitably long service lifetime, hydrophobic frameworks sustained by robust metal–ligand bonds are preferred. Further, ANG sorbents should also feature mild conditions for scale-up synthesis, facile regeneration at relatively low temperature (energy footprint) and minimal structural fatigue over multiple adsorption–desorption cycles, that is, good recyclability, as well as low hysteresis and fast kinetics. Overall, the implementation of ANG storage lies in the development of sorbents that not only deliver high working methane capacities but also meet all of these demands for practical utility. Unfortunately, many top-performing sorbents lack at least one of these critical attributes, creating a barrier to the real-world deployment of ANG technologies. Crystal engineering approaches can enable systematic fine-tuning of materials such as MOFs, which are inherently modular in nature. Structure–function relationships can thereby be elucidated, especially in the context of gas adsorption and separation16. Here we use a crystal engineering approach to design a family of metal–dipyrazolate frameworks that can be regarded as second-generation variants of Co(bdp) (Fig. 1). Two flexible MOFs, namely, Zn(dpt) and Co(dpt) (H2dpt = 2,5-di(1H-pyrazol-4-yl)thiophene), exhibit the same connectivity and similar methane sorption profiles to those of Co(bdp). Further, despite lower surface area and higher density than Co(bdp), they set benchmarks for on-board methane storage from 5 bar to 35/65 bar at 298 K. This performance is enabled by more facile (lower-pressure) phase transformations. Crystal engineering approach to a family of metal–dipyrazolate frameworks. a–g, On the basis of the parent materials BUT-53 (a) and Co(bdp) (b), structural diversification of metal–dipyrazolate frameworks Zn/Co(dpn) (c), Zn(dpf) (e), Zn/Co(dpi) (f) and Zn/Co(dpt) (g) is achieved through ligand tuning (d). Single-crystal X-ray diffraction structure Although Co(bdp) offers benchmark performance in methane storage, its gate opening occurs at high pressure (20 bar) due to strong π–π stacking in the np phase13,17. To disrupt these interactions and access flexible materials with more suitable gate-opening pressures, four dipyrazolate ligands were synthesized by extending the central phenyl of H2bdp to a naphthalene core (H2dpn) or varying the bending angle (H2dpf, H2dpi and H2dpt, 120°–150°) with different central rings (Fig. 1d and Supplementary Figs. 1–4). Solvothermal reactions between these ligands and Co2+ or Zn2+ salts yielded single crystals of metal–dipyrazolate frameworks (Fig. 1c,e–g and Supplementary Figs. 5 and 6). Single-crystal X-ray diffraction (Supplementary Table 1) revealed that Co(dpn) adopts space group P43 (Fig. 1c and Supplementary Fig. 7), whereas Zn(dpf) (Fig. 1e and Supplementary Fig. 8), Zn/Co(dpi) (Fig. 1f and Supplementary Fig. 9) and Zn/Co(dpt) (Fig. 1g and Supplementary Fig. 10) adopt space group I41/amd or I41md. Zn(dpn) is isostructural with Co(dpn), and both can be regarded as reticular expansions of Co(bdp) (ref. 13); Zn(dpf) and Zn/Co(dpi) are isostructural with BUT-53 (ref. 18), whereas Zn/Co(dpt) features a distinct network structure. Zn(dpt) is a three-dimensional framework comprising the thiophene-centred angular ligand dpt2− and Zn-based rod building blocks (RBBs) (Fig. 1g). Zn2+ cations tetrahedrally coordinate pyrazolate (pz) moieties to generate RBBs of formula Zn(pz)2 (Supplementary Fig. 10). Similar RBBs were reported in Co(bdp) (ref. 13), BUT-53, BUT-54 and BUT-55 (ref. 18); however, the orientation/arrangement of metal ions varies, yielding different topologies and pore shapes/sizes. The linear analogue, bdp2−, allows Co(pz)2 RBBs in Co(bdp) to adopt an approximately linear geometry (Fig. 1b). Co(bdp) is a flexible framework with single-layered bridging ligands and large square or rhombic channels13. When bdp2− linkers are packed in a criss-cross fashion, the resulting Co(pz)2 RBBs are helical, forming double-walled BUT-55 as a rigid supramolecular isomer of Co(bdp) (ref. 18). BUT-53 and BUT-54 comprise angular ligands with a bending angle of 120° and the same helical Co(pz)2 RBBs (Fig. 1a). They also exhibit rigid double-walled structures with linkers packed in a face-to-face fashion. The helicity of Zn(pz)2 RBBs in Zn(dpt) falls between the linear RBBs in Co(bdp) and helical RBBs in BUT-53. The combination of such Zn(pz)2 RBBs with an angular ligand, dpt2− featuring a bending angle of 150°, enables Zn(dpt) to exhibit flexibility like Co(bdp). The dpt2− ligands adopt orientations that promote multiple C–H⋯π interactions with adjacent ligands in edge-to-face fashion, as well as multiple C–H⋯N interactions with surrounding ligands (Supplementary Fig. 10). Co(dpt) is isostructural with Zn(dpt) (Fig. 1g, Supplementary Fig. 10 and Supplementary Table 1). Overall, a bending angle of 120°–138° favoured double-walled networks sustained by π⋯π stacking, whereas flexible single-walled structures were generated by more linear ligands (bending angle, 150°–180°). This family of sorbents exhibits very different pore sizes: 14.0 Å for Zn/Co(dpn); 7.2 Å for Zn(dpf); 7.6 Å for Zn/Co(dpi); 12.1 Å (pore window 7.0 Å) for Zn/Co(dpt). Characterization and gas sorption The bulk-phase purity of the seven as-synthesized materials (denoted here as the α phases) was verified by powder X-ray diffraction (PXRD) experiments (Supplementary Figs. 11–17). Thermogravimetric analysis of activated samples (denoted here as β phases, degassed at 60 °C under vacuum after acetone exchange in a Soxhlet extractor) indicated no weight loss until decomposition at 400–500 °C (Supplementary Figs. 18 and 19). CO2 and N2 were used to probe gas-induced phase transformations (Supplementary Figs. 20–23). Among the seven materials, Zn(dpt) and Co(dpt) feature two-step isotherms whose steps appear at a relatively low pressure (Supplementary Figs. 21 and 22), indicating potential for methane storage. The specific surface area of the lp phase of Zn(dpt) was determined to be 1,900 m2 g−1 by Langmuir fitting to the N2 adsorption isotherms19. Hysteresis and steps also occur in the N2 and CO2 desorption isotherms. Steps for CO2 are at higher pressures (0.016 and 0.32 bar) than those for N2 (5.3 × 10−5 and 0.011 P/P0), which we attribute to the lower temperature of data collection for N2 given that CO2 has a larger quadrupolar moment and probably a stronger affinity for pore surfaces20. A small step is present in the CO2 isotherm at low pressure (<0.016 bar) and is not present for N2. Similar multistep N2 and CO2 sorption isotherms with hysteresis were recorded for Co(dpt) with a specific surface area of 2,115 m2 g−1 by Langmuir fitting to the N2 adsorption isotherms (Supplementary Fig. 22). Zn(dpt) and Co(dpt) exhibit different inflection points in their N2 and CO2 isotherms (Zn(dpt): 5.3 × 10−5 and 0.011 P/P0 for N2, 0.016 and 0.32 bar for CO2; Co(dpt): 8.5 × 10−4 and 0.023 P/P0 for N2, 0.020 and 0.32 bar for CO2 (Supplementary Figs. 21 and 22)), as reported previously for metal-substituted isostructural sorbents13,21. To investigate their potential utility for ANG storage, high-pressure CH4 adsorption isotherms were measured on a Xemis instrument at 298 K (Fig. 2a and Supplementary Figs. 24 and 25). A low CH4 uptake was observed at lower pressures before sharp steps at 17.5 and 27.5 bar for Zn(dpt). Although there is hysteresis in the desorption isotherm, the loop ends by 6.5 bar with CH4 uptake of <0.2 mmol g−1 below 5 bar (Fig. 2a). A low CH4 uptake was similarly observed for Co(dpt) at lower pressures before steps at 22.5 and 42.5 bar, with the hysteresis loop in the desorption isotherm closed by 18 bar (Supplementary Fig. 24). Whereas Zn(dpi) exhibits one step at 5 bar, none were observed for Zn/Co(dpn) and Co(dpi) (Supplementary Figs. 26 and 27). Excess CH4 uptakes at 65 bar for Zn(dpt) and Co(dpt) (10.74 and 10.95 mmol g−1) are higher than Zn/Co(dpn) and Zn/Co(dpi) (1.92–5.66 mmol g−1), BUT-53 (4.77 mmol g−1), BUT-55 (4.70 mmol g−1) and Co(bdp) (9.77 mmol g−1). For high-pressure CO2 sorption, both Zn(dpt) and Co(dpt) exhibit two steps (Fig. 2c and Supplementary Fig. 25) similar to those observed for CH4. Large hysteresis was observed in the CO2 desorption isotherms of Zn/Co(dpn) and Zn(dpi), unlike Co(dpi) (Supplementary Fig. 26). High-pressure gas sorption isotherms and in situ VP-PXRD patterns of Zn(dpt) recorded at 298 K. a,c, High-pressure CH4 (a) and CO2 (c) sorption isotherms. b,d, In situ VP-PXRD patterns upon loading/unloading of CH4 (b) and CO2 (d). Source data Structural transformation To characterize the structural changes associated with the stepped CH4 and CO2 adsorption isotherms of Zn(dpt) and Co(dpt), we conducted in situ PXRD and high-pressure sorption coincident measurements (Fig. 2b,d and Supplementary Figs. 28–30)22. Activation was performed by exchanging DMF with acetone, packing the exchanged samples into a Lindemann capillary, attaching the capillary to a bespoke environment cell, heating at 60 °C under vacuum, and collecting the PXRD data under vacuum as well as at specific gas pressures during CH4 and CO2 sorption. The data for CH4-induced phase transformations in Zn(dpt) are presented in Fig. 2b. The as-synthesized Zn(dpt)-α exhibited its characteristic low-angle peak at 7.0°. After activation, low-angle peaks of Zn(dpt)-β appeared at 7.8° and 9.0°. The peak at 7.8° disappeared