Skip to Main Content
Skip Nav Destination

This chapter gives a general introduction to NMR interactions in solids. More details on paramagnetic interactions of NMR, often dominating the spectra of paramagnetic solids, are then presented. A brief introduction to the first-principles calculations for NMR spectra of paramagnetic solids is also provided, which play an important role in spectral assignments.

In 1945, American physicists Felix Bloch1  and Edward Mills Purcell2  discovered the phenomenon of nuclear magnetic resonance (NMR) independently, which can be observed when nuclear spins are placed in a magnetic field. With the development of pulsed Fourier transform NMR and multidimensional NMR spectroscopy by Richard Robert Ernst,3,4  NMR has gradually become one of the most powerful analytical methods that provides detailed structural information of matter at atomic/molecular scale, as well as physical/chemical/electronic properties. The major reason behind this fact is that there are many different types of interaction involving the nuclear spins in the sample, which can be exploited to provide such information. The NMR interactions can be treated with quantum mechanical approaches and readers are encouraged to find more details in some valuable books and review papers.5–13  In this section, only a brief introduction on these NMR interactions will be presented.

A nucleus possesses a nuclear spin angular momentum I, leading to a magnetic dipole moment µI,

µI = ℏγII
Equation 1.1

where is Planck's constant divided by 2π and γI is the nuclear gyromagnetic ratio, a constant depending on the nuclear species. The value of γI can be positive or negative, meaning that the nuclear magnetic moment µI is parallel or antiparallel to the spin I, respectively (see Figure 1.1).

Figure 1.1

The relationship of nuclear (electronic) spin and the magnetic moment. (a) For a nucleus with a positive gyromagnetic ratio, the positive sign of γI leads to parallel alignment of spin and the magnetic moment. (b) For a nucleus with a negative gyromagnetic ratio or electron, the negative sign of γI (or γS) leads to antiparallel alignment of spin and the magnetic moment.

Figure 1.1

The relationship of nuclear (electronic) spin and the magnetic moment. (a) For a nucleus with a positive gyromagnetic ratio, the positive sign of γI leads to parallel alignment of spin and the magnetic moment. (b) For a nucleus with a negative gyromagnetic ratio or electron, the negative sign of γI (or γS) leads to antiparallel alignment of spin and the magnetic moment.

Close modal

For paramagnetic systems, where there are unpaired electrons, it is also important to describe the electronic magnetic moment, since these electrons will interact with the nuclear species of interest and have an effect on the NMR spectra (shift, linewidth, relaxation, etc.). Similar to the case for the nucleus, the electron magnetic moment µS is related to the electronic spin S by

µS = ℏγSS
Equation 1.2

where γS is the electron gyromagnetic ratio, which is negative, indicating the electron magnetic moment µS is always antiparallel to the electronic spin S (see Figure 1.1b).

In electron paramagnetic resonance (EPR), the relationship between the electronic magnetic moment µS and electron spin S is also expressed as

µS = −µBgeS
Equation 1.3

where ge and µB are the g-value of the free electron (approx. 2.002319) and Bohr magneton, respectively. Therefore,

ℏγS=µBge
Equation 1.4

A magnetic dipole moment µ interacts with a magnetic field B, and the energy of this interaction, E, is equal to their scalar product, therefore,

E = −µ·B
Equation 1.5

This equation clearly shows that the interaction energy is dependent on the relative orientation of the magnetic moment µ with respect to the magnetic field B. The corresponding expression for the Hamiltonian is

= −·B
Equation 1.6

where is the magnetic moment operator.

In NMR spectroscopy, the sample is placed in an external magnetic field B0, thus the nuclear magnetic moment interacts with B0, leading to Zeeman interaction. Its Hamiltonian Z is

Z = −I·B0 = −ℏγI·B0
Equation 1.7

where is the nuclear spin operator.

The external magnetic field B0 in NMR is usually set along the z-direction of the laboratory frame, i.e., B0 = (0, 0, B0). Therefore eqn (1.7) becomes

Z = −ℏγIB0z = ℏω0z
Equation 1.8

where z is the operator for the component of the spin along the z axis.

A single nuclear spin I has (2I + 1) states according its magnetic quantum number, mI, where mI = −I, (−I + 1),…, (I − 1), I. These states, which can be expressed as |mI〉, are the eigenstates of Zeeman Hamiltonian, and the eigenvalues of Z, or the possible energy levels, are

formula
Equation 1.9

In the simplest case, I = 1/2 (e.g., 1H, 13C or 19F), and mI = −1/2 or +1/2. At zero external magnetic field, the two states (e.g., |−1/2〉 and |+1/2〉) are degenerate. However, with a non-zero magnetic field, the energy levels of the two states split (see Figure 1.2). The separation between the two energy levels is γIℏB0, which increases linearly with the external magnetic field. Classically, transitions between |−1/2〉 and |+1/2〉 (or in general, ΔmI = ±1), can occur and an NMR signal can be observed by applying an electromagnetic radiation with a frequency that matches the energy difference. This frequency, ω0, is Larmor or resonance frequency (rad−1), which can be derived from

ω0 = −γIℏB0/ = −γIB0
Equation 1.10

Apparently, ν0, Larmor or resonance frequency in Hz, is

v0 = ω0/2π
Equation 1.11
Figure 1.2

Schematic representations of Zeeman interaction for a spin-1/2 nucleus (positive γI), spin populations and net magnetization.

Figure 1.2

Schematic representations of Zeeman interaction for a spin-1/2 nucleus (positive γI), spin populations and net magnetization.

Close modal

In a real sample containing an ensemble of nuclear spins, it can be expected that the states with a higher energy is less populated. The populations of the two states follow the Boltzmann distribution at equilibrium, which is given by

formula
Equation 1.12

where N is the total number of nuclear spins, N±1/2 are the number of spins at corresponding states, T is the absolute temperature and kB is the Boltzmann's constant. Under high temperature approximation, eqn (1.12) becomes

formula
Equation 1.13

The population difference, N+1/2N−1/2 is therefore

formula
Equation 1.14

The magnetization of the sample, M, which is the net nuclear magnetic moment contributed from each individual spin, and the NMR signal can be further analyzed. The equilibrium magnetization M0 and the NMR signal SNMR, are given by9 

formula
Equation 1.15
formula
Equation 1.16

where Vc is the sample volume. Eqn (1.14)–(1.16) also provide some key insights into the characteristics of NMR spectroscopy. Since is small at common external magnetic field and temperature, the sensitivity of NMR is unfortunately low and it is often required to have a relatively large number of spins (N). In order to obtain stronger NMR signals, the nucleus with a larger gyromagnetic ratio (γI) can be chosen for investigation and usually a higher external magnetic field (B0) is preferred.

Similarly, there is Zeeman interaction between an electron spin S and an external magnetic field B0.

In addition to the external magnetic field, the total energy of a nucleus is also affected by different internal spin interactions between the nucleus and its local environment. Therefore, if the relaxation part is not considered, the total Hamiltonian of nuclear spin, Total, can be expressed as

Total = Ext + Int
Equation 1.17

where Ext is the Hamiltonian for external interactions including Zeeman interactions between the spin and external magnetic fields (static and vertical field B0 and radio-frequency field B1), while Int is the Hamiltonian for internal interactions. The latter can be expressed by

Int = CS + D + J + Q + HFC + …
Equation 1.18

where CS, D, J, Q and HFC represent chemical shielding interactions, direct dipolar interactions, indirect spin–spin interaction, quadrupolar interaction (for quadrupolar nucleus only, I > 1/2) and hyperfine interaction (for paramagnetic samples), respectively.

It has been shown that all the NMR interactions can be described by a Hamiltonian in the following form:

λ = Kλ·Rλ·λ
Equation 1.19

where Kλ, Rλand λ are the constant, the second-rank tensor related to spatial orientation, and the coupling partner of spin vector , which can be another spin vector of the same or different spin, or a magnetic field, for a specific interaction λ, respectively. It clearly shows that internal NMR interactions are anisotropic, meaning that they depend on the relative orientation of the spin vector and its coupling partner. In solution NMR, fast molecular tumbling averages the anisotropic interactions and sharp resonances can be observed. However, in solid-state NMR spectroscopy, these anisotropic interactions lead to broad peaks due to lack of rapid molecular motion. In order to obtain high-resolution NMR data and extract detailed structural information of solid materials, such interactions must be carefully treated with rational design and application of a variety of solid-state NMR techniques.

The electrons surrounding the nucleus circulate around due to the presence of an external magnetic field. This movement produces an induced magnetic field, which modifies the effective field experienced by the nucleus. In a diamagnetic material, it is often the case that the induced field is in the opposite direction to the external field, making the nucleus experience a smaller magnetic field. Therefore, a “shielding” effect arises from the interaction between the nuclear spin and the field induced by the surrounding electrons. Since it reflects the local electronic environments, this interaction is called the “chemical shielding interaction” and its Hamiltonian is given by

CS = ℏγI·σ·B0 = ℏγI·Blocal
Equation 1.20

where σ is the chemical shielding tensor,

formula
Equation 1.21

a second-rank Cartesian tensor with a 3 × 3 matrix defining its relative orientation with respect to the laboratory frame, while Blocal is the local magnetic field induced. Therefore, considering the external magnetic field B0 used in NMR (along the z-direction) and the shielding effect Blocal, the actual magnetic B the nucleus experienced is

formula
Equation 1.22

where 1 is the unit matrix.

In common NMR experiments, strong external magnetic field B0 is applied in the z-direction and Zeeman interaction can be considered dominating. Eqn (1.22) shows that σzz is the only component aligned with B0, therefore only this term gives major contributions and now the Hamiltonian of the chemical shielding interaction becomes

CS = γIσzzB0z
Equation 1.23

and resonance frequency of the nucleus changes accordingly.

The chemical shielding tensor σ can be transformed to a diagonalized tensor σPAS in its principal axis system (PAS), in which σ11, σ22, σ33 (σ33σ22σ11) are the eigenvalues of the tensor, appearing on the diagonal of the matrix and mostly determining the NMR spectra.

formula
Equation 1.24

The value of σzz is related to the three eigenvalues by

σzz = (sin2θ cos2ϕ)σ11 + (sin2θ sin2ϕ)σ22 + (cos2θ)σ33
Equation 1.25

where θ and ϕ are the polar angles relating the orientation of the laboratory frame with PAS (see Figure 1.3). In an isotropic, liquid sample, σzz can be replaced by isotropic chemical shielding σiso, given by

formula
Equation 1.26

owing to rapid molecular tumbling, which averages out the anisotropic effects of chemical shielding interactions. In solids, however, lack of this type of motion leads to anisotropic effect which is related to the relative orientation of the tensor of the nucleus.

Figure 1.3

(a) A diagonalized chemical shielding (electric field gradient) tensor (black) oriented in the laboratory frame which can be represented by an ellipsoid (grey). (b) The orientation of the chemical shielding (electric field gradient) tensor with respect to the external magnetic field B0, with θ and ϕ represent polar angles.

Figure 1.3

(a) A diagonalized chemical shielding (electric field gradient) tensor (black) oriented in the laboratory frame which can be represented by an ellipsoid (grey). (b) The orientation of the chemical shielding (electric field gradient) tensor with respect to the external magnetic field B0, with θ and ϕ represent polar angles.

Close modal

In NMR experiments, it is conventional to use chemical shift to describe the local environment of a nucleus. Chemical shift and chemical shielding are related by

formula
Equation 1.27

where δ, ν and σ are the chemical shift, resonant frequency and chemical shielding (δreference = 0). Therefore, a diagonalized chemical shift tensor with eigenvalues of δ11, δ22 and δ33 (δ11δ22δ33) can be obtained according to the chemical shielding tensor. The isotropic average of this tensor is thus

formula
Equation 1.28

The parameter's anisotropy, span (Ω), and asymmetry parameter, skew (κ), can be defined as

Ω = δ11δ33
Equation 1.29
formula
Equation 1.30

respectively, to describe the chemical shift anisotropy (CSA) of the chemical shielding interactions.

In a solid sample, all the possible orientations of the crystal (or the possible values of θ and ϕ) contribute to the observed NMR signal, leading to the powder pattern (see Figure 1.4a), which can be used to extract the values of δ11, δ22 and δ33, as well as δiso, Ω and κ. δiso is often used as a “finger print” for the examination of the environment of the nucleus, while CSA parameters Ω and κ can also provide key information on the asymmetry of the local structure the nucleus. However, it is difficult to obtain δiso when different sites are present and their resonances overlap to produce featureless spectra. Since there is a common term (3cos2θ − 1) describing the dependence of the interactions (chemical shielding, direct dipolar and first-order quadrupolar) on the polar angle (vide infra), it has been shown that rotating the sample at a “magic angle” of 54.736° with respect to the external magnetic field can significantly reduce the effects of anisotropic interactions on the spectrum.14  This commonly used method to acquire high-resolution data in solid-state NMR is called magic angle spinning (MAS), which can be considered as introducing fast motion to the molecules in the solid sample. With a slow MAS rate, the spinning sideband manifolds resemble the powder pattern. With increasing MAS rate, anisotropic interactions are reduced to a larger extent, leading to simpler spectrum (see Figure 1.4b).

Figure 1.4

(a) Powder pattern for spin 1/2 nuclei with axially asymmetric (δ11δ22δ33) and axially symmetric (δ11 = δ22δ33). δiso = (1/3)(δ11 + δ22 + δ33). Reproduced from ref. 12 with permission from Elsevier, Copyright 2010. (b) The removal of shift anisotropy by magic angle spinning 111Cd NMR spectra of polycrystalline cadmium metal, with static condition and rotating about the magic axis at 2.1, 2.6 and 3.6 kHz. Reproduced from ref. 15 with permission from Elsevier, Copyright 1974.

Figure 1.4

(a) Powder pattern for spin 1/2 nuclei with axially asymmetric (δ11δ22δ33) and axially symmetric (δ11 = δ22δ33). δiso = (1/3)(δ11 + δ22 + δ33). Reproduced from ref. 12 with permission from Elsevier, Copyright 2010. (b) The removal of shift anisotropy by magic angle spinning 111Cd NMR spectra of polycrystalline cadmium metal, with static condition and rotating about the magic axis at 2.1, 2.6 and 3.6 kHz. Reproduced from ref. 15 with permission from Elsevier, Copyright 1974.

Close modal

The magnetic dipole moment of spin I1 and I2 interact with each other directly. It can be considered as the interaction between the magnetic moment of a nuclear spin and the local magnetic field generated by a spin nearby. The Hamiltonian of this direct dipolar interaction (direct dipole–dipole interaction) is given by

D = 1·D·2 = ℏb12(3(1·)(2·)/r21·2)
Equation 1.31

where D is the dipolar coupling tensor, a symmetric and traceless tensor in Cartesian coordinates, while is the dipolar vector between spins I1 and I2, r is the internuclear distance and b12 is the dipolar coupling constant, which is defined as

formula
Equation 1.32

where γ1 and γ2 are the gyromagnetic ratio of spins I1 and I2, respectively and µ0 is the permeability constant (4π × 10−7 kg m s−2 A−2).

In common NMR experiments, the high-field approximation can be applied and the Hamiltonian can be simplified as

formula
Equation 1.33

where θ is the polar angle related to the orientation of the dipolar vector r and the external magnetic field (see Figure 1.5a), while I = I1x ± iI1y and I = I2x ± iI2y.

Figure 1.5

(a) Schematic representation of a pair of spins (I1 and I2) in a vertical magnetic field B0 with polar angles θ and ϕ describing the relative orientation of the vector r with respect to the laboratory frame. (b) Pake doublet. The broken line in the upper curve shows the calculated distribution of component line centres for the proton resonance in powdered CaSO4·2H2O. The continuous line on the same plot is the calculated line shape obtained by superposing Gaussians of width 1.54 gauss according to this distribution function. Reproduced from ref. 16 with permission from AIP Publishing, Copyright 1948.

Figure 1.5

(a) Schematic representation of a pair of spins (I1 and I2) in a vertical magnetic field B0 with polar angles θ and ϕ describing the relative orientation of the vector r with respect to the laboratory frame. (b) Pake doublet. The broken line in the upper curve shows the calculated distribution of component line centres for the proton resonance in powdered CaSO4·2H2O. The continuous line on the same plot is the calculated line shape obtained by superposing Gaussians of width 1.54 gauss according to this distribution function. Reproduced from ref. 16 with permission from AIP Publishing, Copyright 1948.

Close modal

The term I1zI2z represent the secular dipolar coupling, which can be treated as perturbations to Zeeman interaction, leading to broad resonances in NMR. (I1+I2− + I1−I2+) is the “flip-flop” term, which is associated transitions between different states in this spin pair. In order to have a significant probability for the “flip-flop” process, it must be energy conserving, which requires that the spins I1 and I2 are the same nucleus (i.e., homonuclear dipolar coupling). For heteronuclear spin pairs, the “flip-flop” term vanishes and eqn (1.33) can be further simplified as

D = ℏb(1 − 3cos2θ)I1zI2z
Equation 1.34

Eqn (1.31)–(1.34) clearly show that the direct dipolar interactions are angular dependent. In solution NMR, fast random motion of molecules removes direct dipolar interactions. In solid state NMR, however, the fixed dipolar vectors in the solid are randomly oriented with respect to the external magnetic field, generating broad resonances (Pake doublet, see Figure 1.5b)16  arising from the same species. Therefore, techniques such as MAS should be used to obtain high resolution spectra.

