Visualizing Structural Transformation and Guest Binding in a Flexible Metal-Organic Framework under High Pressure and Room Temperature.

Understanding the effect of gas molecules on the framework structures upon gas sorption in porous materials is highly desirable for the development of gas storage and separation technologies. However, this remains challenging for flexible metal-organic frameworks (MOFs) which feature "gate-opening/gate-closing" or "breathing" sorption behaviors under external stimuli. Herein, we report such a flexible Cd-MOF that exhibits "gating effect" upon CO2 sorption. The ability of the desolvated flexible Cd-MOF to retain crystal singularity under high pressure enables the direct visualization of the reversible closed-/open-pore states before and after the structural transformation as induced by CO2 adsorption/desorption through in situ single-crystal X-ray diffraction experiments. The binding sites of CO2 molecules within the flexible MOF under high pressure and room temperature have also been identified via combined in situ single-crystal X-ray diffraction and powder X-ray diffraction studies, facilitating the elucidation of the states observed during gate-opening/gate-closing behaviors. Our work therefore lays a foundation to understand the high-pressure gas sorption within flexible MOFs at ambient temperature, which will help to improve the design efforts of new flexible MOFs for applications in responsive gas sorption and separation.

Introduction

Metal–organic frameworks (MOFs), as an emerging type of crystalline porous material, have been of escalating interest due to their potential in a broad range of applications.[1] One of the striking features for MOFs in comparison with conventional porous materials, such as zeolites[2] and activated carbons,[3] lies in their structural flexibility/dynamics, as observed from the “gate-opening/gate-closing” or “breathing” sorption behaviors under external stimuli (e.g., adding or removing guest molecules, pressure, temperature).[4−7] Such flexible/dynamic MOFs hold particular promise for applications in selective gas adsorption,[8−13] gas storage,[14−19] and sensing.[20]

It is believed that the observed “gate-opening/gate-closing” or “breathing” sorption behaviors are associated with the structural transformations and the interplay between the host framework and the guest molecules during the adsorption/desorption processes.[19,21−28] It is of fundamental importance to understand the structural transformations and identify the interactions between the gas molecules and the host framework in these flexible/dynamic MOFs, which have been sought after using various techniques.[29−39] Given the difficulty in maintaining crystal singularity for desolvated flexible/dynamic MOFs under high pressures and ambient temperatures, the studies to monitor their structural transformations have exclusively relied on in situ powder crystal X-ray diffraction (PXRD) tools, which however cannot provide unequivocal structural information as in situ single-crystal X-ray diffraction (SCXRD) experiments can.[19,35,36] This makes it even more challenging to crystallographically identify the binding sites of gas molecules within flexible/dynamic MOFs, which thus far are elucidated mainly through computational studies.[8,13,14,16,25]

In this contribution, we report a flexible three-dimensional (3D) microporous cadmium(II) MOF (Cd-MOF) with 4-fold interpenetration and diamondoid (dia) topology, which after desolvation shows an interesting “gating effect” process and “breathing” behavior upon CO2 sorption. In addition, this flexible Cd-MOF can retain its crystal singularity in the absence of guest solvent molecules under high pressure, which therefore allows us, for the first time, to directly visualize the reversible closed/open end states of the structures induced by gas molecules for a flexible MOF through in situ SCXRD experiments. The elucidated structural information in combination with in situ PXRD facilitates the production of a clear map for the binding of guest gas molecules within a flexible MOF under high pressure and at room temperature.

Results and Discussion

Crystallographic Visualization of Structural Transformation in Flexible Cd-MOF

The flexible Cd-MOF (hereafter denoted 1a) was synthesized via a previously reported route with a slight modification.[40] Single-crystal X-ray diffraction analysis reveals that 1a crystallizes in the tetragonal space group P4212. In the structure of 1a, each Cd(II) site displays distorted octahedral geometry, with coordination bonds to four oxygen and two nitrogen atoms from four different 4-(1H-2-methylimidazol-1-yl)benzoic acid (miba) ligands (Figure a, Figure S1). Each μ2-miba ligand bonds to two adjacent Cd(II) centers via two carboxylate O atoms and one imidazolyl N atom (Figure a, Figure S1). The dihedral angle formed by the benzoate plane and the imidazole plane of miba is 56.4° (Figure a, Figure S3). The planarity of the ligand, as quantified by the N (imidazole)-phenyl centroid-C (carboxylate) angle, is 154° (Figure a, Figure S2). The Cd···Cd distance is 12.46 Å. Cd(II) metal centers are connected by the miba ligands to form a neutral three-dimensional (3D) framework with larger hexagonal and small rhombic 1D channels along the crystallographic c axis (Figure c, Figure S4). Each channel is occupied by structurally disordered solvent DMF molecules. The total solvent-accessible volume for 1a is estimated to be 37.8% using the PLATON program.[41] The framework topology is identified as a 4-connected net with 4-fold interpenetration and symbol (dia) by considering each Cd site as the 4-connected node (Figure S5). The experimental PXRD of 1a matches that of the calculated one from single-crystal data (Figure ).