as a new peak at 8.1° appeared upon loading of CH4 at 10 bar, by which point the peak at 9.0° became stronger and another new peak appeared at 10.0°. These data are indicative of the formation of an intermediate Zn(dpt)-γ1 phase. The CH4 uptake of Zn(dpt)-γ1 is 1.00 mmol g−1 at 10 bar (Fig. 2a). This phase is consistent with that of the step occurring between 2 and 7 bar in the high-pressure CO2 sorption isotherm (Fig. 2d). The PXRD pattern did not change markedly at 20 bar (8.1°, 8.9° and 9.9°) versus that at 10 bar but a step was observed in the sorption isotherm between 10 and 20 bar, implying pore filling of the γ1 phase and the appearance of a more open Zn(dpt)-γ2 phase with an uptake of 4.88 mmol g−1 at 20 bar. When the CH4 pressure was increased to 30 bar, the characteristic low-angle peak of the high-porosity Zn(dpt)-α′ phase (7.1°) reappeared, whereas the γ2 phase peaks at 8.2°, 8.8° and 9.9° remained. This pattern is consistent with a mixture of the α′ and γ2 phases. The second CH4 adsorption step at around 30 bar is consistent with the transformation from γ1 to α′, as seen at 10 bar for CO2 (Fig. 2). Framework expansion to the α′ phase is complete at 40 bar with a CH4 uptake of 9.47 mmol g−1. The sample was then reactivated, affording a PXRD pattern consistent with the β phase, indicating reversibility. The two steps in the 77-K N2 and 195-K CO2 sorption isotherms of Zn(dpt) can, therefore, be attributed to structural transformations as the framework switches from β to γ1, and then from γ2 to α′ (Supplementary Figs. 21 and 22). The structures of the np and lp phases of Zn(dpt) were solved by indexing in situ diffraction data at 0 bar, 10 bar, 20 bar and 40 bar, and constructing preliminary structures by adopting the indexed cell parameters, followed by Le Bail refinements (Supplementary Section 5, Supplementary Figs. 31–34 and Supplementary Table 2)13. The crystal structures of Zn(dpt)-β (0 bar), intermediate Zn(dpt)-γ phases (10 and 20 bar) and Zn(dpt)-α′ (40 bar) were optimized by density functional theory (DFT) calculations (Supplementary Section 6, Supplementary Figs. 35–38 and Supplementary Table 3) and are presented in Fig. 3. The structures of β, γ1 and γ2 adopted space group Fddd, whereas the α′ phase adopted the same space group, I41/amd as Zn(dpt)-α (Supplementary Tables 1 and 2). Even though the porosity changes dramatically from Zn(dpt)-β to Zn(dpt)-α′, the Zn2+ cations exhibit pseudo-tetrahedral geometry in all phases and connectivity of the Zn(pz)2 RBBs is retained. The only differences lie with the orientation and configuration of the dpt2− ligands (Supplementary Figs. 39–42). Non-covalent C–H⋯π and/or C–H⋯N interactions between adjacent ligands sustain these structures throughout the whole transformation process. Computational structure solutions illustrate the phase transformation of Zn(dpt) under continuous CH4 loading. Top: structure evolution from np Zn(dpt)-β (0 bar) to intermediate phase Zn(dpt)-γ2 (20 bar) and lp Zn(dpt)-α′ (40 bar). Middle: dihedral angle between terminal pz groups (blue and green planes) and the central thiophene moiety (red plane). Bottom: non-covalent interactions (C–H⋯N in blue, π⋯π in green and C–H⋯π in pink) between adjacent ligands. The experimental crystal structures of Zn(dpt)-β′, Zn(dpt)-γ2′ and Zn(dpt)-α″ (Zn(dpt) soaked in chlorobenzene) exhibit similar unit-cell parameters to those of the simulated Zn(dpt)-β, Zn(dpt)-γ2 and Zn(dpt)-α′ and are, therefore, consistent with the simulations (Supplementary Tables 4 and 5 and Supplementary Figs. 43–45). Interestingly, the central thiophene moiety of dpt2− ligands in Zn(dpt)-α″ adopts a slightly different orientation from the α′ phase (Supplementary Figs. 46 and 47), indicating that thiophene is able to rotate upon guest loading23. This can explain why Zn(dpt)-γ1 and Zn(dpt)-γ2 have similar PXRD patterns even though the transformation from Zn(dpt)-γ1 and Zn(dpt)-γ2 leads to a step at 20 bar. DFT calculations further revealed that relative to Co(bdp), Zn(dpt) has a more flexible structure that can more readily transform on methane loading/unloading (Fig. 2a,b, Supplementary Figs. 48–52, Supplementary Tables 6 and 7 and Supplementary Section 6). As methane does not induce the transformation of Co(bdp) to its largest pore phase below 65 bar, the second transformation (γ2 to α′) is precluded. These findings suggest that Zn(dpt) and Co(dpt) should outperform Co(bdp) in the working pressure region for ANG storage despite the lower surface area and higher density. The calculations also suggest that the sulfur-containing dpt2− linkers provide additional framework stability through non-covalent interactions. Such interactions are