The direct dipolar interaction, however, depends strongly on the internuclear distance r as shown in eqn (1.31). Therefore, investigations of direct dipolar coupling provide key structural information such as the distance between two atoms or through space connectivity. It is common practice to reintroduce direct dipolar coupling under MAS in order to extract structural information at high spectral resolution. Details of such techniques can be found elsewhere.8–10 

In addition to direct dipolar coupling (through-space), the magnetic moment of spins can also be coupled via the electrons in the chemical bond connecting nuclei. This type of through-bond interaction is indirect spin–spin interaction, also known as J-coupling or scalar coupling. The Hamiltonian for this interaction is

J = 1J·2
Equation 1.35

where J is the tensor that reflects the orientation dependence of the J-coupling with respect to the external magnetic field. Unlike the direct dipolar interaction, the trace of the tensor J is not zero. Therefore, the isotropic J-coupling (scalar coupling constant) is given by

formula
Equation 1.36

and in the heteronuclear case, the Hamiltonian of J-coupling is

J = ℏJiso1z2z
Equation 1.37

Because the J-coupling is not averaged to zero by fast tumbling of molecules, it may have an observable effect in both solution and solid-state NMR spectra. The J-coupling between two spins leads to the splitting of the peak in the spectrum to generate lines evenly spaced (see Figure 1.6a). This “fine structure” shown in the spectrum can provide important information in chemical bonding. For example, small doublets with a separation of 550 Hz (corresponding to J-coupling of 1H–31P) can be observed in non-decoupled 31P MAS NMR spectrum of zeolite HY adsorbed with (CH3)3P, indicating the formation of chemisorbed (CH3)3P–H+, while the doublets disappear in decoupled spectrum (Figure 1.6b).17  In many cases, however, J-coupling is only tens of Hz and much weaker than other NMR interactions in solid, and is not observed.

Figure 1.6

Doublet with 1 : 1 in intensity (a), triplet with 1 : 2:1 intensity (b) and quartet with 1 : 3:3 : 1 intensity (c) can be observed for an isolated spin I coupled to 1, 2 and 3 spin-1/2 nuclei, respectively. (d) High-power, proton-decoupled (top) and proton coupled (below) 31P MAS NMR spectrum of trimethylphosphine adsorbed on H–Y zeolite. Reproduced from ref. 17 with permission from American Chemical Society, Copyright 1984.

Figure 1.6

Doublet with 1 : 1 in intensity (a), triplet with 1 : 2:1 intensity (b) and quartet with 1 : 3:3 : 1 intensity (c) can be observed for an isolated spin I coupled to 1, 2 and 3 spin-1/2 nuclei, respectively. (d) High-power, proton-decoupled (top) and proton coupled (below) 31P MAS NMR spectrum of trimethylphosphine adsorbed on H–Y zeolite. Reproduced from ref. 17 with permission from American Chemical Society, Copyright 1984.

Close modal

The nucleus with a spin quantum number I > 1/2 (70% in the periodic table) is quadrupolar, indicating that the electric charge distribution of the nucleus is not spherical, leading to a non-zero nuclear electric quadrupole moment (eQ). This quadrupole moment further interacts with the electric field gradient (EFG) at the nucleus, which arises from the distribution of surrounding nuclei and electrons, resulting in quadrupolar interaction. The Hamiltonian for this interaction is given by

formula
Equation 1.38

where V is the EFG tensor, a symmetric traceless Cartesian tensor in its PAS, described by only three components in the diagonal, VXX, VYY and VZZ (Figure 1.1.3a). The magnitude these components are ordered as |VZZ| > |VYY| > |VXX|, while VXX + VYY + VZZ = 0. The size of EFG is defined as

eq = VZZ
Equation 1.39

while the shape of the tensor can be described by the asymmetry parameter, η, which is given by

formula
Equation 1.40

A quadrupolar coupling constant, CQ (in Hz), which a measure of the size of quadrupolar interaction, can be defined as

formula
Equation 1.41

CQ can be used to describe local distortion around the nucleus, while η provides the information of the local symmetry, specifically, the deviation of EFG from axial symmetry of the nucleus of interest.

In NMR spectroscopy, quadrupolar interaction is often treated as a small perturbation of Zeeman interaction (first-order and second-order quadrupolar interactions). Details on the treatment of quadrupolar interaction (i.e., the energy levels, powder pattern, etc.) can be found elsewhere in this book. In general, MAS is often used to remove first-order quadrupolar coupling to achieve high resolution observation. However, second-order quadrupolar coupling results in characteristic line shape and additional shift. The shift of the center of gravity (δCG) of the central line is at a lower frequency than the isotropic chemical shift δiso (field-independent) because of second-order quadrupolar coupling and they are related by

formula
Equation 1.42

where ν0 is the Larmor frequency in Hz.18,19 Eqn (1.42) clearly shows that the experimentally observed shift is more negative with a higher CQ and at a smaller external magnetic field. Since quadrupolar interaction is usually the largest line broadening factor in solid-state NMR spectrum of quadrupolar nuclei and second-order quadrupolar coupling is inversely proportional to the external field, a high B0 is preferred to obtain high resolution NMR spectra. Figure 1.7 shows the 27Al MAS NMR spectra of aluminoborate 9Al2O3 + 2B2O3 (A9B2) at different external magnetic fields. Since the Al ions in A9B2 are associated with large CQs, the linewidths are greatly reduced at a higher external field, providing spectra with better resolution.

Figure 1.7

27Al MAS spectra of A9B2 compound from 14 to 40 T. Reproduced from ref. 20 with permission from American Chemical Society, Copyright 2002.

Figure 1.7

27Al MAS spectra of A9B2 compound from 14 to 40 T. Reproduced from ref. 20 with permission from American Chemical Society, Copyright 2002.

Close modal

Many materials in energy applications are paramagnetic, in which the hyperfine interaction between the nucleus and the unpaired electrons plays an important role in determining the features in the NMR spectrum. The two major contributions to the hyperfine interaction are Fermi-contact interaction and dipolar coupling, which can be considered as the J-coupling and direct dipolar interaction, respectively, between a nucleus and the unpaired electrons, instead of another nuclear spin.

Fermi-contact interaction is dependent on the unpaired electron spin density transferred to the nuclear spin of interest, ρ(r = 0), where r is the distance to the center of the nucleus. Because the electron spin density can be transferred through chemical bond, Fermi-contact interaction provides important information on the chemical bonding of the paramagnetic center. Unlike dipolar coupling between two nuclear spins, the nuclear spins can only interact with the time-averaged magnetic moment of the electrons or the paramagnetic ions, leading to a line shape similar to CSA rather than a Pake doublet. Again, the size of hyperfine dipolar coupling is inversely proportional to the cube of the distance between the nuclear and electronic spins.

The Hamiltonian of hyperfine coupling can be written as

HFC = ·A·
Equation 1.43

where A is the hyperfine coupling tensor and the contributions from dipolar coupling and Fermi-contact interaction can be more clearly seen in the following expression

formula
Equation 1.44

where χ is the magnetic susceptibility, and ρ(r) is the total spin-unpaired electron density at position r.

Fermi-contact interaction and dipolar coupling broaden the peaks and cause additional shifts, making the spectra of paramagnetic materials more difficult to analyze than diamagnetic materials. Eqn (1.44) indicates that the size of hyperfine interaction is proportional to both the gyromagnetic ratio γI and the external magnetic field B0, and its dependence on B0 is different from the direct dipolar and indirect spin–spin interactions between nuclear spins. Therefore, in order to decrease the size of hyperfine interaction and simplify the spectra, it is better to perform NMR of paramagnetic materials with the nucleus associated with a lower gyromagnetic ratio, at lower external field with faster MAS rate.21 Figure 1.8 shows the comparison of 6Li and 7Li MAS NMR spectra of paramagnetic LiMn2O4 material. 6Li, which has a smaller gyromagnetic ratio thus less sensitivity, is associated with smaller hyperfine interactions. Therefore, the 6Li NMR spectrum is better resolved than the corresponding 7Li NMR spectrum.

As discussed in the previous section, electron-nuclear spin interactions strongly impact the NMR spectra of paramagnetic materials, including battery electrodes that contain redox-active transition metal species. To correctly interpret paramagnetic NMR spectra, an understanding of the spin, orbital and spin–orbit contributions to the electronic magnetic moment, which in turn dictates the magnetic properties of the material, is required. In this section, electron–nuclear spin interactions in the case of magnetically isotropic systems are described before the more complicated case of materials with magnetic anisotropy is examined. A comparison of the spectral signatures of paramagnetic interactions in the case of magnetic isotropy and anisotropy is also provided.

Figure 1.8

Comparison of the spectra of LiMn2O4 synthesized at 50 °C obtained with 6Li (a) and 7Li (b) MAS NMR at a MAS spinning speed of ∼9 kHz and a magnetic field strength of 4.7 T (200 MHz for 1H). Reproduced from ref. 22 with permission from American Chemical Society, Copyright 1998.

Figure 1.8

Comparison of the spectra of LiMn2O4 synthesized at 50 °C obtained with 6Li (a) and 7Li (b) MAS NMR at a MAS spinning speed of ∼9 kHz and a magnetic field strength of 4.7 T (200 MHz for 1H). Reproduced from ref. 22 with permission from American Chemical Society, Copyright 1998.

Close modal

NMR is a non-invasive and quantitative technique which allows sequences of events in rechargeable battery materials to be followed by studying electrochemical cells operando,23–29  or, more commonly, electrode samples stopped at different stages of charge and discharge ex situ.30–34  As a local structure probe, it is well suited for the investigation of amorphous or disordered systems, such as charged/discharged electrode materials. NMR is a site-specific technique, and crystallographically-unique sites or local environments in a material can be distinguished on the basis of their resonant frequencies or chemical shifts, provided that these frequency/shift differences are larger than the linewidth of the signals. This is usually not an issue for battery electrodes containing open-shell transition metals, as the paramagnetic chemical shift of a species, such as 7Li, 23Na, 31P or 19F, is typically large and highly sensitive to the local geometry and to the number and oxidation state of nearby transition metals (TMs), resulting in well separated signals. Hence, NMR is ideal for the study of local distortions and variations in the electronic structure and in the oxidation states of redox-active TM species during charge and discharge.21  Paramagnetic NMR can also monitor charge ordering transitions and the migration of electrochemically active or inactive species in electrode materials.

We begin this section by presenting the fundamental magnetic properties of TM-containing compounds, thereby laying the conceptual foundations for understanding paramagnetic interactions in common battery electrodes.

The contribution of nuclear spins to the magnetism of substances is negligible except at very low temperatures. At most temperatures, the macroscopic magnetic properties of atoms, molecules and solids depend on the number and distribution of paired and unpaired electrons.35  In the case of battery electrode materials, these properties are by and large determined by the redox-active TMs, and more specifically, by microscopic magnetic moments associated with unpaired electrons borne by the TMs. The next few sections describe the various contributions to the magnetic moment of open-shell TM ions.

Open-shell TM ions exhibit a magnetic moment originating from the orbiting motion of unpaired electrons with a finite charge around the nucleus. The microscopic magnetic moment associated with each electron can be viewed classically as the magnetic field generated by a circulating current. Namely, a current i flowing around an elementary oriented loop of area |dS| will create a magnetic moment

dµ = idS
Equation 1.45

with units of A m2.36  The length of the dS vector is equal to the area of the loop, and its direction is normal to the loop and set by the right-hand rule with respect to the direction of the current around the loop, as shown in Figure 1.9.36 

Figure 1.9

(a) Elementary magnetic moment dµ created by a current i flowing around an elementary oriented loop of area |dS|. (b) The magnetic moment µ due to a current loop is obtained by summing over a large number of infinitesimal current loops. Adapted from ref. 36 with permission from Oxford University Press, Copyright 2001.

Figure 1.9

(a) Elementary magnetic moment dµ created by a current i flowing around an elementary oriented loop of area |dS|. (b) The magnetic moment µ due to a current loop is obtained by summing over a large number of infinitesimal current loops. Adapted from ref. 36 with permission from Oxford University Press, Copyright 2001.

Close modal

For a loop of finite size, the magnetic moment µ is obtained by summing up the magnetic moments associated with a large number of infinitesimal current loops distributed throughout the area of the loop,36  as shown in Figure 1.9b, and

µ = ∫dµ = ∫idS
Equation 1.46

The orbiting motion of electrons around the nucleus not only leads to a charge current, but also to the motion of particles with a finite mass. As such, a magnetic moment is always associated with angular momentum. The orbital angular momentum is described by the quantum number L, which takes integer values. The application of an external magnetic field leads to a Zeeman interaction between the angular momentum and the field, and each level L splits into 2L + 1 non-degenerate states labelled by a unique magnetic quantum number mL, which takes integer values between −L and +L.

In TM ions, the magnetic moment µL associated with an orbiting electron lies along the direction of the angular momentum L of that electron and is proportional to it:

µL = ℏγSL =µBL
Equation 1.47

where γS is the gyromagnetic ratio of the electron, is Planck's constant divided by 2π, and µB is the Bohr magneton. Since the electron has a negative charge, µL is antiparallel to L and the electron gyromagnetic ratio, γS, is negative:

formula
Equation 1.48

where −e and me are the electron charge and electron mass, respectively. The Bohr magneton,

formula
Equation 1.49

represents the magnetic moment of an electron orbiting around the nucleus of a hydrogen atom in its ground state, i.e., with an angular momentum of and in a circular orbit of radius r = a0.36 

Electrons possess a second and intrinsic source of angular momentum due to their spin S. The spin magnetic moment of a free electron is given by:

µS = ℏγSS = −µBgeS
Equation 1.50

where ge is the free electron g-factor (ge ≈ 2.002319), and S always takes the value of ½. Upon application of an external magnetic field, the Zeeman interaction splits the S state into 2S + 1 = 2 non-degenerate states with magnetic quantum numbers .

While µS results in spatially isotropic magnetic properties, µL introduces anisotropy, whereby the magnetic response of the system to the application of an external magnetic field depends on the orientation of the field with respect to that of the principal axis of the molecule or crystal under consideration. Importantly, since L = S = 0 for closed-shell orbitals, the magnetic properties of TM ions arise from the interactions between the magnetic moments of unpaired electrons in open-shell valence orbitals.

In multi-electron systems, electronic angular momenta interact via couplings between individual orbital angular momenta (LiLj), between individual spin angular momenta (SiSj), and between orbital and spin angular momenta (LiSj). These so-called spin–orbit coupling (SOC) interactions are a major source of magnetic anisotropy and greatly impact the NMR properties of TM-containing systems.

While the spin and orbital angular momenta of isolated electrons are conserved within the spherically symmetric field of the nucleus, such that 2 and 2 commute with the Hamiltonian ([2, ] = 0, [2, ] = 0), interactions between them break this symmetry and individual angular momenta are no longer conserved individually but the angular momentum of the system as a whole is conserved. A total angular momentum J is therefore defined, where J is a good quantum number as 2 commutes with . As for the L and S angular momenta, upon application of an external magnetic field, each level J splits into 2J+1 non-degenerate states labelled by a unique magnetic quantum number mJ that takes integer values between +J and −J.

The presence of radially-contracted 3d orbitals leads to strong interelectron interactions. Hence, in battery systems containing 3d TM ions, LiSi couplings are generally much smaller than LiLj and SiSj couplings and can be considered as a perturbation. The Russell–Saunders coupling scheme can therefore be applied, where individual electron spin momenta (Si) couple strongly to give a total spin , and individual orbital momenta (Li) couple strongly to give a total orbital contribution . In this limit, the total angular momentum J is determined through coupling of the S and L vectors: J = L + S, as shown in Figure 1.10a. The quantum number J varies in integer steps between |L + S| and |LS|, resulting in (2L + 1)(2S + 1) possible electronic states labeled with term symbols of the form 2S+1Lj.

Figure 1.10

Angular momentum properties of a transition metal ion subject to Russell-Saunders coupling. (a) The total angular momentum J is the vector sum of the spin and orbital angular momenta, S and L. (b) The total magnetic moment µ is the sum of the spin and orbital magnetic moments, µS and µL. Adapted from ref. 11, https://doi.org/10.1016/j.pnmrs.2018.05.001, under the terms of the CC BY 4.0 license, https://creativecommons.org/licenses/by/4.0/.

Figure 1.10

Angular momentum properties of a transition metal ion subject to Russell-Saunders coupling. (a) The total angular momentum J is the vector sum of the spin and orbital angular momenta, S and L. (b) The total magnetic moment µ is the sum of the spin and orbital magnetic moments, µS and µL. Adapted from ref. 11, https://doi.org/10.1016/j.pnmrs.2018.05.001, under the terms of the CC BY 4.0 license, https://creativecommons.org/licenses/by/4.0/.

