Solid-State ¹⁷O NMR of Paramagnetic Coordination Compounds

Abstract

We demonstrate that high-quality solid-state ¹⁷O (I = 5/2) NMR spectra can be successfully obtained for paramagnetic coordination compounds in which oxygen atoms are directly bonded to the paramagnetic metal centers. For complexes containing V(III) (S = 1), Cu(II) (S = 1/2), and Mn(III) (S = 2) metal centers, the ¹⁷O isotropic paramagnetic shifts were found to span a range of more than 10000 ppm. In several cases, high-resolution ¹⁷O NMR spectra were recorded under very fast magic-angle spinning (MAS) conditions at 21.1 T. Quantum chemical computations using density functional theory (DFT) qualitatively reproduced the experimental ¹⁷O hyperfine shift tensors.

NMR signals from paramagnetic substances are generally more difficult to detect than those from diamagnetic compounds. This is because the hyperfine interactions between magnetic dipoles of unpaired electrons and atomic nuclei are substantially stronger than the typical nuclear spin interactions such as magnetic shielding, nuclear quadrupolar and dipolar couplings, etc. As a result, the NMR signals from paramagnetic compounds are significantly shifted and broadened compared with diamagnetic compounds. Despite experimental difficulties, solid-state NMR studies of paramagnetic compounds can be traced back to the early NMR studies on single crystals of CuSO₄·5H₂O,^[1] CuCl₂·H₂O,^[2] and MnF₂.^[3] In more recent years, there have been considerable interests in solution^[4] and solid-state^[5-16] NMR studies of organic and biological systems containing paramagnetic metal ions. To date most NMR studies of paramagnetic compounds have relied on detection of ¹H and ¹³C nuclei, because hydrogen and carbon atoms are generally remote from the paramagnetic metal centers, thus experiencing relatively weak hyperfine interactions. In contrast, as oxygen atoms are often directly bonded to the paramagnetic metal centers, ¹⁷O NMR for paramagnetic coordination compounds is expected to be more challenging than the corresponding ¹H and ¹³C NMR studies. In addition, the only NMR-active oxygen isotope, ¹⁷O, has very low natural abundance (0.037%) and its nuclear spin is quadrupolar (I = 5/2). In light of the recent advances in solid-state ¹⁷O NMR studies of diamagnetic molecules including biological macromolecules,^[17] we decided to explore the possibility of using solid-state ¹⁷O NMR to study paramagnetic coordination compounds. We should note that solid-state ¹⁷O NMR has been used previously to study ionic high T_c superconductors^[18] and simple paramagnetic inorganic complexes.^[19] The focus of our study is on solid-state ¹⁷O NMR of paramagnetic coordination complexes containing organic ligands (i.e., containing H, C, N, and O atoms).

We first examined two vanadium(III) (S = 1) complexes: V([¹⁷O₂]acac)₃ and K₃V([¹⁷O₄]oxalate)₃·3H₂O. Synthetic details for the preparation of ¹⁷O-labeled paramagnetic coordination compounds are given in the Supporting Information. As seen in Figure 1, the static solid-state ¹⁷O NMR spectrum for V([¹⁷O₂]acac)₃ exhibits signals centered at around −1300 ppm and spanning over 1800 ppm. These spectral features are drastically different from those obtained for the diamagnetic analog, Al([¹⁷O₂]acac)₃, which has its ¹⁷O NMR signal at 270 ppm with a spectral span of only 400 ppm at 21.1 T.^[20] For V([¹⁷O₂]acac)₃, we were able to obtain a very fast MAS ¹⁷O NMR spectrum in which all six crystallographically non-equivalent oxygen sites^[21] are resolved. The ¹⁷O isotropic shifts among the six directly bonded oxygens differ by more than 500 ppm. In comparison, the isotropic ¹⁷O chemical shifts for the six oxygen atoms in the diamagnetic Al([¹⁷O₂]acac)₃ differ by only 5 ppm.^[20] K₃V([¹⁷O₄]oxalate)₃·3H₂O is an interesting compound in which both direct chelating (O1, O2, O3) and non-bonding (O4, O5, O6) oxygen atoms are present.^[22] As seen from Figure 1, the ¹⁷O NMR signals from O1-O3 are found around −1200 ppm whereas those from O4-O6 appear at 350 ppm. In diamagnetic metal oxalates, the ¹⁷O NMR signals from chelating and non-bonding oxygen atoms typically differ by 40-70 ppm.^[23]

Figure 1.

