NMR spectroscopy of minerals and allied materials Free

There are over 4600 formally recognised types of mineral, i.e., elements or chemical compounds that occur naturally as a result of geological processes. Minerals are usually defined as naturally-occurring, stable solids with a specific chemical composition (within some defined limits) and exhibiting an ordered atomic structure. In the past, minerals were typically considered to be inorganic and abiogenic, with biological substances, e.g., bones and shells, excluded from classification, although this latter point has always been the subject of some debate. However, today, many classification schemes include all biominerals, and a specific class of organic minerals is also recognised. The distinction between minerals and rocks, however, should be noted – the latter typically being aggregates containing one or more minerals and exhibiting structural and chemical heterogeneity.^1–3 

The study of minerals has long been recognised to be of considerable importance – not only for understanding the fundamental physical and chemical properties of the materials that make up the surface and inner depths of our planet, but to understand the effects of variations in pressure or temperature upon these properties, and the changes that can occur due to weathering or chemical processes. Many minerals also find industrial use in, e.g., ceramics, cements, fertilisers, catalysts and glasses, making an understanding of their structure, composition and reactivity vital. A large number of materials are structurally and/or chemically related to minerals, and can be produced either by chemical modification/substitution of a mineral or by an entirely synthetic approach. While not strictly minerals (as they are not naturally formed), they nonetheless provide additional possibilities for application, and their study may well also provide insight into that of the parent/related mineral.

All of the 90 natural elements have some geochemical interest,³  but the bulk (∼90%) of the Earth's crust is composed of silicate and aluminosilicate minerals, with elements such as Fe, Ca, Na, K and Mg also of importance, as shown in Fig. 1a. The inner regions of the Earth, i.e., where pressures and temperatures increase, are also typically composed of silicate minerals, but with increased Mg and Fe content, as shown in Fig. 1b. Figure 1 also shows the changes in the major mineral component of the Earth with increasing depth results in the designation of “layers”, e.g., the change from α-(Mg,Fe)₂SiO₄ to β-(Mg,Fe)₂SiO₄ at ∼410 km, signifying the boundary between the upper mantle and the upper transition zone, with further transitions to γ-(Mg,Fe)₂SiO₄ at the boundary with the lower transition zone, and to (Mg,Fe)SiO₃ perovskite in the lower mantle.⁴ 

The requirement for an ordered atomic structure in a mineral has resulted in much previous mineralogical study being carried out using crystallographic diffraction. However, many minerals form extensive solid solutions (i.e., they exhibit a variation in chemical composition) where the exact ordering/position of substitution is not known. Diffraction provides information on the average structure, but is rarely able to provide the atomic-level detail required to understand how and why the structure and/or properties of a mineral vary with composition. This is particularly true where the difference in scattering factors is small (e.g., Al³⁺ and Si⁴⁺), concentrations are low, or dynamics play a significant role. The sensitivity of NMR spectroscopy to the local structure, through the variation of interactions such as the chemical shielding, J-coupling or quadrupolar coupling, provides an ideal tool for structural investigation of minerals, and the recent developments in hardware and software, enabling high-resolution NMR spectra of solids to be acquired with good sensitivity, have considerably widened the application of this technique. Despite these advances, complex spectral lineshapes can be observed for disordered materials. However, over the last 10 years, the approach of combining experiment with theoretical calculations of NMR parameters (typically using density functional theory, DFT) has grown to enable the assignment of spectral resonances and the prediction of spectra for many possible models when a structure is less well defined.

In this chapter we review the NMR spectroscopy of minerals published in the period 2010–2014. We assume a basic working knowledge of the methods used to obtain high-resolution NMR spectra of solids (e.g., MAS, decoupling, MQMAS, etc.,) and some knowledge of prior significant work on minerals, e.g., the use of ²⁹Si NMR to study Si/Al ordering in aluminosilicates. More complete reviews on these can be found in ref. 2, 3, 5 and 6. We shall initially concern ourselves with silicate minerals, dividing these according to their structural features, e.g., materials containing isolated units, chains, layers or frameworks of silicate tetrahedra, before turning to non-silicate minerals (which we shall categorise according to their chemical type). Although we shall focus primarily on materials in the more formal classification of minerals described above, we shall also mention wholly or partly synthetic allied materials, where significant mineralogical insight has been shown.

As Si and O dominate the Earth's crust and much of the mantle, silicates are the most important class of rock-forming minerals, and exhibit great structural variation owing to the stability of Si–O bonds.¹  Most crustal silicates are based on SiO₄⁴⁻ tetrahedra, which may occur in isolation or combine to form more complicated structures. Although rare in nature, six-coordinate Si may also be observed in high-pressure minerals.⁷  Silicate minerals are commonly classified according to the way the silicate polyhedra are linked and the degree of polymerisation, as shown in Fig. 2a. Minerals containing isolated tetrahedra are termed ortho- or nesosilicates, while those with two corner-sharing tetrahedra, i.e., Si₂O₇⁶⁻, are referred to as pyro- or sorosilicates. Tetrahedra may also form rings (cyclosilicates), chains (inosilicates), sheets (phyllosilicates) or 3D networks (tectosilicates).¹  The ease of ²⁹Si NMR spectroscopy (I=1/2), and the sensitivity of the ²⁹Si chemical shift to coordination number, type of coordinating atoms, degree of polymerisation (denoted Qⁿ, where n is the number of coordinated oxygens that “bridge” to other silicons), and even the substitution of next nearest neighbour (NNN) atoms, has led to the widespread application for the study of silicate minerals. The ²⁹Si MAS NMR spectrum of the aluminosilicate mineral analcime in Fig. 2b6 exhibits a change in the chemical shift of ∼6 ppm for each NNN Al substituted. More recently, the study of ²⁷Al (I=5/2), ¹⁷O (I=5/2) and other substituted cations has become more widespread.⁸ 

The simplest silicate minerals contain isolated SiO₄⁴⁻ tetrahedra, corner-sharing tetrahedra in Si₂O₇⁶⁻, or small cyclic clusters of tetrahedra.¹  The most important of these is olivine (Mg,Fe)₂SiO₄, an orthosilicate existing as a solid solution from Mg-rich forsterite to Fe-rich fayalite, which dominates the upper mantle. Olivine contains SiO₄⁴⁻ linked through six-coordinate Mg²⁺/Fe²⁺. This gives a single Si site, two distinct cation sites and three distinct O. As it provides an Fe-free model for olivine and can readily be synthesised at ambient pressure, forsterite has been extensively studied by NMR. Early work determined a ²⁹Si isotropic chemical shift, δ_iso, of −62 ppm,⁹  typical of Q⁰ orthosilicates, and the principal components of the Si shielding tensor were determined.¹⁰  More recently, Palke and Stebbins¹¹  carried out NMR measurements on ²⁹Si-enriched forsterite and showed that a series of peaks between −28 and −60 ppm, and one at −128.5 ppm (each accounting for 0.1–0.2% of the spectral intensity) result from trace paramagnetic impurities, with a strong linear correlation of shift with T⁻¹. The low natural abundance (0.037%) of the only NMR-active isotope of oxygen, ¹⁷O, has restricted most oxygen NMR studies of forsterite to ¹⁷O-enriched material.^12–15  Isotropic ¹⁷O NMR spectra have been obtained with composite spinning (DAS/DOR) techniques,¹⁵  and 2D multiple-quantum (MQ) MAS experiments,^13,14  and have been assigned using periodic DFT calculations.^16  25Mg NMR parameters were determined by early single-crystal studies,¹⁷  and later high-field (21.1 T) MAS experiments.¹⁸  More recent work¹⁹  refined the parameters using high-field MAS, MQMAS and CPMG (Carr-Purcell Meiboom-Gill) experiments on static samples. The work confirmed C_Q for Mg2 (4.31 MHz) and showed that the value for Mg1 (5.33 MHz) was larger than determined in previous work.

Since 2010, most NMR investigations of forsterite have focussed on its carbonation, a reaction relevant to geologic carbon sequestration. Kwak et al.,²⁰  studied the process ex situ, by ²⁹Si and ¹³C NMR. Hydrolysis of forsterite (at 80 °C and 96 atm) produced Q¹ (−84.8 ppm) and Q² (−91.8 ppm) surface species, while reaction with supercritical CO₂ and H₂O produced Q⁴ species (−111.6 ppm) and small amounts of Q³ (−102 ppm) and Q² (−91.8 ppm), suggesting that the formation of magnesite (MgCO₃) was more rapid than forsterite hydrolysis. An intermediate dypingite (Mg₅(CO₃)₄(OH₂·5H₂O) phase was also identified by ¹³C MAS NMR. In further work,²¹  the authors considered the effect of water content on carbonation, and showed that no reaction occurred for trace amounts of water, while below the saturation level a layer of partially-hydrated/hydroxylated magnesium carbonates and hydrous amorphous silica species formed on the forsterite surface. Above the saturation level the reaction products were magnesite and amorphous polymerised silica. ¹³C MAS NMR showed reaction products at shifts between 160 and 175 ppm, with weaker peaks at 164.1 and 166.4 ppm attributed to dypingite, hydromagnesite (3MgCO₃·Mg(OH)₂·3H₂O), and nesquehonite (MgCO₃·3H₂O). Hu and co-workers²²  studied the carbonation of forsterite in situ, using a high-pressure MAS rotor, capable of an internal pressure of 150 bar. Reaction with enriched CO₂ was followed by ¹³C NMR at ∼2 kHz MAS. In addition to CO₂ at 126.0 ppm, a sharp resonance at 161.5 ppm was attributed to mobile HCO₃⁻, an intermediate that disappeared before MgCO₃ formed. Felmy et al.²³  investigated the carbonation reaction at lower temperature (35/50 °C) and determined that initial products were nesquehonite and magnesite after 3–4 days, with magnesite and amorphous silica formed at longer times (14 days). Work in 2014²⁴  revealed that the particles formed in this reaction exhibited a uniform submicron grain size, with rhombohedral morphologies, not consistent with growth on the forsterite surface. Additional work²⁵  considered the impact of diffusive transport in the carbonation, with in situ ¹³C NMR measurements of static samples at 105–120 bar and 80 °C. The carbonate solid produced a broad, axially-symmetric powder pattern with peaks from HCO₃⁻ and CO₂ also observed.

Liu et al.²⁶  studied two polymorphs of ZnSiO₄ (willemite), synthesised at 6.5 GPa (phase III) and 8 GPa (phase IV). The structure of phase III resembled olivine, but with four-, rather than six-coordinate, Zn. Phase IV has four-coordinate Zn and Si, but contained Zn₂O₆ dimers. ²⁹Si MAS NMR confirmed that the two polymorphs were different from those already known, and that all Si was four coordinate.

Over the years there has been much interest in the hydration of the minerals in the Earth's mantle, which is thought to contain at least as much water as the Earth's surface.^4,7  Mantle silicates are formally anhydrous, with no hydrous minerals stable below the upper mantle, and melts and fluids assumed to be absent owing to the increased pressure. Therefore, low concentrations of water are thought to be incorporated as structurally-bound hydroxyl defects in anhydrous minerals. However, the difficulty with identifying low concentrations (ppm) of defects has led to the study of model systems containing stoichiometric proportions of hydrogen. The humite minerals, consisting of n forsterite-like layers separated by layers of Mg(OH,F)₂, have been widely studied using solid-state NMR. There are four minerals in the family – norbergite (n=1), chondrodite (n=2), humite (n=3) and clinohumite (n=4). These minerals were studied in early work by ¹H and ²⁹Si NMR,²⁷  and in 2010 Davis et al. reported the ²⁵Mg CPMG NMR spectrum of clinohumite.¹⁹  This was compared to a ¹H–²⁵Mg CP CPMG NMR spectrum of acid-leached forsterite, indicating the formation of humite-like layers when forsterite was exposed to acidic conditions for 310 h. More detailed insight into the structure of the hydroxyl end members has been obtained from ¹⁷O MQMAS NMR of enriched materials.²⁸  Work by Ashbrook and co-workers²⁹  used ²H NMR to investigate motion of the hydroxyl groups in clinohumite (motivated by an earlier observation using ¹⁷O satellite-transition (ST) MAS experiments of microsecond timescale dynamics in chondrodite and clinohumite³⁰ ). Diffraction reveals two distinct H sites for each hydroxyl group, each with an occupancy of 50%.³¹  DFT calculations confirmed that ¹⁷O MQMAS and STMAS spectra could only be simulated assuming dynamic exchange of the hydroxyl protons between H₁ and H₂.^32  2H NMR was then able to confirm the presence of microsecond timescale dynamics.²⁹  The rate constant obtained (1.0–1.3×10⁵ s⁻¹ at 298 K) was in good agreement with that from ¹⁷O NMR, and an activation energy of 40 kJ mol⁻¹ was estimated for H₁/H₂ exchange.

The substitution of OH⁻ for F⁻ in the humites occurs commonly in nature, and is thought to favour occupancy of H1, owing to the formation of O–H…F⁻ hydrogen bonds. Diffraction studies of 50% fluorinated humites have been interpreted in terms of full occupancy of a single H site.³³  This seemed to be confirmed by ²H MAS NMR of 50% fluorinated clinohumite,²⁹  where a single sharp resonance was observed with a temperature-independent linewidth. However, subsequent ¹⁹F NMR^34,35  revealed four distinct resonances (−166.4, −169.3, −175.1 and −177.7 ppm), indicating multiple F environments. DFT calculations showed that the different shifts resulted from different anions (i.e., OD⁻/F⁻) on neighbouring sites. The assignments were supported by DQ experiments, which revealed unexpected J-couplings between some species.^34,35  J-couplings are usually thought to imply covalent bonding, whereas the fluorines in clinohumite are coordinated only by Mg. However, DFT calculations of these ¹⁹F–¹⁹F “through-space” J-couplings were in good agreement with experiment. A natural clinohumite sample (with composition Mg_8.85Fe_0.01Ti_0.2(Si_3.94O₁₆)O_0.4F_0.97(OH)_0.63) exhibited a very similar ¹⁹F spectrum, with an additional low-intensity resonance attributed to F close to Ti, suggesting that anion disorder is not related to the sample preparation, but intrinsic to the mineral.

Zircon (Zr₂SiO₄) is an orthosilicate, with a tetragonal (I4₁/amd) structure, containing chains of alternating edge-sharing ZrO₈ and SiO₄. The chemical shift of the Q⁰ Si site was shown in early work to be −82 ppm.⁹  Zircon has many attractive physical properties, including high thermal and chemical stability, low thermal expansion and low conductivity, resulting in many applications in ceramics, enamels and glazes. In 2010, Burrows investigated the effect of co-doping zircon with Fe³⁺ and Al³⁺.^36  29Si NMR showed that zircon was formed at 1215 °C after 1 h, with only Q⁰ Si (and Si–O–Zr bonds) present, with an increased formation of zircon at higher temperatures. Alba and co-workers³⁷  studied the substitution of Zr for Hf using ²⁹Si MAS NMR spectroscopy. A solid solution was thought to be obtained between the two end members, and near ideal mixing was suggested by DFT calculations. ²⁹Si shifts for the end members (−81.7 ppm (zircon) and −78.2 ppm (Hf₂SiO₄, hafnon)), agreed with earlier observations. However, while many peaks were seen at intermediate compositions, their relative intensities did not match those predicted assuming a random cation distribution. Fits were improved by including some short-range ordering, but the discrepancies were still not fully explained.