Close modal

Similar to the gyromagnetic ratio, the g-factor relates the magnetic moment to the angular momentum. While the free electron (S = 1/2) g-factor ge ≈ 2.002319, g-factors of individual particles and TM ions can deviate substantially from ge, due to the presence of orbital angular momentum and SOC. For a multi-electron TM ion, the total magnetic moment operator , depicted in Figure 1.10b, is given by

= −µB(gL + ge)
Equation 1.51

where gL = 1 and ge ≈ 2 are the g-factors of the orbital and spin angular momenta, respectively.34 

As shown in Figure 1.10, the vectors µ and J are not parallel because ge ≠ 1. Yet, only the component of µ parallel to J is conserved, and the component of the magnetic moment operator that appears in the interaction Hamiltonian is that which commutes with and its matrix elements are proportional to those of .11  Applying the projection theorem,11  the total magnetic moment operator can be rewritten as:

= −gJµB
Equation 1.52

where gJ is the Landé g-factor given by

formula
Equation 1.53

gJ takes values between 1 and 2, such that gJ = 1 for S = 0 and gJ = 2 for L = 0. A complete derivation of gJ can be found in ref. 11 and 36.

As discussed in Section 1.1, the magnetic spin dipole moment operator of a nucleus, I, is given by:

I = ℏγI
Equation 1.54

where is the spin angular momentum operator and γI is the nuclear gyromagnetic ratio. The product

ℏγI = µNgI,
Equation 1.55

where gI is the dimensionless nuclear g-factor, which depends on the isotope and takes the same sign as γI. µN is the nuclear magneton:

formula
Equation 1.56

where e is the elementary charge, and mp is the rest mass of the proton. Substituting the expression for ℏγI, an alternative expression for the nuclear spin magnetic moment, in terms of the nuclear g-factor, is obtained:

I = µNgI
Equation 1.57

The magnetic properties of isolated TM ions, or free ions, can be fully described in terms of the spin, orbital, and SOC contributions to the magnetic moments of unpaired electrons occupying degenerate valence d orbitals. As described by ligand field theory,37  in molecules and solids the presence of nearby ligand atoms leads to strong interactions between TM d orbitals and ligand orbitals. These interactions, in turn, result in a splitting of the degenerate d manifold of the free ion into non-degenerate sets of d orbitals.

For open-shell TM species, lifting of the orbital degeneracy results in partial or complete quenching of the orbital angular momentum, L, and strongly influences their magnetic properties. For example, for a high spin d3 TM ion octahedrally- and tetrahedrally-coordinated to ligand atoms, the ligand field alters the d orbital energies in such a way that full quenching (whereby the effective orbital angular momentum, Leff, is zero) and partial quenching (Leff = 1) of L is observed, respectively, as shown in Figure 1.11.

Figure 1.11

Orbital angular momentum quenching for a d3 TM ion in the presence of an octahedral (Oh) and a tetrahedral (Td) ligand field. Upon application of an Oh field, the five degenerate d orbitals of the free ion split into two sets of degenerate orbitals, labelled t2g and eg. For a d3 electronic configuration, this leads to complete quenching of the orbital angular momentum (Leff = 0). Upon application of a Td field, the five degenerate d orbitals of the free ion split into two sets of degenerate orbitals, labelled e and t2, leading to partial quenching of the orbital angular momentum for a d3 ion (Leff = 1). The theoretical basis of d orbital splitting in the presence of a ligand field can be found in inorganic chemistry textbooks, e.g., by Kettle38  or by Shriver and Atkins.39 

Figure 1.11

Orbital angular momentum quenching for a d3 TM ion in the presence of an octahedral (Oh) and a tetrahedral (Td) ligand field. Upon application of an Oh field, the five degenerate d orbitals of the free ion split into two sets of degenerate orbitals, labelled t2g and eg. For a d3 electronic configuration, this leads to complete quenching of the orbital angular momentum (Leff = 0). Upon application of a Td field, the five degenerate d orbitals of the free ion split into two sets of degenerate orbitals, labelled e and t2, leading to partial quenching of the orbital angular momentum for a d3 ion (Leff = 1). The theoretical basis of d orbital splitting in the presence of a ligand field can be found in inorganic chemistry textbooks, e.g., by Kettle38  or by Shriver and Atkins.39 

Close modal

For systems where the orbital angular momentum is, to a first approximation, quenched (Leff = 0), the resulting magnetic moment depends solely on the spin angular momentum S, which remains unaffected by the ligand field. However, if the spin and orbital magnetic moments are coupled by SOC, the ligand field is unable to effect a perfect separation of electronic states with the same spin multiplicity. As a result, the ground state may gain some orbital angular momentum through mixing in of excited states of the same spin multiplicity, as shown in Figure 1.12c. The extent of state mixing depends on the energy separation of the excited state from the ground state, and on the strength of SOC, quantified by the SOC constant λ.

Figure 1.12

Zeeman energy levels of a TM ion with spin , as a function of the external magnetic field B, for an orbitally non-degenerate ground state (L = 0): (a) in the absence of ZFS, (b) in the presence of ZFS, and (c) in the presence of ZFS and low-lying excited states. In (b) and (c), the effect of ZFS is shown in the presence of an axially symmetric ligand field with D > 0. The degeneracy of the free ion d states is partially lifted by ZFS before the application of a magnetic field B, here set parallel to the principal axis of the ligand field. Adapted from ref. 41 with permission from Elsevier, Copyright 2002.

Figure 1.12

Zeeman energy levels of a TM ion with spin , as a function of the external magnetic field B, for an orbitally non-degenerate ground state (L = 0): (a) in the absence of ZFS, (b) in the presence of ZFS, and (c) in the presence of ZFS and low-lying excited states. In (b) and (c), the effect of ZFS is shown in the presence of an axially symmetric ligand field with D > 0. The degeneracy of the free ion d states is partially lifted by ZFS before the application of a magnetic field B, here set parallel to the principal axis of the ligand field. Adapted from ref. 41 with permission from Elsevier, Copyright 2002.

Close modal

In multi-electron systems, the presence of a non-cubic ligand field and strong spin–spin interactions can lead to zero-field splitting (ZFS) effects. ZFS results from interactions between electronic magnetic moments and occurs in the absence of a magnetic field, as its name suggests. An energy level diagram illustrating ZFS effects is presented in Figure 1.12, where, for convenience, we set and L = 0. In an axially-symmetric ligand field, the fourfold degeneracy of the Zeeman sublevels (Figure 1.12a) is partially lifted, with the states separated by an amount 2D (in units of energy) from the states (Figure 1.12b).42  The population of the energy levels can be described by a Boltzmann distribution and depends on the relative magnitudes of the axial ZFS parameter D and the thermal energy kT. Figure 1.12b and c represent a situation where D > 0, resulting in ±½ states that are lower in energy than the ±3/2 states.40 

While departure from a spherically-symmetric electronic spin density around the TM ion, as in a non-cubic ligand field, is a necessary condition for ZFS, orbital angular momentum and SOC are important contributors to the splitting and make it fairly large. Similarly to the orbital angular momentum and SOC, ZFS results in magnetic anisotropy and has profound effects on magnetic properties, and therefore on paramagnetic NMR properties.

In typical systems composed of a large number of nuclear and electronic spins, the application of a magnetic field (denoted H for reasons that will become clear shortly) results in a Boltzmann distribution of the spins across the Zeeman energy levels. The lowest energy state with a magnetic moment aligned with the field becomes more populated than the highest energy state with a magnetic moment anti-aligned with the field. This Boltzmann distribution leads to a finite induced magnetic moment µind along H.

Let us consider materials with isotropic magnetic properties, i.e., containing spin-only TMs in a cubic ligand field. The total magnetic moment per unit volume is the net bulk magnetization M which, for many materials (termed linear materials), is proportional to the applied magnetic field H:42 

formula
Equation 1.58

where χV is the dimensionless magnetic susceptibility per unit volume. A magnetic flux density B can be defined as:

B = µ0(H + M) = µ0(1 + χV)H
Equation 1.59

where µ0 = 4π × 10−7 H m−1 is the permeability of free space. Apart from ferromagnets, χV ≪ 1 and

B = µ0H.
Equation 1.60

Hereafter, we restrict ourselves to materials for which eqn (1.60) is valid, i.e., linear materials with χV ≪ 1, and refer to B as the external static magnetic field. The Hamiltonian of a system of nuclear and electronic spins in the presence of the constant B field is36 

formula
Equation 1.61

where 0 is the unperturbed Hamiltonian (without the magnetic field), the second term is the paramagnetic term, which vanishes when all electron spins are paired (L = S = 0), and the third term is the diamagnetic term present in any material with paired and/or unpaired electrons.

We proceed to evaluate the various terms in the Hamiltonian above. From statistical mechanics, the Helmholtz free energy is defined as dF = d(ETS) = −SdTpdVMdB. F can be expressed in terms of the partition function as F = −NkBT·lnZ, and . By convention, B is set along the z axis: B=(0, 0, B).

To obtain the paramagnetic part of the susceptibility, we consider the partition function of a system with total angular momentum magnetic quantum number mJ and g-factor gJ. The magnetic moments associated with individual J states take values of µJ = mJgJµB, with corresponding energies EJ = mJgJµBB.36  Hence, the partition function becomes: . Setting , the partition function is a geometric progression with initial term a = eJx and multiplying term r = ex. Using the formula ,where b is the number of terms in the series (here, b = 2J + 1), and after some maths, the partition function can be written in compact form:36 .

The average magnetic moment µJ is given by:

formula
Equation 1.62

Using the expression of Z derived above,

formula
Equation 1.63

The magnetization M is then:

formula
Equation 1.64

with n the number of individual <µJ> moments in the system. Setting , we find

M = MsatBJ(y)
Equation 1.65

where Msat is the saturation magnetization when all n paramagnetic centres occupy the lowest energy J state with mJ = J, hence Msat = ngJµBJ. BJ(y) is the Brillouin function:36,40 

formula
Equation 1.66

The expressions above are valid for any temperature and magnetic field and can be simplified in two limiting cases. In the saturation regime, i.e., when the field is sufficiently high or the temperature is sufficiently low that mJgJµBBkT, the coth function and BJ(y) tend to unity and M = Msat. Conversely, in the high-temperature limit typical of NMR experiments, EJ = mJgJµBBkT, i.e., y ≪ 1, and the Brillouin function can be simplified using a Maclaurin expansion for coth(y):36 

formula
Equation 1.67

Hence, for low magnetic fields, the susceptibility is given by

formula
Equation 1.68

where

formula
Equation 1.69

is the effective magnetic moment.36  The paramagnetic susceptibility, χpara, takes the form of Curie's law, with a 1/T dependence. For materials with interacting magnetic moments µJ, such as ferromagnets and antiferromagnets, the susceptibility takes a Curie form at temperatures above the magnetic ordering temperature, i.e., in the paramagnetic regime. For ferromagnets, this is the Curie temperature TC, and , and for antiferromagnets, this is the Néel temperature TN, and . Thus, a general expression for the magnetic susceptibility is given by the Curie–Weiss law:

formula
Equation 1.70

where

formula
Equation 1.71

is the Curie constant, Θ the Weiss temperature and α is a temperature-independent term. Θ can, in practice, be very different from the magnetic transition temperature, TC or TN. For Curie paramagnets (no internal interactions between magnetic spins), the Weiss temperature Θ = 0, while Θ > 0 for ferromagnets and Θ < 0 for antiferromagnets.

For the diamagnetic term, B × ri = B(−yi,xi,0) and (B × ri)2 = B2(xi2 + yi2). Within first-order perturbation theory, the change in the ground state energy due to the diamagnetic term is

formula
Equation 1.72

where |ϕ0〉 is the ground state wavefunction and Z is the number of electrons per atom.36  The last equality assumes a spherically-symmetric atom, such that . For a diamagnetic solid composed of N atoms (each with Z electrons of mass m) in volume V, the magnetization at T = 0 K can be derived:36 

formula
Equation 1.73

The diamagnetic susceptibility, χdia, can be extracted from eqn (1.58) and (1.60): (assuming that χdia ≪ 1). We obtain:15 

formula
Equation 1.74

This expression demonstrates that the diamagnetic susceptibility is largely temperature independent.

Van Vleck paramagnetism is a small, second-order perturbation to the magnetic susceptibility of a system with total angular momentum quantum number J = 0 in the ground state|ϕ0〉. In this case, first-order paramagnetic effects vanish since 〈ϕ0||ϕ0〉 = gJµB〈ϕ0| |ϕ0〉 = 0. However, mixing of excited states with J ≠ 0 can affect the ground state energy E0 according to second-order perturbation theory. The change in E0 for a TM ion with J = 0 is36 

formula
Equation 1.75

The sum in the first term is taken over all excited states of the system, and the second term is the diamagnetic term discussed earlier. The Van Vleck susceptibility is obtained from the first term as

formula
Equation 1.76

and is positive because EnE0 > 0.36  Van Vleck paramagnetism, like diamagnetism, is temperature-independent.

For paramagnetic systems, the diamagnetic susceptibility term is very small compared to paramagnetic terms and is hereafter neglected. Four cases can be distinguished:

Case 1: For paramagnetic compounds containing spin-only TM ions in a perfectly cubic environment (A1 ground term symbol), the magnetic susceptibility is purely isotropic and the Van Vleck term vanishes. As a result, the temperature-independent α term in eqn (1.70) for the Curie–Weiss susceptibility becomes negligible and depends on the spin-only effective magnetic moment,

formula
Equation 1.77

Case 2: For paramagnetic systems where the orbital angular momentum of the ground electronic state is quenched (Leff,gs = 0), e.g., by a ligand field, but SOC mixes in low-lying excited states of the same spin multiplicity with a non-zero orbital angular momentum (Leff,es ≠ 0), the ground state gains some orbital angular momentum and the magnetic moment deviates from the spin-only formula.

Case 3: The presence of SOC in the ground electronic state (Leff,gs ≠ 0) adds an anisotropic component to the susceptibility. In the Russell–Saunders coupling regime (weak LS coupling), and in the absence of ZFS, the magnetic moment depends on the J quantum number and the Landé g-factor, gJ, such that . As in the spin-only case, the magnetic susceptibility is of the Curie–Weiss form, with, .

Case 4: In the presence of ZFS effects, an anisotropic, temperature-independent Van Vleck paramagnetic susceptibility term is added to χCW, which will be discussed in greater detail later.

The SOC constant λ, which quantifies the strength of spin–orbit interactions, depends on the one-electron SOC constant ζ proportional to (Z*)4, where Z* is the effective nuclear charge. For 3d TM ions, Z* is relatively low, and ζ3d constants are fairly small, varying from ∼150 cm−1 for Ti3+ to ∼870 cm−1 for Cu2+, as shown in Table 1.1 (for comparison, one-electron SOC constants for lanthanide ions, ζ4f, are typically in the range of 650–3000 cm−1). SOC constants generally decrease when an ion is placed in a ligand field. Hence, even in the presence of residual orbital angular momentum (Leff = 1), SOC effects are typically small compared to ligand field interactions for systems containing open-shell 3d TM species, and experimental magnetic moments measured on TM complexes do not always agree with effective moments predicted from Russell–Saunders coupling of L and S angular momenta.

Table 1.1

Overview of the magnetic properties of high spin 3d octahedral (Oh) TM complexes, and comparison with effective magnetic moments computed (1) using the Russell-Saunders (RS) coupling scheme (µeff,RS), (2) by combining the theoretical spin and orbital magnetic moments (µeff,S+L), and (3) using the spin-only (SO) model (µeff,so). One-electron atomic SOC constants, ζ3d, were obtained from the Handbook of Atomic Data by S. Fraga, J. Karawowski and K. M. S. Saxena, Elsevier, New York, 1976.44  Experimental magnetic moments were obtained from Kettle40 

3d iondnζ3d cm−1−1SLeffJeffgJµeff,RS/µBµeff,S+L/µBµeff,so/µBµexp/µB
V4+, Ti3+ d1 253, 158 ½   0.58 2.24 1.73 1.6–1.8 
V3+ d2 220 — 3.17 2.83 2.7–2.9 
Cr3+, V2+ d3 296, 187   3.87 3.88 3.88 3.7–3.9 
Mn3+, Cr2+ d4 338, 256 4.90 4.90 4.90 4.7–5.0 
Fe3+, Mn2+ d5 499, 343   5.92 5.92 5.92 5.6–6.1 
Fe2+ d6 441  5.77 5.10 4.90 5.1–5.7 
Co2+ d7 561    4.73 4.13 3.88 4.3–5.2 
Ni2+ d8 703 2.83 2.83 2.83 2.8–3.5 
Cu2+ d9 870 ½ ½ 1.73 1.73 1.73 1.7–2.2 
Zn2+ d10 — — 
3d iondnζ3d cm−1−1SLeffJeffgJµeff,RS/µBµeff,S+L/µBµeff,so/µBµexp/µB
V4+, Ti3+ d1 253, 158 ½   0.58 2.24 1.73 1.6–1.8 
V3+ d2 220 — 3.17 2.83 2.7–2.9 
Cr3+, V2+ d3 296, 187   3.87 3.88 3.88 3.7–3.9 
Mn3+, Cr2+ d4 338, 256 4.90 4.90 4.90 4.7–5.0 
Fe3+, Mn2+ d5 499, 343   5.92 5.92 5.92 5.6–6.1 
Fe2+ d6 441  5.77 5.10 4.90 5.1–5.7 
Co2+ d7 561    4.73 4.13 3.88 4.3–5.2 
Ni2+ d8 703 2.83 2.83 2.83 2.8–3.5 
Cu2+ d9 870 ½ ½ 1.73 1.73 1.73 1.7–2.2 
Zn2+ d10 — — 

To demonstrate this, effective magnetic moments predicted from (1) the Russell–Saunders (RS) coupling scheme (), from (2) the combined, albeit non-interacting, theoretical spin and orbital magnetic moments (), and from (3) the spin-only (SO) model, which assumes complete quenching of the orbital angular momentum (Leff = 0 and ), are compared to experimental moments obtained on high spin Oh 3d TM complexes in Table 1.1. Notably, µeff, S+L corresponds to the situation where the orbital angular momentum makes its full contribution to the total magnetic moment,43  while µeff, RS also accounts for SOC.

