Higher Magnetic Fields, Finer MOF Structural Information: ¹⁷O Solid-State NMR at 35.2 T

Abstract

The spectroscopic study of oxygen, a vital element in materials, physical, and life sciences, is of tremendous fundamental and practical importance. ¹⁷O solid-state NMR (SSNMR) spectroscopy has evolved into an ideal site-specific characterization tool, furnishing valuable information on the local geometric and bonding environments about chemically distinct and, in some favorable cases, crystallographically inequivalent oxygen sites. However, ¹⁷O is a challenging nucleus to study via SSNMR, as it suffers from low sensitivity and resolution, owing to the quadrupolar interaction and low ¹⁷O natural abundance. Herein, we report a significant advance in ¹⁷O SSNMR spectroscopy. ¹⁷O isotopic enrichment and the use of an ultrahigh 35.2 T magnetic field has unlocked identification of many inequivalent carboxylate oxygen sites in the as-made and activated phases of the metal–organic framework (MOF) α-Mg₃(HCOO)₆. The subtle ¹⁷O spectral differences between the as-made and activated phases yield detailed information about host–guest interactions, including insight into nonconventional O…H─C hydrogen bonding. Such weak interactions often play key roles in applications of MOFs, such as gas adsorption and biomedicine, and are usually difficult to study via other characterization routes. The power of performing ¹⁷O SSNMR experiments at ultrahigh magnetic field of 35.2 T for MOF characterization is further demonstrated by examining activation of the MIL-53(Al) MOF. The sensitivity and resolution enhanced at 35.2 T allows partially- and fully-activated MIL-53(Al) to be unambiguously distinguished, and also permits several oxygen environments in the partially-activated phase to be tentatively identified. This demonstration of the very high resolution of ¹⁷O SSNMR recorded at the highest magnetic field accessible to chemists to date illustrates how a broad variety of scientists can now study oxygen-containing materials and obtain previously inaccessible fine structural information.

Graphical Abstract

1. Introduction

The element of oxygen is ubiquitous across nearly all scientific fields. Therefore, characterization of oxygen local electronic and geometric environments is very important. ¹⁷O solid-state NMR (SSNMR) spectroscopy has become an ideal site-specific characterization tool for probing oxygen local environments, as ¹⁷O is sensitive to the chemical shift and quadrupolar interactions,^1-11 has a large diagnostic chemical shift range,^12-25 and is influenced by coupling to neighboring NMR-active nuclei (e.g., ¹H, ¹³C, and ¹⁵N).^26-31 There has been tremendous progress made in NMR methodology and technology in recent years, yet the potential of ¹⁷O SSNMR for uncovering detailed structural and bonding information in oxygen-containing compounds has been limited by the inherently low sensitivity and resolution resulting from the very low natural abundance (0.038%), relatively low gyromagnetic ratio (γ = −5.774 MHz·T⁻¹), and quadrupolar nature (spin I = 5/2) of ¹⁷O.³²

The sensitivity problem associated with the low ¹⁷O natural abundance can be mitigated by isotopic enrichment.^16-22 To address the relatively low γ and quadrupolar nature of ¹⁷O, NMR measurements can be performed at high magnetic fields; this not only inherently enhances sensitivity but also reduces spectral line broadening associated with the second-order quadrupolar interaction. A new series-connected resistive/superconducting hybrid magnet operating at a record-high magnetic field strength of 35.2 T (¹H Larmor frequency of 1.5 GHz) has recently entered service,³³ which promises very high ¹⁷O SSNMR resolution in biomolecules and minerals.^33-35 In this work, taking advantage of the state-of-the-art magnet and rf technology, we targeted microporous α-Mg₃(HCOO)₆ to demonstrate that very high spectral resolution of ¹⁷O SSNMR spectra can be achieved at 35.2 T, providing an excellent opportunity for characterizing promising materials such as metal-organic frameworks (MOFs).

MOFs are a fascinating family of hybrid organic-inorganic porous materials with many practical applications.^36-37 SSNMR spectroscopy of MOFs has proven to be a powerful tool for characterizing the immediate environment about metal centers and probing the local structure of organic linkers.^38-40 SSNMR can also provide information on the behavior of adsorbed guests, which is critically important for many applications.^41-42 For example, MOFs are promising materials for the removal of greenhouse gas such as CO₂ and storage of fuels such as H₂ and CH₄. SSNMR can provide information on the location of guest species,^{10, 43-47} which is critically important for practical applications, as the location of guest gas molecules can be directly linked to binding site positions and strength, respectively. Similarly, ascertaining the location of guest species in MOFs is key to understanding their applications in sensing and drug delivery.^48-51 Oxygen present in various carboxylate ligands, which are the most extensively used organic linkers, is a key constituent of many important MOFs.^52-56 Oxygen anions (O²⁻) are also associated with the metal clusters of the frameworks (e.g., MOF-5).⁵² Hydroxyl groups are common linkers bridging metal clusters (e.g., MIL-53)⁵³ and exist as part of the secondary building units (e.g., UiO-66).⁵⁴ Water molecules can directly bond to the metal center, with well-known examples include as-made MOF-74 and HKUST-1.^55-56 These oxygen species play critical roles in applications such as guest adsorption/separation,⁵⁷ sensing,⁴⁸ catalysis,⁵⁸ solid-state conductors,⁵⁹ and biomedicine,^49-51 rendering oxygen a key target for MOF characterization. Although ¹⁷O SSNMR has been utilized to examine some MOFs,^{10, 15-17} potential successes in molecular-level characterization and site assignment have been limited by spectral resolution.

Microporous α-Mg₃(HCOO)₆ is an attractive target for ¹⁷O SSNMR characterization for several reasons.