Zircon is also important as a natural analogue for the study of the behaviour of nuclear wasteforms over time. The Zr site can host actinides and other elements – natural zircons contain ∼5000 ppm U and/or Th, and elements such as Pu can also be incorporated. Study of such samples can provide insight into the short-term and much longer-term behaviour of wasteforms. In 2014, Smye et al.³⁸  attempted to quantify the α-particle radiation damage in zircon, using ²⁹Si and ⁷Li NMR. The zircon was mixed with a small amount of boron, which emitted an α-particle upon neutron irradiation. In addition to the sharp peak at −81.6 ppm, with increased radiation dose the ²⁹Si spectrum contained an additional much broader resonance at lower shift, attributed to the formation of amorphous regions associated with the recoil, enabling the proportion of damaged material to be directly quantified. In similar ²⁹Si NMR experiments, Zietlow et al. studied a natural metamict titanite (CaTiSiO₅).³⁹  Crystalline titanite contains chains of corner-sharing TiO₆ cross-linked by SiO₄ tetrahedra, producing a titanosilicate framework that can incorporate Ca²⁺ into the large cavities. Instead of the sharp signal at −79.3 ppm observed for crystalline titanite, the metamict nature of the natural sample produced a broad, Gaussian-like resonance at −81 ppm, with a full width half maximum of ∼24 ppm.

Mullite (Al₆Si₂O₁₃) is an important technical ceramic with good chemical and thermal stability. It has an unusual structure, with chains of edge-sharing AlO₆, cross-linked by tetrahedral (Al,Si)O₄ chains. Some of the bridging O atoms are removed for charge compensation, giving rise to oxygen vacancies and the formation of a so-called “tricluster”, clearly observed using ²⁷Al NMR spectroscopy.⁴⁰  The commercial importance of mullite has led to significant interest in its synthesis, and most recent NMR studies have used ²⁹Si and ²⁷Al to investigate the local structure of aluminosilicate-based precursors including inorganic polymers,⁴¹  sol–gel materials⁴²  and minerals.⁴³  For sol–gel materials, ²⁷Al MAS NMR spectra showed a progressive decrease in Al^VI with increasing temperature, as Al is forced to adopt a tetrahedral coordination due to the increasing amount of tetrahedral Si.⁴²  Mullite itself formed at 1400 °C. Chen et al.⁴³  employed mechanical activation of the solid precursors, with the formation of Al–O–Si bonds demonstrated by a change in the Al coordination number (with increasing amounts of Al^IV and Al^V observed) and the appearance of a ²⁹Si resonance at −85.3 ppm. Formation of mullite from an ammonium-exchanged sodium aluminosilicate polymer was observed by heating at 1100 °C.⁴¹  Cation exchange had no effect upon the ²⁷Al MAS spectrum (with only Al^IV observed) or the ²⁹Si MAS spectrum (δ ≈ −88 ppm).

The three polymorphs of Al₂SiO₅, sillimanite, andalusite and kyanite have been widely studied using multinuclear NMR spectroscopy.⁸  The three materials have different Al coordination, with only Al^VI present in kyanite, Al^IV and Al^VI in sillimanite and Al^V and Al^VI in andalusite. The variation in ²⁷Al C_Q (from 5.8 to 15.3 MHz), and the presence of multiple Al species in each material has resulted in their use as model compounds for NMR method development, e.g., MQMAS, STMAS and DQ-filtered experiments.^44–47 

Topaz (Al₂SiO₄(OH,F)₂) contains tetrahedral SiO₄ linking AlO₄(OH,F)₂ units, and the mineral has previously been studied using multinuclear NMR.⁸  Recently, Xue et al. described the crystal chemistry of two polymorphs of hydroxyl topaz, with the second, new polymorph produced during synthesis at higher pressures and temperatures.^48  27Al NMR revealed topaz-OH I contained Al^VI (δ_iso=8.2 ppm, C_Q=5.3 MHz and η_Q=0.4), while the ²⁹Si MAS NMR spectrum contained a single peak at −83.4 ppm. For topaz-OH II, however, ²⁷Al MAS NMR showed a broad lineshape characteristic of disorder. A small amount of Al^IV (3.2%) was also observed. The ²⁹Si MAS NMR spectrum was also broader, and a small peak at −178 ppm suggested some Si^VI was present. ¹H MAS NMR confirmed the difference in the phases.

The garnet group minerals are common in metamorphic rocks and the mantle. They have a general formula X₃Y₂Si₃O₁₂, with eight formula units per unit cell, and contain alternating SiO₄ and YO₆, with X cations within the cubic cavities formed. In the related mineral hydrogrossular (sometimes termed a hydrogarnet) up to four OH⁻ per formula unit replace silicate groups. Most NMR interest has been focussed on the diamagnetic minerals pyrope (Mg₃Al₂Si₃O₁₂) and grossular (Ca₃Al₂Si₃O₁₂), where early ²⁹Si NMR work,⁴⁹  reviewed recently by Stebbins,²  provided evidence for non-random cation distributions in binary systems. ²⁹Si and ²⁷Al NMR was used to confirm the nature and purity of synthetic grossular, with a sharp ²⁹Si resonance at −83.9 ppm and a single (octahedral) ²⁷Al resonance.⁵⁰  Recent NMR studies of garnets have focussed on the effect of low levels of paramagnetic cations on the ²⁹Si and ²⁷Al MAS NMR spectra. Palke and Stebbins¹¹  studied Fe-bearing pyrope garnets, having identified “anomalous” ²⁹Si and ²⁷Al resonances at unexpected shifts.⁵¹  Variable-temperature NMR confirmed that the resonances were from species close to paramagnetic ions, as their shift showed a linear dependence on T⁻¹.

In other work on orthosilicates, Kriskova et al.⁵²  investigated the effect of mechanical and chemical activation on the hydraulic behaviour of synthetic merwinite (Ca₃MgSi₂O₈). ²⁹Si NMR showed peaks at −79.3 and −85.2 ppm upon activation, attributed to Q¹ and Q² Si in a calcium silicate hydrate gel. Thorogood et al.,⁵³  studied the dehydration of the titanosilicate sitinakite (ideal composition Na₂Ti₂O₃(SiO₄)·2H₂O), by ²³Na and ²⁹Si MAS NMR. Upon dehydration, a phase transition was observed, evidenced in ²³Na NMR spectra by an upfield shift of the resonance and an increase in linewidth. The asymmetric lineshape indicated a distribution of NMR parameters, suggesting increased disorder. This was also seen in the ²⁹Si NMR spectrum, with a shift from −81.7 to −77.2 ppm. In the latter case a complex resonance, with a number of overlapped components was observed, indicating a loss of symmetry in the silicate units with increased Na⁺ disorder. Evans et al.⁵⁴  studied the borosilicate dumortierite ((Al)Al₆(BO₃)Si₃O₁₃(O,OH)₂) from two locations. Five peaks were observed in the ²⁹Si MAS spectrum (at −95.2, −92.6, −91.3, −89.1, and −86.5 ppm), with areas of 57, 19, 7, 10, and 7%. Theoretical calculations on small clusters around Si were used to assign the peaks at −95.2 and −92.6 ppm to the Q⁴ Si₂ and Si₁, respectively, adjacent to fully-occupied Al sites, with the remaining peaks assigned to Q³ Si adjacent to vacant Al sites.

Sorosilicates (or pyrosilicates) contain isolated Si₂O₇⁶⁻ groups, with typical Q^1 29Si chemical shifts between −72 and −95 ppm.^8,9  Recent work includes that on thortveitite ((Sc,Y)₂Si₂O₇), a primary source of Sc.⁵⁵  There has been considerable debate over the structure of compositions in this solid solution, as the proposed C2/m structure contains unusual 180° Si–O–Si bond angles. Other possible space groups (Cm and C2) show a deviation of the bond angle from 180°, but Cm predicts two Si sites, which are not observed experimentally. ⁸⁹Y CPMG MAS experiments of Allix et al.⁵⁵  indicated the presence of two distinct sites. This suggests that lowering of the symmetry to C2 is probable in intermediate compositions.

The melilite group of sorosilicates are solid solutions of several end members with general formula A₂B(T₂O₇), and common compositions being described by (Ca,Na)₂(Al,Mg,Fe)((Al,Si)₂O₇). The structure is unusual, with potential Al/Si disorder in the “pyrosilicate” unit. This was investigated in detail for gehlinite (Ca₂Al(AlSiO₇),^56–58  using multinuclear NMR and DFT calculations. ²⁹Si spectra confirmed the relative populations of the two Si sites, and hence the amount of Al–O–Al linkages present. The seven possible Al sites (Al–(OAl)_4−p(OSi)_p and Al–(OAl)_3−p(OSi)_p) were resolved and assigned using ²⁷Al MAS, MQMAS and 2D heteronuclear correlation experiments. The results were consistent with randomly disordered Al and Si. This conclusion was supported by DFT calculations of 50 supercells with varying cation positions, which produced composite ²⁷Al lineshapes in excellent agreement with experiment. In addition, calculated ²J(T–O–T) couplings were found to be linearly dependent upon the T–O–T angle. Melilite group materials have also been extensively investigated over the years for optical applications, owing to their ability to incorporate lanthanide ions, with Ca₂Al₂SiO₇:Eu²⁺ a potential red phosphor. Luo and Xia studied the modification of this material by the substitution of Si–N for Al–O.^59  29Si NMR spectroscopy verified the incorporation of N into the material, with a shift of ∼18 ppm observed. The formation of melilite minerals was also shown to be of importance in the phase separation of blast furnace slag, a waste product that is a potentially valuable source of oxides.⁶⁰ 

Lawsonite (CaAl₂(Si₂O₇)(OH)·H₂O) is a hydrated mineral regarded as a potential candidate for the transport of water to mantle depths greater than 200 km. Kozlova and Gabuda⁶¹  used ¹H NMR spectroscopy to study disorder in lawsonite. They concluded, from second moment analysis, that there was a time-averaged disorder, with H₂O and OH groups oscillating between two equivalent sites.

Sorosilicates can have more complex structures – the epidote group for example, contains both Si₂O₇ and SiO₄ units. One member of the group, zoisite (Ca₂Al₃O(SiO₄)(Si₂O₇)(OH)), has been of interest over the years in the development of NMR methodology as it has two octahedral Al sites with large C_Q values (8.0 and 18.5 MHz).⁶²  Recently, zoisite was used to demonstrate the use of double frequency sweeps for sensitivity enhancement of MAS NMR spectra of quadrupolar nuclei.⁶³ 

Cyclo- or ring silicates have structures with linked tetrahedra, with a T : O ratio of 1 : 3. Various ring sizes exist, including T₃O₉⁶⁻, T₄O₁₂⁸⁻ and T₆O₁₈¹²⁻. Recent work on cyclosilicates includes the use of ¹¹B and ²⁷Al NMR to characterise an elbaite mineral from the tourmaline group.⁶⁴  Tourmaline is a crystalline borosilicate, which often contains Fe, Mg, Na or Li. ¹¹B MAS spectra were able to determine the proportions of three- and four-coordinate B (as both δ_iso and C_Q are sensitive to the coordination geometry). ²⁷Al NMR was used to determine the Al coordination number. In other work, Yeom and Lim studied the relaxation of ⁹Be, ²⁷Al and ²⁹Si in Cr³⁺-doped Beryl (Be₂Al₂Si₆O₁₈) crystals,⁶⁵  showing linear variation of T₁⁻¹ with temperature for each nuclear species. ⁹Be relaxation was ∼3 times more rapid than ²⁷Al, reflecting the dependence of relaxation rate upon C_Q².

Chain (or ino-) silicates are composed of corner-sharing chains of silicate tetrahedra, resulting in Q² Si species, and both bridging and non-bridging oxygens. Materials can be made up of single chains (with a 1 : 3 ratio of Si : O), or double chains (with a ratio of 4 : 11).¹  The largest group of single-chain inosilicates are the pyroxenes, which are found as stable phases in almost every type of igneous rock. They can be typically described by the general formula M1M2(Si,Al)₂O₆, and fall into two main types, with orthorhombic and monoclinic structures. The orthopyroxenes primarily consist of (Mg,Fe)SiO₃, while the clinopyroxenes have a much wider compositional range. For many clinopyroxenes M1 and M2 sites contain Mg, Ca and Fe, but substitution with Al, Na, Cr and Li is also common. There are three polymorphs of MgSiO₃ (ortho-, clino- and protoenstatite), which have been studied by ²⁹Si and ¹⁷O MAS,^8,9,66  and, more recently, high-resolution NMR.^67,68  In 2007, periodic DFT calculations⁶⁹  enabled full assignment of the spectra. In more recent work, Griffin et al.⁷⁰  observed resonances corresponding to the two ²⁵Mg sites in MgSiO₃ polymorphs using a wideline frequency-stepped CPMG echo experiment. In earlier MAS spectra, only a single resonance was detected.⁸  This observation was explained by DFT calculations, which predicted a very large C_Q for Mg2 (13–15 MHz), suggesting experiments on a static sample were required.⁷⁰ 

While NMR has been very useful for studying diamagnetic pyroxenes, the incorporation of Fe into these minerals considerably complicates spectral acquisition. However, at low concentrations (e.g., ∼1%) paramagnetic ions can aid acquisition, increasing relaxation rates. Work by Begaudeau et al.⁷¹  investigated spectral acquisition for natural ortho- and clinopyroxenes, where Fe–Ti–Cr rich inclusions were identified by EDS. The paramagnetic ions broadened the ²⁹Si lines and an additional resonance appeared for the orthopyroxene. ²⁷Al MAS NMR spectra revealed both four- and six-coordinate Al. ¹H NMR spectra were dramatically broadened, beyond the typical shift range for diamagnetic solids.⁷¹ 

Evidence for the incorporation of BO₃ into a clinopyroxene diopside was considered by Hålenius in 2010.⁷²  Although the concentration of boron in the mantle is low (<0.1 ppm) it is much greater in sediments and crustal materials. The authors co-doped diopside with B (0.1–0.6 wt%) and Al (0.2 wt%) or Sc (3–6%), and studied the resulting material using multinuclear NMR. The ²⁹Si NMR spectrum showed a sharp Q² signal at ∼84.3 ppm, as expected for diopside. No effect of B substitution was detectable at the levels used, but an additional resonance at −87.1 ppm in one sample was attributed to the partial incorporation of Sc at M1. ¹¹B MAS NMR confirmed that the B was incorporated primarily as BO₃ (C_Q ≈ 2.6 MHz, η_Q ≈ 0.65) and a small (<5%) amount of BO₄ (C_Q ≈ 0.6 MHz, η_Q ≈ 0.8).