As shown in Table 1.1, for high spin d3, d4, d5, d8, d9 and d10 ions, the application of an ideal Oh ligand field quenches the orbital angular momentum (Leff = 0), and effective moments predicted using the spin-only, S + L, and RS models become equal and broadly consistent with experimental values. The small discrepancies observed between the theoretical and experimental µeff values may be due to deviations from the ideal Oh TM coordination geometry (low symmetry ligand fields and possible ZFS), and/or to mixing in of low-lying excited states with Leff ≠ 0.

The moments tabulated above further indicate that complexes containing octahedral V3+/4+ and Ti3+, with a theoretical partially-quenched angular momentum of Leff = 1, appear to choose a ground state where Leff = 0, such that the spin-only formula provides a better picture for the ground state effective magnetic moment. In reality, for ions with a less than half-filled d shell (d1 to d4), SOC leads to spin and orbital angular momenta opposing one another and to a reduction of the net magnetic moment. In cases where the net moment reduction due to SOC effects roughly cancels out the orbital contribution, the resulting moment is fortuitously close to its spin-only value. For d6 to d9 ions with a more than half-filled d shell, the moment is increased due to spin and orbital contributions reinforcing each other, and in Table 1.1, complexes containing octahedral Fe2+ (d6) and Co2+ (d7) exhibit magnetic moments closer to the theoretical values predicted using the RS and ‘S + L’ schemes. It is important to note that the experimental moments presented here were obtained on TM complexes in solution, where interactions between paramagnetic TM centres are fairly weak. In the case of battery materials composed of an extended network of coupled TM ions, interactions between magnetic momenta on neighboring ions are often exacerbated and SOC effects can become significant.

Overall, the above analysis indicates that, for the purpose of describing the magnetic properties of battery electrode materials comprising 3d TM ions, a simple “spin-only” approach is appropriate in the case of complete orbital quenching in a ligand field (Leff = 0), with little mixing in of excited states with a finite orbital contribution. The spin-only approach is also a good approximation for early 3d open-shell TM ions (3d1-3d4) with a partially quenched orbital angular momentum (Leff = 1). The paramagnetic NMR properties of spin-only systems are described in Section 1.2.2.2. On the other hand, for systems containing 3d6-3d9 ions, SOC must be accounted for in the magnetic response, and the paramagnetic properties of such systems are presented in Section 1.2.2.3. The situation is much less clear cut for the 4d and 5d TM series, because the heavier ions have a larger spin–orbit splitting, and the effects of the ligand field, ZFS and of the spin–orbit interaction can be comparable.36 

In the NMR experiment, a sample is placed in an external magnetic field B0, by convention applied along z (B0=(0, 0, B0)). The Zeeman interaction between the nuclear magnetic moments µI and the B0 field leads to a splitting of the I nuclear energy levels into 2I + 1 non-degenerate Zeeman states distinguished by their magnetic spin quantum number mI, which can take any integer value between −I and +I (−I, −I+1, …, I−1, I), as shown in Figure 1.13a. For a paramagnetic TM ion with total spin S obtained by summing over all n unpaired electron spins, , the Zeeman interaction leads to a ladder of electron spin energy levels, as shown in Figure 1.13b.

Figure 1.13

Ladders of nuclear and electron spin energy levels due to Zeeman interactions in an external magnetic field B0. (a) Case of a nuclear spin I with a positive gyromagnetic ratio γI, leading to a mI = +I ground state. The energy separation between adjacent mI Zeeman sublevels (ΔmI = ±1) is the Larmor frequency ω0 = −γIB0, in units of . (b) Energy levels for an arbitrary electronic spin S. The negative gyromagnetic ratio of the electron, γS, leads to a ground state with ms = −S. Note that the energy separation between spin sublevels is not drawn to scale, and ΔEmI = ±1) ≪ ΔEmS = ±1). Adapted from ref. 11, https://doi.org/10.1016/j.pnmrs.2018.05.001, under the terms of the CC BY 4.0 license, https://creativecommons.org/licenses/by/4.0/.

Figure 1.13

Ladders of nuclear and electron spin energy levels due to Zeeman interactions in an external magnetic field B0. (a) Case of a nuclear spin I with a positive gyromagnetic ratio γI, leading to a mI = +I ground state. The energy separation between adjacent mI Zeeman sublevels (ΔmI = ±1) is the Larmor frequency ω0 = −γIB0, in units of . (b) Energy levels for an arbitrary electronic spin S. The negative gyromagnetic ratio of the electron, γS, leads to a ground state with ms = −S. Note that the energy separation between spin sublevels is not drawn to scale, and ΔEmI = ±1) ≪ ΔEmS = ±1). Adapted from ref. 11, https://doi.org/10.1016/j.pnmrs.2018.05.001, under the terms of the CC BY 4.0 license, https://creativecommons.org/licenses/by/4.0/.

Close modal

On closer inspection, each electron Zeeman spectroscopic term can be split further according to the total angular momentum J, due to coupling of the L and S angular momenta, leading to fine structure of the signal lineshape in electron paramagnetic resonance (EPR). In addition, the nuclear spin I and the electron total angular momentum J also couple and result in hyperfine structure. Hence, the expressions hyperfine and paramagnetic interactions are used interchangeably to designate the magnetic interactions between nuclei and electrons.

The total Hamiltonian representing NMR interactions can be written as

Total = Ext + Int.
Equation 1.78

The external contribution, Ext, comprises the Zeeman term due to the interaction of the NMR nucleus and the external static B0 field along z, Z, as well as a term resulting from the interaction with the radiofrequency (RF) field B1.

Z = −ℏγIB0z = ℏω0z
Equation 1.79

where z is the operator representing the component of the spin along the z axis and

ω0 = −γIB0
Equation 1.80

is the Larmor frequency. ω0 is the nuclear precession frequency and corresponds to the energy separation (in units of ) between the mI Zeeman sublevels, as shown in Figure 1.13a. The transient RF field B1 causes transitions between mI spin states when the RF frequency ω1 = ω0.

As discussed in Section 1.1, the internal Hamiltonian,

Int = CS + D + J + Q + HFC
Equation 1.81

has contributions from: chemical shielding interactions, CS; dipolar interactions between two nuclei, D; scalar couplings between two nuclei, J; quadrupolar interactions, Q; and hyperfine interactions between an electron and a nucleus, HFC. Typical NMR experiments are performed in the high field limit, where the interaction between spins with the external magnetic field dominate all internal interactions. In this limit, interactions between nuclear spins or between electron spins and nuclear spins can be treated as perturbations of the Zeeman levels, greatly simplifying the corresponding Hamiltonians.

In strongly paramagnetic materials, the internal part of the NMR Hamiltonian of nuclei with I = ½ (e.g., 1H, 13C, 19F, 31P), Int, is largely dominated by the hyperfine HFC term. For quadrupolar nuclei with I ≥ 3/2 (e.g., 17O, 23Na, 27Al), the contribution from the quadrupolar Q term can also be significant, although many quadrupolar nuclei present in battery systems (e.g., 6/7Li, and in some cases, 23Na) exhibit a small quadrupole moment and/or occupy cubic environments so that the Q term and their NMR properties are dominated by the paramagnetic effects discussed here.

The presence of redox-active, open-shell TM species in battery electrodes leads to hyperfine interactions between nuclear spins and unpaired d electron spins on the TMs. While an overview of the relevant coupling terms is provided below, the reader is referred to reviews by Pell et al.,11,45  Bertini et al.41  and Kaupp and Köhler46  for additional details. This section focuses on magnetically isotropic paramagnets containing spin-only TM ions in a perfectly-cubic ligand field environment, with unpaired electrons occupying a non-degenerate (singlet) orbital state (e.g., high spin d5 Mn2+ and Fe3+), and no added complication from orbital (L = 0), SOC or ZFS effects.

The hyperfine Hamiltonian can be derived either as the interaction between the nuclear magnetic moment µI with a magnetic field due to the unpaired electron, or equivalently as the interaction of the electronic magnetic moment µS with a magnetic field due to the nucleus.16  Abragam and Bleaney47  show that

formula
Equation 1.82

where is the vector derivative operator del, . (x, y, z) are the coordinates of a point at distance from the nucleus.11 

The hyperfine Hamiltonian can be divided into two parts

formula
Equation 1.83

where the first term describes the electron-nuclear spin dipolar coupling and the second term describes the Fermi contact interaction. Let us assume that the unpaired electron is spatially delocalized while the nucleus is localized at r = 0, as appropriate for 3d electrons of first-row TM ions that are delocalized into ligand orbitals.11  The magnetic interaction is calculated by multiplying HFC with the unpaired electron density at position r,

ρ(r) = |ψ(r)|2
Equation 1.84

where ψ(r) is the electron wavefunction, and then integrating over all space.11 

To derive an expression for the spin dipolar coupling term, we consider the situation where the unpaired electron and nucleus are well separated (r ≠ 0), i.e., the electron occupies an orbital other than an s orbital, and the second term in eqn (1.83) integrates to zero. The resulting hyperfine coupling term

formula
Equation 1.85

where r is the vector displacement of the unpaired electron with respect to the nucleus and e the corresponding unit vector, describes the spin dipolar interaction.11 

To derive an expression for the Fermi contact interaction term, we instead consider the situation where the unpaired electron is in the immediate vicinity of the nucleus, i.e., in an s orbital, such that r = 0 and the spin dipolar coupling is equal to zero. Integrating the second term in eqn (1.83) leads to

formula
Equation 1.86

which is the Fermi contact term. This term is purely isotropic and arises due to delocalization of the unpaired electron onto the nucleus, as indicated by the dependence of the size of the interaction on |ψ(0)|2, which is the unpaired electron density at the nuclear position (r= 0).11 

The full Hamiltonian is obtained by summing over D and FC:

formula
Equation 1.87

In paramagnetic battery materials, we generally consider the interaction between a nucleus and a paramagnetic TM ion containing multiple unpaired electrons, with total electronic spin S. The Hamiltonian in eqn (1.87) is modified so that we sum over all electrons:

formula
Equation 1.88

where s,i and ψi(ri) are the magnetic moment and wavefunction of electron i at position ri. This expression can be simplified by writing the sum over the electrons i in terms of the average spin density per electron , where ρα−β(r) is the total unpaired spin electron density at position r, and 2S is the number of unpaired electrons. The Fermi contact term can then be rewritten as

formula
Equation 1.89

where is the magnetic moment operator for the total electron spin .11  Similarly, the spin dipolar term can be rewritten as

formula
Equation 1.90

The total hyperfine interaction Hamiltonian is thus

formula
Equation 1.91

As shown in Figure 1.14 and as evidenced by eqn (1.91), paramagnetic interactions depend on the spatial orientation of µI and µS, and therefore on the orientation of I and S. In practice, electron spin relaxation dynamics (i.e., transitions between ms levels) are fast (on the order of 10−14 to 10−8 s)48  on the timescale of nuclear spin relaxation (in the range 10−4 to 101 s),49  and the nuclear spin I (µI) does not interact with the unpaired electron spin S (µS), but rather with its average value 〈S〉 (〈µS〉).11  As a result, hyperfine coupling between I and S does not lead to a splitting of the NMR resonance, as would be expected for an interaction between two different magnetic dipoles (e.g., heteronuclear dipolar coupling). It instead takes the form of a chemical shift between I and 〈S〉 and causes a shift of the resonant frequency, as well as chemical shift anisotropy.11 

Figure 1.14

Hyperfine interactions in NMR. The nuclear magnetic moment µI (spin I) interacts with the time average of the electron magnetic moment 〈µS〉(spin 〈S〉) due to unpaired d electrons present on neighboring paramagnetic transition metals. 〈µS〉 (or 〈S〉) arises from the unequal population of the different electron spin energy levels in the presence of B0 and fast transitions between these levels on the timescale of NMR. (a) The Fermi contact interaction occurs through bonds and results in delocalization of unpaired electron spin density from the TM d orbitals to s orbitals on the nucleus A under observation. (b) The electron–nuclear spin dipolar interaction occurs through space and depends on the orientation of the vector connecting the nuclear and electron magnetic moments with respect to the external B0 field, as well as the electron-nuclear distance R.

Figure 1.14

Hyperfine interactions in NMR. The nuclear magnetic moment µI (spin I) interacts with the time average of the electron magnetic moment 〈µS〉(spin 〈S〉) due to unpaired d electrons present on neighboring paramagnetic transition metals. 〈µS〉 (or 〈S〉) arises from the unequal population of the different electron spin energy levels in the presence of B0 and fast transitions between these levels on the timescale of NMR. (a) The Fermi contact interaction occurs through bonds and results in delocalization of unpaired electron spin density from the TM d orbitals to s orbitals on the nucleus A under observation. (b) The electron–nuclear spin dipolar interaction occurs through space and depends on the orientation of the vector connecting the nuclear and electron magnetic moments with respect to the external B0 field, as well as the electron-nuclear distance R.

Close modal

To obtain an expression for the expectation value or time average of the electron spin vector, 〈S〉, we assume that it is equivalent to the average over the entire ensemble of paramagnetic centres.11  In this case, 〈S〉 is equal to the Boltzmann average of the components Si of the spin:

formula
Equation 1.92

where k is the Boltzmann constant and T is the temperature. In eqn (1.92) we use the fact that the electron Zeeman term is by far the dominant contribution to the total Hamiltonian in eqn (1.78) and approximate the energy of the system to that of the spin-only electron Zeeman level defined by quantum numbers (S,mS), ES, mS = geµBmSB0. It then becomes evident that 〈Sx〉 = 〈Sy〉 = 0, because these components are perpendicular to the applied field and there is no driving force for inducing a magnetic moment in these directions.11  Yet, the magnetic moment along z is non-zero and

formula
Equation 1.93

which becomes:

formula
Equation 1.94

By analogy with eqn (1.65), 〈Sz〉 can be expressed in terms of the Brillouin function. For spin-only systems, 〈Sz〉 = −SBS(y), with .

In the high-temperature limit typical of NMR experiments ES, MS = geµBMSB0kBT and the expression for 〈Sz〉 becomes

formula
Equation 1.95

This value of 〈Sz〉 is referred to as the Curie spin, as it exhibits a Curie temperature dependence of 1/T. In the high-temperature limit, both 〈Sz〉 and the corresponding hyperfine Hamiltonian, 〈〉.A., depend linearly on B0. 〈Sz〉 can be related to the expectation value of the electronic magnetic moment 〈µS〉 along B0. However, instead of considering the spin S of a single free electron (as in eqn (1.54)), we consider the total effective spin of a spin-only paramagnet containing N unpaired electrons. The energy levels of the electrons are then conveniently described by the single effective spin , which has the same multiplicity 2 + 1 as the true states.47  Here, we consider 3d TMs, for which the pseudo-spin is the same as the true spin S, so we drop the tilde.

The components of the expectation value of the electronic magnetic moment, 〈µS,i〉, are proportional to those of the average spin, 〈Si〉, such that 〈µS,x〉 = 〈µS,y〉 = 0 and

µS,z〉 = 〈µS〉 = −µB ge Sz
Equation 1.96

where we have related the average induced magnetic moment 〈µS〉 to the expectation value of µS,z.11 µS〉 can also be written as the ensemble average of the induced magnetic moment per paramagnetic metal ion:

formula
Equation 1.97

where NA is Avogadro's number, Vm is the molar volume. Using eqn (1.58) and (1.60), we find that . 〈S〉 can then be related to the magnetic susceptibility by substituting this expression into eqn (1.97):

formula
Equation 1.98

where χm = VMχV is the magnetic susceptibility per mole (m3 mol−1).

Combining eqn (1.96) and (1.98), the average electron spin along B0 can be expressed in terms of χm:

formula
Equation 1.99

The magnetization of the sample, M, can be rewritten as

formula
Equation 1.100

And from eqn (1.99) and (1.100), the molar Curie susceptibility is derived:

formula
Equation 1.101

This expression is similar to that in eqn (1.68) and (1.69), albeit for a spin-only system where , with .

We proceed to derive an expression for the hyperfine interaction Hamiltonian in terms of the spin operator and the external magnetic field B0, which in turn allows us to make predictions of NMR data. For this, we start from eqn (1.91) and account for the fact that the nuclear magnetic moment couples to the average electronic magnetic moment operator, 〈S〉:11 

formula
Equation 1.102

S〉 is the operator equivalent of 〈µS〉 in eqn (1.98). Expressed in terms of the magnetic susceptibility per TM ion, χ (where χ = V = NAχm in m3), and for an arbitrary direction of the magnetic field B0,

formula
Equation 1.103

Using this expression and I = ℏγI, we obtain the expression for the hyperfine interaction Hamiltonian presented in Section 1.1, namely

formula
Equation 1.104

The hyperfine Hamiltonian HFC can be written in compact form as

HFC = 〈〉.A.
Equation 1.105

where A is the hyperfine coupling tensor and expresses how much the nuclear spin I and the average electron spins 〈S〉 sense each other.