1. Microporous α-Mg₃(HCOO)₆ is a commercially available and low-cost MOF with good molecular selectivity, such as a preference for C₂H₂ over CO₂.^60-61 It is a representative small pore MOF suitable for gas adsorption.

2. This MOF presents a very challenging case for characterization by ¹⁷O SSNMR, as the crystal structure features twelve inequivalent carboxylate oxygen sites across two bonding modes. Using our highest available field of 21.1 T at the time, only two ¹⁷O NMR signals corresponding to the two different oxygen bonding modes of formate anions could be observed (vide infra).¹⁶ As increasing the magnetic field from 21.1 to 35.2 T leads to an improvement in resolution by a factor of 2.8 (the second-order quadrupolar broadening in ppm varies as the inverse ratio of the fields squared),¹ resolving many of these inequivalent oxygen sites at 35.2 T should be possible.

3. The applications of MOFs require the removal of the solvent molecules from inside the pores of as-made MOFs in a process known as “activation”. Activation often leads to changes in the framework structure. While significant changes can be detected by X-ray diffraction, activation-induced changes for MOFs including α-Mg₃(HCOO)₆ can be subtle (i.e., small changes in unit cell parameters), and the specific molecular-level alterations cannot be detected by diffraction-based methods. Thus, it is important to develop SSNMR as a method complementary to XRD. The high spectral resolution that can be achieved at 35.2 T holds promise for providing fine details on oxygen local environments, reflecting the changes resulting from activation.

4. MOFs have many potential applications across fields such as biology and medicine.^62-63 Thus, nonconventional O…H─C hydrogen bonding involving guest species and bioactive components in the framework can play important roles for host–guest interactions in BioMOFs, a new subclass of MOFs,^49-50 or MOF-based drug delivery systems.⁵¹ Since nonconventional O…H─C hydrogen bonding has been observed in as-made α-Mg₃(HCOO)₆ (C₆H₆Mg₃O₁₂·C₃H₇NO, C₃H₇NO is N,N´-dimethylformamide or DMF),⁶⁴ this system has been selected as a model compound to explore the possibility of using ¹⁷O SSNMR at 35.2 T to directly probe this weak host–guest interaction in MOF systems.

Our results show that the ¹⁷O SSNMR spectra of α-Mg₃(HCOO)₆ samples acquired at 35.2 T indeed exhibit very high resolution, allowing many inequivalent framework oxygen sites to be identified. The high resolution and sensitivity realized at 35.2 T not only lead to ultrafine information on oxygen local environment and corresponding subtle changes upon activation, but also make it possible to detect weak host–guest interactions such as nonconventional O…H─C hydrogen bonding. To further illustrate the benefits of performing ¹⁷O SSNMR at an ultrahigh field of 35.2 T, we examined two activated MOF MIL-53(Al) samples; the high spectral resolution and sensitivity allow us to unambiguously distinguish partially-activated from fully-activated MIL-53(Al) samples. The very high resolution and sensitivity of ¹⁷O SSNMR achievable at 35.2 T detailed in this work illustrates the great potential of SSNMR for unlocking fine structural information in solids as higher magnetic fields become increasingly available.

2. Experimental Methods

2.1. Synthesis and characterization of MOF samples.

As-made ¹⁷O-enriched α-Mg₃(HCOO)₆ was synthesized according to the reported procedure.¹⁶ The starting materials were used as received without further purification. A mixture of Mg(NO₃)₂·6H₂O (3 mmol, Aldrich) and HCOOH (6 mmol, Aldrich) was dissolved in 10 mL of DMF and 0.25 mL of ¹⁷O-enriched H₂O (6 mmol, CortecNet, 35 atom%) in a 23 mL Teflon-lined autoclave. The container was sealed and heated at 383 K for 2 days. After cooling the autoclave to room temperature, the white powder product was collected, washed with DMF and dried overnight at 363 K. Activated ¹⁷O-enriched α-Mg₃(HCOO)₆ was obtained by heating as-made ¹⁷O-enriched α-Mg₃(HCOO)₆ at 423 K overnight under dynamic vacuum. The PXRD patterns of ¹⁷O-enriched α-Mg₃(HCOO)₆ samples are shown in Figure S1, Supporting Information (SI).

In the experimental PXRD patterns of as-made and activated α-Mg₃(HCOO)₆, the two low angle reflections at 2θ = 9.7° (−1 0 1) and 9.9° (1 0 1) are much weaker than the reflection at 10.7° (0 1 1), which strays from the comparable intensities apparent in the simulated PXRD spectra. A previous study demonstrated that the PXRD patterns of α-Mg₃(HCOO)₆ crystals prepared under different synthetic conditions may exhibit different intensity patterns.⁶⁵ In particular, the three low-angle reflections at 9.7°, 9.9°, and 10.7° have different relative intensities if water is involved. For α-Mg₃(HCOO)₆ samples prepared in the presence of water, the relative intensity pattern for the three above-mentioned low-angle diffractions looks very similar to that seen in Figure S1, but distinct from the simulated PXRD pattern based on the structure determined from a single crystal prepared in non-aqueous DMF solvent.⁶⁰ In the present work, ¹⁷O-enriched α-Mg₃(HCOO)₆ was prepared in the presence of ¹⁷O-enriched H₂O. Although the synthesis of α-Mg₃(HCOO)₆ is very straightforward and highly reproducible, and we have made this MOF routinely in many studies with and without water,^{16, 43-46, 64, 66-68} to unambiguously confirm the identity of the samples used in the present study we repeated the sample preparation several times. These additional samples were prepared under the exact same conditions used for preparing ¹⁷O-enriched MOF, except that normal water was used rather than ¹⁷O-enriched water. Figure S1 indicates that when the samples were prepared in the presence of a small amount of water, the relative intensities of the first two low-angle reflections are significantly lower than that of the third one, which is consistent with the literature.⁶⁵ We performed a Le Bail fit of the PXRD data using the GSAS II package (Figure S2, see the SI for details),⁶⁹ and obtained unit cell parameters of the samples prepared in the presence of water that are comparable to those reported in the literature (Table S1); this data indicates that although the relative intensities of reflections may differ, the samples indeed share the same crystal structure.