Spodumene (LiAlSi₂O₆) is a clinopyroxene and a natural source of Li. The high thermal stability, good chemical durability and low thermal expansion coefficient of lithium aluminium silicates have resulted in recent interest in their synthesis. Conventional production requires expensive high-temperature processing, and recent work has focussed on alternative syntheses from Li-geopolymer precursors.^41,73–75  The preparation of precursors, their structure and the subsequent formation of spodumene (or other lithium aluminium silicates) upon heating can be followed by ²⁷Al and ²⁹Si MAS NMR. Li-containing geopolymers can be prepared using cation exchange⁴¹  or solid-state synthesis.⁷⁶  In the first approach, NMR showed that cation exchange did not affect the structure of the geopolymer (which contained Al^IV and Q² Si at −88 ppm). In the latter case, the synthesis produced Li-containing zeolites, evidenced by the Al^IV (δ ≈ 61 ppm) and ²⁹Si peaks at −81.1 ppm (Li-ABW) or 85.3 to −85.7 ppm (Li-EDI). Heating at low temperatures formed β-eucryptite, with spodumene forming at higher temperatures. Nourbakhsh et al. further investigated Li-geopolymer formation and the subsequent formation of spodumene and β-eucryptite,⁷⁴  using ²⁹Si and ²⁷Al NMR. The type and amount of Si used in the silico-thermal reaction did not affect the phases produced, but altered their crystallinity.

The pyroxenoid minerals have a general formula X_nSi_nO_3n, where X is a larger divalent cation, typically Ca, Mn or Fe.¹  Although they contain chains of corner-sharing SiO₄, pyroxenoids are not structurally related to pyroxenes, but contain complex chains with longer repeat units. Two major polymorphs of wollastonite (CaSiO₃) exist, para-wollastonite, containing corner-sharing silicate chains, and pseudo-wollastonite, containing Si₃O₉ rings. A number of polytypes of para-wollastonite, with differing stacking sequences of the chains have been identified, but no complete structures exist for many forms. Florian and Massiot⁵⁷  were able to distinguish between the two polymorphs using ²⁹Si MAS NMR, with peaks at −89.0, −89.6 and −87.8 ppm in para-wollastonite and −83.3, −83.7 and −83.9 ppm for pseudo-wollastonite. The chemical shift anisotropies (CSAs) for pseudo-wollastonite were also larger (−134 ppm) than for para-wollastonite, while the ²J(Si–O–Si) coupling was smaller for the latter phase (1.5 Hz cf. 3.6–8.0 Hz). Low-intensity ²⁹Si signals for para-wollastonite were thought to result from small amount of polytypes with different stacking sequences.

Work on wollastonite was also carried out by Schott et al.,⁷⁷  with an investigation into the formation of surface layers and structural transformation during acid dissolution. Dissolution is strongly pH dependent – with reactions between pH 5 and 12 showing stoichiometric release of Ca and Si, but reactions at lower pH showing a preferential release of Ca. The reconstruction of the mineral was followed by ²⁹Si NMR, with the Q² species of unreacted wollastonite, replaced by Q³ and Q⁴ species as dissolution progressed and amorphous silica was produced. Miller et al.⁷⁸  also studied the in situ carbonation of wollastonite (supercritical CO₂ at 160 bar) using high-pressure MAS equipment. ²⁹Si NMR showed the growth of low-intensity peaks, consistent with Q³ and Q⁴ from amorphous silica, with Q³ species more prevalent.

Chain silicates can also have more complex structures, e.g., the amphibole minerals have double chains of SiO₄⁴⁻ tetrahedra that share corners within and between chains. As with pyroxenes, amphiboles can exist in monoclinic and orthorhombic forms, with a general composition of A₀₋₁B₂C₅(Al,Si)₈O₂₂(OH,F)_2.¹  Four broad divisions are known – calcic (Ca), alkali (Na), sodic–calcic (Na, Ca) and (Fe, Mg, Mn) amphiboles. A range of NMR has been carried out over the years,^79–83  including studies of cation disorder using ²⁹Si and ²⁷Al NMR, disorder and dynamics of hydroxyl groups and Ga substitution, work recently reviewed by Stebbins.¹¹  In more recent work, Lussier and Hawthorne studied the chemical and structural variations of bavenite, a double-chain calcium beryllium aluminosilicate.⁸⁴ 

Phyllosilicates, or sheet silicates, form parallel sheets of SiO₄ tetrahedra with a Si : O ratio of 2 : 5. All of these minerals have hydrated silicate sheets, with either water or hydroxyl groups present, usually joined by sheets of octahedrally-coordinated cations. The physical and chemical properties of the minerals are determined by the stacking of the sheets, e.g., 1 : 1 or 2 : 1 stacking of tetrahedral and octahedral layers, and by the cations (e.g., Al³⁺, Mg²⁺, Fe³⁺, Ca²⁺, etc.,) present.¹ 

The serpentine group contains three major minerals, antigorite, chrysotile and lizardite, all with approximate composition Mg₃Si₂O₅(OH)₄. ²⁹Si NMR spectra of natural serpentine contain a resonance at −94 ppm, consistent with Q³ Si.⁹  Lizardite is the most abundant mineral in the group. The synthesis of Al-containing lizardite was studied using ²⁷Al MAS NMR.⁸⁵  Resonances at 68 and 6 ppm were observed (at 9.4 T), corresponding to Al^IV and Al^VI, respectively. The latter (∼75% of the signal intensity) was decomposed into components attributed to lizardite and kaolinite. However, the low resolution and the neglect of any distributions of parameters (or the use of MQMAS) casts some doubt on this interpretation. The least abundant of the major serpentines is chrysotile, but this is probably the best known as it is an important source of commercial asbestos, owing to its fibrous-like structure. It is composed of one tetrahedral (silicate) layer and one octahedral layer containing Mg²⁺. Anbalagan et al.⁸⁶  studied natural chrysotile from India, using XRD, Raman and NMR spectroscopy. The ²⁹Si NMR spectrum contained a sharp peak at −89 ppm, assigned to Q³(1Al) species, i.e., indicating substitution of Al into the silicate layers. The presence of Al was confirmed by ²⁷Al MAS NMR, where a broad Al^IV resonance was observed at ∼57 ppm. More recently, attempts to form geopolymers from chrysotile were followed using ²⁹Si and ²⁵Mg NMR.⁸⁷  The initial chrysotile exhibited a broad ²⁹Si resonance at −87.6 ppm. After heating, a new resonance at −76.7 ppm was attributed to an amorphous dehydroxylated phase. Although of poor sensitivity, ²⁵Mg MAS spectra of the initial sample contained a resonance at −30 ppm, consistent with octahedral Mg, and a small MgO impurity. After dehydroxylation, low-intensity resonances at 65 and 39 ppm appeared, while after heating a broad resonance at 75 ppm was observed, all tentatively attributed to four-coordinate Mg. The carbonation of antigorite, was studied in situ,⁸⁸  using ¹³C NMR in a high-pressure MAS rotor. The result of reaction at 150 bar and 50 °C with water-saturated supercritical CO₂, was shown to be nesquehonite and HCO₃⁻. Owing to paramagnetic impurities in the natural sample used, the products could only be resolved from the spinning sideband manifold at higher (6 kHz) spinning rates – a technical challenge for a high-pressure MAS experiment.

The mica family of minerals are 2 : 1 layered silicates in which cation substitution in either the tetrahedral (e.g., Al³⁺) or octahedral (e.g., M⁺ for M²⁺) sheets produces negatively-charged layers, charge-balanced by interlayer cations (e.g., K⁺, Na⁺, Ca²⁺). The relative arrangement of consecutive layers (determined by an interlayer stacking angle) leads to six major polytypes.¹  There have been many NMR studies of natural and synthetic micas over the years, with particular emphasis on cation disorder, as discussed in a recent review.²  More recently, Eckert and co-workers⁸⁹  used multinuclear NMR spectroscopy to probe local structure in a synthetic fluoromica. Two Q^3 29Si resonances (at −95.3 ppm (92%) and −98 ppm (8%)) were observed, attributed to hectorite- and talc-like Si, with higher and lower negative charge, respectively. Three types of ²³Na were resolved, corresponding to non-exchangeable (∼40 ppm) and exchangeable (−2.3 and −20 ppm) species. The peak at −2.3 ppm was assigned to hydrated Na⁺ trapped in the interlayer space, while that at −20 ppm was assigned to non-hydrated species. ²³Na{¹⁹F} REDOR experiments provided information on Na–F distances and F–Na–F angles, and showed that the non-exchangeable Na⁺ ions are found exclusively in the Mg(i) sites. ¹⁹F NMR showed that F chemical shifts were very sensitive to the local environment, with many resonances resolved, which could be partly assigned to charged Mg–Mg–Na, charged Mg–Mg-vacancy, or neutral Mg–Mg–Mg octahedral environments.⁸⁹ 

Phlogopite (KMg₃(AlSi₃)O₁₀(OH)₂), a trioctahedral mica with almost complete occupation of the octahedral layer, has been extensively studied using NMR spectroscopy.²  In nature there can be substantial replacement of Mg/Si for Al, according to Tschermak's substitution (where Mg^VI and Si^IV are replaced by Al^VI and Al^IV). Cation ordering in Al-rich synthetic phlogopites was recently studied by Langer et al.,⁹⁰  who combined ¹H and ²⁹Si NMR (MAS, CP MAS and HETCOR) with Monte Carlo simulations. ²⁹Si NMR spectra resolved four Si environments (Q³ with 0–3 Al NNN) and the Si/Al ratio for each of the samples was determined (assuming Löwenstein's rule was obeyed). Comparison of the proportions of environments predicted by the Monte Carlo simulations with the experimental Si spectra suggested some segregation into Al-rich regions, with a preference for Al to occupy neighbouring six- and four-coordinate sites. Phlogopite can also incorporate larger amounts of F than most other minerals, and F can have a significant influence on the cation ordering, as shown recently by Fechtelkord et al.,^91,92  with a series of Al-rich fluorophologopites investigated using ¹H, ²⁹Si and ¹⁹F NMR. It was observed that the presence of F significantly reduced the capacity of phlogopite to incorporate Al, while ¹⁹F NMR spectra showed that F was located in Mg-rich octahedral and Si-rich tetrahedral regions, with hydroxyls found preferentially in Al-rich regions.

There has also been considerable recent interest in a series of synthetic sodium fluorophlogopites, known as Na-n-mica (or Na-mica-n), where the interlayer charge, n, ranges from 2 to 4.⁹³  Unlike phlogopite, these materials have significant expansion capabilities and so have applications as decontaminants and storage media. Although NMR studies of Na-n-micas have been known for some time, elegant work by Delevoye and co-workers⁹⁴  recently showed violation of Löwenstein's rule in this material using natural abundance high-field (18.8 T) ¹⁷O NMR spectra. Signal was observed between 30 and 50 ppm, typical of Si–O–Si and Si–O–Al, but also at 20 to 30 ppm, attributed to T–O–Mg and Al–O–Al. The difficulty of producing pure bulk phase Na-n-micas has led to an interest in understanding their synthesis, and Alba et al.⁹⁵  used ²⁹Si, ²⁷Al and ²³Na NMR spectroscopy to investigate the products obtained at various heating times. An initial rapid reaction (<2 h) produced sodalite and a 2 : 1 layered phyllosilicate (characterised by three Q³ Si species with 0–2 Al NNN, the presence of Al^VI and Al^IV and a ²³Na signal at −9 ppm, demonstrating that Na is surrounded by interlayer water). All NMR spectra changed significantly between 2 and 3 h, with the appearance of a Q³ (3 Al) species and loss of the sodalite signals. Breakdown of Na-4-mica was observed at times >30 h. The observation of a sodalite intermediate prompted Naranjo et al.⁹⁶  to attempt a synthesis of Na-4-mica from this material directly, using ¹H and ²⁷Al NMR to confirm the similarity of the product.

There has also been significant recent interest in the modification of micas with organic molecules, typically by cation exchange, forcing an increase in interlayer spacing.^97–101  For the high-charge, expandable Na-n-micas, ¹H NMR showed that water was displaced from the interlayer space upon incorporation of RNH₃⁺, while ¹H and ²⁹Si NMR revealed that the NH₃⁺ head groups are incorporated in the hexagonal holes in the tetrahedral layer.^97,98  No variation in the intensity of the four ²⁹Si peaks (attributed to Q³(0–3 Al)), was observed upon modification, suggesting that the Si/Al disorder in the layers remained unaffected.^{97,98,101  13}C NMR shows the alkyl chains form a bilayer between the layers, and the molecules show much more significant dynamics.^97,98  The interaction of various cations and interlayer water was shown (using the ¹H chemical shift) to depend upon the corresponding solution pK_a of the cations and the interlayer charge.^100,101  Eckert and co-workers⁸⁹  also considered organocation intercalation in fluoromicas, demonstrating an increase in interlayer spacing, but no modification of the tetrahedral layer structure or ordering.

Muscovite is one of the most common micas, with good electrical and thermal insulating properties making it of industrial importance. In muscovite 25% of the Si sites are occupied by Al, and interlayer cation sites are occupied almost exclusively by K⁺. Palin et al.¹⁰²  recently reviewed the application of Monte Carlo methods to understand cation disorder in a range of 2 : 1 phyllosilicates, including muscovite, comparing conclusions to earlier experimental NMR studies. Other recent NMR work on muscovite has concerned its formation from the high-pressure decomposition of kaolin,¹⁰³  followed by ²⁹Si and ²⁷Al MAS NMR spectroscopy.

Talc (Mg₃Si₄O₁₀(OH)₂) is a trioctahedral phyllosilicate with uncharged layers, leading to very weak interlayer bonding. Natural talc is often associated with other minerals (e.g., chlorite, chrysotile and amphiboles) and can contain a number of substitutions. The inability to grind natural talc homogenously to a particle size below 1 μm without significant amorphisation has limited its applications and initiated an interest in the production of synthetic talcs with controlled particle size. In 2013, Dumas et al. described the spectroscopic characterisation of talcs synthesised using a new protocol.¹⁰⁴  The ¹H spectrum of the submicron talcs showed three sharp signals (at 0.5, 1.8 and 4 ppm), in contrast to natural talc, which gave a broad resonance at 0.5 ppm. The peak at 1.8 ppm was attributed to silanol groups (resulting from the small particle size) and that at 4 ppm to physisorbed water at the particle edges. ²⁹Si NMR also revealed differences, with natural talc exhibiting a single Q³ resonance at −97 ppm,⁹  while the synthetic talcs had a second signal at −95 ppm, which decreased in intensity with increasing synthesis time, suggesting it can be assigned to Q² species. Chabrol et al.¹⁰⁵  used ²⁹Si MAS NMR to study the structure of talc-like phyllosilicates prepared hydrothermally at different temperatures. At low temperatures, the minerals showed a high degree of hydration, structural flaws and low crystallinity, as evidenced by the presence of five ²⁹Si peaks (−97.3 and −95.1 (Q³) and −91.8, −87.5 and −85.2 ppm (Q²)), in contrast to the single resonance in talc. The signal at −97.3 ppm, however, confirmed the presence of talc-like environments. MacKenzie et al. used ²⁹Si and ²⁵Mg NMR to follow the synthesis of inorganic geopolymers from talc,⁸⁷  although this was unsuccessful, with ²⁹Si NMR revealing the presence of enstatite and silica after dehydroxylation. ²⁵Mg NMR confirmed only six-coordinate Mg was present, and no polymer had been formed. Alba and co-workers¹⁰⁶  investigated the reaction of layered silicates (including talc) with lutetium nitrate solution, for applications in waste remediation. Lu₂Si₂O₇ was formed in the reaction, with ²⁹Si NMR confirming a change from the Q³ (−98.1 ppm) species in talc, to Q¹ (−91.7 ppm) in Lu₂Si₂O₇. There has also been interest in the modification of talc by incorporation of organic molecules.^{105,107–109}  The presence of covalently-bound organic molecules is evidenced by ²⁹Si signals between −50 and −70 ppm, confirming Si–C bonds are formed.

