Introduction

Adsorption into the nanospace of porous solids and dissolution into solvents are widely used in daily life and industrial activities, such as separating and refining pure substances from mixtures and facilitating the handling of volatile and hazardous substances. Although both phenomena are similar in that guest molecules occupy space in a medium, there is a crucial difference in the change in space during the accommodation process of the molecules.

In adsorption, the dispersion force mainly contributes to the binding strength between host and guest molecules, which can be modulated by the pore sizes and elements constituting the pore surfaces. The space where the adsorbate molecule is located essentially preexists, and most of the frameworks of conventional porous materials, such as porous carbons and zeolites, are unchanged during adsorption. Recently, metal-organic frameworks (MOFs)1 constructed by regular coordination of metal ions and organic linkers have emerged as flexible adsorbents that undergo structural transformation upon the accommodation of guest molecules2,3,4,5,6. Typically, the guest-free state of a MOF with a narrower pore structure transforms to that with a wider pore, accompanied by a rapid increase in the adsorption amount called gate-opening adsorption7,8,9,10. Such dynamic adsorption behavior in flexible MOFs sometimes results in increased selectivity of adsorption in molecular separation applications11,12,13,14,15,16,17,18. However, even in such materials, the structures maintain long-range regularity and cannot be seamlessly changed to best-fit guest molecules. Therefore, deliberate framework design is required to develop ideal adsorbents for target guest molecules, which is practically quite difficult in many cases.

Unlike adsorption, there is no defined space for dissolution; a solute molecule intervenes in temporary cavities between solvent molecules to secure its own space. Through various intermolecular interactions, such as dielectric interactions and hydrogen and coordination bonds, the dynamic reorganization of solvent and solute molecules spontaneously reduces the free energy; i.e., the dissolution process can self-maximize the guest accommodation ability. However, liquids are often difficult to handle due to their fluidity, volatility, and solidification at coagulation temperatures. Furthermore, elucidation of the molecular dissolution mechanism is quite difficult due to the limited methods applicable for structural analysis, thus preventing the strategic design of liquids with superior molecular trapping ability.

Herein, we fixed solvent molecules on the pore surface of solid materials to integrate the advantages of adsorption and dissolution (Fig. 1a). We selected long flexible perfluoroalkyl chains which are immobilized onto the organic linkers of a MOF, forming nanospaces where chains densely filled the nanospace (Fig. 1b). The perfluoroalkyl chains have a weak dispersion force that can help maintain the mobility of the chains even in a confined environment. Indeed, weak intermolecular interactions lead to the high solubility of gas molecules in perfluoroalkyl solvents. The solubility of oxygen (O2) in perfluorocarbons is much higher than that in the corresponding hydrocarbons (e.g., n-C7F16: 54.8 vol%, n-C7H16: 15.2 vol%)19. In addition, interestingly, O2 has a higher solubility than Ar, which has almost the same boiling point as O2, even at room temperature (0.52 and 0.50 mol%, respectively, for FC-75; Supplementary Fig. 1)20. It is hypothesized that the hybridization of perfluoroalkanes with nanopores will further enhance such characteristics. Based on these assumptions, we created a MOF nanospace filled with dynamic perfluoroalkyl groups and determined its selective gas accommodation properties towards a potential O2 and Ar separation material.

Fig. 1: Conceptual image of this research.
figure 1

a Schematic representation of the adsorptive-dissolution concept, integrating the adsorption on solid surfaces and dissolution in solvents. b Design strategy for the MOF filled with perfluoroalkyl chains.

