Probing Molecular Motions in Metal–Organic Frameworks by Three-Dimensional Electron Diffraction

Flexible metal–organic frameworks (MOFs) are known for their vast functional diversities and variable pore architectures. Dynamic motions or perturbations are among the highly desired flexibilities, which are key to guest diffusion processes. Therefore, probing such motions, especially at an atomic level, is crucial for revealing the unique properties and identifying the applications of MOFs. Nuclear magnetic resonance (NMR) and single-crystal X-ray diffraction (SCXRD) are the most important techniques to characterize molecular motions but require pure samples or large single crystals (>5 × 5 × 5 μm3), which are often inaccessible for MOF synthesis. Recent developments of three-dimensional electron diffraction (3D ED) have pushed the limits of single-crystal structural analysis. Accurate atomic information can be obtained by 3D ED from nanometer- and submicrometer-sized crystals and samples containing multiple phases. Here, we report the study of molecular motions by using the 3D ED method in MIL-140C and UiO-67, which are obtained as nanosized crystals coexisting in a mixture. In addition to an ab initio determination of their framework structures, we discovered that motions of the linker molecules could be revealed by observing the thermal ellipsoid models and analyzing the atomic anisotropic displacement parameters (ADPs) at room temperature (298 K) and cryogenic temperature (98 K). Interestingly, despite the same type of linker molecule occupying two symmetry-independent positions in MIL-140C, we observed significantly larger motions for the isolated linkers in comparison to those reinforced by π–π stacking. With an accuracy comparable to that of SCXRD, we show for the first time that 3D ED can be a powerful tool to investigate dynamics at an atomic level, which is particularly beneficial for nanocrystalline materials and/or phase mixtures.


S3
Solid state nuclear magnetic resonance (NMR). Magic-angle-spinning (MAS) NMR experiments were performed at a magnetic field of 14.1 T (Larmor frequencies of 600.12, and 150.92 for 1 H, and 13 C, respectively) on a Bruker Avance-III spectrometer. Spectra were acquired using a 1.3 mm probehead and a 60 kHz MAS rate. 1 H MAS NMR acquisition involved a use of a rotor-synchronized, double-adiabatic spin-echo sequence with a 90° excitation pulse of 1.25 µs followed by a pair of 50.0 µs tanh/tan short high-power adiabatic pulses (SHAPs) with 5 MHz frequency sweep 2,3 . All pulses operated at the nutation frequency of 200 kHz, and 128 signal transients were acquired using a relaxation delay of 5 s. Crosspolarization (CP) 1 H-13 C CPMAS NMR spectrum involved Hartmann-Hahn matched radiofrequency fields applied for a contact interval of 1.5 ms and 85 kHz spinal64 proton decoupling. 16384 signal transients were collected using a relaxation delay of 5 s. To observe spinning sideband pattern resulting from 13 C NMR chemical shift anisotropy (CSA), an additional 1 H-13 C CPMAS acquisition was performed with a 7 mm probehead at a 7 kHz MAS rate and using 65 kHz spinal64 proton decoupling. 2048 signal transients were collected using a relaxation delay of 5 s. Chemical shifts were referenced with respect to neat tetramethylsilane (TMS).

Section 2. Synthesis of UiO-67 and MIL-140C
Synthesis of pure UiO-67. UiO-67 (Zr6O4(OH)4(O2C(C6H4 )2CO2)6) was synthesized using previously reported procedure 4 . 2.0 g of zirconium chloride (1 eq) was added to the mixture of 0.93 mL distilled water (6 eq) and 33.3 mL of DMF (50 eq) at room temperature on a stirring plate. This solution was heated and 3.14 g of BA (3 eq) was added, stirred until completely dissolved and then 2.92 g of bpdc (1 eq) was added. The solution was transferred to a round bottom flask and heated at 130 °C overnight with stirring and a condenser. The resulting product was filtered and washed with hot DMF and acetone, and dried overnight at 150 °C.