The full Hamiltonian, in the high field limit, is then11 

iso = ℏω0,Iz + µBgeB0z +〈〉.A.
Equation 1.106

where the first two terms are the nuclear and electron Zeeman interactions, respectively, and the third term is the hyperfine coupling.

The hyperfine coupling tensor A is composed of an isotropic Fermi contact term (AFC) and a traceless and symmetric spin dipolar coupling tensor (AD):

A = AFC1 + AD
Equation 1.107

where 1 is the identity matrix.

If the electrons are delocalized onto ligand orbitals, as appropriate for unpaired electrons in 3d TM orbitals, AFC and the Cartesian components of AD are given by11 

formula
Equation 1.108
formula
Equation 1.109

where i and j are equal to x, y or z, and δij is the Kronecker delta.

We now describe the Fermi contact interaction depicted in Figure 1.14a. Starting from the hyperfine coupling tensor A presented in eqn (1.107), the isotropic part of the tensor, AFC, leads to a Fermi contact interaction between S and I. Since AFC is a constant (eqn (1.108)), the corresponding Fermi contact Hamiltonian can be written as FC = AFC〉.. For spin-only systems, this interaction is purely isotropic and leads to a shift of the resonant frequency of the nucleus. The Fermi contact shift, δFC, in ppm, is obtained by dividing the contact coupling energy by the nuclear Zeeman energy:

formula
Equation 1.110

with

formula
Equation 1.111

where S is the total spin of the paramagnetic (TM) centre, and γI and ω0 are the gyromagnetic ratio and Larmor frequency, respectively, of the nucleus. A lower γI results in lower sensitivity, but also in smaller paramagnetic shifts. The size and sign of AFC determine the magnitude and direction, respectively, of the chemical shift observed. Through its dependence on AFC, the Fermi contact shift is proportional to ρα−β(0). Hence, δFC necessarily results from a non-zero unpaired electron spin density at the position of the nucleus, via electron delocalization and polarization mechanisms discussed in the next section.

Replacing 〈SZ〉 in eqn (1.110) by its expression in (1.99), we derive an expression for δFC in terms of the isotropic molar susceptibility, χm:

formula
Equation 1.112

The equation above indicates that paramagnetic TM ions that exhibit a stronger tendency for their electronic magnetic moments to align with the external magnetic field (large susceptibility) also give rise to larger Fermi contact shifts.11 

Equivalently, using the expression for 〈SZ〉 in eqn (1.96):

formula
Equation 1.113

The through-bond Fermi contact interaction leads to a chemical shift proportional to the magnitude of the unpaired electron spin density at the position of the nucleus of interest, ρα−β(0). For battery materials containing a paramagnetic TM ion, unpaired spin density transfer occurs from the TM d orbitals to the s orbitals of the species A of interest. This transfer occurs through direct TM(d)-A(s) orbital overlap or involves additional orbitals on bridging atoms.

An example of a local environment for nucleus A is shown in Figure 1.15. Each paramagnetic TM ion in the vicinity of the A nucleus (generally within a ≈5 Å radius from A, based on recent first principles studies34,50,51 ) transfers a finite amount of unpaired electron spin density to the s orbital on A, via a spin density transfer pathway Pi. This spin density transfer results in a finite contribution to the overall Fermi contact shift, δPi. In Figure 1.15, three distinct Pi pathways give rise to three distinct Fermi contact shift contributions.

Figure 1.15

Example local environment around an A nucleus. Each open-shell TM ion in a ≈5 Å sphere around species A transfers a finite amount of unpaired electron spin density from its d orbitals to s orbitals on A, via bridging O 2p orbitals. This in turn results in a finite contribution to the overall Fermi contact shift of A. In this case, three TM–O–A spin density transfer pathways, Pi, can be distinguished, leading to different amounts of spin density transfer and therefore to different shift contributions.

Figure 1.15

Example local environment around an A nucleus. Each open-shell TM ion in a ≈5 Å sphere around species A transfers a finite amount of unpaired electron spin density from its d orbitals to s orbitals on A, via bridging O 2p orbitals. This in turn results in a finite contribution to the overall Fermi contact shift of A. In this case, three TM–O–A spin density transfer pathways, Pi, can be distinguished, leading to different amounts of spin density transfer and therefore to different shift contributions.

Close modal

A number of solid-state NMR studies on paramagnetic materials22,50–54  have shown that individual shift contributions from nearby TM ions are additive, and that the overall isotropic Fermi contact shift observed experimentally can be obtained by summing over all individual TM⋯A bond pathway shift contributions δPi (in ppm), where the TM ion is within the first few coordination shells around A:

formula
Equation 1.114

The δPi contributions can be positive or negative. The additivity of Fermi contact shift contributions entails that a careful assignment of the features in paramagnetic NMR spectra can provide a wealth of information on the system, ranging from detailed insight into the local environments experienced by the NMR nucleus (including the number and nature of the TM ions within ≈5 Å from the nucleus) to the extent of spin density transfer.45  In practice, individual TM⋯A shift contributions are first determined experimentally or computationally on simple model compounds with well-known structures. They are then used to interpret the shifts of more complex materials.

We take the example of alkali transition metal oxides, which account for a large fraction of the cathode materials used in commercial Li-ion battery systems. In these compounds, unpaired electron spin density transfer from the TM d orbitals to the alkali (A) s orbitals involves bridging oxygen (O) p orbitals. The limiting 90° and 180° TM–O–A spin density transfer mechanisms, shown in Figure 1.16, can be rationalized using the Goodenough–Kanamori (GK) rules.55–58  These rules, originally designed to predict the sign and approximate magnitude of the 90° and 180° isotropic super exchange coupling between TM d electrons, indicate that the sign and magnitude of spin density transfers depend on the geometry (bond lengths and angles) and covalency of the relevant TM–O–A bond pathways. The sign and magnitude of spin density transfers, and therefore of the Fermi contact shift contributions, also depend on the magnetic susceptibility of the material and are sensitive to both orbital occupation and the oxidation state of the paramagnetic TM ions. As such, the Fermi contact shift not only provides detailed information on the local structure of electrode materials, but also on the redox processes taking place on charge and discharge.41,50,51,59,60  Importantly, predicting the sign and magnitude of Fermi contact shift contributions using the simple empirical model outlined here rapidly becomes difficult when TM–O–A bond pathways deviate from the limiting 90° or 180° geometries. For this reason, density functional theory-based computational approaches have been developed for the determination of accurate Pi values.51,52  These will be discussed in section 1.3.

Figure 1.16

Limiting linear (180°) and orthogonal (90°) cases for unpaired spin density transfer pathways from a half-filled TM d orbital to an s orbital on A via an intervening 2p orbital on O. Spin density transfers are depicted with blue arrows, while spin polarization interactions are depicted with black double arrows. (a) and (b) depict delocalization-type majority spin density transfers from the TM d orbital to the nucleus of species A, which are typically strong interactions. Since the TM d orbital is half-filled and the O 2p orbital is completely filled, an interaction between the eg (a) or t2g (b) electron and the O 2p orbital can only occur if minority spin density is transferred from O to the TM. This then results in majority spin density transferred to the empty s orbital on A. (c) and (d) represent weak polarization-type minority spin density transfers from the TM to A via an intervening O. In (c), the half-filled t2g TM orbital interacts with the empty eg TM orbital via exchange coupling. Exchange coupling occurs between spin densities of the same sign on two orthogonal orbitals of the same atom, and favours transfer of majority spin density from the filled O 2p orbital to the empty eg orbital. As a result, minority spin density is transferred to the A s orbital.21  In (d), exchange coupling takes place between a filled and an empty 2p orbital on the bridging O. In this case, net spin densities of the same sign on the two orthogonal O 2p orbitals result from the transfer of majority spin density from the TM eg orbital to the empty O 2p orbital, concurrently with the transfer of minority spin density from the filled O 2p orbital to the s orbital on A.

Figure 1.16

Limiting linear (180°) and orthogonal (90°) cases for unpaired spin density transfer pathways from a half-filled TM d orbital to an s orbital on A via an intervening 2p orbital on O. Spin density transfers are depicted with blue arrows, while spin polarization interactions are depicted with black double arrows. (a) and (b) depict delocalization-type majority spin density transfers from the TM d orbital to the nucleus of species A, which are typically strong interactions. Since the TM d orbital is half-filled and the O 2p orbital is completely filled, an interaction between the eg (a) or t2g (b) electron and the O 2p orbital can only occur if minority spin density is transferred from O to the TM. This then results in majority spin density transferred to the empty s orbital on A. (c) and (d) represent weak polarization-type minority spin density transfers from the TM to A via an intervening O. In (c), the half-filled t2g TM orbital interacts with the empty eg TM orbital via exchange coupling. Exchange coupling occurs between spin densities of the same sign on two orthogonal orbitals of the same atom, and favours transfer of majority spin density from the filled O 2p orbital to the empty eg orbital. As a result, minority spin density is transferred to the A s orbital.21  In (d), exchange coupling takes place between a filled and an empty 2p orbital on the bridging O. In this case, net spin densities of the same sign on the two orthogonal O 2p orbitals result from the transfer of majority spin density from the TM eg orbital to the empty O 2p orbital, concurrently with the transfer of minority spin density from the filled O 2p orbital to the s orbital on A.

Close modal

The anisotropic part of the A tensor in eqn (1.107), AD, leads to an orientation-dependent coupling interaction between nuclear and electron spin dipoles, as shown in Figure 1.14b. In spin-only systems, this interaction can be described by a CSA and leads to broadening of the NMR resonance with no shift of the resonant frequency.

The spin dipolar coupling is a through-space interaction that is proportional to the unpaired electronic spin density at position r from the nucleus, ρα−β(r), and to the isotropic magnetic susceptibility χ, as shown in eqn (1.104). Unlike the short-range (∼5 Å) Fermi contact interaction, which requires the nucleus to be within the coordination environment of the paramagnetic centre for electron density to be delocalized onto the s orbital of the nucleus, the spin dipolar coupling is longer range and nuclei exhibit a shift anisotropy even when located several 10s of Å away from the paramagnetic centre.11  Hence, the point dipole approximation, which assumes that unpaired electrons are completely localized in the form of point dipoles situated at the TM nuclear positions,50  is often applied to simplify the evaluation of the spin dipolar interaction, and electron delocalization need only be considered to compute the Fermi contact term.11  In the point dipole approximation, the Cartesian components of AD are given by

formula
Equation 1.115

where i and j are equal to x, y or, z ei and ej represent the x, y or z components of a unit vector pointing from the nuclear spin to the electron spin in a chosen coordinate system, and δij is the Kronecker delta. R is the position of the paramagnetic centre where all electrons are localized (|ψi(r)|2 = δ(rR), and δ is the Dirac delta function), hence, R is the distance between the nucleus and the paramagnetic centre. The expression above indicates that spin dipolar couplings have a distance dependence of .11  The errors in the point dipole approximation are not great at large distances from the paramagnetic centre, typically when R > 4 Å, but can lead to inaccuracies for nuclei that are closer.11,61 

Rather than using the dipolar hyperfine coupling tensor, AD, whereby

D = 〈〉.AD.
Equation 1.116

the electron–nuclear spin dipolar Hamiltonian is often expressed in terms of the dipolar coupling tensor 〈D̃en〉 (in frequency units), such that62,63 

formula
Equation 1.117

For an isotropic, spin-only TM system, eqn (1.98) and (1.101) can be used to show that . In the point dipole approximation, the dipolar tensor is defined in terms of its matrix elements by5,13 

formula
Equation 1.118

The geometric part of the dipolar coupling tensor in eqn (1.117) is represented in spherical polar coordinates as:62,63 

formula
Equation 1.119
dxx = (1 − 3sin2θ cos2φ)/R3,
dxy = (−3sin2θ cosφ sinφ)/R3,
dxz = (−3sinθcosθcosφ)/R3,
dyx = (−3sin2θcosφsinφ)/R3,
dyy = (1 − 3sin2θsin2φ)/R3,
dyz = (−3sinθcosθsinφ)/R3,
dzx = (−3sinθcosθcosφ)/R3,
dzy = (−3sinθcosθsinφ)/R3,
dzz = (1 − 3cos2θ)/R3,

where θ and φ denote the polar and azimuthal angles that the vector connecting the electron and nuclear magnetic dipoles makes with the x, y, z, axes of a Cartesian axis system.

Assuming that the nuclear spin is coupled to many electron spins which act as many independent local fields (Curie moments), the total coupling matrix D̃en can be written as the sum over all electron spins k within a chosen cutoff radius of the nucleus:

formula
Equation 1.120

provided that all tensors are expressed in a common axis system.

The total dipolar coupling Hamiltonian, D, is then:63 

formula
Equation 1.121

In the high-field approximation, and using the convention that B0 lies along the z Cartesian axis, this expression simplifies to:

formula
Equation 1.122

The equation above clearly indicates a 1 − 3cos2θ angular dependence of the electron–nuclear dipolar interaction, in a similar manner to the nuclear–nuclear dipolar interaction discussed in Section 1.1. In solution, fast molecular tumbling completely averages the D interaction to zero. In the solid state and under static conditions, the dipolar vectors are randomly oriented with respect to B0, leading to broadening of the NMR lines. Hence, to increase resolution, NMR experiments on paramagnetic solids are generally performed under MAS conditions, whereby the sample is spun at an angle of θ ≈ 54.736° from the direction of the B0 field. Since typical electron–nuclear interaction strengths are greater than currently-accessible spinning speeds, anisotropic dipolar interactions are only partially averaged by MAS, and broad paramagnetic resonances are broken into a series of spinning sidebands separated by the spinning speed.

The traceless dipolar tensor in eqn (1.120) can be transformed into its principal axis frame (PAF) by diagonalization of the final coupling matrix to obtain the eigenvalues of the shift anisotropy, δzz, δxx and δyy. Here, we use the convention |δzzδiso| ≥ |δxxδiso| ≥ |δyyδiso| to define the principal components of the tensor. The shift anisotropy, Δδ, and the asymmetry, η, are obtained from the eigenvalues as follows:63 

formula
Equation 1.123

and

formula
Equation 1.124

The sign and magnitude of Δδ depend on the site symmetry and the strength of the dipolar interaction, respectively, while η takes values between 0 and 1 and quantifies the deviation of the spectral pattern from axial symmetry. Starting from the case of axial symmetry (η = 0), an oblate-type arrangement is represented by a negative value of Δδ, and a prolate arrangement by a positive Δδ value.

We have so far assumed that the electronic magnetic moment arises from the electron spin angular momentum S. This approach is only strictly correct when the paramagnetic centres behave as spin-only ions, but can provide a satisfactory description of more complex systems, including paramagnetic battery materials with isotropic shifts dominated by the Fermi contact contribution, and shift anisotropies largely dictated by spin dipolar interactions.50,51,64,65  In such systems, the interpretation of paramagnetic NMR data is relatively straightforward, with the isotropic shift and shift anisotropy providing information on the spin transfer pathways from the unpaired electrons to the nucleus and on the geometry of the system, respectively.11 

Systems that contain TM ions exhibiting SOC effects, and in coordination environments of non-cubic symmetry, are magnetically anisotropic. This is the case for certain battery electrodes, as has been shown for LiMPO4 (M = Fe, Co)66,67  and LixV2(PO4)3.68,69  The theory by which orbital, SOC and ZFS terms affect paramagnetic NMR properties is by no means simple, and we shall give here only a superficial and pragmatic account of the subject. The reader is referred to work by Bertini et al.,21  Vaara and coworkers,70,71  Kaupp and Köhler46  and Pell et al.11  for a more detailed account of magnetically anisotropic systems.

As mentioned earlier, in the presence of both spin and orbital angular momentum, the total electronic magnetic moment operator of a TM subject to Russell-Saunders (LS) coupling, is given by = −µB(gLL̂ + ge) (eqn (1.55)), where gL = 1 and ge ≈ 2 are the g-factors for the orbital and spin angular momenta, respectively. In a crystalline solid, the orbital contribution to the magnetic moment of a TM ion has its own orientation with respect to the crystal axes. The ligand field experienced by the TM determines the crystal axis along which the orbital magnetic moment is oriented. As a result, the projection of the total magnetic moment along B0 depends on the orientation of the crystal and is therefore anisotropic.72 

For the purpose of interpreting paramagnetic NMR data, the Hamiltonian, and therefore the total magnetic moment operator, are expressed in terms of spin operators. To this effect, the magnetic moment anisotropy due to L is represented by replacing the isotropic free-electron value, ge, by a g-tensor, g, acting on , such that

= −µBg.
Equation 1.125

The difference between eqn (1.55) and (1.125) lies in the description of the spatial dependence of the magnetic moment operator. Eqn (1.55) contains , giving an intrinsic spatial dependence. In eqn (1.125), the explicit dependence on is removed and only spin operators are retained, while the spatial dependence is encoded in the g-tensor anisotropy.11 

Magnetic anisotropy can be represented by replacing the isotropic magnetic susceptibility (χ) and average magnetic moment (〈µ̂ 〉) values by χ and 〈〉 tensors (in this section, all values are given per TM ion). By analogy with eqn (1.98), 〈〉 depends on the orientation of the χ tensor with respect to the direction of the B0 field,

formula
Equation 1.126

as depicted in Figure 1.17.

