Solid-State 1H and 27Al NMR Studies of DMSO-Kaolinite Intercalates

Nuclear magnetic resonance (NMR) provides a powerful tool to describe local nuclear environments. In this work, unique structural information on kaolinite and on kaolinite dimethylsulfoxide (DMSO) intercalate were provided by solid-state 1H and 27Al magic-angle spinning (MAS) NMR. The interlayer chemistry of kaolinite (K) was examined by intercalating a select group of highly polar organic molecules or salts into kaolinite as a first step. Once the interlayer space is expanded, the intercalated compounds can be replaced in a second step. Intercalating DMSO into kaolinite to form the DMSO-K intercalate is, thus, a particularly useful first step toward the intercalation of a large variety of molecules, including polymers and ionic liquids. Well developed characterization methods are essential to define the structural modifications of kaolinite, and MAS NMR is a useful complement to other techniques. The use of 1H and 27Al MAS NMR for this purpose has been relatively rare. 1H NMR, nevertheless, can give unique information about kaolinite hydroxyls. Because quadrupolar interactions are sensitive to the local octahedral Al(III) geometry, solid-state 27Al NMR can follow subtle structural modifications in the octahedral sheet. In the present work, the 1H MAS NMR chemical shifts of KGa-1b were unambiguously attributed to the internal surface hydroxyls at 2.7 ppm and to the internal hydroxyls at 1.7 ppm. The 1H MAS NMR chemical shifts of the two methyl groups in DMSO-K are not equivalent and can be attributed to the 2.9 and 4.2 ppm peaks. The 27Al MAS NMR spectra of KGa-1b obtained under different magnetic fields revealed that most of the quadrupolar effects were highly reduced at 21.1 T, whereas the spectra at lower field, 4.7 T, were dominated by quadrupolar effects. The two octahedral Al(III) sites are not equivalent and can be distinguished in the low-field spectral simulation. Increased quadrupolar constants were observed and showed the major perturbations of the local Al symmetry that resulted from DMSO intercalation. Both the 1H and 27Al MAS NMR studies at different magnetic fields afforded important information about the local environments of the kaolinite hydroxyl groups and structural Al(III).


INTRODUCTION
Clay minerals are known for many unique chemical and physical properties that make them attractive starting materials for a vast variety of applications (Johnston, 2010;Bergaya and Lagaly, 2013). Kaolinite is very abundant in nature and large quantities are used in classical applications (e.g. paper coating, paint, ceramics, etc.) (Murray and Keller, 1993;Murray, 2007). Kaolinite plays a prominent role in human activities (Schroeder and Erickson, 2014). Hence, it is important to understand the kaolinite structural modifications and surface functionalization used to control a vast spectrum of applications and processes, such as catalysis and organic reactivity (Laszlo, 1986;Mccabe andAdams, 2013, Ngnie et al., 2016), clay-polymer nanocomposites Galimberti et al., 2013), bitumen extraction from oil sands (Detellier et al., 2015;Lin et al., 2016), soil permeability and soil transport processes (Yong and Mourato, 1990), or the abiotic origin of life (Brack, 2013). In that context, well developed characterization methods are needed to precisely monitor kaolinite structural modifications that are related to intercalation. Kaolinite has potential uses in more specialized and advanced application niches if the interlayer chemistry that results from grafting or intercalating organic moieties can be better understood, mastered, and developed . The kaolinite structure is one of the kaolin sub-group polytypes and is made of dioctahedral 1:1 layers with the Al 2 Si 2 O 5 (OH) 4 general composition (Brigatti et al., 2013). The crystallographic structure of kaolinite was determined by Bish (1993) and belongs to the C1 space group. Two types of hydroxyl groups are bonded to the octahedral Al cation: (1) the inner surface hydroxyls, which interact with the siloxane groups of an adjacent layer; and (2) the inner hydroxyl O-H bonds, which are parallel to the layer. The stoichiometric ratio of these two types of kaolinite hydroxyls is 3 to 1, respectively. Consequently, the kaolinite structure is highly asymmetric, which results in strong dipolar interactions between the layers, in addition to the network of H-bonds between the inner surface hydroxyls of the aluminol groups in the octahedral sheet and the siloxane groups in the tetrahedral sheet (Giese, 1982;Detellier and Schoonheydt, 2014). These interactions between the sheets of different layers result in the stacking of the layers (Mansa et al., 2017), which make direct intercalation of guest molecules into the interlayer space quite challenging. Only a select group of guest compounds can directly intercalate kaolinite . Because of the high dipole moment of DMSO, it can successfully intercalate kaolinite as was demonstrated by Olejnik et al. (1968). Later, Thompson (1985) observed two, non-equivalent 13 C NMR signals for the DMSO-K intercalate. Further, X-ray and neutron diffraction studies indicated that one DMSO methyl group was keyed into the ditrigonal cavity of the kaolinite layer, the other