Synthesis of mixed phase UiO-67 and MIL-140C.
A systematic study was performed to examine the effect of addition of small amounts of water in the synthesis of UiO-67 using ZrCl4: 4,4'-biphenyldicarboxylic acid (H2bpdc): benzoic acid: DMF as 1:1:3:50 molar equiv. In this series the reactions with intermediate amount of water revealed a mixed phase of MIL-140C and UiO-67. Three samples were synthesized, namely mixed phase (5 eq), mixed phase (4 eq) and mixed phase (3 eq) where 0.77 mL (5 eq), 0.62 mL (4 eq) and 0.46 mL (3 eq) of water was added, respectively, in the reaction mixture. These experimental procedures were performed the same as described in the above section.
Synthesis of pure MIL-140C. Synthesis of MIL-140C (ZrO(O2C12H8-CO2)) was done using the reported protocol with modifications 5 . Typically, 1.0 g of zirconium chloride (1 eq) was added to 16.6 mL of DMF (50 eq) in a 50 mL autoclave. 1.57 g of BA (3 eq) was added under stirring, followed by 1.46 g of bpdc (1 eq). The autoclave was sealed and heated overnight at 220 °C. A white product was obtained which was filtered and washed with hot DMF and acetone, and finally dried overnight at 150 °C.

Discussion.
Understanding the role of the reagents and their amount are vital as they can affect the reproducibility of the MOFs. In the synthesis of UiO-67, the reagent ratios play important S4 roles to obtain defect free crystals 4 . In this work, we study the effect of amount of water. The synthesis was conducted by using molar ratio of ZrCl4:H2bpdc:BA:H2O:DMF (anhydrous) = 1:1:3:x:50 (x= 0, 3,4,5,6). When anhydrous DMF is used, the addition of water in the synthesis becomes an important role as it is the only source of water. The PXRD patterns of the obtained products show that pure UiO-67 and MIL-140 were obtained using 0 and 6 equivalent of water, respectively, and phase mixtures of UiO-67 and MIL-140C were obtained using 3, 4, and 5 equivalent of water ( Figure S1). The additional peaks (marked by asterisk) in the PXRD patterns are attributed to MIL-140C. Moreover, SEM images show octahedral morphology of UiO-67 and plate-like morphology of MIL-140C ( Figure S2).

Section 3. Structural analysis of UiO-67 and MIL-140C
Space groups and unit-cell parameters were initially obtained from the software REDp 6  The data were subsequently processed through XDS 7,8 to extract the reflection intensities. The unit cell parameters of UiO-67 and MIL-140C are refined as a = b= c= 26.880(3) Å and α = β = γ = 90°, and a = 32.360(7) Å, b = 15.800(3) Å, c = 7.910(2) Å, α = 90°, β = 103.00(3)° and γ = 90°, respectively. Although some diffuse scattering along the a*-axis is observed ( Figure  S4c), it is very weak and therefore not considered in the structure determination. The structures were solved ab initio with dual-space algorithm implemented in SHELXT and refined with a full-matrix least squares technique using SHELXL 9,10 . During the refinement, atomic scattering factors and wavelength for electrons were used. For both MOFs, all the non-hydrogen atom positions could be directly located from the initial structure solution. Hydrogen atoms were added with HFIX 43. For UiO-67, the structure was solved and refined from an individual dataset. During the refinement of UiO-67, soft geometrical restrains DANG were applied on the linker. The refinement converged with the agreement values R1=0.215 for 404 reflections with Fo > 4σ(Fo) and 0.249 for all 703 reflections (Table S1). For MIL-140C, because of the preferred orientation, three datasets per temperature from individual crystals were merged for structure determination. The merging was performed with XSCALE, which is part of the XDS package. The correlation coefficients of the common reflection intensities (CCI) were used to assess the quality of merging (Table S4) 11 . During the refinement of MIL-140C, soft S5 geometrical restrains (DFIX, DANG, and FLAT) were applied on the non π-stacked linker molecule. For the data collected at room temperature, SIMU was applied on carbon atoms along the molecular axis of the π-stacked linker. The refinement converged to R1=0.169 for 2083 reflections with Fo > 4σ(Fo) and 0.210 for all 3079 data for the crystals at room temperature (298 K), and R1=0.173 for 1892 reflections with Fo > 4σ(Fo) and 0.200 for all 2685 data for the crystals at cryogenic temperature (98 K) (Table S1). At the end of the refinement, the SWAT 12 parameter was applied to all structures to compensate the effects of solvents in the pores. The relatively high R1 values are mainly caused by dynamical effects that make the intensities deviated from kinematic intensities. For the analysis of the ADPs, the principal mean square atomic displacements U given as an output by SHELXL were used. These values correspond to the three eigenvalues λ1, λ2 and λ3. The values from three different merged datasets, each one obtained by merging three datasets from individual crystals, were averaged and standard deviations for the eigenvalues were calculated.