Full size image

Results

Synthesis and structure

PFAC-2-as (PFAC = perfluoroalkyl-chain filled MOF, [Zn2(F3tpa)2(bpy)]‧DMF) was synthesized by the solvothermal reaction of 2,5-bis(n-heptafluoropropyl)terephthalic acid (H2F3tpa), 4,4’-bipyridine (bpy) and zinc(II) nitrate in N,N-dimethylformamide (DMF) at 120 °C for 72 h. The resulting colorless crystals were subjected to single-crystal X-ray diffraction (SXRD) structural analysis (Fig. 2a, Supplementary Table 1). Zn2(COO)4 paddlewheel clusters as secondary building units (SBUs) were bridged by F3tpa and bpy in the equatorial and axial positions, respectively, forming a noninterpenetrated pillared-layer framework. PFAC-2-as contained DMF molecules in a potential solvent-accessible void (2757 Å3, 40.7% of the unit cell calculated with a probe diameter of 3.46 Å (kinetic diameter of O2)). Thermogravimetric analysis of PFAC-2-as showed weight loss up to 100 °C as a result of the removal of crystal solvents (Supplementary Fig. 2a). Heating PFAC-2-as at 80 °C for 72 h gave PFAC-2 ([Zn2(F3tpa)2(bpy)]) via a single-crystal-to-single-crystal transformation (Fig. 2b, Supplementary Fig. 2b, c). The SXRD structural analysis of PFAC-2 revealed that the [Zn(F3tpa)] layer was distorted from a rectangle to a rhomboid and that the bpy pillar was inclined from that in PFAC-2-as, as observed in some pillared-layer MOFs21,22. As a result, the cage became completely filled by perfluoroalkyl chains. The solvent-accessible void disappeared (0.00 Å3, calculated using a probe diameter of 3.46 Å (kinetic diameter of O2)). The estimated density of perfluoroalkyl chains that occupy the nanospace was ~1.6 g/cc. This value is comparable to those of perfluoroalkyl solvents (e.g., 1.67 g/cc for n-C6F14). Compared to other fluorine-containing MOFs23,24,25,26,27,28,29,30,31, PFAC-2 has much larger weight and volume densities of fluorine atoms, which are the next largest values after our PFAC-126, and is the MOF with the largest fluorine atom density for which the crystal structure was successfully determined without solvent (Supplementary Fig. 3, Supplementary Table 2). We also synthesized HAC-2-as ([Zn2(H3tpa)2(bpy)]·nDMF) as a reference using 2,5-di-n-propylterephthalic acid (H2H3tpa) instead of H2F3tpa (Fig. 2c). HAC-2-as was almost isostructural to PFAC-2-as, except for the dihedral angle between the carboxylate and phenylene rings of the terephthalate because of the difference in the bulkiness of the side chains. Unlike PFAC-2-as, HAC-2-as did not undergo a structural change upon desolvation (Supplementary Fig. 2b). The contrasting difference may be related to the weak attractive interaction of a fluorine atom on C-F bonds with other molecules, as previously reported32,33,34. In the crystal structure of HAC-2-as, the shortest distance between the oxygen atom of the carboxylate and the hydrogen atom of the C3H7 group is 1.98 Å (Supplementary Fig. 2d). The distance is much smaller than the sum of the van der Waals radii of oxygen and hydrogen atoms (2.72 Å), indicating that they interact strongly with each other. It can be assumed that the interaction effectively stabilizes the Zn2(COO)4 paddlewheel SBU, which helps to maintain the framework even after the guest molecule is removed. On the other hand, in the crystal structure of PFAC-2-as, the shortest distance between the oxygen atom of the carboxylate and the fluorine atom of C3F7 group is 2.71 Å (Supplementary Fig. 2d), which is close to the sum of van der Waals radii of oxygen and fluorine atoms (2.99 Å). This indicates that C3F7 group does not supports the structure strongly compared with C3H7 group. Therefore, when the guest molecule is removed, the entire skeleton is deformed so that C3F7 groups fill the space. We also checked the stability of PFAC-2 in water. PFAC-2 exhibited water-repellent property. The powder X-ray diffraction (PXRD) pattern of the sample soaked in water for 14 h showed a slight shift in the peak around 7° from PFAC-2, but the peak intensity was maintained, suggesting that the structure may be retained in water for a short time (Supplementary Fig. 2e). However, it changed to a different crystalline phase after 70 h, indicating that PFAC-2 does not have thermodynamic stability in water.

Fig. 2: Crystal structures.
figure 2

a PFAC-2-as, b PFAC-2, and c. HAC-2-as. The left and right figures are side and top views with respect to [Zn(F3tpa)] layers, respectively. Green, fluorine; grey, carbon; blue, nitrogen; red, oxygen; white, hydrogen. Hydrogen atoms on bpy were omitted for clarity. Solvent molecules are disordered and treated with solvent masks in the structural refinement for PFAC-2-as and HAC-2-as.