likely to enable the reversible phase transformations seen for Zn(dpt). Co(dpt) exhibits similar two-phase transformations but at higher loading pressures. These results indicate that linker substitution is more impactful than metal substitution with respect to threshold pressures. CH4 storage HKUST-1 exhibits the highest volumetric CH4 uptake reported so far, 267 cm3 (STP) cm−3 at 65 bar and 298 K. Unfortunately, its relatively high uptake at 5 bar reduces deliverable CH4 working capacity to 190 cm3 (STP) cm−3 (ref. 24). The deliverable CH4 working capacity of Zn(dpt) reaches 225 cm3 (STP) cm−3 between 5 and 65 bar at 298 K (Fig. 4a,b and Supplementary Figs. 53 and 54). Owing to its negligible uptake at 5 bar, it outperforms other porous materials in terms of the CH4 working capacity13,25–30 (Fig. 4b and Supplementary Table 8). The cobalt analogue of Zn(dpt), Co(dpt), has a similar CH4 working capacity—227 cm3 (STP) cm−3. Moreover, 5–35 bar is also relevant as 35 bar—the pressure typical of a single-stage compressor3,6. The CH4 working capacity of Zn(dpt) (173 cm3 (STP) cm−3) exceeds the previous benchmark, Co(bdp) (155 cm3 (STP) cm−3), even though Zn(dpt) has a smaller surface area than Co(bdp) and several rigid MOFs (Figs. 5a and 4b and Supplementary Table 8)10,11,13,14,20,24–29,31. The deliverable CH4 working capacity of Co(dpt) is 121 cm3 (STP) cm−3 for 5–35 bar, much lower than that of Zn(dpt) because of its different threshold pressures. The volumetric capacity of Zn(dpt) is 262 cm3 (STP) cm−3 at 298 K and 80 bar, comparable with the rigid COF benchmark, three-dimensional TFB-COF-Et (264 cm3 (STP) cm−3 at 100 bar)10. These results were validated by repeating measurements on independently produced Zn(dpt) samples, and testing reproduced HKUST-1 (representative of rigid structures) and Co(bdp) (representative of flexible structures) samples using the same instrument and conditions (Supplementary Figs. 27 and 55). Further, the adsorption behaviour of Zn(dpt) was independently confirmed by volumetric measurements (Supplementary Fig. 56). The calculated net adsorption32 capacities also highlight the advantages of materials reported here (Supplementary Figs. 57 and 58). High-pressure CH4 sorption properties of Zn/Co(dpt) at 298 K. a, High-pressure CH4 sorption isotherms of Zn/Co(dpt) provide deliverable CH4 capacity between 5 and 35 bar and between 5 and 65 bar. ads, adsorption; des, desorption. b, Comparison of deliverable CH4 capacity between 5 and 35 bar (abscissa) and between 5 and 65 bar (ordinate) among top-performing MOFs and COFs calculated from sorption isotherms at 298 K unless otherwise stated (unit: cm3 (STP) cm−3 (crystal volume)), crystallographic density basis; the deliverable CH4 capacities of MFU-4l-Li and NU-1500-Al were calculated from adsorption isotherms recorded at 296 K; UTSA-110, Co(bdp) and Fe(bdp) were determined from 5.8 to 35/65 bar. c,d, 100 cycles of CH4 pressure swing between 5 and 65 bar for Zn(dpt) (c) and Co(dpt) (d). Source data Scale-up synthesis and methane storage performance of Zn(dpt). a, Images of scaled Zn(dpt) and HKUST-1 samples. b, Tanks were loaded with Zn(dpt) sorbent and elastic filler in two ways (Zn(dpt)-1, Zn(dpt) in the middle with elastic filler covering both ends; Zn(dpt)-2, Zn(dpt) layers evenly spaced by thin elastic filler layers (yellow, Zn(dpt); green, elastic filler)). In addition, a reference tank was loaded with HKUST-1. c, Comparison of the average volumetric deliverable capacities of the three tanks based on crystallographic density (cm3 (STP) cm−3 (crystal volume)) (top) and actual packing density (cm3 (STP) cm−3 (packing volume)) (bottom) over five cycles. d, Gas composition analysis of the released stream in different charging/discharging cycles of Zn(dpt)-1 and HKUST-1 tanks, using methane gas containing 5% ethane between 5 and 65 bar. e, Radar chart comparing Zn(dpt), Co(bdp) and HKUST-1. Source data As mentioned earlier, the potential utility of rigid adsorbents for ANG storage is mitigated by their type-I isotherms and resulting high uptake at 5 bar (ref. 7), while flexible MOFs that exhibit stepped sorption isotherms promise higher deliverable working capacity13,14,20. In the case of Zn(dpt), there are two steps between 10 bar and 35 bar, as well as hysteresis. The pressure range in which they occur is key to the high working capacity of 173 cm3 (STP) cm−3 between 5 and 35 bar at 298 K (Fig. 4a), which also contributes to the high CH4 working capacity of 225 cm3 (STP) cm−3 between 5 and 65 bar at 298 K (Fig. 4a). Both are benchmarks for physisorbents (Fig. 4b and Supplementary Table 8). Pressure swing experiments revealed that Zn(dpt) retained 95% of its deliverable