Section 4. NMR analysis
The 1 H solid-state NMR spectrum of MIL-140C revealed five proton resonances at 8.2, 7.5, 6.6, 5.9, and 5.4 ppm, which reflects differences in distances to inorganic fragments and distinct linker conformations ( Figure S8a). Signal at 1.7 ppm is attributed to water molecules present in the pores of the material 13 . The 1H-13C CPMAS spectrum shows resonances in the range of 120-150 ppm, which originate from carbon atoms in the phenyl rings of the bpdc linkers, whereas those of carboxyl groups give signals in the 170-180 ppm region ( Figure S8b). In addition, the 1 H-13 C CPMAS spectrum at a slow MAS rate of 7 kHz was recorded to allow observation of the 13 C chemical shift anisotropy (CSA) pattern ( Figure S8c). The estimate of the chemical shift anisotropy (δaniso) as well as the shape of the CSA pattern provide insights into the potential dynamics and motion of the linkers within the MIL-140C framework. Three pairs of spinning sidebands are observed in Figure S8c. The first sideband at c.a. 180 ppm exhibits substantially higher intensity than its counterpart at c.a. 80 ppm, whereas the remaining two pairs of sidebands have comparable intensities. This indicates an axially asymmetric NMR shielding tensor and CSA pattern (δxx≠δyy≠δzz), which is expected for carbon atoms involved in the phenyl group with restricted motion. In the case of π-flips, the partial averaging of the shielding tensor would occur, and the averaged shielding tensor is expected to display axial symmetry and narrower CSA pattern 14,15 . Moreover, since three pairs of sidebands are present, the shift anisotropy in frequency units can be estimated to be around three times the MAS rate, which translates to an average δaniso=~140 ppm on our experimental setup. Because of relatively wide range of isotropic chemical shifts from phenyl rings of MIL-140C linkers, it was not possible to acquire a spectrum at even lower MAS rate in order to provide better accuracy for chemical shift anisotropy determination by involving more sidebands. The estimated average value of δaniso=~140 ppm is moderately lower than δaniso=180 ppm reported for the benzene at 20 K 16 . However, the low-temperature asymmetric CSA pattern of benzene collapses into an axially symmetric one at 223 K due to rapid molecular reorientation 17 , which is not the case for the MIL-140C linkers. Furthermore, the 13C chemical shift anisotropies for phenyl groups of the MIL-140C linkers were also calculated on the models derived from our 3D ED structural model. Calculations were performed with the robust DSD-PBEP86/pcSseg-2 level of theory 18 . Obtained δaniso values were in the range of 110-190 ppm with the mean chemical shift anisotropy of δaniso=162 ppm, which is close to our estimate for the MIL-140C. This together with the axially asymmetric CSA tensor suggests the absence of π-flips, while librations are probably present in the MIL-140C structure at the room temperature.