The structure of pyrophyllite is similar to talc, with Al rather than Mg occupying the octahedral layers, and the increased charge leading to the formation of a dioctahedral, rather than trioctahedral, material with formula Al₂Si₄O₁₀(OH)₂. Recent NMR studies have focused on its reactivity and transformation. The work of Alba and co-workers studying the reaction of layered silicates with Lu³⁺ solutions described above¹⁰⁶  also considered pyrophyllite. The ²⁹Si NMR spectrum contained a peak at −95.2 ppm, (Q³(0 Al)). After hydrothermal treatment two minor ²⁹Si signals were observed (attributed to Lu₂Si₂O₇ and H₂Si₂O₅, with intensities of 4.3% and 1.8%, respectively). No changes were observed in the ¹H and ²⁷Al spectra after hydrothermal treatment, indicating the stability of pyrophyllite to reaction. Thermal transformation and alkaline dissolution of pyrophyllite was considered by Li et al.¹¹⁰  Thermal transformations were followed using ²⁹Si MAS NMR, with a change from the initial spectrum (single peak at −95.1 ppm) to an intense resonance at −100.9 ppm after calcination at 800 °C, from dehydroxylated pyrophyllite. At 1100 °C, the resonance from pyrophyllite disappeared, while that at −100.9 ppm remains. Two new peaks appear at higher temperatures, at −109 ppm (SiO₂) and −86 ppm (mullite).

An extremely important and widely-used group of phyllosilicates are the clay minerals.¹  A clay is fine-grained rock/soil material, which combines one or more clay minerals with accessory material, including metal oxides and organic matter. Clay minerals all contain tetrahedrally- and octahedrally-coordinated cations, but are differentiated (by their layer spacings) into four major groups: the kaolinite group, the illite group, the smectite group and the vermiculites. There have been far too many NMR studies on clays and clay minerals, even in the last five years, to list here. In general, investigations focus on understanding the structure of the clay (e.g., ordering of cations within the layers and the interlayer material present), how clays change upon thermal treatment or reaction, the intercalation or grafting of material, typically small organic molecules or organometallic complexes (and subsequent thermal treatment or reactions of modified materials), and the formation of composite/nanocomposite and heteromaterials. In most cases, NMR is used in one or more basic ways: (i) to characterise the type of clay present (and the speciation of Si), (ii) to determine whether intercalation between the clay layers has occurred, (iii) to probe whether a small molecule or ion has been grafted onto the surface (i.e., a covalent bond is formed), or (iv) to probe whether a composite or nanocomposite has been formed. In many examples, the NMR spectra are not used to determine structural information, merely to assess which compounds/components are present. Below, we highlight work where NMR has provided more significant structural insight, rather than giving a truly comprehensive list of all recent NMR spectra of clays. See ref. 2 and 111 for more detailed reviews.

Kaolinite is the most important member of the kaolinite group (which also contains the rarer polymorphs, dickite and nacrite, and the hydrated halloysite).¹  Allophane and imogolite are related, but poorly ordered, hydrous silicates that occur in soils. Kaolinite has composition Al₂Si₂O₅(OH)₄, and is used to produce paper, ceramics, cosmetics, whitewash and in paint. Ryu et al. demonstrated that synthetic kaolinite (produced from reaction of Al(OH)₃·xH₂O and SiO₂) exhibited similar NMR spectra to the natural mineral, with a single ²⁹Si single resonance at −91.4 ppm (typical for a Si-rich Q³ system), and a sharp ²⁷Al peak at ∼2 ppm (9.4 T) confirmed the presence of Al^VI.¹¹⁰  The ²⁷Al MAS NMR spectrum was studied in more detail by Paris,¹¹²  who used the satellite-transition spinning sideband manifold to determine accurate parameters for the two Al species expected (but not easily resolved, even at higher field), with best fits achieved when δ_iso=7.5 ppm, C_Q=3.4 MHz, η_Q=0.8 and δ_iso=8.0 ppm, C_Q=3.0 MHz, η_Q=0.9. Experimental measurements were supported by theoretical electric field gradient (EFG) calculations. Work by Begaudeau et al.⁷¹  showed the ²⁷Al NMR spectrum of kaolinite displayed three resonances, at 6, 58 and 72 ppm (17.6 T). The peak at 6 ppm is attributed to Al in the octahedral layers, that at 72 ppm to Al substituting for Si in the tetrahedral layers. The peak at 58 ppm results from a feldspar impurity. In this latter work, little effect on the spectra was observed when the kaolinite was ground with paramagnetic magnetite particles. However, a more significant effect was observed in 2012,¹¹³  for samples containing kaolinite and goethite (α-FeOOH). Small shifts (of 1–2 ppm) were observed in the ²⁹Si and ²⁷Al MAS NMR spectra of kaolinite both mixed and associated (i.e., coated) with goethite, with the interaction stronger in the latter case. In 2010, Mueller and co-workers studied the surface hydroxyl species in kaolinite with a fluorinated probe molecule, which bound selectively to non hydrogen-bonded Q³ Si.¹¹⁴  Quantitative ¹⁹F NMR was used to show the number of reactive hydroxyl sites per gram.

Much of the interest in kaolinite is concerned with its thermal behavior and transformation. Heat treatment initially leads to dehydroxylation, eventually producing a material known as metakaolinite. As kaolinite does not contain interlayer cations or water, the dehydroxylation temperature depends upon the stacking of the layers. Metakaolinite is amorphous but retains some long-range order in the layer stacking. The transformation (and structure of the metakaolinite) can be followed by ²⁹Si and ²⁷Al NMR.^115–118  Metakaolinite is characterized by a ²⁹Si resonance at −101.9 ppm, and the presence of four-, five- and six-coordinate Al. It was shown that the reaction temperature significantly affects the ratio of the three Al species present,¹¹⁶  and ref. 117 showed, using ²⁷Al DQ NMR, that the four- and six-coordinate Al were spatially close. Subsequent thermal reaction (with varying pH), followed by NMR, produces a range of materials including illite/muscovite,^103,119  mesoporous silicas,^120,121  zeolites^122,123  and silicoaluminophosphate frameworks.^124,125  The addition of silicates under alkaline conditions during thermal treatment can also form geopolymers, with the reaction and the final product studied using ²⁹Si and ²⁷Al NMR,^126–130  and other hybrid materials.¹³¹  Other reactions studied include the dissolution of kaolinite¹³²  and the incorporation of Ln³⁺.¹³³  In the first case, NMR was used to count the number of reactive surface sites during dissolution,¹³²  showing a decrease as a function of dissolution time (and preferential dissolution of the reactive edge sites). In the latter, sorption of Eu³⁺/Y³⁺ on natural kaolinite was investigated.¹³³  Little effect was seen in ²⁷Al MAS spectra, but loss of ¹H signal was attributed to Ln³⁺ binding to isolated surface Al–OH groups.

There are also many studies of intercalation and grafting, typically of organic species, between the kaolinite clay layers.^134–137  As described above, NMR is often used simply to prove the organic species is present, with little detailed structural information obtained. In some cases, NMR is able to determine whether intercalation or grafting occurs, with more significant spectral changes in the latter case.¹³⁵  Some studies, however, have used NMR to investigate the clay modification in greater detail. For example, Senker and co-workers¹³⁴  undertook a comprehensive NMR investigation of the covalent grafting of ethylene glycol onto the μ-bridged aluminol groups in the octahedral layer of kaolinite, as shown in Fig. 3a. NMR was used to prove that the ethylene glycol was grafted rather than intercalated, with evidence from ²⁷Al MQMAS spectra (which showed a considerable change in the local Al environment) and ¹³C/²⁷Al REAPDOR measurements (see Fig. 3b), which showed a C–Al distance of 3.1 Å. Variable-temperature wideline ¹H experiments determined the grafted molecule rotated around the covalently-bonded hydroxyl group in the interlayer space.¹³⁴ 

Halloysite is a hydrated form of kaolinite, with a single layer of water molecules in the interlayer space. The layers in halloysite are often rolled into cylinders, tubes or spheres, of significant interest in nanoscience. Halloysite gives a ²⁹Si resonance at −93 ppm (Q³) and a relatively narrow ²⁷Al resonance at ∼0 ppm (Al^VI).¹³⁸  Many recent NMR studies of halloysite have focused on its transformation, with reactions to form silica nanotubes,¹³⁹  geopolymers,⁷³  zeolites^140,141  and organohybrid materials¹⁴²  followed using ²⁹Si and ³¹P NMR spectroscopy. Yah et al.¹⁴³  used multinuclear NMR to show that halloysite nanotubes could be selectively modified by phosphonic acid, with the molecule binding to the inner alumina-rich surface of the tube, rather than the outer silica surface, creating a micelle-like hybrid material. ¹³C NMR showed the increased order of the acid molecules within the tube, while ³¹P NMR indicated the formation of Al–O–P bonds.

Imogolite and allophane are poorly crystalline materials, with general formulae of Al₂SiO₃(OH)₄ and Al₂O₃·(SiO₂)_1.3−2·(2.5–3)H₂O, respectively, and typically occur in clays, soils and volcanic ash. As with halloysite, the propensity of imogolite to form tubular and cylindrical structures (i.e., single-walled nanotubes) has generated considerable interest. However, most recent NMR studies have been performed on synthetic imogolite-based systems only, and are not discussed further. Imogolite and allophane are often found together in soil, as shown by recent NMR studies of the effect of long-term fertilisation.¹⁴⁴  The authors showed only Al^VI was present in the control soil sample, but increasing amounts of four- and five-coordinate species were observed as fertilisation increased. A recent ²⁹Si NMR study determined that andosol (formed from the natural weathering of volcanic ash) contained up to 10% allophane.^145  27Al MAS NMR spectra of imogolite exhibit a peak at ∼6 ppm (14.1 T) and a much smaller peak at 60 ppm, from Al^VI and Al^IV, respectively,¹¹⁵  while ²⁹Si NMR spectra contain a single peak at −79 ppm (Q³(3 Al)).¹⁴⁶  Upon heat treatment, imogolite was shown (using ²⁷Al MQMAS NMR) to transform to an amorphous phase by ∼673 K, with Al^IV and Al^V present. For allophane, ²⁷Al MAS NMR spectra also showed Al^IV and Al^VI, while ²⁹Si MAS NMR spectra contained an intense peak at −78 ppm (Q³(3 Al)), and smaller peaks at −93 and −104 ppm, indicating deformation and separation of the thin aluminosilicate layer.¹⁴⁷ 

The illite group of clay minerals has general chemical composition (K,H₃O)(Al,Mg,Fe)₂(Si,Al)₄O₁₀((OH)₂,(H₂O)), but many substitutions are possible. Structurally, illite is similar to the micas (particularly muscovite), but has fewer interlayer cations, reducing the bonding between layers and resulting in less regular stacking. The formation of illite from thermal treatment of kaolinite was followed using ²⁷Al MAS NMR spectroscopy in 2010,¹⁰³  while work in 2011 discussed the thermal reactivity of illite itself.^148  27Al MAS NMR of the illite clay showed mostly Al^VI but confirmed a small amount of substitution of Al for Si in the tetrahedral layers. Heating (at 600 °C) produced a material with primarily Al^IV, suggesting dehydroxylation had occurred. Silva et al. demonstrated the functionalistion of illite clays with organosilating agents, with new ²⁹Si NMR signals at −68 and −57 ppm for the modified materials demonstrating the formation of covalent bonds.¹⁴⁹ 

The smectite clay minerals have considerable industrial importance owing to their high expandability, with applications in the production of paper, rubber and building materials. The group includes 2 : 1 dioctahedral (montmorillonite, beidellite, nontronite) and trioctahedral (saponite, hectorite and sauconite) minerals, with chemical compositions of (1/2Ca,Na)_0.7(Al,Mg,Fe)_y[(Si,Al)₈O₂₀](OH)₄·nH₂O, where y=4 or 6. Clays rich in smectites are usually termed bentonite clays. NMR studies of smectites are extremely common in the literature, with a few hundred papers between 2010 and 2014. As described above, in many cases NMR is used to verify the nature of the initial clay used or to study subsequent reactions of modified materials. In the interests of brevity we have restricted our attention to studies with significant amounts of NMR investigation or where new structural insight is shown.

Saponite, M_x⁺(Mg₃)(Si_4−xAl_x)O₁₀(OH)₂·nH₂O is a 2 : 1 trioctahedral smectite with exchangeable interlayer M⁺ cations. He et al.¹⁵⁰  recently studied saponites with varying Si/Al ratios, and demonstrated (using ²⁷Al MAS NMR) that Al preferred to occupy the tetrahedral, rather than octahedral, sites. The ²⁹Si MAS NMR spectrum of saponite contained two signals at −95 and −91 ppm, corresponding to Q³(0Al) and Q³(1Al), which merged into a broad signal as the Si/Al ratio varied. Cation exchange in saponite was studied by Hunger et al.^151  23Na MAS NMR showed a broad signal centred at −12 ppm (9.4 T) that gradually narrowed upon hydration, reflecting the decrease in C_Q (and increased site symmetry). MQMAS revealed that dehydrated saponite contained two Na environments (with ∼1 : 1 intensities). Maheshwari et al.¹⁵²  showed how the ¹H chemical shift of nanoconfined water molecules in saponite varied with temperature, and attributed this to a change in the hydrogen bonding, identifying two new phase transitions at temperatures above and below the bulk freezing point of water. Thien et al.,¹⁵³  used NMR to demonstrate the formation of a saponite/hectorite-like phyllosilicate from the reaction of nuclear waste containment glass with water. ²⁷Al NMR confirmed both six- and four-coordinate Al was present (in a 2 : 1 ratio), while the ²⁹Si spectrum showed a peak at −93 ppm (Q³). The reaction of saponite with Lu³⁺ was followed using ²⁹Si NMR spectroscopy.¹⁰⁶  The untreated material showed an intense peak at −95.8 ppm (Q³(0Al)) and two weaker peaks (at −90.8 and −85.0 ppm), corresponding to Q³(1Al) and Q³(2Al). Upon treatment, only the peak at −95.8 ppm remained (showing a leaching of Al from the tetrahedral sheets), with new peaks demonstrating the formation of Lu₂Si₂O₇. As with other clays, NMR has also been used to understand the interaction between the clay mineral and organic molecules. Jaber et al., used ¹³C NMR to show that l-DOPA zwitterions were accommodated vertically in the interlayer region as a monolayer,¹⁵⁴  but that the basic pH of the layers catalysed oxidation by dissolved O₂ to indolic species and polymerisation to pseudomelanin. Work in 2010 used ²⁹Si NMR to investigate clay-propranolol dispersions used for controlled drug release,¹⁵⁵  demonstrating an electrostatic interaction between the clay and the organic species.