capacity after 100 adsorption/desorption cycles between 5 and 65 bar at 298 K (Fig. 4c and Supplementary Fig. 59a). We attribute the decrease in working capacity to changed hysteresis, as also observed for CH4 isotherms recorded after cycling (Supplementary Fig. 60a). To our knowledge, this is the only demonstration of an MOF’s methane storage performance validated over 100 cycles of 65-bar adsorption and 5-bar desorption. The weaker performance of Co(dpt) between 5 and 35 bar arises from the second step that occurs above 35 bar (42.5 bar, Fig. 4a), whereas the CH4 working capacity between 5 and 65 bar is less affected by it (Fig. 4b and Supplementary Table 8). Co(dpt) also demonstrated recyclability in pressure swing tests between 5 and 65 bar for 100 cycles, with 96% retention of working capacity (Fig. 4d and Supplementary Fig. 59b). The inflection point of the adsorption steps shifts to higher pressure, leading to the slight decrease in working capacity (Supplementary Fig. 60b). Both increased hysteresis and shifted inflection point can be attributed to reduced particle size over successive pressure swing measurements (Supplementary Figs. 61–65)30, a phenomenon commonly seen for flexible materials (Supplementary Figs. 66–68). The crystallinity of Zn(dpt) and Co(dpt) was observed to be retained after long-term cycling measurements, as evidenced by unchanged PXRD patterns (Supplementary Fig. 15). The substitution of the linear bdp2− linker in Co(bdp) to the slightly angular dpt2− variant weakens the π⋯π interactions that stabilize the np phases of Zn(dpt)/Co(dpt). As a result, their more flexible structures feature multiple isotherm steps in the working pressure region appropriate for methane storage. Stability, recyclability and practical utility The inherent water vapour in NG streams corresponds to a relative humidity (RH) range of 0.04%–0.97% in the working pressure range of 5–65 bar. Although these RH values are low, long-term exposure to such a level of water vapour is expected to gradually degrade moisture-sensitive MOFs. In addition, materials are inevitably exposed to the higher RH levels of ambient air when filling the tank. Water vapour sorption isotherms, accelerated humidity stability tests (long term) and ambient humidity stability (short term) should be considered when evaluating the hydrolytic stability of a methane sorbent. Accelerated humidity stability testing of the β phases of Zn(dpt) and Co(dpt) revealed that they retained crystallinity after exposure to 75% RH at 40 °C for 1 month or air exposure for 6 months (Supplementary Figs. 69 and 70)33,34. By contrast, Co(bdp) decomposed when exposed to air for 24 h after activation, and PXRD peaks broadened after soaking in DMF (Supplementary Fig. 16a). The hydrolytic stability of Zn(dpt) and Co(dpt) could be related to their hydrophobic nature, as suggested by the low water uptake of 0.41 wt% and 0.74 wt% at 298 K and 95% RH (Supplementary Fig. 71). Humidity-exposed Zn(dpt)-β and Co(dpt)-β transformed to the corresponding α phases when soaked in DMF at 298 K (Supplementary Fig. 15). Zn(dpn) and Zn/Co(dpi) are also hydrophobic, whereas Co(dpn) exhibits S-shaped isotherms with a water vapour uptake of 11.2% at 60% RH (Supplementary Figs. 72 and 73). Zn(dpn) and Zn(dpi) retained their PXRD patterns after water vapour sorption experiments; however, patterns for the cobalt counterparts deteriorated (Supplementary Figs. 11–14). Zn(dpt) was scaled to 5 g per reaction with retained crystallinity and porosity (Supplementary Figs. 74–76). The scaled Zn(dpt) exhibits the same CH4 sorption isotherm as a vial-synthesized sample (Fig. 5a and Supplementary Fig. 72). Effects of minor components of NG such as C2H6 were also examined (Supplementary Fig. 77). As observed for CH4, the high-pressure isotherm of C2H6 at 298 K exhibits two steps. Both steps appear earlier (1 bar and 5 bar) than for CH4 (20 bar and 30 bar), presumably due to stronger interactions between C2H6 and the sorbent (Supplementary Fig. 77a,b). For a 95/5 mixture of CH4/C2H6, 5–65 bar corresponds to a C2H6 partial pressure of 0.25–3.25 bar. Cycling C2H6 sorption experiments (0.25–15 bar at 298 K; Supplementary Fig. 77c,d) suggest that C2H6 does not accumulate during cycling. Although rigid materials are amenable to shaping into pellets to enhance packing density and recyclability, flexible materials can undergo large volume changes, which tend to render them unsuitable for pelletization. To address this matter, we looked for an alternative formulation approach by using elastic filler to support the flexible sorbent. Such formulations can provide space for the phase transformations of the sorbent, enabling the