Figure 1

Cd···Cd distances, bond angles, and dihedral angles of the miba ligand and the coordination environment of the Cd(II) atoms of (a) 1a and (b) 1b. (c) Perspective along the c axis view of the 3D framework of 1a showing 1D channels. (d) Perspective view along the c axis of the 3D structure of 1b.

Figure 2

Comparison of the PXRD patterns of 1a and 1b, and the experimental patterns were recorded for the sample incrementally exposed in situ up to 40 bar CO2 pressure.

To examine the possible structural transformation of the Cd-MOF, 1a was activated by soaking freshly prepared crystals in dry dichloromethane to exchange guest DMF molecules followed by degassing under dynamic vacuum. The desolvated sample, named 1b hereafter, still retained crystal singularity, allowing for SCXRD studies. The crystal data of 1a and 1b are summarized in Table S1. SCXRD analysis demonstrates that 1a experiences drastic structural transformation upon desolvation to a dense (or closed) phase for 1b. 1b crystallizes in the orthorhombic space group P21212 with a smaller (pseudotetragonal) unit cell and much lower solvent-accessible void volume (4.4%). The framework of 1b undergoes changes in the coordination geometry of Cd(II), from six-coordinated octahedral geometry to five-coordinated hexahedral geometry (Figure b, Figure S6). During the transformation from 1a to 1b the Cd···Cd distance decreases significantly from 12.46 to 11.77 Å. The dihedral angle between the benzoate and imidazole planes increases from 56.4° to 91.3°, and the planarity of the ligand, as quantified by the N (imidazole)-phenyl centroid-C (carboxylate), decreases from 154° to 131.4° (Figure b, Figures S7 and S8). Moreover, the 1D channel becomes closed in 1b as a result of “breathing effect” contortions (Figure d, Figure S9).[6,12,15,18,21,24,26,37] This observation is also supported by the PXRD data (Figure ). Compared to 1a, all of the characteristic peaks of 1b in the PXRD pattern are shifted toward higher 2θ angles, and match well with the calculated ones (calculated from the crystallographic information file for 1b with PXRD lattice parameters, see details in the Supporting Information, Figure ).

To further probe the structural transformation from the open form 1a to the closed form 1b, we performed a detailed computational first-principles dispersion-corrected density-functional theory (DFT-D) investigation (see details in the Experimental Section). The structures of 1a and 1b were first fully optimized. Then, nudged elastic band calculations were performed to find the minimum energy path of the structural transition. An animation consisting of structure snapshots along the minimal energy path shows clearly how the flexible framework evolves from a fully open-pore phase to a nearly closed-pore phase (see the movie in the Supporting Information). As expected, the imidazole and phenyl rings of the miba ligand remain relatively rigid. However, the bond connecting the two rigid rings and those coordinating the ligand to the Cd center are rather flexible, particularly in terms of bond angles. Such flexibility enables both the metal center and the organic linker to undergo a large degree of rotation and distortion upon crystal desolvation, leading to a configuration with dense framework packing, diminished pore volume, and low system energy.

CO2 Sorption and in Situ Crystallographic Visualization of Structural Transformation under High Pressure

Capitalizing on the single-crystal to single-crystal “breathing effect” of 1a and 1b, we carried out CO2 sorption experiments to assess the “gating effect” performance. A low-pressure CO2 sorption isotherm was first measured at 195 K for a single sample of 1b. The isotherm shows a broad gate-opening step with an onset pressure of ca. 0.34 bar, followed by a gradual increase in uptake at higher relative pressure (0.34 bar < P < 0.95 bar), indicating that a structural transformation occurs (Figure a). Upon desorption, a pronounced hysteresis loop is observed at intermediate pressures (0.5–0.08 bar); a sudden decrease implies that a gate-closing phase transition occurs. The maximum uptake of CO2 under the conditions used is 130 cm3 g–1, which is equivalent to 6 CO2 molecules per unit cell of 1b_open.