Hectorite is a clay mineral with formula Na_0.3(Mg,Li)₃Si₄O₁₀(OH)₂, that is relatively rare in nature, but has generated industrial interest and is often produced synthetically. Cation exchange in hectorite was studied by Hunger and co-workers.^151  23Na MQMAS NMR showed three signals in dehydrated hectorite, attributed to Na in well-defined sites in the interlayer space. Recent interest has focussed the water within the clay layers and the effect of interlayer cations. ¹H and ⁷Li variable-temperature NMR was used to study the water and interlayer cations in flurohectorite, as a function of humidity.¹⁵⁶  NMR and XRD suggested a model with 1.5 layers of water in the interlamellar space, and provided evidence for water dynamics. The effects of charge-balancing cations upon the mineral-water interface in Na-hectorite were also studied using ²H and ²³Na NMR.¹⁵⁷  Two ²³Na resonances were observed fordry hectorite, (P_Q=3.0 and 2.7 MHz), but a much narrower resonance was observed for the hydrated material. ²H wideline NMR revealed anisotropic motion of water close to the clay surface at lower temperatures, with spectra dominated by diffusion and exchange at higher temperatures. The influence of water on cation motion was shown to decrease from Na⁺ to K⁺ to Cs⁺. Bowers¹⁵⁸  investigated Ca-hectorite, with ⁴³Ca NMR showing that at low temperatures Ca²⁺ exists in a two-layer hydrate, while at high temperatures a water/solid paste forms. Lineshapes indicate diffusion-based motion, rather than interactions with the clay surface. ²H NMR revealed a similar picture for Na-hectorite,¹⁵⁷  suggesting water dynamics mechanisms are only weakly influenced by the metal cation.

The exact structure of beidellite, a 2 : 1 dioctahedral clay mineral, is not known. Loera et al.¹⁵⁹  investigated the synthesis of aluminosilicates under high pressure, using sulphur as a directing agent. The materials produced were shown to contain significant amounts of beidellite (Na_0.3Al₂(Si,Al)₄O₁₀(OH)₂·2H₂O), characterised by ²⁹Si signals at −94, −86 and −82 ppm, corresponding to Q³ species with differing numbers of Al^IV and Al^VI NNN, and both Al^IV and Al^VI in the ²⁷Al MAS NMR spectrum, in a ∼1 : 2 ratio. In other work, beidellite was used as a model system to gain insight into the formation of metastable hydrous aluminosilicates,¹⁶⁰  with REDOR used to probe the Al–H distances for Al^VI and Al^IV.

Laponite is an industrial synthetic clay with applications in hybrid materials and coatings. Work in 2013¹⁶¹  used ²⁹Si DNP (dynamic nuclear polarization), i.e., transferring polarisation from the electrons of a polarising agent to the nuclei in the clay, to study the surface and local structure of laponite particles. Both indirect DNP (where ¹H spin diffusion distributes polarisation over the whole sample) and direct DNP (which only enhances the signals of ²⁹Si sites nearer to the polarizing agent), were used, providing complementary information. The reaction of laponite to form a hybrid or intercalated materials was also studied using NMR spectroscopy.^{154,162–164} 

One of the most studied smectite is montmorillonite, a 2 : 1 phyllosilicate with greater than 50% octahedral charge owing to the substitution of Mg for Al, giving a general formula of (Na,Ca)_0.33(Al,Mg)₂(Si₄O₁₀)(OH)₂·nH₂O. A detailed structural investigation into synthetic and natural montmorillonite was undertaken by Cadars et al.^165  29Si NMR spectra contained an intense peak at −93.7 ppm (Q³) and a smaller shoulder at −88.6 ppm (Q³(1Al)). The observation of impurity peaks in the Si spectrum (not detected by XRD) also enabled a more accurate determination of the composition of the synthetic sample. ²⁷Al NMR showed both Al^IV and Al^VI, while 2D ²⁹Si/²⁷Al HMQC experiments confirmed the spectral assignment. ²⁵Mg NMR revealed a distribution of NMR parameters (average C_Q=3.3 MHz), reflecting the Mg/Al disorder in the octahedral sheets. The spectral assignments were supported by DFT calculations of a series of model structures. Sanders et al.¹¹⁴  used a fluorinated probe molecule to study the reactive surface area of montmorillonite, with quantitative ¹⁹F spectra able to provide information on the (mass normalized) number of reactive hydroxyl sites. Proof that the probe molecule was attached was obtained from ²⁹Si CP MAS spectra (δ=13 ppm). Reinholdt et al.¹⁶⁶  used NMR to characterise hydrothermal Ni–Al montmorillonites. NMR (²⁹Si and ²⁷Al) indicated substitution of Al for Si in the tetrahedral sheet, while ¹⁹F NMR highlighted clustering both of the metal cations and of the vacancies in the octahedral sheets.

The thermal treatment of montmorillonite clays is also of considerable interest, particularly for the pozzolanic activity of the products and their use in cement and concrete. Recent studies include those using NMR to monitor the effect of thermal and mechanical treatment.^{122,148,167,168}  In ref. 148, thermal treatment of montmorillonite was compared to similar treatment of illite and kaolinite. Upon heating, a change in Al coordination from six to four was observed in ²⁷Al MAS spectra, although the order of the structural layers appeared more strongly conserved for montmorillonite than kaolinite. Fernández et al.¹⁶⁷  considered both thermal and mechanical treatment of montmorillonite, and the effect of various cations. Structural changes were followed by ²⁹Si and ²⁷Al NMR, with the behaviour observed dependent upon the charge and size of the interlayer cations. A detailed NMR study of the thermal activation of montmorillonite was undertaken in ref. 168, using ²⁹Si MAS/CPMAS and ²⁷Al MAS/MQMAS NMR. Distortions of the SiO₄ and AlO₆ sites occurred on heating, with a significant degree of disorder exhibited upon dehydroxylation. Breakdown of the clay layers to form inert crystalline phases was observed at 1000–1100 °C. Thermal reaction of montmorillonite (under alkaline conditions) to form zeolites was followed using ²⁹Si and ²⁷Al NMR in ref. 122, with zeolite HS formed after 6 days. The reaction of montmorillonite with water and supercritical CO₂ was studied in situ using ¹³C MAS NMR (at 50 °C and 90 bar).¹⁶⁹  Water and CO₂ were shown to be present simultaneously in the interlayer space, with the ¹³C spectrum having two components – one narrow line (bulk CO₂) and one broader, attributed to rotationally-confined CO₂ between layers. NMR has also been widely applied to study the modification of montmorillonite and the reactivity of the modified materials. Work includes studies of pillared clays,^170–174  grafting of organic materials^{149,175–179}  and the formation of nanocomposites/hybrid materials.^155,180,181 

Like illites, vermiculites are similar to micas (particularly phlogopite), as they contain an octahedral layer of (Mg, Fe) ions between two tetrahedral silicate layers. The layer charge is balanced primarily by interlayer Mg²⁺ (rather than K⁺, as in phlogopite), although vermiculites with other cations can be prepared. There are also (in contrast to phlogopite) water molecules between the structural layers. Recent NMR studies have focussed on the mobility of the water in the interlayer space,¹⁸²  using ¹H NMR spectroscopy and T₁ relaxometry. Hongo et al.¹⁸³  considered the effect of mechanochemical treatment of vermiculite clays, using ²⁹Si NMR to show how this increases both the proportion of amorphous silica, and the number of surface hydroxyls. NMR spectroscopy was also used by Alba and co-workers to study the reaction of vermiculite with Lu³⁺ solutions.¹⁰⁶  The ²⁹Si spectrum of the treated material showed the formation of Lu₂Si₂O₇, while ²⁷Al MAS NMR showed leaching of Al from the tetrahedral sheets. Modification of vermiculite clays has been followed using NMR,¹⁸⁴  with the grafting of organosilanes confirmed by the appearance of new peaks (at −60 and −67 ppm) in the ²⁹Si MAS NMR spectrum.

Palygorskite and sepiolite are fibrous minerals with much in common with the clay minerals (and are often categorised as such), with tetrahedral silicate sheets joined by ribbons, rather than sheets, of (Al, Mg, Fe) octahedra, leaving channels that can easily accommodate small molecules. Early NMR studies¹⁸⁵  confirmed ²⁹Si resonances at −92, −95 and −98.5 ppm for sepiolite, and −92 and −98.5 ppm for palygorskite, all assigned to Q³(0 Al) species (both also exhibit Q² signals from “external edge” sites). Work in 2014 studied the effect of the Si precursor (sodium silicate or tetraethoxysilane) on the structure and catalytic properties of synthetic sepiolite.^186  29Si NMR showed the products were similar, but the Q³/Q² ratio was greater for that synthesised from sodium silicate, indicating a greater degree of polymerisation. The two materials had different morphologies and catalytic activity.

Most recent NMR studies of palygorskite and sepiolite have focussed on their modification and reactivity. A number of studies^187–189  have been concerned with the Maya blue pigment, which has major components of palygorskite/sepiolite and indigo. Work on sepiolite-based pigments¹⁸⁸  showed that indigo adsorption affected the ¹H NMR signals from bound water, suggesting the indigo was within the channels, but no evidence of strong hydrogen bonding was observed. ¹³C NMR showed small shifts for some resonances when the pigment was formed, suggesting the presence of an interaction (albeit weaker than for palygorskite pigments) between the dye and mineral. Sepiolite pigments were also studied by Raya et al.¹⁸⁷  who used 2D ¹H/²⁹Si correlation experiments to assign the ²⁹Si resonances and provide direct evidence that indigo molecules are inserted into the sepiolite structure (as no interaction between indigo and the external silanols was observed). Natural and synthetic palygorskite pigments were studied using ²⁹Si, ²⁷Al, ¹³C and ¹²⁹Xe NMR, and the effect of aging also investigated.^189  29Si and ²⁷Al NMR showed little difference between materials produced from natural or synthetic dyes, whereas ¹³C NMR revealed that the natural material contained indoxyl molecules in the tunnels, which oxidised over time. Additional studies have focused on of clay/polymer nanocomposites.^190–192  In ref. 192, ²⁹Si MAS NMR showed loss of Q² silanols in the composite, and a new peak (at −100 ppm) from Q⁴ Si, demonstrating strong cross linking. In ref. 190–192, wideline and DQ ¹H NMR detected three fractions within the clay – a rigid part, an interface component and a mobile amorphous fraction.

An important property of palygorskite/sepiolite minerals is their ability to incorporate heavy metals, of importance for their application in decontamination processes. Many authors have attempted to modify the clay to control or improve the adsorption properties, using NMR to understand the structural changes. These include acid treatment,¹⁹³  and grafting of organosilanes and amino groups.^149,194,195  In a similar approach, Wicklein et al.¹⁹⁶  studied sepiolite modified by phosphatidylcholine, as an immobilisation host for biological species, with evidence of controlled formation of lipid bilayers provided by solid-state ³¹P MAS NMR.

Tectosilicates, (or framework silicates), with infinite 3D networks of silicate tetrahedra and general formula SiO₂, form the largest rock-forming mineral group, comprising ∼75% of the Earth's crust.¹  The simplest tectosilicates are the silicas, electrically neutral minerals, with no additional cation or anion species. There are many polymorphs of SiO₂, including quartz (α and β), tridymite (α and β), cristobalite (α and β), coesite and stishovite, with stability fields defined by varying temperature and pressure.^1  29Si NMR parameters have been determined in early work.¹⁹⁷  The detailed surface structure of quartz and cristobalite was investigated more recently by Murray,¹⁹⁸  who used ²⁹Si CPMAS NMR to differentiate and characterise geminal (i.e., Q²) silanols, which accounted for ∼15% of the hydroxyl groups. Recent work observed the formation of quartz and cristobalite (characterised by Q^4 29Si chemical shifts of −107 and −109 ppm) from the thermal treatment of silica-germania geopolymers,¹⁹⁹  silica microspheres²⁰⁰  and pyrophyllite clay,¹¹⁰  and also in the walls of organosilica films.²⁰¹  Work in 2012²⁰²  used ²⁹Si and ¹³C NMR to observe the formation of α-cristobalite from the thermal reaction of semicrystalline inorganic–organic polyhedral oligomeric silsesquioxane (POSS). Spektor et al.²⁰³  considered the hydration of stishovite (a high-pressure form of SiO₂ stable from 9 to 50 GPa). Hydrous stishovite was prepared by hydrothermal treatment of silica glass and coesite between 350–550 °C at 10 GPa. Two resonances were observed by ²⁹Si NMR, at −191 ppm (Si^VI, also seen in the anhydrous material), and −188.6 ppm, enhanced using CP MAS. ¹H MAS NMR showed peaks at 0.5, 4.7, and 10.5 ppm. The hydrogen was shown to be incorporated via hydrogarnet defects, where a cluster of four OH⁻ groups replaces SiO₄⁴⁻.

Opal is a hydrated amorphous silica, with a water content ranging from 3 to 21 wt%, and is, therefore, a mineraloid, rather than a mineral, owing to its non-crystalline character. Experimental evidence (including NMR spectroscopy) for the structure of opal was reviewed recently by Wilson.²⁰⁴  Early work on opals was reconsidered, and the mean Si–O–Si angles (138–170°) were calculated from the ²⁹Si chemical shifts (−110 to −112 ppm for Q⁴ species). Opal-CT (named after the disordered cristobalite and tridymite it was thought to contain) was actually shown to resemble tridymite most closely. Stebbins and co-workers²⁰⁵  used ²⁹Si MAS NMR to quantify the network speciation and hydroxyl content of opals. Spectra consisted of two broad, overlapping peaks at −111 and −102 ppm, with a smaller peak at −92 ppm, corresponding to Q⁴, Q³ and Q² Si, respectively. Different intensity ratios were observed for opals from different locations, with the Hawaiian silica (basalt coating) having the highest water content (∼5.4 wt%). Zhou et al.²⁰⁶  used ²⁹Si, ¹H and ¹³³Cs MAS NMR to study geyserite, a Cs-bearing opal. The relative amounts of Q⁴, Q³ and Q² Si varied with the mineralisation stage, with ageing increasing the degree of condensation. Cs acted as a network modifier, associating with Q³ and Q² silanols, and being coordinated by O²⁻, OH⁻ and H₂O.