Full size image

Gas adsorption and separation properties

Gas adsorption measurements for PFAC-2 were conducted at 90.2 K. The O2 adsorption isotherm at 90.2 K showed appreciable guest uptake in the low-pressure region (< 0.12 P/P0), followed by a steep increase in the adsorption at 0.12 P/P0 as a result of gate-opening (Fig. 3a). To elucidate the relation between the host structure and adsorption behavior, in situ PXRD measurements under an O2 atmosphere were conducted. The PXRD pattern was changed together with gate-opening adsorption, which is quite typical for flexible MOFs (Supplementary Fig. 4a). However, what is most surprising is that even though PFAC-2 shows apparent adsorption in the very low-pressure region below the gate-opening pressure, the PXRD pattern was almost the same as that of PFAC-2 in that region (Fig. 3b). This indicates that the O2 molecules are trapped in the crystal without any change in the nonporous skeletal structure, suggesting that the perfluoroalkyl chains are dynamically reorganized to provide space to accommodate O2 molecules, similar to gas dissolution in liquids. In situ Raman spectra of PFAC-2 under an O2 atmosphere at 90 K showed the O2 peaks at 1552 cm−1, which is slightly shifted from that of bulk O2 gas (1556 cm−1), suggesting that the molecules are under the influence of the potential of the nanospace (Fig. 3c).

Fig. 3: Characteristics of PFAC-2 in response to gas adsorption.
figure 3

a Adsorption isotherms of O2 (red) and Ar (black) at 90.2 K. Solid points represent the adsorption isotherms, and hollow points represent the desorption isotherms. P/P0 is the relative pressure where P0 indicates the saturated vapor pressure for each gas at the measurement temperature. b PXRD patterns (right, red: observed patterns, black: simulation pattern) of PFAC-2 under O2 at 90.2 K measured coincidently with the adsorption isotherm of O2 (left). c In situ Raman spectra of PFAC-2 measured under O2 at 90.2 K. The value of O2 relative pressure (P/P0) is shown next to the right of each spectrum. d Adsorption isobars of O2 (red) and Ar (black) at 1 bar. Solid points indicate cooling processes, and hollow points indicate heating processes. The temperature was first decreased from 250 to 90 K and then increased to 250 K.

Full size image

To evaluate the molecular uptake ability of O2, the adsorption measurement of Ar was performed as a comparison. Ar is known to have quite similar properties to O2, such as boiling point and molecular size, making the separation of Ar and O2 an extremely difficult (Supplementary Fig. 5). The adsorption isotherm of Ar for PFAC-2 also showed gate-opening adsorption at the same relative pressure as O2. However, interestingly, the adsorption of O2 below the gate-opening pressure (i.e., 0 ≤ P/P0 ≤ 0.12) was greater than that of Ar (2.1 times at P/P0 = 0.11). This contrasts with the fact that the amounts of O2 and Ar adsorbed after gate-opening are identical. Thus, it is assumed that densely fluorinated nanospaces preferentially accommodate O2 over Ar.

With the sudden application of gas pressure at cryogenic temperatures such as 77 K and 90 K, guest molecules sometimes cannot diffuse into the micropores, and a significantly lower amount of adsorption than that expected from the pore volume is observed, which is known as the pore-blocking effect (PBE)35. Due to the PBE, the adsorption may not reach true thermal equilibrium, leading to erroneous adsorption selectivity. To determine whether PFAC-2 actually selectively adsorbs O2, measurements in the temperature range at which the PBE does not occur are necessary. Thus, we measured the adsorption isobar with a stepwise decrease in temperature from 250 to 90 K and an increase back to 250 K under constant pressure of 1 bar (Fig. 3d). As the temperature decreased, the adsorption gradually increased above ~110 K until it showed a steep increase as a result of the gate-opening process. The ratio of O2 to Ar adsorption amount exceeds the ratio of their inverse of saturated vapor pressures at the same temperatures, indicating that O2/Ar separation on PFAC-2 is more thermodynamically favorable than distillation (Supplementary Fig. 4b). On the other hand, the adsorption isobar for HAC-2 revealed that while the amount of O2 adsorbed was slightly higher than that of Ar (Supplementary Fig. 4c), the ratio was almost the same as that of inverse of saturated vapor pressure (Supplementary Fig. 4b). This suggests that there is no preferential adsorption interaction of HAC-2 toward O2. The unusual O2 accommodation behavior of PFAC-2 was also observed during the desorption process. As the temperature increased from 90 K, the amount of Ar adsorbed decreased instantly at ~120 K. In contrast, the hysteresis in the desorption isobar of O2 was more striking than that of Ar; a gradual decrease in the adsorption amount ranging between 120 and 180 K was observed. This result also suggests that densely fluorinated nanospaces containing O2 are more stable than those containing Ar. Additionally, the CO2 adsorption isotherm was measured at 195 K (Supplementary Fig. 4 d). As CO2 has a smaller kinetic diameter (O2: 3.46 Å, Ar: 3.40 Å, CO2: 3.30 Å) and a higher boiling point (O2: 90 K, Ar: 87 K, CO2: 195 K) compared to O2 and Ar, its adsorption isotherm is less affected by PBE around their boiling points. Therefore, it is more suitable for characterizing the pore volume. In fact, the CO2 adsorption isotherm is similar to that of O2, suggesting that O2 adsorption is not influenced by PBE.