Figure 3

(a) CO2 adsorption and desorption isotherms for 1b at 195 K. (b) High-pressure CO2 adsorption and desorption isotherms for 1b at 298 K.

Encouraged by the above CO2 sorption results, we conducted high-pressure CO2 sorption experiments on 1b to evaluate the effect of the CO2-induced framework structural transformation at 298 K in the pressure range 0–30 bar. The isotherm of 1b exhibits very low uptake at the beginning (below 8 bar), followed by a steep increase at approximately 8 bar, which indicates a gate-opening transition (Figure b). An adsorption capacity commensurate with that at low temperature was reached upon saturation at approximately 30 bar. The total uptake (excess adsorption capacity) is 117 cm3 g–1 (equivalent to a total adsorption capacity of 122 cm3 g–1, 5.6 CO2 molecules per unit cell of 1b_open) at 30 bar. A sudden decrease at 3 bar on desorption indicates reversible switching between the open and closed phases. Variable-pressure in situ PXRD diffraction patterns (Figure , Figure S19) were recorded with incremental CO2 loading to further investigate the gate-opening and gate-closing events at 298 K. The closed-pore phase of 1b was confirmed under vacuum (Figure S19). Upon increasing the pressure, the closed-pore form for 1b is still retained below 8 bar. Significant changes become apparent at 9 bar; the peaks of the closed phase diminish as peaks of the fully open phase appear, which indicates the structural transformation (gate-opening). As expected, this process of gate-opening and gate-closing, as monitored by PXRD, is in agreement with the CO2 sorption results (Figure b, Figure S19). The open structure is retained over the pressure range 9–40 bar. Upon evacuation of the sample, the phase transforms (gate-closing) back to the closed-pore form for 1b when the gas pressure is reduced to 2 bar or lower (Figure S20).

If these transformations can occur in single-crystal to single-crystal fashion, it would enable the characterization of each phase transition in a manner that is highly desired for flexible MOFs, which unfortunately tend to be poorly crystalline following breathing or swelling thereby limiting the collection of single-crystal data.[19,23] Fortunately, 1b can maintain its single-crystal character even under high CO2 pressure, which makes it possible for the first time to visualize the states before and after the structural transformation of a flexible MOF under high pressure through in situ SCXRD studies. Considering that unequivocal structural information for the high-pressure-induced open form of a flexible MOF has never been achieved, in situ SCXRD experiments were conducted on 1b at the selected pressures of 15 and 20 bar, at which 1b should feature the open-pore structure based on high-pressure CO2 sorption isotherms (Figure b) and variable-pressure in situ PXRD diffraction patterns (Figure ). Single-crystal structure analysis of 1b under 15 bar CO2 atmosphere (hereafter 1b_15 bar) reveals that the structure of 1b_15 bar is similar to the DMF-solvated structure of 1a (Figures c and 4). The planarity of the miba ligand, as quantified by the N (pyridyl)-phenyl centroid-C (carboxylate) angle, is 154.5° in 1b_15 bar close to 154.0° in 1a (Figure a, Figures S11 and S18). The benzoate and imidazole planes exhibit dihedral angles of 53.2° and 56.4° for 1b_15 bar and 1a, respectively (Figure a, Figures S12 and S18). The Cd···Cd distances are almost the same in 1b_15 bar (15.46 Å) and 1a (15.45 Å) (Figure a, Figure S18). The opened rhombic and hexagonal channels are occupied by adsorbed CO2 molecules, which are highly disordered and therefore not able to be well refined (Figure ). Similar single-crystallographic data were determined for 1b under 20 bar CO2 atmosphere (the sample named 1b_20 bar) (Table S1). The mechanism of structural flexibility of flexible MOFs such as MIL-53,[6,24] MIL-88,[23] YO-MOF,[25] and RPM3-Zn[8] is mainly based upon hinge motion associated with carboxylate-ligands coordination. The SCXRD structural characterization of the various phases of 1b clearly indicates that the transformation here largely relies upon the metal center and organic linker contortion. Moreover, 1b might be stabilized by the interactions between interpenetrated dia networks.[19] To our knowledge, the rotation and distortion of the metal center and organic linker during adsorption/desorption of CO2 molecules within the MOF pores are rare.[4−7]

Figure 4

Reversible single-crystal to single-crystal transformation upon CO2 adsorption/desorption: perspective view along the c axis, showing the closed-channel (left) and open-channel (right) of 1b_15 bar. Note: the CO2 guest molecules are diffuse (or disordered) in the 1D channels of 1b_15 bar.