Feldspars are a group of rock-forming minerals that make up ∼60% of the Earth's crust. Alkali feldspars have compositions between KAlSi₃O₈ (sanidine, microcline, orthoclase) and NaAlSi₃O₈ (albite), while plagioclase feldspars have compositions between NaAlSi₃O₈ and CaAlSi₃O₈ (anorthite). The considerable amount of early NMR work on feldspars is reviewed in ref. 2. Recent work²⁰⁷  has included a very detailed study (using high-field (19.6 T) MAS, MQMAS, CP MAS, DQ MAS and 2D correlation experiments), of anorthite, enabling a detailed picture of the local cation disorder to be established. Sánchez-Muñoz,²⁰⁸  investigated cation disorder in K-rich feldspars, using ²⁹Si, ²⁷Al, ³⁹K and ²³Na NMR to show that the “ideal” average structures produced by diffraction are often far from reality, with only microcline showing long-range order. Other feldspars exhibited long-range disordered structures with non-random disorder at the medium-range scale, and considerable distortions producing different tetrahedral framework sites and two types of alkali metal sites. More recently ³¹P NMR was used to study a variety of K-feldspars that contain up to 1.5 wt% P₂O₅.²⁰⁹  The feldspar formed depended upon the amount of P present. Other recent NMR studies involving feldspars include work using albite as a model sample for method development,⁶³  and the decomposition of microcline to form kalsilite (KAlSiO₄).²¹⁰ 

Feldspathoid minerals are tectosilicates that resemble feldspars but typically have lower silica content. Extensive studies of Si/Al distribution in these materials have been made in early NMR work.²  More recent studies have used NMR to follow the formation of feldspathoids from a range of precursors, rather than structural studies. This has included the synthesis of cancrinite from thermal activation of kaolin²¹¹  and saponite gels,²¹²  synthesis of leucite from thermal treatment of aluminosilicate geopolymers⁴¹  and the synthesis of nepheline by heating Na-rich zeolites.²¹³  Many authors have also used NMR to study the synthesis of sodalite (Na₈(Al₆Si₆O₂₄)Cl₂), and related feldspathoids such as hauyne or nosean.^{211,214–216}  Rivera et al.²¹⁷  also considered the formation of sodalite and cancrinite during homogeneous nucleation from caustic solutions, for applications in waste remediation. NMR of K- and Rb-exchanged sodalite was considered by Igarashi et al.^{218,219  27}Al NMR spectra showed antiferromagnetic transitions at ∼70 K and ∼80 K for Rb- and K-sodalites.

Analcime (Na[AlSi₂O₆]·H₂O) is a hydrated sodium aluminosilicate framework mineral, with large cavities occupied by water molecules, and a set of smaller channels partially occupied by Na.¹  Although often referred to as a zeolite mineral (see below), it is structurally and chemically more similar to feldspathoids. Kim et al.²²⁰  studied analcimes with varying Si/Al ratios using ²⁹Si NMR spectroscopy. Four peaks were observed (Q⁴(0–3Al)), with changes in chemical shift (as the Al content varied) dependent upon the distances between Si and the tetrahedral atoms on the second- and fourth-nearest neighbour sites. Work in 2012¹⁵⁹  studied analcime synthesised hydrothermally at high pressure. The presence of Na₂S in the reaction mixture resulted in S substitution into the framework, evidenced by new peaks in the ²⁷Al (72 ppm, 7 T) and ²⁹Si (−81 and −84 ppm) NMR spectra. Recent NMR studies have also investigated Mn- and Co-doped analcimes.^221,222  Little effect was observed in the ²⁹Si spectrum upon substitution of the paramagnetic ions, suggesting the materials were heterogeneous on the nanoscale. Analcime was studied by Mason et al. as a model system, to gain insight into the formation of metastable hydrous aluminosilicates,¹⁶⁰  with strong dipolar couplings between the framework and the confined, motionally-restricted water leading to significant REDOR dephasing.

The zeolite minerals are aluminosilicate frameworks, which contain large cavities occupied by cations and water molecules that have considerable motional freedom.¹  Chemically, zeolites are related to feldspars, but have much more open structures resulting in facile ion exchange and dehydration. There are 45 naturally-occurring zeolites with ∼25 different structures. Zeolitic channel structures are formed from different combinations of linked rings of aluminosilicate tetrahedra, with minerals subdivided by structure. In addition, many zeolitic materials can be produced synthetically, although attention is typically focused on materials and structures with the most industrial relevance. The NMR literature on zeolites (and zeotype analogues) is vast and more complete reviews can be found in ref. 2, 223–226.

Natrolite is a member of the zeolite sub-group which bears its name and has ideal composition Na₁₆[Al₁₆Si₂₄O₈₀]·16H₂O.¹  Work in 2013²²⁷  studied a range of ion-exchanged natrolites with both alkali and alkaline Earth cations, using ²⁹Si MAS, ²⁷Al MAS and MQMAS/STMAS experiments. Two ²⁹Si resonances are observed in the Na form (at −87.6 and −95.1 ppm), corresponding to Q⁴(3Al) and Q⁴(1Al), reflecting the complete cation ordering of the framework. For exchanged species ²⁷Al NMR showed more than the single Al site expected, suggesting that the actual crystal symmetry is lower than that reported. The Ca end member (scolecite) was also shown to have lower symmetry than the reported crystal structure.²²⁷ 

The chabazite group consists of minerals with structures containing single or double rings of six tetrahedra, with general formula XAl₂Si₄O₁₂·6H₂O, where X is typically Ca, K, Na or Sr.¹  The simple, yet open, structure of these materials has resulted in many industrial applications as catalysts, and for storage and separation, with much research also focused on phosphate-based analogues.^224–226  Recent work has included ¹H NMR of the water in the chabazite cages, with concentration-driven phase transitions observed for materials with ∼4.2 and ∼5.1 water molecules per formula unit.²²⁸  NMR has been used to study the synthesis of chabazite.^229,230  Chabazites with a range of Si/Al ratios were synthesised,²²⁹  with ²⁷Al NMR demonstrating the absence of extraframework Al (i.e., only Al^IV was observed). A re-assignment of the ²⁹Si spectra in previous work (to take into account Q³(0Al) defects) enabled an accurate determination of the Si/Al ratio. Liu et al.²³⁰  also used ²⁹Si, ²⁷Al and ¹⁹F NMR to study chabazites prepared in fluoride media, with NMR verifying the presence of a single type of F⁻ in the zeolite channels, and following the effect of fluoride levels on the synthesis. NMR has also been used to investigate the reaction of chabazites, including the thermal durability of chabazite-based catalysts,²³¹  reduction in Ag-exchanged chabazite²³²  and the role of water in the catalytic performance.²³³  Chabazite has also been used as a model sample to investigate methods for extracting the distributions of ²⁷Al NMR parameters from MQMAS spectra in disordered materials.²³⁴ 

Clinoptilolite and heulandite are isotypic natural zeolites that differ in Si/Al ratio (clinoptilolite having Si/Al>4 and heulandite Si/Al<4). Variation of the cations gives clinoptilolite-Na, clinoptilolite-K, etc. The ²⁹Si NMR spectrum of natural clinoptilolite²³⁵  showed three peaks, corresponding to Q⁴(0–2Al), and one with lower intensity (at −95 ppm) arising from small amounts of Fe³⁺. ²⁷Al MAS NMR showed only Al^IV was present. This work also showed that repeated washings with aqueous HCl resulted in dealumination, producing extra-framework Al^VI, and an increase in the Q³(0Al) signal.²³⁵  A number of recent studies have focused on the water within the zeolite pores, including work by Gabuda²³⁶  who showed (by ¹H NMR) that penetration of NH₃ within the pores resulted in disorder, and rapid exchange of the water protons. ¹H NMR was also used to study water in single crystals of clinoptilolite and heulandite.²³⁷  Below 170 K water was located at fixed positions (different for the two minerals), while above 290 K translational and orientational diffusion of water was present, with the water structure in the two materials almost identical. ²⁷Al/¹H REDOR showed little dephasing for clinoptilolite,¹⁶⁰  indicating that the zeolitic water is not rigidly bound. The role of water in the catalytic performance of clinoptilolite was investigated using ²⁹Si CP MAS.²³³ 

Mordenite ((Ca,Na₂,K₂)Al₂Si₁₀O₂₄·7H₂O) is one of the most abundant zeolites and has a range of commercial uses. Mordenite has been the subject of many NMR investigations in recent years with work focused on its synthesis,^238,239  structure,^160,240  doping,^221,239,241  dealumination/desilication,²⁴¹  adsorption of small molecules,²⁴²  thermal treatment,²⁴³  formation of hybrid materials²⁴⁴  and catalytic activity.^240,242  Highlights include the synthesis of high-silica mordenite using a dual-templating method, where two organic structure-directing agents were used.^238  13C NMR showed both templates were present in the final material (in a ratio of 65 : 35), while ²⁹Si NMR confirmed the Si/Al ratio. Caldarelli et al., used NMR spectroscopy to study the insertion of Co into mordenite,²³⁹  with pseudocontact shifts in ²⁹Si and ²⁷Al NMR spectra used to estimate the number of framework sites in the proximity of the paramagnetic centre. Results showed that the Co was located preferentially at siloxane bridges, close to Al in the framework. Work by Grey and co-workers²⁴⁰  used ¹H–¹⁷O experiments (HETCOR and REDOR) to study the acid sites in mordenite, and distinguish those with different O–H distances. ¹H resonances at 1.8, 4.0 and 6.2 ppm were assigned to Si–OH, Si–OH–Al and Si–OH…H₂O–Al sites, and ¹⁷O resonances assigned using correlation experiments. The accessibility of the acid sites was quantified by adsorbing known amounts of a probe molecule (trimethylphosphine). Work in ref. 244 used ²⁷Al and ³¹P NMR to explore the ability of mordenite to form an extraframework AlPO₄ phase by reaction of a phosphate precursor with expelled framework aluminum. More crystalline AlPO₄ was found outside of the zeolitic channels, with amorphous AlPO₄ found within.

As shown in Fig. 1b, the inner Earth can be divided into chemically-distinct layers.⁴  The outer layer, or crust, is composed primarily of aluminosilicates and is typically 5–50 km thick. The Mohorovicic discontinuity, at the boundary with the mantle, denotes a significant change in composition. The mantle is further subdivided by discontinuities resulting from changes in the minerals within the layers, with olivine (Mg,Fe)₂SiO₄ the major component in the upper mantle (down to depths of 410 km), before transformations to wadsleyite, β-(Mg,Fe)₂SiO₄, in the upper transition zone (between 410 and 530 km), ringwoodite, γ-(Mg,Fe)₂SiO₄, in the lower transition zone (to depths of 660 km), and (Mg,Fe)SiO₃ perovskite in the lower mantle. Owing to the difficulties associated with obtaining natural mantle minerals, most NMR studies have focused on synthetic samples produced at high pressure, resulting in very small sample volumes. A synthetic approach does, however, enable isotopic enrichment of species such as ²⁹Si, ²⁵Mg or ¹⁷O, significantly improving spectral sensitivity. For a more detailed review on NMR of mantle silicates see ref. 7.

Wadsleyite, β-Mg₂SiO₄, is a pyrosilicate containing a single Q¹ Si species, four O (one bridging (O2), two non-bridging (O3, O4) and an ‘underbonded’ oxygen (O1) coordinated to five Mg²⁺ cations), and three Mg species. Few NMR studies have been reported, probably as a result of the demanding synthesis conditions (16 GPa and 1600 °C) and the small sample volume produced. Early work by Stebbins²⁴⁵  determined a (Q¹) Si shift of −79 ppm, while the four ¹⁷O sites were resolved by Ashbrook using STMAS.²⁴⁶  The experimental NMR parameters were subsequently confirmed by DFT calculations.¹⁴  NMR of wadsleyite was recently reviewed by Griffin and Ashbrook.⁷  Owing to its high capacity for water storage, recent work has focussed on hydrous wadsleyite. In 2013, Griffin et al.²⁴⁷  carried out a combined experimental and computational investigation of the hydration of wadsleyite. As shown in Fig. 4, ¹⁷O NMR spectroscopy revealed a broad resonance with large C_Q at low chemical shift, attributed to hydroxyls, and a decrease in intensity of the O1 signal, suggesting that the majority of protonation occurred here. Broadening of the STMAS spectrum suggested a distribution of parameters, and significant structural disorder. ¹H MAS, DQ NMR spectroscopy, two-dimensional heteronuclear correlation experiments and DFT calculations (see also Fig. 4) confirmed that most protonation occurred at O1, but that protonation of the silicate oxygens was also present (with ¹H chemical shifts from 6 to 10 ppm). Comparison of experimental and calculated ²H C_Q values revealed that silicate protonation was primarily at the non-bridging oxygens, O3/O4, and not at the bridging O2. Best agreement between experimental and calculated ¹⁷O spectra was obtained with an Mg₃ vacancy.²⁴⁷ 

With increasing pressure, wadsleyite transforms to ringwoodite γ-(Mg,Fe)₂SiO₄, an orthosilicate spinel, containing one type of Si, O and Mg. The high pressure (20 GPa) required for synthesis has limited the application of NMR, as described in a recent review.⁷  A low ²⁹Si chemical shift is observed for ringwoodite (−82 ppm), at the extremity of the Q⁰ range, although this has been confirmed many times, including by work on ²⁹Si-enriched material.^{7,14,248,249  17}O NMR parameters were determined by STMAS.^7,14  The C_Q of 4.8 MHz is larger than usual for non-bridging oxygen, but is supported by DFT calculations.^7,14  Ringwoodite can also accommodate significant amounts of water (up to 3 wt%). Ye et al., studied hydrous ringwoodite using XRD and previous NMR by Stebbins et al.,^248–250  to demonstrate that hydration takes place via substitution of Mg²⁺ for 2 H⁺, and that at higher temperatures a small amount of H⁺ moves from Mg to Si sites, with Si moving to the vacant octahedral site.

At the higher pressures and temperatures found in the transition zone the pyroxenes found in the upper mantle transform to majorite (MgSiO₃), which has a garnet structure (described as Mg₃(Mg,Si)Si₃O₁₂). As discussed in a recent review,^7  29Si and ¹⁷O NMR parameters were determined in 2007.⁶⁹  The four distinct Si sites (including six-coordinated Si at −197.1 ppm) were assigned using periodic DFT calculations, and variable field STMAS experiments were used to resolve the O species. As expected, larger C_Q values were found for the three bridging oxygens (4.4–4.8 MHz) than for the three non-bridging oxygens (2.8–3.0 MHz).⁷  The cation ordering in the solid solution between majorite and pyrope (Mg₃Al₂Si₃O₁₂) has been studied using ²⁹Si NMR spectroscopy^2,251  with the proportion of four- and six-fold Si/Al determined. ²⁷Al MAS NMR spectra revealed all Al was six-fold coordinated, while the ²⁹Si spectrum showed the presence of Mg/Si NNN disorder.