sorbent to achieve the full capacity consistent with its structure, but without substantial increase in tank volume, energy input or capital cost. Their weight changes on consecutive CH4 loading/unloading cycles were recorded, with HKUST-1 serving as a reference (Fig. 5b, Supplementary Figs. 78–87 and Supplementary Section 7). The tanks demonstrate comparable volumetric deliverable CH4 capacity over five cycles (Supplementary Figs. 81–83). The average deliverable capacities of three tanks were also compared based on crystallographic density and actual packing density (Fig. 5c). Zn(dpt) outperforms HKUST-1 in both cases (219 versus 139 cm3 (STP) cm−3 (crystal volume) and 133 versus 49 cm3 (STP) cm−3 (packing volume)). Tank experiments involving Zn(dpt) (5–65 bar, 95/5 mixture of CH4/C2H6) also demonstrate consistent deliverable capacity and retention of gas composition over 30 cycles (Fig. 5d and Supplementary Figs. 24–86). In addition to very high isotherm/tank-deliverable CH4 capacities, Zn(dpt) features excellent hydrolytic stability, minimal C2H6 impact, synthesis scalability, mild activation and good recyclability (Fig. 5e and Supplementary Table 9). To our knowledge, Zn(dpt) represents the only high uptake methane sorbent that meets all these criteria. The methane storage performance of Zn(dpt) and Co(dpt) is governed by their methane-induced structural transformations. Zn(dpt) exhibits isotherm steps associated with the β → γ → α′ transition from ~10 to 35 bar whereas Co(dpt) exhibits corresponding steps in the range of ~10–55 bar. These pressure ranges fall within the operating window of ANG systems (5–35/65 bar). Such alignment enables pore expansion upon charging and pore contraction upon depressurization, producing a sharp uptake increase within the operating window. That the threshold pressure for the γ → α′ transformations occurs at much lower pressure than for Co(bdp) is key to the relative higher performance of Zn(dpt) and Co(dpt), which are inherently denser, lower-uptake materials than Co(bdp). In the case of Zn(dpt), its particularly favourable pressure thresholds underpin the high deliverable capacity observed at both material and tank levels. Rational evaluation of performance parameters on the system level allows for engineering a porous compound to have utility as a methane sorbent. In particular, by taking the nature of sorbents into account during device design, there is an opportunity to exploit the full capacity promised by the inherent chemistry of either rigid or flexible sorbents. By regulating the length and bending angle of dipyrazolate ligands, a family of metal-RBB-based frameworks is reported, with diverse structures and different degrees of flexibility. In addition to hydrolytic stability, facile scale-up, mild regeneration and good recyclability, Zn(dpt) was found to exhibit two-step CH4 adsorption isotherms that enable benchmarks for ANG storage performance in the operating window of 5–35/65 bar at 298 K. We also address the unsuitability of pelletization for flexible materials and introduce an appropriate formulation approach for loading tanks. This research further highlights how a crystal engineering approach to materials design can enable systematic fine-tuning of structure–function relationships in flexible MOFs, in this case affording enhanced stability and more facile transformation to the lp phase compared with the parent material, Co(bdp). Methods Synthesis of ligands The synthesis procedures of the four ligands, namely, H2dpt, H2dpn, H2dpf and H2dpi, are described in Supplementary Section 2 and Supplementary Figs. 1–4. Single-crystal synthesis of Zn(dpt) (BUT-5) Light yellow rod-like crystals of Zn(dpt) were synthesized from solvothermal reactions between H2dpt ligand (20.0 mg, 0.0920 mmol) and Zn(NO3)2·6H2O (40.0 mg, 0.135 mmol) in a mixed solvent of DMF:H2O:AcOH (5 ml:3 ml:0.01 ml) (a vial with a volume of 20 ml) at 80 °C for 24 h. The as-synthesized sample was washed with fresh DMF (10 ml × 3) until the supernate becomes colourless, and then exchanged with fresh acetone (10 ml × 6) within 2 days. After exchange, the sample was loaded to a sorption tube, and activated at 60 °C under vacuum for 6 h (22.7 mg, yield 87.8% based on H2dpt). Single-crystal synthesis of Co(dpt) (BUT-6) Purple rod-like crystals of Co(dpt) were synthesized from solvothermal reactions between H2dpt ligand (20.0 mg, 0.0920 mmol) and Co(NO3)2·6H2O (40.0 mg, 0.138 mmol) in a mixed solvent of DMF:H2O (5 ml:0.8 ml) (a vial with a volume of 20 ml) at 150 °C for 24 h. The as-synthesized sample was washed with fresh DMF (10 ml × 3) until the supernate becomes colourless, and then exchanged with fresh acetone (10 ml × 