Molecular structures (a, d, where hydrogen atoms are omitted for clarity), experimental and simulated static (b, e) and MAS (c, f) ¹⁷O NMR spectra (21.1 T) of V([¹⁷O₂]acac)₃ and K₃V([¹⁷O₄]oxalate)₃⋅3H₂O. Simulated sub-spectra for individual sites are also shown. The sample spinning frequency was 62.5 and 55.0 kHz in c) and f), respectively. The signal labeling within either direct chelating or non-bonding oxygens is arbitrary.

In order to properly analyze the experimental solid-state ¹⁷O NMR spectra, a brief background theory is warranted. In paramagnetic compounds, the hyperfine interaction between nuclear and unpaired electron spins can be written as:

$$H_{\mathit{\text{hyperfine}}}=I·A·S$$ (1)

where I and S are nuclear and electron spins, respectively, and A is the hyperfine interaction tensor. In general, the A tensor can be separated into the isotropic hyperfine coupling constant known as the Fermi contact shift, A_iso, and the anisotropic (traceless) “dipolar tensor”, T, i.e.,

$$\mathit{A}={\mathrm{A}}_{\text{iso}}+\mathit{T}$$ (2)

The A_iso and T can be further expressed as:

$$A_{\mathit{\text{iso}}}=\frac{4\pi }{3S}\hslash {\gamma }_{N}g{\mu }_B{\rho }_{\alpha \beta }$$ (3)

$$T_{ij}=-\frac{1}{2S}\hslash {\gamma }_{N}g{\mu }_B\langle \frac{r^2{\delta }_{ij}-3r_i{r}_j}{r^5}\rangle$$ (4)

where γ_N is the nuclear magnetogyric ratio, g is the free electron g-value, μ_B is the Bohr magneton, and ρ_αβ is the electron spin density at the nucleus.

It is also known that, under the condition of very rapid electron relaxation, the hyperfine interaction manifests itself in NMR spectra of a powder sample as a characteristic line shape identical to that due to the chemical shift anisotropy (CSA).^[6] As a result, we can define a paramagnetic shift tensor containing both the orbital (from all paired electrons) and hyperfine (from unpaired electrons) contributions:

$${\delta }_{ii}={\delta }_{ii}^{\mathit{\text{orb}}}+{\delta }_{ii}^{hf}$$ (5)

where

$${\delta }_{ii}^{hf}=\left(\frac{A_{ii}}{\hslash }\right)\frac{g{\mu }_{B}S(S+1)}{3{\gamma }_{N}kT}=\frac{m(S+1)}{T}({\rho }_{\alpha \beta }+\frac{8\pi }{3}{\rho }_{\alpha \beta }^{ii})$$ (6)

In eq. (6), m = μ₀(gμ_B)²/9k = 2.35 × 10⁷ ppm K au^-1 and ${\rho }_{\alpha \beta }^{ii}$ is the anisotropic spin dipolar tensor components in atomic units. For the paramagnetic shift tensor, we used the same convention for tensor components as that for the CS tensor for diamagnetic compounds, i.e., δ₁₁ > δ₂₂ > δ₃₃.

In addition to the magnetic shielding and hyperfine interactions, ¹⁷O nuclei in paramagnetic coordination complexes should also experience the same spin interactions as encountered in diamagnetic compounds (e.g., quadrupolar and dipolar couplings). Therefore, the ¹⁷O NMR spectrum from a paramagnetic compound is generally determined by the interplay of all the tensor interactions involved. A general analysis of the line shape for a powder sample can be quite complicated. Fortunately, at high magnetic fields (e.g., 21.1 T), the spectral contribution from the ¹⁷O paramagnetic shift tensor is much larger than those from other spin interactions. In this study, we were able to analyze the solid-state ¹⁷O NMR spectra by considering only the paramagnetic shift tensor (eq. 5) and quadrupolar coupling tensor. Indeed, as seen from Figure 1, the ¹⁷O NMR spectrum for each direct bonding oxygen site resembles the so-called CSA powder pattern.^[6] For each compound, we typically analyzed ¹⁷O NMR spectra obtained at two magnetic fields; see Figure S1.