The activation process for this MOF has been proven to be robust.^{60, 64, 66, 70} To verify the solvent DMF molecules occluded inside the channels are completely removed upon activation, TGA profiles and ¹H-¹³C CP/MAS SSNMR spectra were obtained (Figure S3 and S4, with experimental details in Section S2). The results unanimously agree that the activation process is very effective and the solvent molecules are completely removed.

The synthesis of ¹⁷O-enriched MIL-53(Al) samples and corresponding PXRD patterns (Figure S5) can be found in Section S3 of the Supporting Information.

2.2. ¹⁷O solid-state NMR measurements.

The 1D rotor-synchronized spin-echo spectrum of activated α-Mg₃(HCOO)₆ was recorded at 21.1 T (¹⁷O Larmor frequency of 122.0 MHz) on a Bruker Avance II spectrometer at the National Ultrahigh-Field NMR Facility for Solids in Ottawa, Canada. A 4 mm H/X MAS Bruker probe and a spinning frequency of 18 kHz were used. The recycle delay was 4 s. A π/2 pulse of 4 μs was used.

1D and 2D ¹⁷O SSNMR experiments at 35.2 T were performed on the series-connected hybrid (SCH) magnet (¹⁷O Larmor frequency of 203.4 MHz) at the National High Magnetic Field Laboratory (NHMFL) in Tallahassee, USA.³³ A Bruker Avance NEO console and a NHMFL home-built single-resonance 3.2 mm low-gamma MAS probe were used, along with a spinning frequency of 18 kHz and a pulse delay of 0.1 s. The pulse delay was optimized to achieve the highest S/N and no significant changes in NMR lineshapes were observed when employing different pulse delays. A π/2 pulse of 5.0 μs was used in 1D rotor-synchronized spin echo experiments. All rotor-synchronized 3QMAS spectra were acquired using the shifted-echo pulse sequence.⁷¹ The 3Q excitation and conversion pulses were 3.0 and 1.0 μs, respectively. The number of t₁ increments was 34. Since the shifted-echo is phase modulated, the number of time increments was also 34, corresponding to a maximum t₁ evolution time of ~1.9 ms. Note that during spectral acquisition, the number of t₁ increments were carefully chosen to include a significant portion beyond the point where the signal dropped to the baseline noise level, ensuring that the shifted-echo 3QMAS spectra were acquired with the highest S/N during the limited SCH magnet time without compromising the resolution. The 3QMAS spectra were acquired using rotor-synchronized t₁ increments to avoid spinning sidebands, which are significant due to the chemical shift anisotropy (CSA) at the very high field and 3Q excitation/conversion modulation along the indirect F1 dimension. However, this restricts the F1 window to the spinning frequency (18 kHz), which is not wide enough to cover the spread of resonances and spinning side bands (SSBs). Fortunately, this problem can be solved by the Q-shearing method to the k = 3 Q-representation.⁷² The F1 spectral window can then be zero-filled and expanded to whatever limits are needed for shearing back to the isotropic representation with a large and unfolded F1 window. In the present case, the F1 spectral window was zero-filled eight times and expanded to give an unfolded F1 window of 8 × 18 kHz after Q-shearing. Consequently, the F1 digital resolution is equivalent to 8 × 34 t₁ increments with the final expanded F1 window. The ¹⁷O spectra were referenced to 18 atom% ¹⁷O-enriched H₂¹⁷O(l) or distilled water at 0 ppm.

2.3. Spectral simulations.

For quadrupolar nuclei such as ¹⁷O (spin I > 1/2), their electric quadrupole moments interact with the surrounding EFG, resulting in broad powder patterns rather than sharper resonances. The interplay between the quadrupolar interaction with the CSA effect makes the shapes of powder patterns more complicated and difficult to simulate. The dmfit software package was used to simulate SSNMR spectra using the Int2QUAD mode, including both the quadrupolar and CSA effects.⁷³ In dmfit, the EFG tensor is described by three principal components in the following order: ∣V_YY∣ ≤ ∣V_XX∣ ≤ ∣V_ZZ∣. The quadrupolar coupling constant (C_Q) and asymmetry parameter (η_Q) describe the spherical and cylindrical symmetry of the EFG tensor, respectively, and are defined as follows: C_Q = (eQV_ZZ/h) × 9.7177 × 10²¹ (in Hz) and η_Q = (V_YY – V_XX)/V_ZZ, where e is the electric charge, Q is the quadrupole moment (−2.558 × 10⁻³⁰ m²),³² and a conversion factor of 9.7177 × 10²¹ V·m⁻² is used during the calculation of C_Q to convert from atomic units to Hz. The chemical shift (CS) tensor is also described by three principal components such that ∣δ₂₂ – δ_iso∣ ≤ ∣ δ₁₁ – δ_iso∣ ≤ ∣ δ₃₃ – δ_iso∣, and the isotropic chemical shift δ_iso= (δ₁₁ + δ₂₂ + δ₃₃)/3 relates to the bonding modes. The CSA parameters are defined by Δ_CS = δ₃₃ – δ_iso and η_CS = ∣( δ₂₂ – δ₁₁)/ Δ_CS∣. Three Euler angles (ϕ, χ, ψ) are employed to describe the orientations of the CS tensor with respect to the EFG principal (fixed) axis frame using the ZYZ convention: the corresponding transformation matrix was used to deduce the new directional characteristics of the CS tensor with respect to the EFG system. As a result, eight independent parameters, C_Q, η_Q, δ_iso, Δ_CS, η_CS, ϕ, χ, and ψ are required to characterize a single ¹⁷O site when both the quadrupolar and the CSA effects are considered. All uncertainties in NMR parameters were estimated by bidirectional variation of the parameter of interest in both directions from the best-fit value while holding all other NMR parameters constant.