6) within 2 days. After exchange, the sample was loaded to a sorption tube, and activated at 60 °C under vacuum for 6 h (21.6 mg, yield 85.5% based on H2dpt). Single-crystal synthesis of Co(dpn) (BUT-182) Purple rod-like crystals of Co(dpn) were synthesized from solvothermal reactions between H2dpn ligand (20.0 mg, 0.0770 mmol) and Co(NO3)2·6H2O (40.0 mg, 0.138 mmol) in a mixed solvent of DMF:H2O:AcOH (5 ml:0.5 ml:0.01 ml) (a vial with a volume of 20 ml) at 135 °C for 12 h. The as-synthesized sample was washed with fresh DMF (10 ml × 3) until the supernate becomes colourless, and then exchanged with fresh acetone (10 ml × 6) within 2 days. After exchange, the sample was loaded to a sorption tube, and activated at 60 °C under vacuum for 6 h (17.6 mg, yield 72.2% based on H2dpn). Synthesis of Zn(dpn) (BUT-183) Colourless microcrystalline powder of Zn(dpn) was synthesized from solvothermal reactions between H2dpn ligand (20.0 mg, 0.0770 mmol) and Zn(NO3)2·6H2O (40.0 mg, 0.135 mmol) in a mixed solvent of DMF:H2O:AcOH (5 ml:1 ml:0.01 ml) (a vial with a volume of 20 ml) at 80 °C for 24 h. The as-synthesized sample was washed with fresh DMF (10 ml × 3) until the supernate becomes colourless, and then exchanged with fresh acetone (10 ml × 6) within 2 days. After exchange, the sample was loaded to a sorption tube, and activated at 60 °C under vacuum for 6 h (20.4 mg, yield 82.0% based on H2dpn). Single-crystal synthesis of Zn(dpf) (BUT-184) Colourless rod-like crystals of Zn(dpf) were synthesized from solvothermal reactions between H2dpf ligand (20.0 mg, 0.100 mmol) and Zn(NO3)2·6H2O (40.0 mg, 0.135 mmol) in a mixed solvent of DMF:H2O:AcOH (5 ml:1.5 ml:0.01 ml) (a vial with a volume of 20 ml) at 80 °C for 24 h. The as-synthesized sample was washed with fresh DMF (10 ml × 3) until the supernate becomes colourless, and then exchanged with fresh acetone (10 ml × 6) within 2 days. After exchange, the sample was loaded to a sorption tube, and activated at 60 °C under vacuum for 6 h (12.3 mg, yield 46.7% based on H2dpf). Single-crystal synthesis of Zn(dpi) (BUT-185) Light yellow rod-like crystals of Zn(dpi) were synthesized from solvothermal reactions between H2dpi ligand (20.0 mg, 0.0920 mmol) and Zn(NO3)2·6H2O (40.0 mg, 0.135 mmol) in a mixed solvent of DMF:H2O:AcOH (5 ml:1.5 ml:0.01 ml) (a vial with a volume of 20 ml) at 80 °C for 24 h. The as-synthesized sample was washed with fresh DMF (10 ml × 3) until the supernate becomes colourless, and then exchanged with fresh acetone (10 ml × 6) within 2 days. After exchange, the sample was loaded to a sorption tube, and activated at 60 °C under vacuum for 6 h (16.9 mg, yield 65.4% based on H2dpi). Single-crystal synthesis of Co(dpi) (BUT-186) Purple rod-like crystals of Co(dpi) were synthesized from solvothermal reactions between H2dpi ligand (20.0 mg, 0.0920 mmol) and Co(NO3)2·6H2O (40.0 mg, 0.138 mmol) in a mixed solvent of DMF:H2O:AcOH (5 ml:0.5 ml:0.01 ml) (a vial with a volume of 20 ml) at 135 °C for 12 h. The as-synthesized sample was washed with fresh DMF (10 ml × 3) until the supernate becomes colourless, and then exchanged with fresh acetone (10 ml × 6) within 2 days. After exchange, the sample was loaded onto a sorption tube, and activated at 60 °C under vacuum for 6 h (18.1 mg, yield 71.6% based on H2dpi). Variable-pressure PXRD High-pressure CO2 (from 0 to 25 bar) and CH4 (from 0 to 40 bar) PXRD experiments were carried out for both Zn(dpt) and Co(dpt) using a PANalytical X’Pert PRO instrument with Debye–Scherrer geometry. Intensity data were recorded using an X’Celerator detector, and 2θ scans in the range of 3°−35° were performed with a step size of 0.017° at a scan speed of 0.02° s−1. During the experiment, the powdered samples were exposed to Cu Kα radiation (λ = 1.5418 Å). The acetone solvates of Zn(dpt) and Co(dpt) were individually packed and sealed within Lindemann glass capillaries, which were attached to proprietary environmental gas cells and evacuated ex situ. These environment cells were then sealed and mounted onto a PANalytical X’Pert PRO instrument where vacuum PXRD was recorded. Following this, each capillary was then progressively loaded with CO2 ex situ using a gas manifold (the pressures selected correspond to events observed in the gas sorption isotherms collected at the same temperature), which allowed to equilibrate at this pressure for at least 6 h and variable-pressure (VP) PXRD patterns were measured at a constant temperature of 298 K. Both samples were then evacuated, and the CH4 gas sorption experiments were performed in the same way. High-pressure gas sorption measurements High-pressure CO2 (99.999%), CH4 (99.999%) and C2H6 (99.99%) sorption experiments were