Figure 1.17

Representation of the magnetic susceptibility (χ) tensor in the presence of an external magnetic field, B0. χ̃xx, χ̃yy and χ̃zz are the components of the χ tensor in its principal axis frame, i.e., in the frame of reference where the χ matrix is diagonal. For an arbitrary orientation of the crystal in the B0 field, the magnetic susceptibility is given by the length of the vector χkk.72 

Figure 1.17

Representation of the magnetic susceptibility (χ) tensor in the presence of an external magnetic field, B0. χ̃xx, χ̃yy and χ̃zz are the components of the χ tensor in its principal axis frame, i.e., in the frame of reference where the χ matrix is diagonal. For an arbitrary orientation of the crystal in the B0 field, the magnetic susceptibility is given by the length of the vector χkk.72 

Close modal

Since the Zeeman interaction energy is E = −B0., the infinitesimal change in the average energy 〈Eper paramagnetic centre resulting from an infinitesimal change in magnetic field (at constant entropy and constant volume) is . After integration, the average energy per paramagnetic centre is obtained as

formula
Equation 1.127

from which we see that the χ tensor must be symmetric. Eqn (1.126) provides a straightforward way to evaluate the components of the second-rank 〈  〉 tensor as

formula
Equation 1.128

where the χij are the components of the symmetric, second-rank χ tensor and (B0x, B0y, B0z) are the three components of the B0 vector. Eqn (1.126) clearly shows that, while 〈〉 is proportional to B0 in the case of an isotropic χ, magnetic anisotropy leads to an orientationally-dependent proportionality constant χij.

The symmetric magnetic susceptibility tensor can be rewritten as

χ = χiso1 + Δχ
Equation 1.129

where χiso is the isotropic part, and Δχ is the traceless and symmetric susceptibility anisotropy. The anisotropy is parametrized according to one of two conventions.11  The first is in terms of the axial and rhombic anisotropies, Δχax and Δχrh, which are defined in terms of the PAF components ii as

formula
Equation 1.130
Δχrh = xxyy.
Equation 1.131

The principal components are defined so that zzyyxx. Alternatively, the principal components can be defined to satisfy|zz− χiso| ≥ |xx − χiso| ≥ |yy − χiso|. The anisotropy Δχ and asymmetry ηχ parameters are given by

Δχ = zz − χiso
Equation 1.132

and

formula
Equation 1.133

If we consider a collection of paramagnetic ions, Δχ is a measure of the degree of spatial anisotropy in the tendency for their electronic magnetic moments to align with the external magnetic field.11 

In systems that are orbitally non-degenerate, magnetic anisotropy can be adequately represented by an anisotropic g-tensor, whereby the ii coefficients of the tensor, for any direction ii of the magnetic field, are solutions of the equation

ϕ|ii + geii|ϕ〉 = iiMS
Equation 1.134

where ϕ stands for the electron Zeeman eigenfunctions.41  The values of 〈〉 along the three main directions of the reference frame are then computed as

formula
Equation 1.135

where En,ii = iiµBMS,nB0 is the Zeeman energy of the nth eigenfunction, and the exponential has been approximated to first order, as appropriate in the high-temperature limit (En,iikBT). The principal values of the magnetic susceptibility (χ) tensor are:21 

formula
Equation 1.136

In cases when the Zeeman energy (En,ii) is not much smaller than the thermal energy (kBT), the first-order approximation used in eqn (1.135) is no longer valid, and different equations can be obtained by analogy with the Brillouin function (eqn (1.66)).41 

The equations above hold rigorously for orbitally non-degenerate systems.50  For multi-electron systems (), they hold as long as the magnetic spin ground state with spin quantum number S is well separated from higher energy states, i.e., thermal population of excited spin states can be neglected, and in the absence of ZFS discussed below.

In the absence of SOC but in the presence of ZFS effects, which act to split the S states in the absence of an external magnetic field, the total energy of the system equals the Zeeman energy plus an E0 ZFS term that does not depend on the applied magnetic field. The E0 term typically dominates, and Zeeman effects can be treated using perturbation theory.41  In such cases, the effect of ZFS on the magnetic susceptibility can be computed using the Van Vleck equation:73,74 

formula
Equation 1.137

The sums are taken over all excited states of the system, with energies Ei0. The anisotropic, temperature-independent Van Vleck paramagnetic susceptibility term is added to χCW, and the overall magnetic susceptibility deviates from that predicted by the Curie–Weiss law in the temperature range where ZFS effects are comparable or greater than the thermal term (kBT). While beyond the scope of this chapter, a more complete derivation of the ZFS interaction Hamiltonian can be found in works by Pell et al.11  and Abragam and Bleaney.47 

In the presence of both SOC and ZFS, the electronic magnetic moment operator and magnetic susceptibility tensor can be expressed in terms of the g and D tensors, which describe the effects of SOC and ZFS, respectively. For an ensemble of TM ions, the electron magnetic moment operator takes the form

formula
Equation 1.138

and we can approximate the susceptibility to second order in :11 

formula
Equation 1.139

This expression describes the susceptibility tensor in the high-temperature limit, where ZFS effects are small compared to the thermal term (kBT).11 

A general form of the Hamiltonian describing a paramagnetic system with magnetic anisotropy is11 

aniso = ℏγIB0.(1σorb). + µBB0.g. + 〈〉.A. + .D.
Equation 1.140

where the first two terms represent the nuclear and electron Zeeman interactions, respectively, the third term is the hyperfine contribution discussed earlier, and the fourth term describes ZFS effects, where D is the ZFS coupling tensor. The nuclear Zeeman term contains the chemical shielding tensor, σorb, which accounts for the fact that the external magnetic field causes electrons to move in such a way that they produce their own induced magnetic field, modifying the actual field experienced by the nucleus.11  In turn, σorb leads to a deviation of the nuclear resonance frequency from ω0. The electron Zeeman term contains the orbital and spin contributions to the total electron magnetic moment encountered in previous sections.

aniso includes electron spin, nuclear spin and electron–nuclear spin–spin interactions to fully describe the magnetic response of the system to the applied B0 field. The electron Zeeman and ZFS interactions do not involve nuclear spins and cannot be probed directly by NMR, yet they have a direct impact on the average electron spin angular momentum, 〈〉, thus on electron–nuclear hyperfine interactions and must be accounted for. In the case of the ZFS interaction, the .D. coupling does not affect the resonant frequency ω0 of the nucleus under observation, as it shifts all the S spin energy levels by the same amount in the same direction,11  but it can impact nuclear spin relaxation75  and therefore NMR lineshapes.

In previous sections, we have used the high-field approximation to simplify the NMR Hamiltonian and retained only those terms that commute with the unperturbed Zeeman interaction. This approximation is only valid if the Zeeman interaction is several orders of magnitude larger than the other interactions, which is not necessarily the case here as ZFS can dominate.11  Hence, the full Hamiltonian is used in the rest of this section.

While magnetic anisotropy stems from the presence of both spin and orbital angular momenta, the Hamiltonian shown above only contains spin operators and can readily be used to interpret paramagnetic NMR spectra. The process of converting the interaction Hamiltonian aniso from a form containing both and operators to one that solely depends on is described in detail by Pell et al.11  This process involves the application of first- and second-order perturbation theory to the orbital ground state of the electron, and results in non-relativistic terms, as well as relativistic terms resulting from SOC. The formalism adopted here is generally used in EPR spectroscopy, although it has been used extensively to describe hyperfine interactions in NMR,46,70,71  and involves g, , A and D tensors to describe magnetic anisotropy. Alternatively, the susceptibility formalism makes use of the susceptibility tensor χ to account for anisotropy.41,48  A comparison of these two formalisms can be found in Pell et al.'s review.11 

The g, A and D tensors can be expanded as Taylor series in terms of the fine structure constant α, which takes the value 1/137.036.11,76  The g tensor is composed of a non-relativistic (NR) and a relativistic SOC term:77 

g = gNR + gSOC
Equation 1.141

where gNR is of order O(α0) and gSOC is of order O(α2). The NR term is simply the isotropic free-electron g-factor,

gNR = ge1 + O(α0)
Equation 1.142

and the SOC term is the sum of an isotropic term and a traceless anisotropic term that is not necessarily symmetric,11 

gSOC = Δgiso1 + Δg + O(α2)
Equation 1.143

Grouping isotropic and anisotropic terms together, the overall expression for the g tensor is:

g = (ge + Δgiso)1 + Δg
Equation 1.144

Similarly, the hyperfine coupling tensor can be expanded as:78 

A = ANR + ASOC
Equation 1.145

The NR contribution,

ANR = AFC1 + AD + O(α2)
Equation 1.146

is the sum of the isotropic Fermi contact term (AFC) and of the symmetric and anisotropic electron–nuclear spin dipolar part (AD) previously introduced for magnetically isotropic systems. The SOC contribution,

ASOC = AFC,21 + AD,2 + Aas + O(α4)
Equation 1.147

contains an isotropic part AFC,2, which is referred to as the second-order Fermi contact coupling constant,11  a symmetric and anisotropic component, AD,2, denoted the second-order dipolar interaction, and an antisymmetric anisotropic contribution Aas.11  Grouping isotropic and anisotropic terms together:

A = (AFC + AFC,2)1 + AD + AD,2 + Aas.
Equation 1.148

The ZFS tensor takes the form:79 

D = DNR + DSOC
Equation 1.149

where the NR part is due to the electron spin–spin interaction, and is usually small compared to the SOC term DSOC. D is symmetric and traceless, and is equal to zero either for electronic spins S < 1, or for TMs in perfectly cubic environments.11 

As summarized in Table 1.2, only non-relativistic terms contribute in the absence of SOC, while additional terms are needed in the presence of SOC. We note that the expressions for the hyperfine interactions derived earlier for magnetically-isotropic systems did not account for the NR contribution to the ZFS, and are therefore only approximations for S > ½ systems.11 

Table 1.2

Relevant g, A and D tensor components in the case of magnetically isotropic or anisotropic systems

Magnetic isotropyg, A and D tensor components
, isotropic g = ge, A = AFC1+AD, D = 0 
, anisotropic g = ge, A = AFC1+AD, D = DNR 
Anisotropic with SOC g = (ge+Δgiso)1+Δg, A=(AFC + AFC,2)1 + AD + AD,2 + Aas, D = DNR + DSOC 
Magnetic isotropyg, A and D tensor components
, isotropic g = ge, A = AFC1+AD, D = 0 
, anisotropic g = ge, A = AFC1+AD, D = DNR 
Anisotropic with SOC g = (ge+Δgiso)1+Δg, A=(AFC + AFC,2)1 + AD + AD,2 + Aas, D = DNR + DSOC 

For magnetically isotropic systems, NR expressions for the isotropic Fermi contact shift and shift anisotropy due to electron–nuclear spin dipolar interactions can be used. While these interactions have been discussed in previous sections, the expressions provided below allow for an easy comparison with the magnetically anisotropic case that follows. In the absence of SOC or ZFS, the isotropic Fermi contact shift, δFC, in Hz, is given by11 

formula
Equation 1.150

which is similar to eqn (1.113) encountered earlier, save from the 106 factor which yields shift values in ppm. We can also define the Fermi contact shielding, σFC, such that11 

formula
Equation 1.151

noting that the only difference between the chemical shift and chemical shielding conventions is a change of sign. Hence, these two formalisms can be used interchangeably.

The anisotropy of the spin dipolar shielding interaction, σD, is parameterized in terms of the shielding anisotropy, given by11 

formula
Equation 1.152

and asymmetry

η = ηD
Equation 1.153

where ΔAD and ηD are the anisotropy and asymmetry of the non-relativistic spin dipolar interaction. The overall shielding tensor therefore has the same PAF and anisotropic properties as the non-relativistic spin dipolar coupling tensor.11 

The impact of SOC on electron–nuclear spin couplings are described below for 3d TM systems, where SOC effects are rather small. Expressions for the various hyperfine interactions are obtained using the g, A and D tensors presented earlier. For comparison, we also provide expressions for selected interactions using the susceptibility formalism. All of these terms are discussed in greater detail in the review by Pell et al.,11  which also deals with more complex ions such as lanthanides and actinides. In the presence of magnetic anisotropy, the expressions for the various isotropic and anisotropic interactions contain tensor products corresponding to cross-terms between SOC, ZFS and other interactions (hyperfine, spin–spin, electron Zeeman). In general, each tensor has its own PAF, which does not necessarily coincide with the PAF of other tensor properties. Hence, the overall anisotropy and asymmetry of the interaction under consideration are not simple combinations of the tensor contributions, but are instead obtained via matrix diagonalization.11 

In the high-temperature limit relevant to NMR experiments, the expression for the isotropic contact shift, δcon, iso, takes the form11 

formula
Equation 1.154

The equation above indicates that both terms involving ge, which includes the SOC contribution geAFC,2, as well as the SOC contribution to the Fermi contact interaction ΔgisoAFC, appear to first order in (1/kBT). Conversely, the g-anisotropy and ZFS splitting only appear to second order in .11  In the absence of ZFS, the isotropic contact shift becomes

formula
Equation 1.155

which is a simple modification of the non-relativistic expression, with changes to the isotropic g-factor and to the contact coupling interaction due to SOC.11 

In the susceptibility formalism, the isotropic contact shift depends on the isotropic part of the magnetic susceptibility tensor, χiso and on the unpaired electron spin density at the nuclear position:11 

formula
Equation 1.156

In the high-temperature limit, the anisotropic contact shielding is given by11 

formula
Equation 1.157

where Δgsym and {Δg.D}sym are the symmetric anisotropic parts of those tensors. These terms are given by11 

formula
Equation 1.158
formula
Equation 1.159

The first term in eqn (1.159) depends on the ZFS tensor, with the anisotropy and asymmetry being proportional to those of the D tensor, ΔD and ηD, respectively, and the same PAF as the ZFS. The second term depends on the anisotropic g-tensor, with an anisotropy proportional to Δg, an asymmetry parameter equal to ηg, and the same PAF as the symmetric g-tensor. The third term is more complicated, with an anisotropy, asymmetry parameter, and PAF that depend on the product of the anisotropic parts of D and g.11 

The ZFS only contributes to second order in (1/kBT), while the g-anisotropy contributes to first order. If there is no ZFS, the following expression is obtained for the contact shielding anisotropy:11 

formula
Equation 1.160

In this case, σcon,aniso depends on the anisotropic properties of the g tensor and can be parameterized in terms of the anisotropy (Δσcon,aniso) and asymmetry (ηcon) parameters

formula
Equation 1.161

and

ηcon = ηg
Equation 1.162

In the susceptibility formalism, the form of the contact shielding anisotropy is similar to that of the isotropic contact shift, and σcon,aniso depends on the susceptibility anisotropy, Δχ, and on the unpaired spin density at the nuclear position. The anisotropy and asymmetry parameters are given by:

formula
Equation 1.163

and

ηcon = ηχ
Equation 1.164

The first expression indicates that TM ions with a greater degree of magnetic anisotropy lead to larger contact shift anisotropies.11 

The isotropic pseudo-contact shift (PCS) arises from the electron–nuclear spin dipolar interaction and has a very different form to that of the contact shift. Mathematically, it results from the matrix product of the spin dipolar tensor (AD or AD,2) with at least one other anisotropic and symmetric tensor.11  In the high-temperature limit, the following expression is obtained for the PCS:

formula
Equation 1.165

The first term depends on the orientation of the PAF of the non-relativistic spin dipolar tensor relative to the PAF of the symmetric part of the g-tensor. The second and fourth terms depend on the orientation of the PAF of the non-relativistic spin dipolar tensor relative to the PAF of the ZFS tensor. The third term is similar, but depends on the orientation of the PAF of the SOC spin dipolar tensor relative to the PAF of the ZFS tensor. The fifth term depends on the orientation of the non-relativistic spin dipolar PAF relative to the PAF of the symmetric part of the product Δg·D. The g-anisotropy, Δg, is present in terms with both a first- and second-order temperature dependence (1/kBT), whilst the ZFS only affects the PCS to second order.11 

In the absence of ZFS, the PCS takes a simple form:11 

formula
Equation 1.166

In the susceptibility formalism, a simple expression for δPCS, iso can be obtained by using the fact that, at TM ion-nucleus distances greater than 4 Å, the point dipole approximation can be employed. The PCS then takes the form11 

formula
Equation 1.167

where θ and ϕ are the polar and azimuthal angles relating the PAF of the susceptibility tensor relative to the TM ion-nucleus vector.11  The expression above indicates that, similarly to the contact shift, the PCS depends on the interaction with the unpaired spin density. Yet, while the contact shift results from the spin density at the nuclear site, the PCS is a longer range effect that depends on the spatial position of the nucleus with respect to the paramagnetic centre. Eqn (1.167) also shows that the PCS depends only on the susceptibility anisotropy, hence it only arises for TM ions subject to SOC.11 

The anisotropic shielding tensor that arises from the spin dipolar interaction, in the high temperature approximation, takes the form11 

formula
Equation 1.168

The first four terms depend on the dipolar coupling tensors and g-anisotropy, and have a temperature dependence of (1/kBT). The other terms, which also depend on the ZFS tensor, exhibit a second order dependence on (1/kBT).11  In the absence of ZFS, σD,sym becomes:

formula
Equation 1.169

The antisymmetric hyperfine shift anisotropy, σas,sym, does not contribute to the isotropic paramagnetic shift but contributes to the shielding anisotropy. In the high-temperature limit,11 

formula
Equation 1.170

The temperature dependence is second order in (1/kBT). The anisotropic properties depend on the tensor product D·Aas, while the g-anisotropy does not play a role. Hence, in the absence of ZFS, this term disappears altogether.11 

The main differences between NMR spectra collected on spin-only and spin–orbit coupled paramagnetic systems are illustrated in Figure 1.18 below. For magnetically isotropic TM systems, the through-bond Fermi contact interaction is purely isotropic and leads to a shift δFC of the resonant frequency of the nucleus under consideration away from the reference chemical shift, conventionally set at 0 ppm. Concurrently, the through-space electron–nuclear spin dipolar interaction is purely anisotropic and leads to a shift anisotropy. In the presence of SOC, both the Fermi contact and electron–nuclear spin dipolar interaction have an isotropic and anisotropic component. The chemical shift due to isotropic contact interactions is δcon,iso, and the shift due to the isotropic spin dipolar coupling is the pseudo-contact shift δPCS,iso.