Another polymorph of MgSiO₃, of relevance to the transition zone, is akimotoite, which has an ilmenite structure with one O and one Si^VI species. As described in ref. 7, ²⁹Si NMR shows a single peak at −181 ppm, with no evidence for significant cation disorder.^69,245  The ¹⁷O MAS spectrum reveals one lineshape with C_Q=3.4 MHz, low for a bridging oxygen species, but in agreement with calculation.^7,69 

As the pressure increases at greater depths, there is a change in the composition of the mantle, with MgSiO₃ perovskite the major component in the lower mantle. ref. 7 recently reviewed NMR of this phase. The ²⁹Si spectrum contains a single Si species (−191.7 ppm), in a highly-symmetric environment with no CSA, and no evidence for any significant cation disorder.²⁵²  Later ¹⁷O MAS and STMAS spectra were unable to resolve the two O resonances expected, but DFT calculations predicted very similar NMR parameters for the two sites, with almost no difference in C_Q and a very small difference in η_Q.^7,69  There has been significant interest over the years in the substitution of Al into perovskite. Although Al is much less abundant in the mantle than the crust, Al-rich minerals may be found in subducted slabs, with the substitution of minor levels of Al into the mantle having an important effect on its physical and chemical properties. In 2012, Palke et al. studied Al- and Fe-bearing perovskites with compositions (Mg_1−xFe_x)(Si_1−xAl_x)O₃, where x=0.01, 0.025 and 0.05,²⁵³  building on earlier work reviewed in ref. 2 and 7. Al was found to substitute onto the (tetrahedral) Si site and the (octahedral) Mg site, with Fe substitution resulting in broadening of the signals and loss of ²⁹Si and ²⁷Al spectral intensity. For ²⁹Si, there was a significant decrease in signal intensity as x increased from 0.01 to 0.025, with a less significant decrease with further increases in x. In contrast, the ²⁷Al signal was affected only slightly as x increased from 0.01 to 0.025, but was almost completely lost for x=0.05. The authors concluded that Al³⁺ and Fe³⁺ occupy adjacent sites in the structure, i.e., a non-random cation distribution.

Owing to the uncertainty surrounding the exact role of water in the mantle, there has been significant effort aimed at understanding the stability of stoichiometrically hydrous magnesium silicate phases. There have been a number of such phases produced synthetically, known as dense hydrous magnesium silicates or “alphabet phases”, but they have not been identified naturally. These phases could be of importance in subducted material, they may be present as impurities in synthetic silicates, and their study may provide more general insight into hydrous silicates. Most NMR measurements of the dense hydrous phases (phase B, superhydrous B, phase A, phase E and phase D) were carried out a number of years ago,^27,245,254  but have been discussed in more recent reviews.^2,7  In most cases, minerals were studied in mixed phase samples, leading to some uncertainty in the NMR parameters extracted. Griffin et al. studied superhydrous phase B (Mg₁₀Si₃O₁₄(OH)₄), to ensure this phase was not confusing the analysis of hydrous wadsleyite samples.²⁴⁷  Superhydrous phase B contains both Si^IV and Si^VI, although precise H positions (and the exact space group) have been the subject of some debate. ref. 247 showed ²⁹Si resonances at −74.8 and −166.9 ppm, in good agreement with previous work, while the ¹H spectrum contained signals at 4.2 and 3.2 ppm. A ¹H DQMAS spectrum confirmed the two ¹H sites were close in space.

Although silicates are the largest group of rock-forming minerals, many non-silicate minerals are of geological, industrial or commercial importance. Non-silicates are diverse in their origins and physical properties, and are typically divided into classes based on their chemical composition (more specifically, their anionic chemistry).¹ 

Oxide minerals are a diverse class of minerals that contain the O²⁻ bonded to one or more metals. Many simple metal oxides used routinely in the chemical industry (e.g., TiO₂, SnO₂, MgO) have natural mineral analogues. Addition of H produces both hydroxides and oxyhydroxides. Previous NMR investigation of these minerals tended to focus on ¹⁷O NMR,²⁵⁵  although recent improvements in magnetic field strengths and MAS rates have enabled the NMR study of more metal centres.

Corundum (α-Al₂O₃) is the naturally-occurring and most stable of the ∼8 polymorphs of alumina (known as transition aluminas). Although not natural, γ-Al₂O₃ is of considerable industrial interest as a catalyst support material. Recent ²⁷Al NMR studies of aluminas have focussed on the synthesis of α-Al₂O₃ from other alumina polymorphs and hydroxides,^256–258  including investigation of mechanical transformation²⁵⁶  and the effect of particle size on the transition temperatures.²⁵⁸  Other work used ³¹P NMR to study the sorption of phosphate onto corundum, with ²⁷Al/³¹P REAPDOR used to demonstrate that the majority is bound in a bidentate manner.^259,260  DFT calculations of aluminas, including corundum, have shown good agreement with experimental measurements.²⁶¹  For γ-Al₂O₃, four structural models were considered, with the Fd–3m structure showing the best agreement.

Aluminium hydroxide, Al(OH)₃, is found naturally as the mineral gibbsite, although rarer polymorphs (including bayerite) exist. Also closely related are boehmite and diaspore, aluminium oxyhydroxide (AlO(OH)) polymorphs. Early work utilised MAS NMR spectroscopy to probe the coordination number and local environment of Al and the number/type of ¹H sites.⁸  More recently, high-resolution ¹H CRAMPS experiments were used to resolve six ¹H sites in gibbsite, and to ascribe them to OH groups with intralayer or interlayer hydrogen bonds.²⁶²  This work also used DFT calculations to assign the ²⁷Al spectrum, and to explain the different C_Q values (4.6 and 2.2 MHz) exhibited. Many authors have used ²⁷Al NMR to study the thermal or mechanical transformation of gibbsite, often via boehmite, to form transition aluminas and α-Al₂O₃.^43,257,263  The sorption of phosphate on gibbsite, bayerite and boehmite was studied in ref. 260 using ³¹P NMR, revealing bidentate binding, with similar ³¹P shifts for gibbsite and bayerite, but significantly different shifts for boehmite. The uptake of phosphate in Ca-substituted boehmite (to form biologically-relevant phases) has also been studied using ³¹P NMR.²⁶⁴ 

There have also been a number of recent NMR studies of other simple oxide and hydroxide minerals. This has included ²⁹Si NMR of natural silicified pyrochlores²⁶⁵  and ³¹P/⁷Li MAS NMR to study phosphate and lithium adsorption on the iron oxyhydroxides, goethite, akaganite and lepidocrocite.^{266,267  43}Ca high-field NMR spectra of portlandite (Ca(OH)₂) and CaTiO₃ perovskite²⁶⁸  were acquired by Bryce and co-workers, with supporting DFT calculations able to demonstrate clear correlations between NMR observables and structural parameters, such as Ca–O bond distances, as shown in Fig. 5. This work also demonstrated the first example of ¹H/⁴³Ca CP in natural abundance portlandite, also shown in Fig. 5. Recent NMR studies of brucite (Mg(OH)₂) have included the determination of ²H NMR parameters (and comparison to those in the brucite-like layers of the humite minerals)²⁹  the study of reaction with supercritical CO₂ (in relevant conditions for geological CO₂ sequestration) using in situ ¹³C NMR²⁶⁹  and the determination of the relative orientation of the OH dipolar tensor and the ¹⁷O quadrupolar tensor using a 3D experiment.²⁷⁰  Sun et al. used ¹⁷O NMR to investigate isotope exchange in TiO₂ (anatase and rutile),²⁷¹  with activation energies determined, and preferred exchange of surface sites suggested.

Spinels are cubic minerals with general formula AB₂O₄, where A is Mg (in spinel itself) or Be, Zn, Mn, Fe, Cu, Ni and B is Al (in spinel) or Fe, Cr or V. In normal spinels, A and B cations occupy tetrahedral and octahedral sites, while in inverse spinels all A cations and half of the B cations are octahedral, while the remaining B cations occupy tetrahedral sites. NMR has been used to probe cation coordination number and disorder.²  Recent work has used ²⁷Al NMR to investigate the mechanical transformation of MgAl₂O₄, ZnAl₂O₄ and Li_0.5Al_2.5O₄ using high-energy milling.²⁷²  Mechanical action randomised cations over the two sublattices. The effect of Co doping on the thermal behavior of MgAl₂O₄ spinel was investigated using ²⁷Al NMR in ref. 273. Work by Gaudon et al.²⁷⁴  also studied doping of aluminate spinels (with Co²⁺ and Ni²⁺), for applications in blue/cyan pigments, with the cation distribution shown to have a significant effect on the colour produced.

Phosphates contain orthophosphate, PO₄³⁻, which, as for silicates, can condense to pyrophosphate (P₂O₇)⁴⁻ or form chains, rings and cages.¹  However, unlike SiO₄⁴⁻ tetrahedra, PO₄³⁻ cannot condense to form infinite frameworks analogous to SiO₂ zeolites, except in the presence of group 3 cations (e.g., Al³⁺ and Ga³⁺, as in synthetic aluminophosphates and gallophosphates). ³¹P (I=1/2, 100% abundance) is particularly amenable to study by NMR spectroscopy, as its large (∼400 ppm) shift range and high receptivity make it an excellent structural probe.

While no aluminophosphate zeolites occur naturally, the dense hexagonal quartz-like phase of AlPO₄, berlinite, is a natural mineral. Early work on berlinite and the other dense polymorphs of AlPO₄, cristobalite and tridymite, showed that ³¹P and ²⁷Al NMR could distinguish between phases, and proposed a relationship between the average Al–O–P bond angle and ²⁷Al and ³¹P shifts.²⁷⁵  More recently, attempts have been made to measure the ²J_Al–O–P couplings in berlinite,^276–278  with Xue²⁷⁸  providing a comprehensive theoretical insight and showing that ²J_Al–O–P≈25 Hz.

The monazite/xenotime series is of interest for many applications, including radioactive waste storage, proton conduction and catalysis. Both materials have the formula MPO₄, where M is a rare-earth element, with xenotime containing mainly the later, smaller rare earths (Tb to Y) and having a tetragonal zircon structure, whereas monazite contains the earlier, larger rare earths (La to Gd), which distort the structure, reducing the symmetry to monoclinic. Fair et al.²⁷⁹  deposited La-rhabdophane (a monohydrated form of monazite, LaPO₄·H₂O) on the surface of Al₂O₃ fibres. ³¹P MAS and ³¹P–²⁷Al TRAPDOR experiments confirmed the formation of rhabdophane (³¹P δ_iso=−6 ppm) as well as amorphous AlPO₄. Ramesh et al.²⁸⁰  prepared porous La monazite and other LaP_xO_y materials for use as ethanol dehydration catalysts. Materials were characterised by ³¹P MAS NMR and all exhibited resonances at ca. −6 ppm. However, with the exception of La³⁺ and Y³⁺, all trivalent Ln³⁺ cations are paramagnetic and can induce significant ³¹P paramagnetic shifts. Stebbins and co-workers have investigated the effects of LaPO₄ and YPO₄ doped with Nd, Ce, Pr, Dy and V.^281,282  Owing to the compact nature of the electronic f orbitals, paramagnetic shifts are relatively small (compared to those from d electrons), ca. −200 ppm (Y_0.99Nd_0.01PO₄) to +250 ppm (Y_0.99Eu_0.01PO₄). Consideration of the effects of paramagnetic substitution of the seven cations closest to ³¹P allowed Palke et al. to demonstrate that the cation substitution is random in La_1−xCe_xPO₄, Y_1−xCe_xPO₄ and Y_1−xNd_xPO₄.²⁸²  More recently, Maron et al. studied La-monazite doped with small amounts of Nd and Gd, showing that the ³¹P spin–lattice relaxation rate (1/T₁) varies linearly with doping level, enabling its use for accurately quantifying amount of present, particularly below 1% doping.²⁸³ 

With the increased importance of alternative and cleaner energy technologies, olivine-type lithium phosphates such as triphylite (LiFePO₄) and lithiophilite (LiMnPO₄) have emerged as potential electrode materials for lithium ion batteries. Despite the high sensitivity of ³¹P and ⁷Li, the experiments can be challenging, owing to the presence of paramagnetic Fe species. While many NMR spectra of these materials have been reported in recent years (see ref. 284 and 285 for more discussion), these studies typically provide insight into the electrochemical devices rather than the phosphate electrode material, and are not discussed further. However, Grey and co-workers have also carried out more fundamental work on paramagnetic phosphates, including heterosite (FePO₄), strengite (orthorhombic FePO₄·2H₂O), phosphosiderite (monoclinic FePO₄·2H₂O) and triphylite, investigating the ³¹P and ⁷Li NMR parameters^286,287  and carrying out a detailed investigation into the spin transfer pathways giving rise to the paramagnetic shifts in LiFe_xMn_1−xPO₄.^287,288 

Phosphates are also biologically important, particularly apatites, which are present in bones and teeth. Apatites have the nominal formula Ca₁₀(PO₄)₆(X)₂, where for hydroxyapatite, X=OH, fluorapatite, X=F and chlorapatite, X=Cl. However, the mineral displays significant compositional variation, particularly in synthetic apatites, with substitution of CO₃²⁻, SiO₄⁴⁻ and Mg²⁺ into the structure investigated in recent years.^289–300  Despite its unfavourable NMR properties, natural-abundance (although typically high-field) ⁴³Ca NMR has proved a useful probe of cation substitution in apatites in recent years.^{289,291,293,301}  The ⁴³Ca NMR spectrum of hydroxyapatite contains resonances at 11.4 and −0.6 ppm, assigned to the seven-coordinate Ca(ii) and nine-coordinate Ca(i) sites, respectively.²⁸⁹  Owing to the different coordination, preferential substitution has been observed for several cations. Of particular note is the work by Laurencin et al., who demonstrated, using ⁴³Ca NMR, XANES, EXAFS and computational methods, that Mg²⁺ preferentially substitutes on Ca(ii) rather than Ca(i).²⁹³  The substitution of CO₃²⁻ into apatites is more complicated, with two possible mechanisms: replacing X⁻ (A type) or PO₄³⁻ (B type). The two types of CO₃²⁻ can be identified by ¹³C chemical shifts of 167 and 170 ppm, respectively, in hydroxyapatite²⁹⁰  and 169 and 170 ppm, in fluorapatite,²⁹⁵  with both types of CO₃²⁻ typically observed. Apatite can also accommodate H₂O at defect sites and the H₂O within the structure has been probed by ¹H, ²H and ³¹P NMR.^{292,294,302,303}  Yoder et al., recently used ²H NMR to confirm the presence of structural water in apatites, probably within the channels parallel to the c axis.²⁹⁴ 

The surface termination of apatite crystals can affect their properties, governing how the material interacts with other components of biological or chemical systems. Owing to the sensitivity of NMR to the local structure, it is the ideal technique to investigate surface modifications of apatite^{292,304–308}  Without treatment, the surface comprises a layer of water,²⁹²  but many species, particularly biocompatible organophosphonates, have been grafted to apatite to improve properties such as biocompatibility^306,307  or ion exchange.^304,306  NMR can simply confirm successful grafting, or provide more detailed information. For example, Wu et al. used 2D correlation experiments to probe the binding of citrate ions to fluorapatite, demonstrating that citrate directs the particle morphology by forming selective hydrogen bonds to phosphate.³⁰⁷ 

Several apatite-based composite materials have also been characterised by solid-state NMR,^{300,307–311}  often with the aim of determining how the apatite surfaces interact with biological or biomimetic species. Chen et al. prepared apatite in liposomes to investigate the early stages of biogenic apatite crystallisation, with NMR providing structural insight inaccessible by other techniques such as TEM.³¹²  Vyalikh et al. used ¹⁹F MAS and ¹H–³¹P correlation experiments to investigate fluorapatite in a gelatine matrix, showing that the surface of the crystalline apatite is coated with a hydrated amorphous layer, which interacts with the protein.³¹¹ 

While inherently more complicated than their synthetic models, the apatite in biogenic samples has also been characterised by NMR.^{291,297,299,301,313–315}  In one particularly detailed study, Laurencin et al. used multinuclear solid-state NMR to investigate the local Ca and Na environments in natural bone and teeth samples.²⁹¹  By demonstrating the close spatial proximity of ²³Na and ³¹P, the study was able to provide the first direct evidence of the presence of Na within natural apatite.