Next we examined two copper(II) (S = 1/2) complexes: Cu([¹⁷O₂]-DL-alanine)₂·H₂O and K₂Cu([¹⁷O₄]oxalate)₂·2H₂O. As seen from Figure 2, the ¹⁷O NMR spectra of Cu([¹⁷O₂]-DL-alanine)₂·H₂O are surprisingly simple, in which only the signal from the non-bonding oxygen, O2, can be detected; see Figure S2 for the full spectrum. The O2 signal appears at 655 ppm, which is significantly shifted from the typical chemical shift, 250 ppm, observed for diamagnetic analogs. Similarly, the chelating oxygen atoms of the square-planar Cu(II) complex K₂Cu([¹⁷O₄]oxalate)₂·2H₂O were also invisible in the spectra. However, one of the oxalate oxygen atoms, O1, which is directly coordinated to the Cu(II) centre as an axial ligand, exhibits an ¹⁷O NMR signal at 240 ppm, not so different from those observed for diamagnetic metal oxalates.^[23]

Figure 2.

Molecular structures (a, d, where hydrogen atoms are omitted for clarity), experimental and simulated static (b,e) and MAS (c,f) ¹⁷O NMR spectra (21.1 T) of Cu([¹⁷O₂]-DL-alanine)₂·H₂O and K₂Cu([¹⁷O₄]oxalate)₂·2H₂O. Simulated sub-spectra for individual sites are also shown in e) and f). The sample spinning frequency was 45.0 kHz in c) and f). Note that the signal integration for O3,O7,O8 is nearly three times of that for O1.

Here it is necessary to address the issue of observability of NMR signals in paramagnetic compounds. As Abragam explained in detail,^[24] NMR signals may be observable only for nuclei for which the condition of 2πAτ_e ≪ 1 holds, where A is the hyperfine coupling constant and τ_e is the averaged lifetime of electron spin in a given quantized state (generally equivalent to the electron spin-lattice relaxation time). This is the so-called fast relaxation/exchange limit. For mononuclear Cu(II) complexes, τ_e is on the order of 10⁻⁹ s.^[25] Typical A values for chelating oxygen atoms in square-planar Cu(II) complexes are ca. 50 MHz.^[26] Thus, the fast exchange condition usually does not hold for a direct bonding oxygen in Cu(II) complexes, making its signal too broad to be detected. Then why was the ¹⁷O NMR signal for the direct chelating oxygen at the axial position in K₂Cu([¹⁷O₄]oxalate)₂·2H₂O observed? This is because, in square-planar Cu(II) complexes, the A value for the axial oxygen is distinctly small, ca. < 0.2 MHz,^[26] thus satisfying the fast exchange condition.

The above success in detecting ¹⁷O NMR signals in small Cu(II) complexes led us to test whether the same approach can be extended to studies of solid materials. To this end, we chose to examine [Cu₃(BTC)₂(H₂O)₃]_n (BTC= benzenetricarboxylate) (also referred to as HKUST-1),^[27] which is perhaps one of the most studied metalorganic frameworks. In our case, all carboxylate oxygens in BTC are ¹⁷O-labeled. As shown in Figure 3, the core structural motif is a di-nuclear tetracarboxylate linker where the two Cu(II) ions, separated by 2.628 Å,^[27] are antiferromagnetically coupled. As a result, although the ground state of the system has S = 0, thus being diamagnetic, there exists a low-lying excited state with S = 1. At 300 K, HKUST-1 exhibits considerable paramagnetism.^[28] The ¹⁷O NMR spectrum of HKUST-1 exhibits an isotropic paramagnetic shift of 3530 ± 100 ppm with the span being approximately 3100 ppm. This is the first time that ¹⁷O NMR signals are detected in antiferromagnetically coupled di-nuclear coordination complexes. This observation also suggests that the electron relaxation time in HKUST-1 must be much shorter than the typical value of 10^-9 s, which is consistent with the previous observation of Bertini and co-workers^[29] for di-nuclear Cu(II) compounds in solution.