2.4. Theoretical calculations.

The unit cell parameters were set to the single-crystal XRD parameters⁶⁰ and kept fixed during geometry optimizations to ensure consistency between experimental and optimized structures. Proton positions were then optimized using the VASP (Vienna Ab-initio Simulation Package) code⁷⁴ based on the Kohn-Sham Density Functional Theory (DFT) and using a plane-wave pseudopotential approach. The NMR parameters were then calculated within Kohn-Sham DFT using the QUANTUM-ESPRESSO code.⁷⁵ The PBE generalized gradient approximation⁷⁶ was used and the valence electrons were described by norm-conserving pseudopotentials⁷⁷ in the Kleinman Bylander form.⁷⁸ The wave functions were expanded on a plane wave basis set with a kinetic energy cut-off of 80 Ry. The integral over the first Brillouin zone was performed using a Monkhorst–Pack 2 × 2 × 2 k-point grid for the charge density and chemical shift tensor calculation. The magnetic shielding tensor was computed using the Gauge-Including Projector Augmented Wave (GIPAW) approach,^79-81 which enables the reproduction of the results of a fully converged all-electron calculation. The isotropic chemical shift δ_iso is defined as δ_iso = [σ_iso – σ_iso(ref)], where σ_iso is the isotropic magnetic shielding and σ_iso(ref) is the isotropic magnetic shielding of the same nucleus in a reference compound. In the present case, the fit of the linear correlation between the experimental δ_iso and the calculated σ_iso values of ¹⁷O for Na₂SiO₃, α-Na₂Si₂O₅, α- and γ-glycine, and α-SrSiO₃ enabled the determination of the relation between δ_iso and calculated σ_iso for the ¹⁷O nucleus, as described previously.²¹ It is worth noting that for most MOFs, the solvent molecules are disordered inside the framework. Very often, disordered solvent molecules have to be removed before calculation. In the present case, the DMF molecules are orderly distributed within the channels. This rare situation makes the calculation not only simpler, but also more accurate.

3. Results and Discussion

The three-dimensional framework of microporous α-Mg₃(HCOO)₆ is formed by corner- and edge-sharing MgO₆ octahedra interconnected by formate ligands (Figure 1), and features zig-zag channels measuring 4.5 × 5.5 Å.⁶⁰ Among the twelve crystallographically distinct framework oxygen sites, six are associated with the carboxylate group and adopt a μ²-O bonding mode (sites O1, O3, O5, O7, O9, and O11), and the other six are associated with the carboxylate group and are in a μ¹-O bonding mode (sites O2, O4, O6, O8, O10, and O12). The C─μ¹-O bonds are of shorter length and more double-bond character than the C─μ²-O bonds. However, the local environments of all six μ¹-O sites are almost identical, while the six μ²-O sites are also very similar. As a result, only two signal groups were resolved in previous ¹⁷O 1D magic-angle spinning (MAS) spectrum of as-made α-Mg₃(HCOO)₆ at 21.1 T (Figure 1).¹⁶ Those two ¹⁷O signals were simulated reasonably well by two ¹⁷O powder patterns at 21.1 T and were assigned to groups of μ¹- and μ²-O sites, respectively (Figure S6, Supporting Information).

Figure 1.

Top (left to right): Representations of the framework of activated α-Mg₃(HCOO)₆, twelve framework oxygen sites, and two different oxygen bonding modes. Color coding: Mg, turquoise; C, grey; H, white; O, red; μ¹-O, orange; μ²-O, pink. Bottom: ¹⁷O 1D MAS NMR spectra of ¹⁷O-enriched α-Mg₃(HCOO)₆ at fields of 35.2 T (blue) and 21.1 T (black) acquired at a spinning frequency of 18 kHz. The asterisk (*) denotes spinning sidebands (SSBs). The ¹⁷O 1D MAS NMR spectrum of as-made α-Mg₃(HCOO)₆ at 21.1 T is adapted with permission from the American Chemical Society.¹⁶

As-made MOFs typically consist of solvent molecules occupying their pores and channels; thus the creation of permeable spaces in MOFs by evacuating the solvent (i.e., the activation process) is a pre-requisite for many applications. Therefore, it is of fundamental importance to understand the effect of activation on the MOF structure, especially regarding the subtle local environment changes that are invisible in diffraction-based techniques but very important for applications. The single-crystal XRD data of as-made and activated α-Mg₃(HCOO)₆ phases indicate that the local oxygen environments only undergo very minor changes upon removal of DMF solvent during activation, as the Mg ions are coordinately saturated and this MOF framework is fairly rigid.⁶⁰ Consequently, detecting the very small activation-induced structural changes via ¹⁷O SSNMR is quite challenging, as evidenced by the nearly identical ¹⁷O 1D MAS spectra of two phases at 21.1 T (Figure 1). It is apparent that higher spectral resolution is necessary to detect the subtle difference in oxygen local environment. The activation method employed in this work for DMF solvent removal from α-Mg₃(HCOO)₆ is well established.^{60, 70} Based on our previous experience, we are certain that this activation process completely removes all residual DMF guests from the as-made MOF.^{64, 66} To confirm, we carried out TGA alongside ¹H─¹³C CP/MAS experiments, and the results unambiguously prove that the activation is complete (see Section S2 for details).