There are ∼60 naturally-occurring carbonate minerals, containing the trigonal carbonate (CO₃²⁻) anion, although many of these are rare.^1  13C is suitable for routine study by NMR although CP is generally not possible for pure carbonates (as these do not contain H), but can be used to identify signals from non-carbonate species in mixed-component systems.³¹⁶  The low natural abundance and long relaxation times mean that enrichment in ¹³C is often desirable for the study of carbonate minerals and, particularly, their chemical reactions.^20,21 

Recently, Sevelsted et al.³¹⁷  measured the ¹³C chemical shift tensors of several model cementitious carbonates, which showed little variation in the isotropic (∼9 ppm) and anisotropic (∼10 ppm) components upon changes in the surrounding cations. However, the asymmetry, η_CS, showed more variation (from 0.0 to 0.5), due to deviation from axial symmetry in some cases. This parameter can also distinguish between the CaCO₃ polymorphs, aragonite (η_CS=0.2) and calcite (η_CS=0.0).

As discussed above, geologic sequestration of CO₂ has received much attention recently, with the reaction of Mg-containing minerals such as forsterite^{20,21,23–25}  or brucite²⁶⁹  with supercritical CO₂ being followed by in situ and ex situ ¹³C NMR under a variety of reaction conditions.^{20,21,23–25,269}  The products (magnesite and nesquehonite) and intermediates (hydromagnesite and dypingite) can be identified by their ¹³C chemical shifts,²¹  although the small shift range means that even small inconsistencies in referencing may lead to ambiguity.²⁶⁹  Calcium carbonates are also of interest, particularly for biomineralisation, as the shells and exoskeletons of many organisms contain CaCO₃. Understanding the structure of these materials will, therefore, assist in the development of new biocompatible and biomimetic composites. CaCO₃ has three common crystalline polymorphs, calcite (β-CaCO₃), aragonite (λ-CaCO₃) and vaterite (μ-CaCO₃). The hexagonal calcite structure is normally thermodynamically favoured, whereas the denser, orthorhombic aragonite is stabilised by the inclusion of Mg²⁺.³¹⁸  Huang et al. precipitated Mg-doped CaCO₃ (60% enriched in ⁴³Ca) from aqueous solution and showed that calcite precipitated initially, followed by re-dissolution and subsequent precipitation of aragonite.³¹⁸  The two polymorphs have small ⁴³Ca C_Q values, and can be distinguished by their isotropic chemical shifts, 19.3 ppm (calcite) and −27.2 ppm (aragonite) or CSAs, with aragonite having a larger anisotropy.

Some shellfish store amorphous CaCO₃ in “gastroliths” prior to conversion to calcite or aragonite, and recent work has focused on understanding the interactions between the inorganic and organic components of crayfish³¹⁹  and lobster³²⁰  gastroliths, and the local structure of the amorphous CaCO₃. The components of the shells of chicken eggs have also been studied by ¹³C NMR, with calcite identified by a resonance at 168.1 ppm.³¹⁶  In such organic-inorganic hybrid systems, the “spectral editing” ability of CP may be exploited to selectively study the organic components.

Sulphide minerals contain S²⁻ as the major anion, although the sulphide class is often discussed alongside selenides, tellurides, arsenides and antimonides. Conventional NMR measurements are relatively sparse, owing to the difficulty associated with ³³S NMR (I=3/2, low γ, 0.76% abundance), particularly if a large EFG is present (i.e., at sites with lower symmetry).⁸  Sulphate minerals contain the SO₄²⁻ anion, and can be anhydrous or contain hydroxides/water. However, the symmetrical S environment of sulphates has resulted in a small, but growing, number of studies.⁸ 

The difficulty in acquiring high-resolution ³³S NMR spectra has resulted in most recent work on sulphides using Nuclear Quadrupole Resonance (NQR), where samples with large quadrupolar interactions are studied outside of the magnetic field. Work in ref. 321 used ⁶³Cu, ⁶⁵Cu, ¹²¹Sb and ¹²³Sb NQR to study minerals including chalcocite, Cu₂S, covellite, CuS (and intermediate compositions), stephanite (Ag₅SbS₄) and bournonite (CuPbSbS₃). Chalcopyrite (CuFeS₂), an important commercial source of metallic copper was also studied, although the presence of antiferromagnetic ordering generates an internal magnetic field, resulting in a ⁶³Cu “NMR”, rather than NQR, spectrum. Similar “zero field” NMR measurements of chalcopyrite were carried out in ref. 322. ⁷⁵As NQR measurements have also been made recently on arsenide minerals, including enargite (Cu₃AsS₄), niccolite (NiAs), arsenopyrite (FeAsS), and loellingite (FeAs₂).³²³  This approach was improved in 2014,³²⁴  where broadband WURST pulses ensured an improved excitation profile of the ⁷⁵As NQR lines in natural FeAsS, FeAs₂, As₂S₃, and As₄S₄. Multinuclear NMR spectroscopy (⁶⁷Zn, ¹¹⁹Sn and ⁶⁵Cu) of Cu₂ZnSnS₄, a synthetic material closely related to the natural minerals kesterite, Cu₂(Zn,Fe)SnS₄, and stannite, Cu₂FeSnS₄, with applications in solar cells, was carried out in ref. 325 and 326. Static and MAS NMR was used to determine which of the two structures were adopted, with the two differing in the order of the cationic layers along the c axis.

Although alum (NH₄Al(SO₄)₂·12H₂O) has been used as a model compound for ³³S MAS NMR (as it exhibits a very narrow resonance), most recent work has been carried out using static experiments and relaxation measurements for single crystals.^{327–329  1}H and ²⁷Al NMR showed a structural phase transition, with the loss of H₂O, with increasing temperature. Ca-based sulphate minerals such as gypsum (CaSO₄·2H₂O) and anhydrite (CaSO₄) have been studied in early ⁴³Ca (and ¹H) NMR work.⁸  More recently, a novel method for processing gypsum, forming a composite of anhydrite crystallites bound by a water-resistant phosphate matrix, was introduced, of interest for construction applications. The composite product was studied using XRD and ³¹P NMR,³³⁰  with an interfacial layer between the anhydrite crystallites and phosphate coating identified. NMR spectroscopy has also been used to study the sulphate mineral jarosite AFe₃(SO₄)₂(OD)₆ (A=K⁺, Na⁺ or D₃O⁺),³³¹  a textbook example of a 2D Kagomé lattice, where antiferromagnetic coupling of the Fe³⁺ leads to geometrical frustration. Variable-temperature ²H NMR identified Fe₂-(OD) and D₃O⁺ groups in stoichiometric regions of the sample (where chemical shifts followed a Curie-Weiss law above 150 K) and Fe-OD₂ and D₂O close to Fe³⁺ vacancies. Motion of the D₃O⁺ was followed by investigating the diamagnetic analogue alunite, (D₃O)Al₃(SO₄)₂(OD)₆, and the activation energy determined.

The halide minerals have dominant halide anions (e.g., F⁻, Cl⁻, Br⁻ or I⁻) with the lighter halides most prevalent in nature. While ¹⁹F (I=1/2, 100% abundance) is suited to solid-state NMR spectroscopy, the other halogens are more challenging to study, owing to their higher spins, lower sensitivities and lower natural abundances.

The most common fluoride mineral is fluorite (CaF₂). In 2012, Schmedt auf der Günne et al. reported the first observation of elemental F₂ in nature in “antozonite”, a variant of fluorite containing radioactive defects (typically U or Th).³³²  Radiation damage leads to the formation of clusters of metallic Ca and fluorine radicals (F˙), which rapidly combine to form F₂. The F₂ released from the crushed mineral was challenging to identify owing to its rapid reaction with most analytical apparatus and reagents. ¹⁹F NMR was able to observe F₂ in the intact mineral, as a singlet with δ_iso=425 ppm, compared to −108.8 ppm for CaF₂. The authors also searched for F˙, but found no evidence for this species. Fluorite-based solid solutions have potential application as chemically-stable ion conductors, and many of these, including CaF₂-YF₃,³³³  BaF₂-LaF₃,³³⁴  KF-BiF₃,³³⁵  PbF₂-MF-BiF₃ (M=Na, K, Rb, Cs),^335,336  have been studied by ¹⁹F MAS NMR. Abdellatief and co-workers^337,338  have carried out a study of the structural defects induced during mechanochemical and sol–gel syntheses of nanocrystalline fluorite, showing that ¹⁹F CSAs and T₁ relaxation times are sensitive to the particle size and contamination with Fe if steel-containing ball mills are used. Krahl et al.³³³  used ¹⁹F MAS and ¹⁹F-⁸⁹Y CP MAS NMR to investigate the fluoride defects in Ca_1−xY_xF_2+x, showing that, at x<0.01, F⁻ occurs as point defects, giving rise to resonances between −120 and −150 ppm (from FCa₆) and −50 to −90 ppm (FCa₅Y). At higher Y doping, clustering of F−, accompanied by lattice vacancies was observed, although the decreasing phase purity indicates that the solid-solution limit is below x=0.4.³³³  In addition, static variable-temperature ¹⁹F NMR was shown to be a good probe of the local F⁻ dynamics.^334–336 

The atacamite minerals came to the attention of the physics community in 2007, when it was discovered that synthetic herbertsmithite (ZnCu₃Cl₂(OH)₆) was a structurally-perfect Kagomé antiferromagnet.³³⁹  At low temperatures, the distorted pyrochlore structure of the atacamites leads to the formation of a “spin liquid”, where degeneracies lead to fluctuating rather than ordered magnetic spins. Low-temperature solid-state ¹H NMR was used to probe these spin liquids in atacamite (Cu₂Cl(OH)₃)^340,341  and its Ni and Mn analogues,^340,342  whereas herbertsmithite has been studied by low-temperature ¹⁷O NMR,³⁴³  and its metastable polymorph, kapellasite, has been investigated by ³⁵Cl NMR.^344,345  These studies have all used the effects of the internal magnetic fields to gain insight into the nature and behaviour of the spin liquid, including the magnetic ordering at very low temperature in Cu₂Cl(OH)₃,³⁴¹  the coexistence of frozen and liquid spins in Ni₂Cl(OH)₃,³⁴⁰  field-induced freezing of the spins in herbertsmithite,³⁴³  and short-range spin correlations and Zn/Cu disorder in kapellasite.^344,345 

Borate minerals contain the BO₃³⁻ anion, which may be polymerised (as discussed for silicates), resulting in e.g., B₂O₅, B₃O₆ and B₂O₄ units, with more complex structures also including halide and hydroxide anions. ¹¹B NMR spectroscopy is sensitive to the B coordination environment, with B^III and B^IV having chemical shifts of 17 to 20 and −1 to 2 ppm, respectively, and differences in C_Q (<1.0 and 2.5–3 MHz, respectively).⁸  In recent work, ¹¹B NMR has been used to study a number of boron-containing minerals, including kernite, Na₂B₄O₆(OH)₂·3H₂O, and borax, Na₂B₄O₅(OH)₄·8H₂O.³⁴⁶  Correlations between ¹¹B NMR parameters and local structure (e.g., site symmetry, number of bridging O, bond valences and NNN cations) was investigated using NMR and ab initio calculations. Work by Kroeker and co-workers³⁴⁷  used ¹¹B and ²⁵Mg NMR to refine the structure of inderite and kurnakovite (MgB₃O₃(OH)₅·5H₂O polymorphs). Variable-field MAS measurements and DFT calculations showed the positions of H atoms two bonds away had a significant impact on the ¹¹B and ²⁵Mg quadrupolar parameters. Kroeker and co-workers also combined DFT and ¹¹B NMR to study the crystal structure of ulexite (NaCa[B₅O₆(OH)₆]·5H₂O), demonstrating that structural optimisation yielded NMR parameters in much better agreement with experiment, and that density-of-states calculations on the optimized structures allowed insight into the origins of the ¹¹B EFG.³⁴⁸ 

Many other minerals, containing complex, mixed or biogenic anions, exist outside the above chemical categories.¹  There are few examples of the application of solid-state NMR to such minerals in recent years, but notable cases include the work of Colas et al., who probed the precise local structure of whewellite (CaC₂O₄.H₂O) by ⁴³Ca MAS and ¹³C CP MAS NMR³⁴⁹  and two studies of the low-temperature magnetism of vesignieite (Cu₃BaV₂O₈(OH)₂) by ⁵¹V and ^63/65Cu NMR/NQR.^350,351 

While the field of solid-state NMR applied to mineral-like materials is too vast to cover, even in a full review, we wish to highlight here a few key areas where solid-state NMR has contributed significantly to the understanding of the structure, properties or behaviour of mineral-related materials. Some specific examples have been discussed above, but here the reader is referred only to more extensive reviews.

Of particular geological significance are glasses and melts derived from mineral compositions and designed to model chemical speciation under the conditions found within the Earth. Owing to their amorphous nature, characterisation of the chemical species in glasses can be challenging and NMR is ideally placed to probe features such as coordination number and chemical environment of network formers (Al, Si, etc.) and the bridging and terminal anions (typically O in mineralogical melts).^3,5,352 

Many important ceramics are isostructural with natural minerals, with the perovskite and pyrochlore families, in particular, finding many practical applications. Solid-state NMR has proven particularly sensitive to features such as cation distribution and composition-induced phase transitions in pyrochlores and the precise tilting of the BX₆ octahedra in ABX₃ perovskites.^35,353,354 

As well as the naturally-occurring aluminosilicate zeolites, many synthetic analogues, including pure silicates, aluminophosphates, gallophosphates and mixed-metal phosphates are known.^223–226  Typically, NMR parameters are very sensitive to the precise local structure of these materials^275,355  – a fact that has been used to great effect to solve the structure of pure silicates from ²⁹Si NMR spectroscopy alone.³⁵⁶  These materials have been extensively probed by NMR owing to the prevalence of nuclei amenable to routine study and the wealth of information available from the technique. For further details see, for example, ref. 223–226.

Closely related to minerals are the complex phase mixtures of cements. As cements often contain amorphous phases, variable or disputed amounts of water, or cation/anion disorder, solid-state NMR is often a suitable technique to study these phenomena and provide quantification of the different species present.³⁵⁷ 

As discussed above, many lithium ion batteries are based on olivine-type phosphates, and perovskite materials are employed in solar cells. In recent years, many of the technological challenges involved in studying operating electrochemical devices in situ by NMR spectroscopy have been overcome and these experiments can now provide unprecedented insight into battery materials.^284,285