Identification of Carbon Dioxide Binding within the Framework under High Pressure

To understand the interactions involved for the high-pressure gas-induced structural transformation of the flexible MOF, it is essential to elucidate how the gas molecules occupy the MOF pores. While both in situ SCXRD and in situ PXRD show clearly how the framework of 1b evolves to open the pore upon CO2 adsorption under high pressure, we found it difficult to directly resolve the atomic coordinates of the adsorbed CO2 guest molecules from the SCXRD data. Therefore, we turned to DFT-D calculations for clues.

We first optimized the open-pore structure 1b_open. CO2 molecules were then introduced into the MOF pore, followed by a full relaxation (on both CO2 positions and orientations). From the calculations, three major CO2 adsorption sites were identified. Guided by the calculated MOF/CO2 structure model, we further attempted to refine the PXRD data. We expected that the MOF pores would be more fully occupied at high pressure, and that the CO2 molecules are likely more ordered within the pore (and could hopefully be better modeled). Therefore, we focused on the PXRD data of 1b_40 bar (Figure , Table S1), and carried out a Rietveld structural refinement. To our delight, the refinement converged satisfactorily, and a reasonable goodness of fit was achieved (Figure , Table S1). From the refinement, we obtained an experimental crystal structure of 1b_40 bar(R), with detailed information on the adsorbed CO2. We note that, at room temperature, there is inevitably some disorder associated with the adsorbed CO2 (in terms of both positions and orientations) which was not modeled in the refinement. The structure derived from PXRD is regarded as an “average” structure, which we believe is still of value to understand the CO2–framework interaction.

Figure 5

Rietveld refinement of the PXRD data of 1b_40 bar(R). Experimental (circles), calculated (line), and difference (line below observed and calculated patterns) powder diffraction profiles are shown. Vertical bars indicate the calculated positions of Bragg peaks. Goodness of fit data: Rp = 0.0382, Rwp = 0.0486.

In the refined structure, there exist three preferential CO2 adsorption locations (sites I, II, and III), as also suggested by the DFT-D calculations. All three sites are partially occupied, which is an indication of CO2 disorder to some extent. Figure shows schematically the 1D channels of the framework filled with CO2 molecules located at these three sites. As in the structure of 1a, there are two different types of channels in the open-pore structure of 1b_40 bar(R). Sites I and II are located in the wider channel while site III is in the narrower channel. The total calculated occupancy of these three sites is 5.9 CO2 molecules per unit cell, which is consistent with the experimental results (5.6 and 6 CO2 molecules per unit cell at 298 and 195 K, respectively). The interactions between the CO2 molecules and the framework miba ligands consist mainly of van der Waals interactions.

Figure 6

Crystal structure of 1b_40 bar(R) from PXRD refinement showing (a, b) preferential binding sites for CO2 uptake (sites I, II, and III), and (c) 3D framework of 1b_40 bar(R) from PXRD refinement with guest CO2 molecules in the channels.

Taken together, the above results demonstrate that the desolvated Cd-MOF exhibits extreme “breathing” behavior under high gas pressure at room temperature. Upon increasing/decreasing CO2 pressure, the reversible closed-/open-pore states before and after the structural transformation have been directly visualized by in situ SCXRD techniques. The observed gate-opening/gate-closing can be ascribed to the rotation and distortion of the metal center and organic linker during adsorption/desorption of CO2 molecules within the MOF pores. Rietveld structural refinements based on the DFT-D calculated structural model were carried out to gain an in-depth understanding of the CO2–framework interaction. Three binding sites for CO2 molecules with two in the larger hexagonal channel and one in the small rhombic 1D channel have been identified.

Conclusion

In summary, we reported a flexible 4-fold interpenetrated diamondoid Cd-MOF which displays unusual dynamic gating behavior upon CO2 sorption. The open and closed states of a breathing framework as induced by CO2 adsorption/desorption under high pressure and room temperature have been crystallographically observed through in situ SCXRD experiments. The interactions involved during the gate-opening/gate-closing have also been elucidated by identifying the binding sites of CO2 molecules within the flexible MOF, thereby, the host–guest interactions via combined in situ SCXRD and PXRD studies. Our work therefore provides a clear map to understand the gas sorption within a flexible MOF particularly at high pressure and ambient temperature, which will help to design new MOFs and other types of porous materials with pressure-induced gate-opening/gate-closing behaviors and lays the foundation for further understanding the mechanisms associated these processes for applications in task-specific gas sorption and separation.