performed using a Hiden Isochema XEMIS microbalance. Temperatures were maintained at 273 K and 298 K using a Grant LT EcoCool 150 temperature controller. Activated samples were further outgassed under secondary vacuum for 3 h in situ before isotherm experiments were conducted. High-pressure cycling experiments were carried out using pressure ramps between 5 and 65 bar, using 30 min for each equilibration stage in each cycle. Online content Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41563-026-02621-7. Supplementary information Supplementary Sections 1–9, Figs. 1–87 and Tables 1–9. Peer Review File Phase transformation of Co(bdp). Phase transformation of Co(dpt). Phase transformation of Zn(bdp). Phase transformation of Zn(dpt). Source data Source data for Fig. 2. Source data for Fig. 4. Source data for Fig. 5. Acknowledgements B.S. acknowledges computing time from High-Performance Computing at NC State and ACCESS (CHE230105). We thank D.-X. Xue from Shaanxi Normal University for assistance with the high-pressure volumetric sorption measurements. Author contributions M.J.Z., J.-R.L., T.H. and X.-J.K. conceived the research idea and designed the experiments. X.-J.K. performed the synthesis experiments. X.-J.K., T.H., A.C.E., D.S., G.-R.S., X.L. and S.M. conducted the bulk characterizations and gas sorption measurements. V.I.N. and L.J.B. collected the in situ VP-PXRD data. X.-J.K., T.H., A.C.E. and D.S. analysed the experimental data. D.Y. and H.H. constructed and optimized the structures of different phases. L.M. and B.S. conducted the density functional theory calculations. M.J.Z., J.-R.L., T.H., X.-J.K., V.I.N., A.C.E., D.S., S.M., L.J.B., B.S., X.-H.B. and Z.-R.N. discussed, wrote and revised the paper. All authors have given their approval to the paper. Peer review Peer review information Nature Materials thanks Banglin Chen, Stefan Kaskel, Lothar Wondraczek and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available. Funding J.-R.L. discloses support from the National Natural Science Foundation of China (22225803 and U25A20621) and Beijing Outstanding Young Scientist Program (JWZQ20240102008). M.J.Z. discloses support from Research Ireland (21/US/3760 and 12/RC/2275_P2) and the European Research Council (ADG 885695). T.H. discloses support from the National Natural Science Foundation of China (22401168), the Postdoctoral Fellowship Program of CPSF (GZC20250645) and the China Postdoctoral Science Foundation (2025M780887). S.M. discloses support from the Research Ireland (21/PATH-S/9454). L.J.B. discloses support from the National Research Foundation of South Africa. B.S. discloses support from the National Science Foundation (2154882) and the National Science Foundation Graduate Research Fellowship Program (DGE-2137100). H.H. discloses support from the National Natural Science Foundation of China (22578328). Data availability The X-ray crystallographic data for the structures reported in this Article are deposited at the Cambridge Crystallographic Data Centre (CCDC), under deposition CCDC codes 2324055 (Zn(dpt)), 2324056 (Co(dpt)), 2448953 (Zn(dpt)-α″), 2448952 (Zn(dpt)-β′), 2448958 (Zn(dpt)-γ2′), 2448954 (Co(dpn)), 2448955 (Zn(dpf)), 2448956 (Co(dpi)) and 2448957 (Zn(dpi)). These data can be obtained free of charge from the CCDC via www.ccdc.cam.ac.uk. All the relevant data supporting the findings of this study are presented in the article or Supplementary Information. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Zuo-Ren Nie, Email: zrnie@bjut.edu.cn. Tao He, Email: hetao@bjut.edu.cn. Jian-Rong Li, Email: jrli@bjut.edu.cn. Michael J. Zaworotko, Email: xtal@ul.ie Supplementary information The online version contains supplementary material available at 10.1038/s41563-026-02621-7. References Associated Data Supplementary Materials Supplementary Sections 1–9, Figs. 1–87 and Tables 1–9. Peer Review File Phase transformation of Co(bdp). Phase transformation of Co(dpt). Phase transformation of Zn(bdp). Phase transformation of Zn(dpt). Source data for Fig. 2. Source data for Fig. 4. Source data for Fig. 5. Data Availability Statement The X-ray crystallographic data for the structures reported in this Article are deposited at the Cambridge Crystallographic Data Centre (CCDC), under deposition CCDC codes 2324055 (Zn(dpt)), 2324056 (Co(dpt)), 2448953 (Zn(dpt)-α″), 2448952 (Zn(dpt)-β′), 2448958 (Zn(dpt)-γ2′), 2448954 (Co(dpn)), 2448955 (Zn(dpf)), 2448956 (Co(dpi)) and 2448957 (Zn(dpi)). These data can be obtained free of charge from the CCDC via www.ccdc.cam.ac.uk. All the relevant data supporting the findings of this study are presented in the article or Supplementary Information. Source data are provided with this paper.