The newly-obtained ¹⁷O 1D MAS NMR spectra of as-made and activated α-Mg₃(HCOO)₆ phases at 35.2 T are shown in Figure 1 alongside the 21.1 T spectra. At 35.2 T, the resonances of both phases are considerably narrower, owing to the reduced second-order quadrupolar broadening. The spectral envelope containing all overlapping signals of the μ¹-O group is now completely separated from that of the μ²-O group at 35.2 T, and several diagnostic spectral features including the “edges” and “horns” of individual ¹⁷O SSNMR resonances have emerged. Thus, this spectral envelope must consist of several overlapping powder patterns corresponding to multiple μ¹-O sites. Similarly, the spectral envelope of the μ²-O group should also be simulated using multiple powder patterns. However, due to the severe overlap of powder patterns, it is very challenging to determine the number of oxygen sites and their corresponding NMR parameters using only a 1D MAS spectrum. Furthermore, the ¹⁷O 1D MAS spectra of as-made and activated α-Mg₃(HCOO)₆ phases are now distinctly different at 35.2 T, implying that the activation-induced structural changes not apparent at 21.1 T are now discernable, owing to the much higher spectral resolution achieved at 35.2 T. It is worth mentioning that the changes in the μ¹-O spectral envelope are also more significant than those in the μ²-O one, suggesting that the local environments of μ¹-O sites are more influenced by activation versus μ²-O sites. The spectra exhibit very intense spinning sidebands due to the chemical shift anisotropy (CSA) enhanced at 35.2 T.

As mentioned earlier, the ¹⁷O 1D MAS spectra at 35.2 T feature overlapping resonances arising from multiple ¹⁷O sites, indicating that the maximum achievable 1D spectral resolution is still not high enough to resolve fine features. This is because the conventional 1D MAS experiments only partially averages the ¹⁷O second-order quadrupolar interaction. To further enhance spectral resolution, 2D 3QMAS experiments were performed:⁸² this technique can eliminate the ¹⁷O second-order quadrupolar broadening along the indirect F1 dimension and thus separate the overlapping signals observed in 1D MAS spectra. The ¹⁷O 2D 3QMAS spectra of as-made and activated α-Mg₃(HCOO)₆ at 35.2 T are shown in Figure 2. There are a number of well-resolved signals along the F1 dimension in the spectra of the as-made and activated phases, respectively. Since the number of resolved signals in the isotropic dimension (seven for the as-made phase and eight for the activated phase) is smaller than the twelve crystallographically distinct framework oxygen sites, some signals along the F1 dimension must correspond to very similar signals arising from multiple oxygen sites with almost identical NMR parameters.

Figure 2.

¹⁷O 2D 3QMAS NMR spectra of ¹⁷O-enriched α-Mg₃(HCOO)₆ at 35.2 T. Black dashed lines correspond to the slices examined. Blue and red solid lines denote experimental and simulated spectra, respectively. The asterisk (*) denotes the SSBs. The 3QMAS spectra were acquired using a shifted-echo MQMAS pulse sequence and rotor synchronized t₁, and processed using the Q-shearing method to avoid spectral folding of the peaks and SSBs.⁷² The 3QMAS spectra without markups are shown in Figure S7 for clarity.

For each F2 cross-section extracted at δ₁ along the F1 dimension, the isotropic chemical shift, δ_iso (in ppm), and the quadrupolar product, P_Q = C_Q(1 + η_Q²/3)^1/2 (in MHz), can be obtained directly from the spectral center of gravity (δ₂) along the F2 dimension¹⁷ using the following equations⁸³

$${\delta }_{\text{iso}}=\frac{17}{27}{\delta }_1+\frac{10}{27}{\delta }_2$$

$$P_{\mathrm{Q}}=\{\frac{170}{81}\frac{[4I(2I-1){]}^2}{[4I(I+1)-3]}({\delta }_1-{\delta }_2){\}}^{1∕2}{v}_0\times {10}^{-3}$$

where ν₀ is the Larmor frequency and I is the spin quantum number.

The δ_iso and P_Q values derived from each peak along the F1 dimension are given in Table 1. These two values are determined accurately from the resonance positions in F1 and F2 dimensions, without the need of fitting the F2 cross-section, under the S/N obtained with the limited magnet time.

Table 1.

Experimental ¹⁷O NMR parameters, calculated^a δ_iso values, and peak assignments of α-Mg₃(HCOO)₆.

Sample	O Type	δ1 (ppm)	PQ (MHz)	δiso (ppm)	δiso, calc (ppm)	Assignment	CQ (MHz)	ηQ
As-made	μ2-O	224(1)	7.2(5)	220(2)	227.1	O3	6.5(4)	0.80(10)
	μ2-O	227(1)	6.6(5)	223(2)	230.5	O9	6.0(4)	0.88(10)
	μ2-O	234(1)	6.6(5)	230(2)	236.1	O1
					238.9	O11
					239.4	O5
	μ2-O	241(1)	7.8(5)	236(2)	256.4	O7	7.0(4)	0.70(10)
	μ1-O	278(1)	7.5(5)	273(2)	288.7	O10
					291.2	O6
					293.2	O4
	μ1-O	293(1)	8.8(5)	286(2)	295.5	O12
					297.2	O8
	μ1-O	296(1)	8.3(5)	290(2)	306.5	O2	7.9(4)	0.55(10)
	DMF	305(1)	9.1(5)	298(2)	308.3	O1S
Activated	μ2-O	225(1)	6.6(5)	221(2)	227.7	O9	6.0(4)	0.85(10)
	μ2-O	230(1)	7.5(5)	225(2)	232.0	O3
					232.0	O5
	μ2-O	232(1)	7.2(5)	228(2)	234.7	O1	6.5(4)	0.80(10)
	μ2-O	235(1)	6.9(5)	231(2)	238.7	O11	6.1(4)	0.85(10)
	μ2-O	239(1)	7.2(5)	235(2)	243.5	O7	6.4(4)	0.80(10)
	μ1-O	290(1)	7.8(5)	285(2)	285.3	O6	7.7(4)	0.20(10)
	μ1-O	295(1)	8.1(5)	289(2)	288.0	O8
					289.9	O10
	μ1-O	302(1)	7.5(5)	297(2)	296.4	O12
					296.5	O4
					298.5	O2

The numbers in the parentheses are the estimated uncertainty of the last significant figure.