Figure 1.18

Effects of contact (through-bond) and electron–nuclear spin dipolar interactions (through-space) on the NMR signal obtained for (a) a magnetically isotropic and (b) a spin–orbit coupled paramagnetic system. Simulated solid-state NMR (powder) spectra are depicted under static conditions. Experimental spectra for the LiMnPO4 and LiFePO4 Li-ion cathodes were collected at 11.75 T at 60 kHz MAS using a double adiabatic spin echo sequence. For LiMnPO4 with an isotropic magnetic susceptibility, the chemical shift is largely determined by the Fermi contact shift. For LiFePO4 with an anisotropic susceptibility tensor, the shift likely contains contributions due to both the Fermi contact and pseudo-contact mechanisms, of which the former is expected to dominate.54  Adapted from ref. 54 with permission from American Chemical Society, Copyright 2012.

Figure 1.18

Effects of contact (through-bond) and electron–nuclear spin dipolar interactions (through-space) on the NMR signal obtained for (a) a magnetically isotropic and (b) a spin–orbit coupled paramagnetic system. Simulated solid-state NMR (powder) spectra are depicted under static conditions. Experimental spectra for the LiMnPO4 and LiFePO4 Li-ion cathodes were collected at 11.75 T at 60 kHz MAS using a double adiabatic spin echo sequence. For LiMnPO4 with an isotropic magnetic susceptibility, the chemical shift is largely determined by the Fermi contact shift. For LiFePO4 with an anisotropic susceptibility tensor, the shift likely contains contributions due to both the Fermi contact and pseudo-contact mechanisms, of which the former is expected to dominate.54  Adapted from ref. 54 with permission from American Chemical Society, Copyright 2012.

Close modal

The presence of magnetic anisotropy leads to complex expressions for hyperfine interactions. Hence, it is helpful to summarize the impact of the g, A and D tensor properties on the form of the isotropic shift and shift anisotropy, and to review the temperature dependence and size of the various contributions for different systems.11 

In systems where SOC effects are negligible, the isotropic shift stems from the interplay between purely isotropic g, ZFS and hyperfine terms through the product geDisoAFC. The shift depends on contact interactions and therefore on through-bond transfer of unpaired electron spin density onto the nucleus under observation. The shift anisotropy is given by geDisoAD, and is similarly dependent on isotropic g and ZFS terms, but it results from anisotropic through-space coupling of nuclear and electron spins. These interactions have a (1/kBT) temperature dependence.11 

SOC mixes thermally-populated electronic energy levels, adding contributions to the g, hyperfine and ZFS tensors. In this case, through-space electron–nuclear spin dipolar terms can contribute to the isotropic hyperfine chemical shift (e.g., through the PCS), while through-bond contact terms contribute to the chemical shift anisotropy. The isotropic paramagnetic shift is the sum of the contact shift and PCS. These two interactions provide complementary information on the through-bond and through-space interactions with the unpaired electrons, respectively.11  In situations where there is good overlap between TM d orbitals and s orbitals on the nucleus of interest, a significant amount of unpaired spin density may be transferred to the nucleus and the contact shift is likely to dominate the overall isotropic shift. On the other hand, in situations where orbital overlap is less favourable, or if the nucleus is outside of the immediate (∼5 Å) coordination environment of the TM ion, the contact shift becomes very small and the PCS is more important. Finally, in the high-temperature limit, the effect of the ZFS tensor on hyperfine interactions exhibits a 1/(kBT)2 temperature dependence. Hence, for TM ions where the ZFS is small, its contribution can be neglected.11 

Complex interactions reflected in NMR spectroscopy of paramagnetic ions can provide rich structural information of the solids, while they also increase the difficulty of the spectral analysis, as the two sides of the coin. Calculation of NMR shifts with first-principles method is very helpful for spectral assignments. In this section, the paramagnetic NMR (PNMR) theory for paramagnetic molecules is presented briefly, then the corresponding implementation methods in the calculation of chemical shifts for paramagnetic solids are induced briefly though a series of phosphate battery materials. More details on the calculation of chemical shifts in paramagnetic solids can be found in some excellent review papers and books.11,46,76 

The PNMR theory is developed and implemented for paramagnetic molecules at the beginning, particularly in transition metal coordination compounds.70,73,81  The most promising rechargeable battery cathode materials usually contain redox-active transition metal ions, thus it attracts abundant NMR study. Nevertheless, state-of-the-art formalism used for paramagnetic solid is more or less a straightforward modification from the theory developed for simple molecules.67  Herein, a brief introduction for the developing of paramagnetic theory for molecules is presented first, although the paramagnetic molecules are not the main interest of NMR study of energy storage materials.

In the analysis of PNMR spectra, PNMR chemical shift is usually decomposed into three parts: the orbital shift δorb, the Fermi contact shift δFC, and pseudo-contact shift δPC:

δ = δorb + δFC + δPC
Equation 1.171

The orbital shift is taken as usual NMR chemical shift in the diamagnetic system, this term is approximately temperature-independent and equal to the chemical shift in a same diamagnetic environment. The orbital shift arises from the induced current of paired electrons in the external magnetic field. According to the Biot–Savart law, a current density j(r) induces a magnetic field at position s:

formula
Equation 1.172

where c is the speed of light in a vacuum.

The δFC comes from interaction between the nuclear magnetic moment and the average spin density at the location of the nucleus. In the simplest case, it is given by:82 

formula
Equation 1.173

µB and µN are the Bohr and nuclear magnetons, respectively. S is the spin quantum number, AisoI is the isotropic hyperfine coupling constant that is determined by spin density calculated though density functional theory (DFT) methods, g is the rotationally averaged electronic g-value, gNI is the nuclear g-value, kB is the Boltzmann constant and T is the absolute temperature. The long-range dipolar interaction between the nuclear magnetic moment and the induced magnetic moment at the paramagnetic centre gives the δPC. In the simplest form, it is given by:83 

formula
Equation 1.174

Here, if paramagnetic centre is taken as coordinate origin, Ω is the angle between its principal symmetry axis and the direction to the observed nucleus. R is the distance between the induced magnetic moment and the nucleus. F(g) is an algebraic function of the g-tensor values, which subsumes the relative magnitudes of various relaxation time. There is an assumption that R is large enough, thus the paramagnetic centre can be treated as a point dipole.

There are also other assumptions73  in the derivation of eqn (1.173) and (1.174) that still restrict their wide application: (1) The paramagnetic molecule has only one thermally populated energy level in the absence of the magnetic field, and the states of this ground level could be assigned a spin quantum number, S; (2) Splitting of the ground level in the absence of the field (zero field splitting, ZFS) could be ignored when S ≥ 1; (3) the orbital contribution to the isotropic NMR shift could be taken into account indirectly through the use of g-tensor components in eqn (1.173) and (1.174). This assumption requires that there be at most only a first-order orbital contribution to the magnetic moment of paramagnetic centre, brought about by spin–orbital mixing of the ground level with nonpopulated excited states. However, abundant transition metal complexes do not satisfy all those assumptions. Therefore, Kurland and McGarvey develop an approximate density matrix treatment to handle these situations.12  It is a milestone of paramagnetic NMR theory, involving both spin–orbit coupling (SOC) and ZFS effect in the formalism, while the formalism is not accessible though first-principles computational properties.

Moon and Patchkovskii76  derived the formalism of PNMR shift for molecules having only one unpaired electron (doublet electronic state). The main difference from the simplest case is that the hyperfine shielding is expressed by the matrix product of g and A tensors, rather than a rotationally averaged electronic g-value. It is given by:

formula
Equation 1.175

Both the A and g tensor could be expressed via the isotropic and anisotropic parts:

A = AFC1 + Adip
Equation 1.176
g = (ge + giso)1 + Δ
Equation 1.177

Then the total isotropic part is given by:

formula
Equation 1.178

The σorb,isoI, geAFCI 1 + ΔgisoAFCI1 and ΔAdipI in brackets correspond to orbital shift, Fermi contact shift and pseudo-contact shift, respectively. Moreover, ΔgisoAFCI1 corresponds to an orbital magnetic moment induced δFC in this way the spin–orbit effect is included. This formalism should also satisfy for the molecules with small ZFS effect.

Vaara70  derived a complete and DFT methods accessible PNMR shift representation for paramagnetic molecules, which not only satisfies the high-temperature region but also the low temperature region by including the ZFS effect correctly. It is given by:

formula
Equation 1.179

graphic
graphic

SS〉 is a spin dyadic with the components 〈SεSτ〉 evaluated in the manifold of eigenstates |q〉 with eigenenergies Eq of the ZFS Hamiltonian, HZFS. The off-diagonal elements of the symmetric matrix Qpq incorporate magnetic couplings between the eigenstates of the ZFS Hamiltonian, which is necessary for the correct behavior when going to low temperatures.

The computation of the NMR shift for a paramagnetic solid is a greater challenge than paramagnetic molecules. It is still in an early stage, the main difficulty for the complete calculations comes from both theoretical and empirical implemental aspects. In contrast to the neglectable interaction between intermolecular paramagnetic centres in molecule systems, the solid multi-paramagnetic centres are closely correlated, usually leading to a Curie–Wiess magnetic susceptibility in the room temperature region. The interaction between them is incorporated by introducing the Wiess constant in the state-of-art formalism, which is a quite straightforward modification of Vaara's formalism. The lack of periodic code and high computational resource demanded for g-tensor, zero-field-splitting tensor and exchange coupling constant between paramagnetic centres, are their limitations. A simple classification of current implementation in paramagnetic battery materials according to shift mechanism is shown in Figure 1.19.

Figure 1.19

Classification for current PNMR calculations in battery materials.

Figure 1.19

Classification for current PNMR calculations in battery materials.

Close modal

A careful comparison of different implementations for the PNMR shift of a series lithium transition metal phosphate shows the importance of including the spin–orbit and ZFS effect. The first direct δFC calculation for lithium-ion batteries (LIBs) cathode materials, which are typical Fe(iii) phosphates, were carried out by Kim et al..50  The formula is given by:

formula
Equation 1.180

µeff is the effective spin moment, Θ is the Weiss constant, both are obtained in the magnetic susceptibility measurement. Comparing with eqn (1.173), S(S + 1) and T are replaced by µeff2 and (TΘ), thus including the effect of SOC and residual exchange couplings respectively. The computational hyperfine constants are implemented in CRYSTAL with B3LYP based hybrid functional, which is based on all electron linear combinations of the atomic orbitals (LCAO) method. It turns out that the δFC is highly dependent on Hartree–Fock (HF) exchange contents, which determine the localized extent of the computational electronic state (Figure 1.20). Moreover, using XRD (SPE) or DFT optimized structures for δFC calculation yields remarkable shift differences, which indicates that the hyperfine constant is highly sensitive to atomic positions. The XRD can usually get cell parameters precisely, whereas the small local structure distortions are sometime missing refinement, thus it is necessary to do the cross check upon different geometry schemes. Nevertheless, the calculated δFC and experimental results are generally consistent, especially for the 31P shift. This shows that the δFC dominates the PNMR shift. A possible reason could be the small SOC and ZFS effect of Fe(iii) 3d-orbital, which is a high-spin state because of the symmetric half filled 3d orbital in the tetrahedral field, leading to a small δPC. It is a clever choice for the purpose of comparing with the experimental shift at the very beginning.

Figure 1.20

Correlations of the calculated 31P (a) and 7Li (b) hyperfine shifts of Fe(iii) phosphates for various exchange-correlation functional with the experimental data. Solid lines denote linear fits: (dashed line) ideal trend; (black circles) 0% SPE; (blue squares) 20% SPE; (red triangles) 35% SPE; (green diamonds) 20% OPT. Adapted from ref. 50 with permission from American Chemical Society, Copyright 2010.

Figure 1.20

Correlations of the calculated 31P (a) and 7Li (b) hyperfine shifts of Fe(iii) phosphates for various exchange-correlation functional with the experimental data. Solid lines denote linear fits: (dashed line) ideal trend; (black circles) 0% SPE; (blue squares) 20% SPE; (red triangles) 35% SPE; (green diamonds) 20% OPT. Adapted from ref. 50 with permission from American Chemical Society, Copyright 2010.

Close modal

Boucher et al.83  has calculated the δFC of a series of LiMPO4 (M = Mn, Fe, Co, Ni) using the same formalism as Kim et al. The hyperfine field calculation is implemented in full-potential linearized augmented plane wave (FP-LAPW) code, WIEN2k. The δFC are tested with PBE, DFT + U and PBE based on-site hybrid functional. Comparing with experiment shift, the on-site hybrid functional applied on TM and oxygen gives the best agreement for TM = Mn, Fe, Ni cases, however the LiCoPO4 shift is a significant overestimate (Figure 1.21).

Figure 1.21

Comparison of experimental FC shifts of Li in olivine LiMPO4 (M = Mn, Fe, Co, and Ni) with the computational results obtained with different approaches (T = 310 K). The effective potential (Ueff) in GGA + U is given in Ry unit, while the hybrid functionals (HyF1 of HyF2) mixing parameters are reported: a = 0.35 (Mn, Co, and Ni); a = 0.5 (Fe). Reproduced from ref. 83 with permission from American Chemical Society, Copyright 2012.

Figure 1.21

Comparison of experimental FC shifts of Li in olivine LiMPO4 (M = Mn, Fe, Co, and Ni) with the computational results obtained with different approaches (T = 310 K). The effective potential (Ueff) in GGA + U is given in Ry unit, while the hybrid functionals (HyF1 of HyF2) mixing parameters are reported: a = 0.35 (Mn, Co, and Ni); a = 0.5 (Fe). Reproduced from ref. 83 with permission from American Chemical Society, Copyright 2012.

Close modal

It suggests that the limitation of neglecting δPC has emerged.

Pigliapochi et al.66  recalculate the PNMR shift of LiMPO4 (M = Mn, Fe, Co, Ni). Again, the residual exchange coupling is included using the experimental Weiss constant, as shown in eqn (1.181).The formalism is modified from the doublet electronic state that also satisfies the small ZFS cases. The g-tensor calculation is implemented in QUANTUM EXPRESSO with PAW and pseudopotential methods. It is the first time that the SOC effect for cathode materials NMR shifts by first-principles EPR parameters rather than the experimental effective magnetic moment has been incorporated. The computational 7Li shift of LiCoPO4 is still an overestimate, leading to a 7Li shift sequence: δLiMnPO4 > δLiFePO4 > δLiCoPO4 > δLiNiPO4, while it disagrees with the experimental 7Li shift results: δLiMnPO4 > δLiFePO4 > δLiNiPO4 > δLiCoPO4. The existing qualitative error suggests that the ZFS effect plays an important role in the PNMR shift of LiCoPO4.

formula
Equation 1.181

Kaupp15  included the ZFS effect using the following formula:

formula
Equation 1.182

It is still a straightforward modification from Vaara's formalism because the residual exchange coupling is incorporated by the experimental Weiss constant. Comparing with molecule systems, another difference is that paramagnetic solids contain multiple paramagnetic centres. Only the hyperfine interaction between the first neighborhood paramagnetic centre and the observed nucleus are involved in the formula, n are the first neighborhood numbers, as shown in Figure 1.22.

Figure 1.22

Visual representation of the influence on the shielding of nucleus I by the surrounding paramagnetic centres (a, b, c, d, e, f). Reproduced from ref. 80 with permission from American Chemical Society, Copyright 2019.

Figure 1.22

Visual representation of the influence on the shielding of nucleus I by the surrounding paramagnetic centres (a, b, c, d, e, f). Reproduced from ref. 80 with permission from American Chemical Society, Copyright 2019.