The O2/Ar separation performance of the densely fluorinated nanospace of PFAC-2 was evaluated by both single-gas and mixed-gas experiments. The single-gas adsorption isotherms of O2 and Ar at 140 K were both Langmuir-type, and the amount of O2 adsorbed at 1 bar was ~1.4 times greater than that of Ar (Fig. 4a). The adsorption isotherms were simulated using the virial equation, employing virial parameters derived from fitting experimental adsorption isotherms. Subsequently, the isosteric heats of adsorption (Qst) were estimated from these simulated isotherms at various temperatures (130, 140, 150, and 160 K) (Fig. 4b, Supplementary Fig. 6, and Supplementary Table 3). Qst of O2 (16.6-13.8 kJ/mol) and Ar (15.9-11.7 kJ/mol) are similar at the initial adsorption, but Qst of Ar decreases more than that of O2 as the gas uptake increases. Overall, Qst of O2 is higher than that of Ar. The O2/Ar adsorption selectivity for 50:50 mixture at 140 K was estimated by ideal adsorption solution theory36 (IAST, Fig. 4c). As similarly to Qst, IAST selectivity becomes larger at higher pressure until it reaches 1.35 at 1 bar. The substantial O2/Ar separation performance was demonstrated by breakthrough curve measurements at 140 K using a 50:50 O2/Ar mixed gas (Fig. 4d). Both O2 and Ar started to elute at 250 seconds after injection of the mixed gas into the fixed bed of PFAC-2. However, the elution of Ar ended rapidly, while O2 eluted gradually, indicating that O2 was selectively adsorbed over Ar. We also performed the cycling breakthrough curve measurements. The breakthrough curves remained consistent over multiple cycles (Supplementary Fig. 7a). The PXRD pattern of the sample after the cycling breakthrough curve measurements was the same as that before the measurement (Supplementary Fig. 7b), indicating that the structure of PFAC-2 was maintained after repeated O2/Ar adsorption and desorption. In contrast, the breakthrough curve of HAC-2 revealed that both O2 and Ar eluted simultaneously like that of the blank sample (Fig. 4e and f); i.e., there was no O2/Ar adsorption selectivity. The adsorption amounts of O2 and Ar in PFAC-2 calculated from the breakthrough curves are 17 mL/g and 9.4 mL/g, respectively. The total adsorption amount (26 mL/g) was close to those obtained from single-gas adsorption isotherms (30 mL/g for O2 and 22 mL/g for Ar). Interestingly, the O2/Ar selectivity (1.8) calculated from the breakthrough curves was much greater than that estimated by IAST (1.35). The difference of selectivity between breakthrough experiments (1.8) and IAST (1.35) can be attributed to the kinetic effect (i.e., the faster adsorption of O2 than Ar). To elucidate the adsorption kinetics of O2 and Ar, the time-course pressure decrease during the adsorption process was analyzed (Supplementary Fig. 8). The obtained adsorption kinetic constants of O2 were the same as those of Ar when the adsorption amount was small but became increasingly larger as the adsorption amount increased, which is consistent with the results of breakthrough curves: Elution of O2 and Ar started simultaneously, while it takes longer time to complete the elution of O2, indicating O2 occupies the adsorption sites more quickly compared to Ar. Thus, densely fluorinated nanospaces were both thermodynamically and kinetically beneficial for O2/Ar separation.