^aThe complete calculated ¹⁷O NMR parameters are shown in Table S2.

For the peaks along the isotropic dimension corresponding to a single oxygen site, C_Q and η_Q values can be extracted by fitting the F2 cross-section. If a peak along the F1 dimension originates from multiple oxygen sites, δ_iso and P_Q represent average values. To assign each resolved signal in the isotropic dimension to a single or multiple oxygen sites, gauge-including projector augmented wave (GIPAW) density functional theory (DFT) calculations were carried out,^79-81 as this approach has proven to be very reliable for ¹⁷O NMR spectral assignments in various systems.^{13-15, 20-21, 24} Although the calculated δ_iso value may not exactly match the experimental value, assignments of multiple signals based on relative calculated δ_iso values are typically valid.^{13-15, 20-21, 24} Accordingly, calculated δ_iso values were utilized for the assignment of each signal in the F1 dimension (Table 1) to the framework oxygen site(s). For example, the order of calculated δ_iso values for μ²-O sites in as-made α-Mg₃(HCOO)₆ is O3 < O9 < O1 ≈ O11 ≈ O5 < O7. The ¹⁷O signal with the lowest measured δ_iso of 220 ppm (δ₁ = 224 ppm) is thus assigned to O3, the ¹⁷O signal with the second-lowest δ_iso of 223 ppm (δ₁ = 227 ppm) is assigned to O9, the ¹⁷O signal with the third-lowest δ_iso of 230 ppm (δ₁ = 234 ppm) and significantly higher intensity is assigned to O1, O5 and O11, and the ¹⁷O signal with the highest δ_iso of 236 ppm (δ₁ = 241 ppm) is assigned to O7. The ¹⁷O signals of μ¹-O sites are assigned in a similar fashion. Since the ¹⁷O signals at δ₁ = 224 ppm, 227 ppm, and 241 ppm each correspond to a single oxygen site, their C_Q and η_Q values can be further extracted by fitting the F2 cross-sections (Table 1). The 2D 3QMAS spectrum of activated α-Mg₃(HCOO)₆ was analyzed with the same approach and the results are also shown in Table 1.

The spinning sidebands in 1D spectra are particularly intense, suggesting a very large chemical shift anisotropy (CSA) is present at the ultrahigh magnetic field of 35.2 T. Efforts were made to simulate ¹⁷O 1D MAS spectra to estimate the ¹⁷O CSA. Theoretically, fitting a 1D MAS spectrum with 12 sites requires 96 independent parameters if both the EFG and CSA effects are considered. Therefore, to practically simulate the spectrum, some approximations must be employed to reduce the number of fitting parameters. Thus, the experimental C_Q, η_Q, and δ_iso values of oxygen sites that were resolved in the F1 dimension (e.g., O2, O3, O7, and O9 sites of as-made phase) were directly used in simulations without adjustment. For oxygen sites that are unresolved in the F1 dimension (e.g., O1, O5, and O11 sites of as-made phase), the P_Q and δ_iso are average values. However, it is evident from the lineshape of cross-sections that these sites give rise to very similar NMR parameters. For the 1D spectral simulations, the calculated η_Q values (Table S2) were used without further adjustment and the corresponding C_Q values were obtained from the known relationship between P_Q and C_Q/η_Q. The δ_iso values were obtained using the average δ_iso as a starting point and making slight adjustments. Keeping the C_Q and η_Q constant during the simulation is reasonable as the anisotropy from the EFG is much smaller compared to the CSA at the ultrahigh magnetic field. The small variation of the EFG parameters among different sites contributes very little to the intensity of the SSBs (Figure S8).

To include the CSA effects, additional NMR parameters must be incorporated: two CSA parameters including the reduced anisotropy Δ_CS and the chemical shift asymmetry parameter η_CS, along with three Euler angles (ϕ, χ, ψ) describing the orientations of the chemical shift tensor with respect to the EFG tensor principal axis frame (see Experimental).⁷³ It is reasonable to assume that the 6 μ¹-O sites have the same Δ_CS, since they reside in very similar chemical environments and the Δ_CS value for all 6 μ²-O sites have the same value yet are distinct from that of μ¹-O sites. The Δ_CS values for μ¹-O and μ²-O were modified during simulations, but the calculated η_CS values and Euler angles (Table S2) were kept constant.

The final simulated 1D MAS spectra are shown in Figure 3, illustrating that the isotropic regions and SSBs can be simulated reasonably well by considering both the quadrupolar and the CSA effects (see Table S3 for the final NMR parameters for simulation). If only the second-order quadrupolar interaction is considered, the isotropic regions of ¹⁷O signals can still be simulated accurately, but the simulated SSBs are far too low in intensity (Figure S8). It is also worth noting that there is a very weak ¹⁷O signal at δ₁ = 305 ppm (P_Q = 9.1 MHz, δ_iso = 298 ppm) in the 3QMAS spectrum. The signal position suggests that it arises from DMF molecules within MOF channels.⁸⁴ Apparently, DMF oxygen atoms are also subject to ¹⁷O exchange under the reaction conditions. The enhanced sensitivity and resolution of ¹⁷O SSNMR at 35.2 T permit detection of the DMF signal, thus it was included in the simulation of 1D MAS spectrum of as-made α-Mg₃(HCOO)₆ (i.e., the O1S site).

Figure 3.

Experimental and simulated ¹⁷O 1D MAS NMR spectra of ¹⁷O-enriched α-Mg₃(HCOO)₆ at 35.2 T. Both the quadrupolar and CSA effects are considered in these simulations, using the parameters shown in Table S3. In each phase, the signal intensity of each individual oxygen in the μ¹-O sites is approximately equal, and the same is true for signals arising from the μ²-O sites. The asterisk (*) denotes SSBs.