Close modal

Periodic GAPW based code cp2k is used for the hyperfine tensor, while the g- and D-tensor are implemented in the non-periodic ORCA program with the single paramagnetic centre incremental cluster model and high level post-HF method, multi-configurational self-consistent field (MCSCF), N-electron valence state perturbation theory (NEVPT) methods.68,80  The reason for this is that the g-tensor and D-tensor are associated with electronic excited states, which is still a challenge for DFT methods. When the relativistic effect is neglectable, the paramagnetic part of the NMR shift is composed of six terms with different physical meanings:

σ = σorbσgeSSAFCσΔgisoSSAFCσΔg̃̃SSAFCσ̃SSAFCσgeSSAdipσΔgisoSSAdip
Equation 1.183

The results in Figure 1.23 show that the 7Li shift of LiCoPO4 is dominated by negative δPC, which is neglected83  or underestimated66  in the previous work. Finally, qualitatively correct computational results of the series of transition metal phosphate are obtained. However, comparing with the situation for molecules, this approach still has some shortcomings: (1) The previously suggested use of the experimental Weiss constant to incorporate magnetic couplings within the Curie–Weiss temperature regime into the computations seems to work well but obviously introduces a semi-empirical aspect and a potential double-counting of single-ion ZFS. This could be done by combining quantum-chemical computations of exchange-coupling constants with statistical treatments. (2) The cluster model employed for the g- and D-tensor computations clearly still had deficiencies that should be improved using periodic approaches. (3) Spin–orbit effects on hyperfine couplings could not yet be computed properly for solids. (4) Use of beyond-DFT approaches or of improved density functionals would be desirable in the computation of hyperfine couplings, but this is also still a problem for the molecules.

Figure 1.23

Analyses of 7Li chemical shifts for LiMPO4 (M = Mn, Fe, Co, Ni) computed for both XRD and optimized (OPT) structures when using 25% EXX admixture (PBE0, PBE25) in the computation of the HFC tensors. g-Tensors and D-tensors obtained at the NEVPT2 level and orbital shielding at the PBE level. Shielding converted to shifts relative to aq. lithium chloride (LiCl). Reproduced from ref. 80 with permission from American Chemical Society, Copyright 2019.

Figure 1.23

Analyses of 7Li chemical shifts for LiMPO4 (M = Mn, Fe, Co, Ni) computed for both XRD and optimized (OPT) structures when using 25% EXX admixture (PBE0, PBE25) in the computation of the HFC tensors. g-Tensors and D-tensors obtained at the NEVPT2 level and orbital shielding at the PBE level. Shielding converted to shifts relative to aq. lithium chloride (LiCl). Reproduced from ref. 80 with permission from American Chemical Society, Copyright 2019.

Close modal

In a short conclusion, computational methods provide a big opportunity to simulate and explanations of the experimental NMR spectra, however, we should be clear whether the utilized computational methods are suitable or not in calculating the PNMR shift of the target systems, this requires not only an understanding of the limitation of the methods, but also obtaining the available magnetic and electronic information of the materials.

1

This author wrote Section 1.1.

2

This author wrote Section 1.2.

3

Those authors wrote Section 1.3.

1.
Bloch
 
F.
Hansen
 
W. W.
Packard
 
M.
Phys. Rev.
1946
, vol. 
69
 pg. 
127
 
2.
Purcell
 
E. M.
Torrey
 
H. C.
Pound
 
R. V.
Phys. Rev.
1946
, vol. 
69
 pg. 
37
 
3.
Ernst
 
R. R.
Anderson
 
W. A.
Rev. Sci. Instrum.
1966
, vol. 
37
 pg. 
93
 
4.
Muller
 
L.
Kumar
 
A.
Ernst
 
R. R.
J. Chem. Phys.
1975
, vol. 
63
 pg. 
5490
 
5.
M.
Mehring
,
Principles of High Resolution NMR in Solids
,
Springer-Verlag
,
Berlin
, 2nd edn,
1983
6.
A.
Abragam
,
The Principles of Nuclear Magnetism
,
Oxford University Press
,
New York
,
1994
7.
C. P.
Slichter
,
Principles of Magnetic Resonance
,
Springer
,
New York
,
1996
8.
M. J.
Duer
,
Solid State NMR Spectroscopy: Principles and Applications
,
Wiley-Blackwell Science Inc.
,
Malden
, 1st edn,
2002
9.
K. J. D.
MacKenzie
and
M. E.
Smith
,
Multinuclear Solid-state Nuclear Magnetic Resonance of Inorganic Materials
,
Elsevier Science Ltd.
,
Kidlington
, 1st edn,
2002
10.
M. H.
Levitt
,
Spin Dynamics: Basics of Nuclear Magnetic Resonance
,
John Wiley & Sons Inc.
,
Chichester
, 2nd edn,
2008
11.
Pell
 
A. J.
Pintacuda
 
G.
Grey
 
C. P.
Prog. Nucl. Magn. Reson. Spectrosc.
2019
, vol. 
111
 pg. 
1
 
12.
Saitô
 
H.
Ando
 
I.
Ramamoorthy
 
A.
Prog. Nucl. Magn. Reson. Spectrosc.
2010
, vol. 
57
 pg. 
181
 
13.
U.
Haeberlen
,
High Resolution NMR in Solids: Selective Averaging
,
Academic Press
,
New York
,
1976
14.
Andrew
 
E. R.
Bradbury
 
A.
Eades
 
R. G.
Nature
1958
, vol. 
182
 pg. 
1659
 
15.
Andrew
 
E. R.
Hinshaw
 
W. S.
Tiffen
 
R. S.
J. Magn. Reson.
1974
, vol. 
15
 pg. 
191
 
16.
Pake
 
G. E. J.
Chem. Phys.
1948
, vol. 
16
 pg. 
327
 
17.
Rothwell
 
W. P.
Shen
 
W. X.
Lunsford
 
J. H.
J. Am. Chem. Soc.
1984
, vol. 
106
 pg. 
2452
 
18.
Samoson
 
A.
Chem. Phys. Lett.
1985
, vol. 
119
 pg. 
29
 
19.
Lippmaa
 
E.
Samoson
 
A.
Magi
 
M.
J. Am. Chem. Soc.
1986
, vol. 
108
 pg. 
1730
 
20.
Gan
 
Z.
Gor'kov
 
P.
Cross
 
T. A.
Samoson
 
A.
Massiot
 
D.
J. Am. Chem. Soc.
2002
, vol. 
124
 pg. 
5634
 
21.
Grey
 
C. P.
Dupré
 
N.
Chem. Rev.
2004
, vol. 
104
 pg. 
4493
 
22.
Lee
 
Y. J.
Wang
 
F.
Grey
 
C. P.
J. Am. Chem. Soc.
1998
, vol. 
120
 pg. 
12601
 
23.
Bhattacharyya
 
R.
Key
 
B.
Chen
 
H.
Best
 
A. S.
Hollenkamp
 
A. F.
Grey
 
C. P.
Nat. Mater.
2010
, vol. 
9
 pg. 
504
 
24.
Blanc
 
F.
Leskes
 
M.
Grey
 
C. P.
Acc. Chem. Res.
2013
, vol. 
46
 pg. 
1952
 
25.
Ogata
 
K.
Salager
 
E.
Kerr
 
C. J.
Fraser
 
A. E.
Ducati
 
C.
Morris
 
A. J.
Hofmann
 
S.
Grey
 
C. P.
Nat. Commun.
2014
, vol. 
5
 pg. 
3217
 
26.
Zhou
 
L.
Leskes
 
M.
Liu
 
T.
Grey
 
C. P.
Angew. Chem., Int. Ed.
2015
, vol. 
54
 pg. 
14782
 
27.
Bayley
 
P. M.
Trease
 
N. M.
Grey
 
C. P.
J. Am. Chem. Soc.
2016
, vol. 
138
 pg. 
1955
 
28.
Kayser
 
S. A.
Mester
 
A.
Mertens
 
A.
Jakes
 
P.
Eichel
 
R.-A.
Granwehr
 
J. J.
Phys. Chem. Chem. Phys.
2018
, vol. 
20
 pg. 
13765
 
29.
Kitada
 
K.
Pecher
 
O.
Magusin
 
P. C. M. M.
Groh
 
M. F.
Weatherup
 
R. S.
Grey
 
C. P.
J. Am. Chem. Soc.
2019
, vol. 
141
 pg. 
7014
 
30.
Leskes
 
M.
Drewett
 
N. E.
Hardwick
 
L. J.
Bruce
 
P. G.
Goward
 
G. R.
Grey
 
C. P.
Angew. Chem., Int. Ed.
2012
, vol. 
124
 pg. 
8688
 
31.
Leskes
 
M.
Moore
 
A. J.
Goward
 
G. R.
Grey
 
C. P.
J. Phys. Chem. C
2013
, vol. 
117
 pg. 
26929
 
32.
Salager
 
E.
Sarou-Kanian
 
V.
Sathiya
 
M.
Tang
 
M.
Leriche
 
J.-B.
Melin
 
P.
Wang
 
Z.
Vezin
 
H.
Bessada
 
C.
Deschamps
 
M.
Tarascon
 
J.-M.
Chem. Mater.
2014
, vol. 
26
 pg. 
7009
 
33.
Chen
 
R.
Witte
 
R.
Heinzmann
 
R.
Ren
 
S.
Mangold
 
S.
Hahn
 
H.
Hempelmann
 
R.
Ehrenberg
 
H.
Indris
 
S.
Phys. Chem. Chem. Phys.
2016
, vol. 
18
 pg. 
7695
 
34.
Clément
 
R. J.
Middlemiss
 
D. S.
Seymour
 
I. D.
Ilott
 
A. J.
Grey
 
C. P.
Chem. Mater.
2016
, vol. 
28
 pg. 
8228
 
35.
Sugden
 
S.
J. Chem. Soc.
1943
pg. 
328
 
36.
S.
Blundell
,
Magnetism in Condensed Matter
,
Oxford University Press
,
Oxford
,
2001
37.
Griffith
 
J. S.
Orgel
 
L. E.
Q. Rev., Chem. Soc.
1957
, vol. 
11
 pg. 
381
 
38.
S.
Kettle
,
Physical Inorganic Chemistry: A Coordination Chemistry Approach
,
Springer, Oxford University Press
,
Oxford
,
1998
39.
P.
Atkins
,
T.
Overton
,
J.
Rourke
,
M.
Weller
,
F.
Armstrong
and
M.
Hagerman
,
Shriver and Atkins' Inorganic Chemistry
,
W. H. Freeman and Company
,
New York
, 5th edn,
2010
40.
R. L.
Carlin
,
Magnetochemistry
,
Springer-Verlag
,
Berlin Heidelberg
,
1986
41.
Bertini
 
I.
Luchinat
 
C.
Parigi
 
G.
Prog. Nucl. Magn. Reson. Spectrosc.
2002
, vol. 
40
 pg. 
249
 
42.
B. I.
Bleaney
and
B.
Bleaney
,
Electricity and Magnetism
,
Oxford University Press
,
Oxford
, 3rd edn,
2013
43.
F. A.
Cotton
and
G.
Wilkinson
,
Advanced Inorganic Chemistry
,
John Wiley & Sons
,
New York
, 5th edn,
1988
44.
S.
Fraga
,
J.
Karwowski
and
K. M. S.
Saxena
,
Handbook of Atomic Data
,
Elsevier Scientific
,
Amsterdam, Oxford and New York
,
1976
45.
Pell
 
A. J.
Pintacuda
 
G.
Prog. Nucl. Magn. Reson. Spectrosc.
2015
, vol. 
84–85
 pg. 
33
 
46.
Kaupp
 
M.
Köhler
 
F. H.
Coord. Chem. Rev.
2009
, vol. 
253
 pg. 
2376
 
47.
A.
Abragam
and
B.
Bleaney
,
Electron Paramagnetic Resonance of Transition Ions
,
Oxford University Press
,
Oxford
,
1970
48.
I.
Bertini
,
C.
Luchinat
and
G.
Parigi
,
Solution NMR of Paramagnetic Molecules
,
Elsevier Science
,
Amsterdam, Oxford and New York
, 2nd edn,
2016
,
vol. 2
49.
Bertini
 
I.
Luchinat
 
C.
Coord. Chem. Rev.
1996
, vol. 
150
 pg. 
77
 
50.
Kim
 
J.
Middlemiss
 
D. S.
Chernova
 
N. A.
Zhu
 
B. Y. X.
Masquelier
 
C.
Grey
 
C. P.
J. Am. Chem. Soc.
2010
, vol. 
132
 pg. 
16825
 
51.
Middlemiss
 
D. S.
Ilott
 
A. J.
Clément
 
R. J.
Strobridge
 
F. C.
Grey
 
C. P.
Chem. Mater.
2013
, vol. 
25
 pg. 
1723
 
52.
Cheetham
 
A. K.
Dobson
 
C. M.
Grey
 
C. P.
Jakeman
 
R. J. B.
Nature
1987
, vol. 
328
 pg. 
706
 
53.
Grey
 
C. P.
Lee
 
Y. J.
Solid State Sci.
2003
, vol. 
5
 pg. 
883
 
54.
Clément
 
R. J.
Pell
 
A. J.
Middlemiss
 
D. S.
Strobridge
 
F. C.
Miller
 
J. K.
Whittingham
 
M. S.
Emsley
 
L.
Grey
 
C. P.
Pintacuda
 
G.
J. Am. Chem. Soc.
2012
, vol. 
134
 pg. 
17178
 
55.
Goodenough
 
J. B.
Phys. Rev.
1955
, vol. 
100
 pg. 
564
 
56.
Goodenough
 
J. B.
J. Phys. Chem. Solids
1958
, vol. 
6
 pg. 
287
 
57.
Kanamori
 
J.
J. Phys. Chem. Solids
1959
, vol. 
10
 pg. 
87
 
58.
J. B.
Goodenough
,
Magnetism and the Chemical Bond
,
Interscience-Wiley
,
New York
,
1963
59.
Carlier
 
D.
Ménétrier
 
M.
Grey
 
C. P.
Delmas
 
C.
Ceder
 
G.
Phys. Rev. B
2003
, vol. 
67
 pg. 
174103
 
60.
Kim
 
J.
Nielsen
 
U. G.
Grey
 
C. P.
J. Am. Chem. Soc.
2008
, vol. 
130
 pg. 
1285
 
61.
Autschbach
 
J.
Patchkovskii
 
S.
Pritchard
 
B.
J. Chem. Theory Comput.
2011
, vol. 
7
 pg. 
2175
 
62.
Nayeem
 
A.
Yesinowski
 
J. P.
J. Chem. Phys.
1998
, vol. 
89
 pg. 
4600
 
63.
Lee
 
Y. J.
Grey
 
C. P.
J. Phys. Chem. B
2002
, vol. 
106
 pg. 
3576
 
64.
Seymour
 
I. D.
Middlemiss
 
D. S.
Halat
 
D. M.
Trease
 
N. M.
Pell
 
A. J.
Grey
 
C. P.
J. Am. Chem. Soc.
2016
, vol. 
138
 pg. 
9405
 
65.
Trease
 
N. M.
Seymour
 
I. D.
Radin
 
M. D.
Liu
 
H.
Liu
 
H.
Hy
 
S.
Chernova
 
N.
Parick
 
P.
Devaraj
 
A.
Wiaderek
 
K. M.
Chupas
 
P. J.
Chapman
 
K. W.
Whitthingham
 
M. S.
Meng
 
Y. S.
Van der Ven
 
A.
Grey
 
C. P.
Chem. Mater.
2016
, vol. 
28
 pg. 
8170
 
66.
Pigliapochi
 
R.
Pell
 
A. J.
Seymour
 
I. D.
Grey
 
C. P.
Ceresoli
 
D.
Kaupp
 
M.
Phys. Rev. B
2017
, vol. 
95
 pg. 
054412
 
67.
Mondal
 
A.
Kaupp
 
M.
Phys
 
J.
Chem. Lett.
2018
, vol. 
9
 pg. 
1480
 
68.
Mondal
 
A.
Gaultois
 
M. W.
Pell
 
A. J.
Iannuzzi
 
M.
Grey
 
C. P.
Hutter
 
J.
Kaupp
 
M.
J. Chem. Theory Comput.
2018
, vol. 
14
 pg. 
377
 
69.
Mondal
 
A.
Kaupp
 
M.
Solid State Nucl. Magn. Reson.
2019
, vol. 
101
 pg. 
89
 
70.
Vaara
 
J.
Rouf
 
S. A.
Mareš
 
J.
J. Chem. Theory Comput.
2015
, vol. 
11
 pg. 
4840
 
71.
Pennanen
 
T. O.
Vaara
 
J.
Phys. Rev. Lett.
2008
, vol. 
100
 pg. 
133002
 
72.
I.
Bertini
and
C.
Luchinat
,
Electron-nucleus Interactions and Their Biophysical Consequences
,
Biophysical Society
,
1998
73.
Kurland
 
R. J.
McGarvey
 
B. R.
J. Magn. Reson.
1970
, vol. 
2
 pg. 
286
 
74.
J. P.
Jesson
,
The paramagnetic shift
, in
NMR of Paramagnetic Molecules
,
Academic Press
,
New York
,
1973
75.
Abernathy
 
S. M.
Sharp
 
R. R.
J. Phys. Chem. A
1997
, vol. 
101
 pg. 
3692
 
76.
V. G.
Malkin
,
M.
Kaupp
and