methyl group was parallel to the sheet, and the sulfonyl group was H-bonded to the aluminol sheet (Thompson and Cuff, 1985;Raupach et al., 1987). When the temperature was increased above 320 K, some of the DMSO methyl groups were released from the ditrigonal cavities (Hayashi, 1997) and the interlayer DMSO that remained was H-bonded through the DMSO sulfonyl group to the aluminol groups. This behavior made DMSO-K a particularly useful and efficient starting material for the intercalation of a large variety of small molecules, polymers, and ionic liquids into kaolinite (Letaief et al., 2008;Tonle et al., 2009;Dedzo and Detellier, 2016). The DMSO-K in these cases functioned as a pre-intercalate material and the interlayer DMSO can be replaced by other molecular groups. An optimum DMSO replacement temperature was shown to be around 180ºC. At this temperature, the DMSO was expelled from the interlayers and simultaneously replaced by other guest compounds. At higher temperatures, the DMSO in the pre-intercalate material was completely removed before it could be replaced by other compounds and kaolinite without intercalated DMSO was recovered . Suitable molecular units can be simultaneously grafted to kaolinite internal surfaces at 180ºC during the intercalation process (Tonlé et al., 2011;Dedzo and Detellier, 2016). An in-depth evaluation of the methods to intercalate DMSO into kaolinite was recently published (Abou-El-Sherbini et al., 2017).
In addition to X-ray diffraction, infrared techniques (Johnston et al., 1984;Yariv and Lapides, 2008;Scholtzova and Smrcok, 2009;Horvath et al., 2010) have been extensively used in the characterization of DMSO-K. Because nuclear magnetic resonance (NMR) is a powerful tool to describe local nuclear environments, especially in the solid state, it is a useful complement to other techniques. For example, the two observed 29 Si MAS NMR signals confirmed the asymmetrical nature of kaolinite layers (Barron et al., 1983), whereas 13 C MAS NMR confirmed the keyed structure of DMSO in the DMSO-K intercalate (Thompson, 1985). The use of 1 H and of 27 Al MAS NMR, however, has been much rarer, even though routine use would be very useful. The use of 1 H MAS NMR can give unique information on kaolinite hydroxyls and as a complementary technique to FTIR. The use of 1 H MAS NMR would be a particularly useful technique to follow as organic units are grafted to aluminol internal surfaces. Given the sensitivity of quadrupolar interactions on the local geometry of octahedral Al(III), solid-state 27 Al NMR could be very useful to follow the subtle structural modification of the octahedral sheet that follows intercalation and/or grafting. While infrared methods have been performed in several DMSO-K intercalate studies (Johnston et al., 1984;Yariv and Lapides, 2008;Scholtzova and Smrcok, 2009;Horvath et al., 2010), few studies have used 1 H NMR. The high magnetic fields combined with the use of high spinning rates allow the acquisition of high resolution 1 H NMR spectra in the solid state using conventional single pulse MAS NMR. Hayashi et al. (1992a) pioneered the use of high-resolution solid-state 1 H NMR to study the structure of kaolinite. They studied the static and CRAMPS (combined rotation and multiple-pulse spectroscopy) spectra for a variety of kaolinites. The CRAMPS technique is a specialized pulse program used to improve the resolution of spectra that are broadened by dipolar and shielding anisotropies in the solid state (Gerstein, 2009). In the static spectra, two components with different shapes were observed and were deconvoluted into broad and narrow components. The components were assigned to water molecules adsorbed on external kaolinite surfaces and to kaolinite structural hydroxyl groups, respectively. The 1 H CRAMPS NMR spectra of kaolinite gave one broad line centered at 2.8 ppm. Spin-lattice relaxation times were measured and reported in an accompanying paper (Hayashi et al., 1992b). The relaxation curve was deconvoluted into two components, long and short, that were ascribed in a similar manner to external water molecules and internal hydroxyls. The spin-lattice relaxation rates obtained under MAS conditions confirmed that spin diffusion plays an important role in 1 H relaxation, which is diffusion-limited (Hayashi and Akiba, 1995). The 1 H NMR single pulse experiments were performed on two kaolinites at 300 MHz and a spinning rate of 10.5 kHz (Wang et al., 2002). Two peaks for each of the two kaolinites were observed at 2.4 and À1.3 ppm and at 3.0 and À0.9 ppm, respectively. The two peaks were attributed to inner surface (2.4 and 3.0 ppm) and inner (À1.3 and À0.9 ppm) hydroxyls (Wang et al., 2002). Cheng et al. (2015) reported the 1 H MAS NMR spectrum for a well ordered kaolinite. A relatively sharp signal was observed at 4.9 ppm and was attributed to inner surface hydroxyl groups, but inner hydroxyls were not observed. Zhu et al. (2016) obtained the 1 H MAS NMR spectra of three kaolinites at 500 MHz and two peaks were observed and deconvoluted. Chemical shifts in the 2.8À3.1 ppm and 1.7À1.9 ppm ranges were attributed to inner surface and inner hydroxyls, respectively. The relative populations of the chemical shifts, however, did not correspond to the expected 3:1 ratio.