Fig. 4: O2/Ar separation performance of PFAC-2.
figure 4

a Adsorption isotherms of O2 (red) and Ar (black) at 140 K. Solid and hollow points represent the adsorption and desorption isotherms, respectively. b Qst of O2 (red) and Ar (black) for PFAC-2. c IAST selectivity (O2:Ar = 50:50) at 140 K for PFAC-2. d–f Breakthrough curves of O2 (red) and Ar (black) for an equimolar O2/Ar mixture at 140 K and 1 bar for PFAC-2, HAC-2, and blank samples, respectively. C and C0 represent outlet and inlet concentrations for each gas, respectively.

Full size image

Theoretical calculations

To elucidate the distinct adsorption behaviors of O2 and Ar, we conducted a detailed analysis of the PFAC-2 structure. By varying the probe diameter to calculate the solvent-accessible void in PFAC-2 (Fig. 5a), although no void space was observed when the probe of the same size as O2 was used, two isolated voids (site A and B, Fig. 5b) near the fluoroalkyl chains at the center of the [Zn(F3tpa)] layer and between two bpy ligands appears, respectively, when the probe diameter is less than ~3 Å. These sites could be adsorption space and diffusion pathways if reorganization of fluoroalkyl chains occurs. To verify the possibility of sites A and B to be adsorption sites, we examined the structure of PFAC-2 accommodating either one O2 or Ar at sites A or B by theoretical calculation (see the Supplementary information for detail). The position and orientation of building blocks including C3F7 chains were rearranged to make the space for a gas molecule with the best fitting in the pore (Fig. 5c–f). The stabilization energy for O2 adsorption at site B (65 kJ/mol) is significantly higher than the others (O2 (site A): 47 kJ/mol, Ar (site A): 48 kJ/mol, and Ar (site B): 40 kJ/mol). Notably, the energies of structural rearrangements (1.2-5.9 kJ/mol) were very small, indicating that fluoroalkyl chains can move flexibly (Supplementary Table 4).

Fig. 5: Theoretical studies for gas adsorption in PFAC-2.
figure 5

a Calculated solvent-accessible void volume depending on probe diameters. b The solvent-accessible void space calculated with a probe diameter of 2.4 Å in the crystal structure of PFAC-2, visualized by the transparent orange color. Green, fluorine; grey, carbon; blue, nitrogen; red, oxygen; white, hydrogen. c–f Optimized gas adsorbed structure of PFAC-2 (c O2 at site A, d O2 at site B, e Ar at site A, f Ar at site B). Green, fluorine; grey, carbon; blue, nitrogen; red, oxygen; white, hydrogen. O2 and Ar are drawn by CPK models and colored red and blue, respectively. Optimized structure of guest-free PFAC-2 (pink) is overlayed for each gas-adsorbed structure.

Full size image

In summary, we developed PFAC-2 in which nanopores were entirely filled by perfluoroalkyl chains. The fact that potentially nonporous PFAC-2 accommodated the gas without changes in its framework indicates that reorganization of the perfluoroalkyl side chains occurred to secure a space for gas molecules to be accommodated. The way gas molecules occupy the spaces in PFAC-2 is reminiscent of the dissolution of molecules in liquids, which could be deemed a new concept, “adsorptive-dissolution”, integrating the rigidity of solid with the fluidity of liquid. Our work will open up the frontier of nanospace material science, enabling molecular separation that has not been realized thus far.

Methods

Synthesis of PFAC-2-as

A solution of Zn(NO3)2‧6H2O (31 mg, 0.10 mmol) and H2F3TPA (50 mg, 0.10 mmol, see Supplementary Fig. 9 for the synthesis) in DMF (1.5 mL) was stirred at room temperature for 1 day. The mixture was added bpy (16 mg, 0.10 mmol) solution in DMF (0.5 mL) and was heated at 120 °C for 36 h to give colorless crystals of PFAC-2-as.

Synthesis of HAC-2-as

A solution of Zn(NO3)2‧6H2O (26 mg, 0.08 mmol) and H2H3TPA (21 mg, 0.08 mmol, see Supplementary Fig. 9 for the synthesis) in DMF (1.5 mL) was stirred at 50 °C for 1 day. The mixture was added bpy (16 mg, 0.10 mmol) solution in DMF (0.5 mL) and was heated at 80 °C for 48 h to give colorless crystals of HAC-2-as.