In general, the calculated δ_iso values are slightly overestimated compared to experimental values. Several factors can be responsible for these discrepancies, including (i) limitations of the GIPAW method, (ii) possible inaccuracy of the crystal structure, (iii) temperature effects on the crystal structure, and (iv) dynamics within the crystal structure.⁸⁵ In this case, the single crystal structures are based on the XRD data obtained at 100 K. In addition, our DFT calculations do not consider molecular motions such as the librational motions of formate anions reported for the related β-Ca(HCOO)₂ at room temperature,⁸⁶ while NMR experiments are subject to their influence.

¹⁷O 2D 3QMAS experiments at an ultrahigh magnetic field strength of 35.2 T has unlocked the identification of twelve inequivalent framework oxygen sites in both the as-made and activated phases of α-Mg₃(HCOO)₆. The experimental P_Q and δ_iso values of μ¹-O sites range from 7.5 to 8.8 MHz and 273 to 297 ppm, respectively; these ranges are 6.6 to 7.8 MHz and 220 to 236 ppm for μ²-O sites. Both sets of these ranges are typical for C═O and C─O environments of carboxylates.^{3, 8}

The very high resolution achieved at 35.2 T permits observation of small changes in ¹⁷O NMR parameters such as δ_iso at each oxygen site upon activation, which were not observable at a lower field of 21.1 T. Such changes reflect the influence and interaction of guest DMF solvent molecules with the MOF host. Activation of the as-made α-Mg₃(HCOO)₆ phase, and corresponding removal of DMF molecules from the pores, only results in subtle changes in local bond angles and distances while long-range order is preserved.⁶⁰ A comparison between the experimental ¹⁷O δ_iso values of as-made and activated α-Mg₃(HCOO)₆ phases reveals that more significant changes occur at μ¹-O sites versus μ²-O sites. This disparity in local structural changes is because each μ²-O site is firmly anchored to the framework by two Mg and one C atoms, thus the degree of perturbation on their local oxygen coordination spheres by guest molecules is not as evident as the coordination spheres of μ¹-O sites, which are only bound to the framework via one Mg and one C atom. The most significant activation-induced changes are associated with the δ_iso values of O4 and O10 (> 15 ppm). Figure 4 illustrates the orderly arrangement of DMF molecules along the zig-zag channels of α-Mg₃(HCOO)₆. Each DMF molecule interacts with two adjacent framework formate anions containing O4 and O10 sites. According to the criteria for the formation of O…H─C nonconventional hydrogen bonds (O…H distance < 2.72 Å and O…H─C angle > 130°),⁸⁷ both the O1S…H5 distance of 2.38 Å and the O1S…H5─C5 bond angle of 157° in as-made α-Mg₃(HCOO)₆ are in favor of weak hydrogen bonding. Thus, the significant change in the δ_iso value of O10, bound to C5 via a C─μ¹-O bond, is due to the O1S…H5─C5 hydrogen bonding interaction. In the case of O4, although both the O1S…H2 distance of 2.83 Å and the O1S…H2─C2 bond angle of 88° are not very favorable for O1S…H2─C2 hydrogen bond formation, the O1S…C2 distance of 3.00 Å is considerably shorter than the summation of their van der Waals radii (3.22 Å),⁸⁸ pointing toward van der Waals forces between O1S and C2 as responsible for the significant change in the δ_iso value of O4 upon activation.

Figure 4.

Schematic illustrations of DMF guest molecules within the zig-zag channels of as-made α-Mg₃(HCOO)₆. Only the Mg nodes are shown for clarity in the bottom diagram. The top inset shows a DMF molecule and two adjacent formate anions. The distances listed were extracted from the DFT-optimized structures. Color coding: Mg, turquoise; N, blue; C, grey; H, white; O1S of DMF, red; μ¹-O, orange; μ²-O, pink.

The formation of O1S…H5─C5 hydrogen bond is also evident from the δ_iso value of the DMF amide oxygen site (O1S) in as-made α-Mg₃(HCOO)₆. As demonstrated in the literature,⁸⁴ the δ_iso value of the DMF amide oxygen in the absence of hydrogen bonding is 323 ppm, and this value decreases with increasing hydrogen bonding strength. The δ_iso value of 298 ppm for O1S corresponds to a hydrogen bonding strength between those of infinitely diluted DMF in ethanol (299.3 ppm) and infinitely diluted DMF in methanol (292.5 ppm).⁸⁴ It is the high sensitivity and high resolution achieved at an ultrahigh field of 35.2 T that make it possible to detect the site-specific O…H─C nonconventional hydrogen bonding involving O10 and O1S and estimate the strength of this interaction. As mentioned earlier, nonconventional O…H─C hydrogen bonding can play important roles in the host–guest interactions in BioMOFs or MOF-based drug delivery systems.^49-51 Thus, the ability to detect this type of weak interaction and estimate its strength by ¹⁷O SSNMR at ultrahigh field, as demonstrated in this study, provides researchers a useful tool for investigating this type of host–guest interaction in MOF systems.

To further demonstrate the power of ultrahigh magnetic field on MOF characterization using ¹⁷O SSNMR, we examined two activated MOF MIL-53(Al) samples. MIL-53(Al) is a well-studied MOF with high thermal and chemical stability, and is very promising in guest separation.⁸⁹ As Figure 5a illustrates, the channels of as-made MIL-53(Al) are occupied by the unreacted linker precursor molecules, 1,4-benzenedicarboxylic acid (H₂BDC). After activation, the channel dimension increases from 7.3 × 7.7 Ų to 8.5 × 8.5 Ų due to the removal of the hydrogen bonding interaction between H₂BDC and the carboxylate of BDC²⁻ linkers in the framework. Upon activation, the framework topology is retained, but the crystal symmetry changes from Pnma in the as-made phase to Imma in the activated phase.⁴⁷ There are two reports on ¹⁷O SSNMR studies of MIL-53 conducted at lower magnetic fields.^16-17

Figure 5.