Figure 3.

a) The two antiferromagnetically coupled Cu(II) ions form the structural building block of the 3D framework in HKUST-1. Hydrogen atoms are omitted for clarity. b) Experimental and simulated static ¹⁷O NMR spectra (14.1 T) of HKUST-1.

Finally we examined Mn(III)([¹⁷O₂]acac)₃ (S = 2). According to eq. 6, this S = 2 system exhibits an exceedingly large paramagnetic shift, ca. 7500 ± 500 ppm, as well as a very large paramagnetic shift anisotropy, ca. 8000 ppm. As seen from Figure S3, since the entire ¹⁷O NMR spectrum for the central transition spans close to 1 MHz, sufficient excitation cannot be achieved with a single RF pulse. Consequently, we acquired a variable offset cumulative spectrum (VOCS)^[30] for this compound.

After obtaining high-quality solid-state ¹⁷O NMR data for paramagnetic coordination compounds, we decided to test whether current computational methodologies can yield reliable results for ¹⁷O hyperfine interactions. Several recent studies have shown that hyperfine shifts can be accurately computed for atoms that are not directly bonded to the paramagnetic metal center.^[12,31] However, calculating ¹⁷O hyperfine interactions for direct bonding oxygens may represent a greater challenge. After an extensive examination of various computational methods (see Supporting Information), we found that the LC-wPBE approach produces the best computational results. As seen from Figure 4, the agreement is reasonable. These results suggest that the computation method can be used as a complementary tool to aid the interpretation of experimental data. For example, the computational results made it possible to assign the ¹⁷O NMR signals between direct chelating and non-bonding oxygens. It is also interesting to note that both positive and negative ¹⁷O paramagnetic shifts were observed for the direct chelating oxygens. As Pritchard and Autschbach explained,^[31g] the sign of the paramagnetic shift is related to the α/β spin balance in oxygen-to-metal σ donation. In Cu(II) d^[9] and Mn(III) d^[4] high-spin complexes, as the α-spin orbitals of the σ symmetry on the metal are occupied, only the β-spin orbital from the oxygen lone pair can contribute to the oxygen-to-metal σ bond, resulting in partial depletion of β-spin density (or α-spin surplus) at the oxygen nucleus. This leads to a positive ¹⁷O paramagnetic shift. In V(III) d² complexes, because the metal σ acceptor orbitals are empty, both α- and β-spin orbitals can accept the oxygen lone pair. However, the contribution from the metal α-spin orbital is slightly larger, which produces an excess of β-spin density at the oxygen nuclear site thus a negative ¹⁷O paramagnetic shift. In this regard, our DFT calculations can in fact be used to provide easy visualization of the electron spin density distribution across the entire molecule (see Figure S5). Finally, while the general agreement between observed and calculated ¹⁷O paramagnetic shift tensors is reasonable over the 10000 ppm range, considerable discrepancies do exist for the direct chelating oxygens, which have prevented us from making unambiguous signal assignment among structurally similar oxygen sites (e.g., O1-O6 in V(acac)₃ and O1-O3 for K₃V(oxalate)₃·3H₂O). Possible sources for such discrepancies may include (a) the neglect of other contributions to the paramagnetic shift tensors (e.g., pseudocontact shift, bulk magnetic susceptibility, magnetic coupling among the paramagnetic metal centers, and zero-field splitting for systems with S > 1/2), (b) inaccurate crystal structures, (c) crystal packing effect, (d) limitation of the current DFT methodologies, and (e) large uncertainties in experimental data (e.g., in Mn(acac)₃). Further investigations are clearly needed to address these issues.

Figure 4.

Comparison between experimental and calculated a) isotropic ¹⁷O paramagnetic shifts and b) anisotropic paramagnetic shift tensor components.

In summary, we have shown that high-quality solid-state ¹⁷O NMR spectra for various paramagnetic coordination compounds can be obtained and experimental ¹⁷O paramagnetic shift tensors were qualitatively reproduced by DFT computations. These results will inspire future interests in using solid-state ¹⁷O NMR spectroscopy to study paramagnetic substances such as metalloproteins and materials.