Experimental Section

No unexpected or unusually high safety hazards were encountered.

Synthesis of 1a

1a was prepared using a reported procedure with a slight modification.[40] A mixture of Cd(NO3)2·4H2O (0.078 g), 4-(1H-2-methylimidazol-1-yl)benzoic acid (Hmiba, 0.065 g), and DMF (6 mL) was sealed in a 23 mL polytetrafluoroethylene lined stainless steel container and heated to 85 °C for 3 days, and then cooled to room temperature. Colorless crystals of 1a were obtained.

Preparation of 1b

A sample of 1a was soaked in 20 mL of dry dichloromethane for 24 h. The solvent was replaced 6 times over 24 h, after which the sample was evacuated at 323 K for 300 min.

Characterization

Powder X-ray diffraction (PXRD) at ambient pressure was performed on a Rigaku X-ray diffractometer with a Cu Kα source. Pressure-resolved PXRD data were collected using Cu Kα radiation (λ= 1.5418 Å, 40 kV and 30 mA) on a PANalytical X’pert PRO instrument operating in Bragg–Brentano geometry. The sample was prepared in an environmental gas cell. CO2 sorption data were collected at 195 K using automatic volumetric adsorption equipment, Micromeritics ASAP2020, after pretreatment under vacuum at 323 K for 300 min. High-pressure CO2 sorption data were collected at 298 K. Thermogravimetric analyses (TGA) for all measurements were carried out on a PerkinElmer TGA thermogravimetric analyzer under air flow at a heating rate of 10 °C min–1 from 20 to 800 °C. Fourier transform infrared (FTIR) spectra were recorded on a ThermoElectron Nicolet high-resolution FT-MIR/FT-FarIR instrument in the range 4000–400 cm–1.

Single-Crystal X-ray Diffraction

Single-crystal X-ray diffraction data for 1a, 1b_15 bar, and 1b_20 bar were collected on a Bruker D8 Venture instrument equipped with a Photon II CPAD detector and an Oxford Cryosystems 800Plus cryostat, using graphite-monochromated Mo–Kα (λ = 0.710 73 Å) radiation. The single-crystal data of 1b were obtained from a Bruker SMART APEX II CCD-based diffractometer with graphite-monochromatized Mo Kα radiation (λ = 0.710 73 Å).

For 1a, the crystal was mounted on a glass fiber using super glue, and data were collected at room temperature. For 1b, 1b_15 bar, or 1b_20 bar the crystal was mounted on a glass fiber inside an environmental gas cell.[35,42] Prior to data collection the gas cell was evacuated at ∼2 × 10–3 mbar for 12 h to remove any potential traces of guest molecules or adsorbed water vapor. The crystal was then subjected to pressures of 15 bar (1b_15 bar) and 20 bar (1b_20 bar), and the data were collected for each of these pressures.

Data reduction was performed via a standard procedure with the use of the Bruker software package SAINT.[43] SADABS[44] was used for absorption corrections and correction of other systematic errors. The crystal structures were solved through direct methods utilizing SHELXS-97 and refined using SHELXL-97.[45] X-Seed[46] was employed as the graphical interface for the SHELX program suite. The contribution of disordered guest molecules was treated as diffuse with the utilization of the Squeeze procedure implemented in the Platon program.[41] All non-hydrogen atoms were refined anisotropically. Hydrogen atoms were placed in calculated positions using riding models.

Density-Functional Theory Calculations

First-principles density-functional theory calculations were performed using a Quantum-Espresso package.[47] A semiempirical addition of dispersive forces to conventional DFT (DFT-D) was included in the calculation to account for van der Waals interactions.[48] We used Vanderbilt-type ultrasoft pseudopotentials and the generalized gradient approximation (GGA) with the Perdew–Burke–Ernzerhof (PBE) exchange correlation. A cutoff energy of 544 eV and a 2 × 2 × 4 k-point mesh (generated using the Monkhorst–Pack scheme) were determined to be sufficient for total energy to converge within 0.01 meV atom–1. To study the structural flexibility, full structural relaxations were first performed on 1a and 1b phases. Then, nudged elastic band calculations were performed to find the minimum energy path of the structural transition between 1a and 1b. To calculate the CO2 adsorption sites, the open-pore structure of 1b_open was optimized first. Then, CO2 molecules were introduced to the optimized MOF structure, followed by full structural relaxations.