(a) ¹⁷O 1D MAS NMR spectra of ¹⁷O-enriched fully- (Sample A) and partially- (Sample B) activated MIL-53(Al) samples at 35.2 T. The blue and red solid lines denote experimental and simulated spectra, respectively. Only the carboxylate oxygen regions are shown for clarity. The full spectra are shown in Figure S9. The asterisk (*) and dollar sign ($) denotes the SSBs of −COO⁻ and μ²-OH, respectively. The structures of as-made and activated MIL-53(Al) phases are shown at top. (b) ¹⁷O 2D 3QMAS NMR spectrum of partially-activated (Sample B) ¹⁷O-enriched MIL-53(Al) at 35.2 T. Black dashed lines correspond to the slices examined for analyses.

Activation of MIL-53 is not very straightforward. Early on, activation of as-made MIL-53(Al) was performed by directly calcination in air at 503 K for 3 days.⁴⁷ The conditions were harsh and led to the reduced crystallinity. To avoid such harsh activation conditions, a milder alternate route was developed by first exchanging the trapped H₂BDC with DMF under solvothermal conditions (e.g., 423 K) for an extended period (e.g., 12 h or longer) and then heating the DMF-exchanged MOF at 473 K under dynamic vacuum (≤ 1 mbar) overnight.⁸⁹ We prepared two ¹⁷O-enriched MIL-53(Al) samples which were activated under identical conditions except the DMF exchange time (see Section S3): 24 h for Sample A and 12h for Sample B. According to literature, an exchange time of 12 h should be sufficient for the activation of MIL-53(Al).⁹⁰ Although the PXRD patterns of Sample A and B look very similar (Figure S5), the ¹⁷O 1D MAS spectra at 35.2 T of Samples A and B look distinctly different in the carboxylate region (Figure 5a). The spectrum of Sample A exhibits a relatively narrow pattern that can be well fitted with a single set of ¹⁷O NMR parameters (Table S4), indicating that the signal corresponds to a single oxygen site. This is consistent with the crystal structure of the activated phase which only has one carboxylate oxygen site in the unit cell.⁴⁷ Thus, the ¹⁷O MAS spectrum clearly indicates that the sample exchanged with DMF for 24 h is fully activated.

The spectrum of Sample B displays a rather broad profile in the carboxylate region that cannot be simulated by a single oxygen site, implying the existence of multiple carboxylate oxygen environments in this sample. It appears that this MOF sample is only partially-activated. The ¹⁷O 3QMAS spectrum of Sample B obtained at 35.2 T provides detailed information. A long spectral “ridge” is observed for the carboxylate oxygen sites, implying that the partial activation can contribute to the distributions on the quadrupolar coupling and chemical shift, consequently line broadening. Such a situation is not unexpected since MIL-53(Al) is a flexible MOF that undergoes a phase transition from Pnma in the as-made phase to Imma upon full activation without breaking any bonds. It appears that Sample B represents an intermediate state between these two phases. Furthermore, the unreacted linker precursors in the channels are disordered and seem to be randomly removed. Nonetheless, five peaks are extracted above the “ridge” at δ₁ = 212, 217, 226, 237, and 244 ppm, respectively. The corresponding P_Q and δ_iso values are shown in Table S4. Based on the δ_iso values, the peaks at δ₁ = 212 and 217 ppm are assigned to the ─OH and C═O of H₂BDC molecules within the MOF channels, respectively.³³ The signal at δ₁ = 237 ppm is attributed to the oxygen site in the empty channels because its δ_iso value (232 ppm) is similar to that of activated MIL-53(Al), while the resonances at δ₁ = 226 and 244 ppm are tentatively assigned to the oxygen sites in the occupied channels with local environments similar to those in as-made MIL-53(Al).^16-17 The high resolution and sensitivity of ¹⁷O NMR gained at 35.2 T not only allow for unambiguously distinguishing partially-activated from fully-activated MIL-53(Al) samples, but also permit observation of several oxygen sites in empty and occupied channels in partially-activated MIL-53(Al).

4. Conclusions

In summary, the very high spectral resolution and sensitivity achieved at an ultrahigh magnetic field of 35.2 T in this work represents an advance in ¹⁷O SSNMR spectroscopy. At 35.2 T, many inequivalent carboxylate oxygen sites have been identified in ¹⁷O SSNMR spectra of both the activated and as-made α-Mg₃(HCOO)₆ MOFs. The very high resolution achieved at 35.2 T enables the observation of subtle changes in ¹⁷O SSNMR spectra of as-made and activated α-Mg₃(HCOO)₆ phases. These alterations arise from weak site-specific interactions between DMF guests and the MOF framework, such as hydrogen bonding and van der Waals forces. The investigation of these weak interactions is important for MOF applications in various fields including gas adsorption and biomedical applications. The advantage of performing ¹⁷O SSNMR experiments at 35.2 T for MOF characterization is further illustrated by activation of MIL-53(Al). The partially- and fully-activated phases of MIL-53(Al) can be unambiguously distinguished. Several oxygen sites with different local environments in the partially-activated phase are tentatively identified.

This work illustrates how a wide variety of organic and inorganic compounds are now viable targets for ¹⁷O SSNMR at an ultrahigh magnetic field of 35.2 T. The sensitivity and resolution afforded at this field strength greatly extend the volume and quality of structural and chemical information available from ¹⁷O SSNMR spectroscopy, as much of this data is unavailable at or below magnetic fields of 21.1 T.