In comparison to 1 H NMR, kaolinite 27 Al MAS NMR has more often been reported (Rocha and Pedrosa de Jesus, 1994;Ashbrook et al., 2000;Zhou et al., 2009;Paris, 2014). Many of these studies have been focused on finding evidence for the two distinct Al sites in the kaolinite crystallographic structure that were identified in X-ray diffraction studies (Bish and Von Dreele, 1989). Small parameter differences are observed between each of the Al species, which makes distinguishing between them difficult using 27 Al NMR. The literature on intercalated and grafted kaolinite complexes is more limited. The published studies, however, have shown a high likelihood to observe multiple Al species, especially in grafted complexes. Detellier (1993, 1994), for example, reported that major changes occurred to the 27 Al NMR spectra of kaolinite grafted with ethylene glycol. They attributed the spectral changes to ethylene glycol covalently grafted to Al in the octahedral sheets in the clay mineral structure. Hirsemann et al. (2011) performed multiple quantum magic-angle spinning (MQMAS) experiments on ethylene glycol grafted to kaolinite and found two Al species with different chemical shifts and quadrupolar parameters. They also performed a REAPDOR (rotational echo adiabatic passage double resonance) experiment, which allowed ethylene glycol grafted to kaolinite internal aluminol surfaces to be conclusively demonstrated. REAPDOR is a specialized pulse program used to measure the distance between spin Ý and quadrupolar nuclei in the solid state (Guillion, 1995).
The purpose of the present study was to obtain precise local structural information at the atomic level on kaolinite and on the kaolinite DMSO intercalate (DMSO-K) using 1 H MAS NMR spectra at 21.1 T and the 27 Al MAS and MQMAS NMR spectra at 21.1 and 4.7 T to (1) differentiate the inner and the inner surface hydroxyl groups, aided by partial deuteration; and (2) isolate the quadrupolar parameters of the different octahedral Al(III) sites in each of these materials by simulation of the 27 Al NMR spectra at low field.

Materials
Kaolinite. The KGa-1b (Georgia) kaolinite is a highly crystalline, low defect kaolinite (Hinckley Index = 1.09) and was obtained from the Source Clays Repository of The Clay Minerals Society. It has the high purity characteristic of Georgia kaolins with anatase as the only significant impurity. The particle size distribution of KGa-1b was determined to be 57.8% <2 mm and 32.0% <0.5 mm (Pruett and Webb, 1993). A poorly crystallized kaolinite sample (KGa-2) was also used in some experiments ( Figure S1 in the Supplemental Materials section (deposited with the Editor-in-Chief and available at http://www.clays.org/JOURNAL/ JournalDeposits.html)). The <2 mm fraction was isolated by sedimentation using a method similar to published methods (Jeffries and Jackson, 1949). A 60-g sample of KGa-1b was dispersed in 4 L of distilled water by mechanical stirring in a beaker and the pH of the suspension was adjusted to 9.0 by the addition of 0.1 M NaOH to create a stable suspension of the clay particles. Stirring of the suspension was then stopped, the suspension was allowed to settle for 6 h, and the top two-thirds of the suspension was siphoned off and set aside. The remaining suspension was re-dispersed in distilled water and the sedimentation procedure was repeated three more times. The four siphoned suspensions were then combined and the pH was adjusted to 6.0 by the addition of 0.1 M HCl, which rapidly flocculated the clay particles. The supernatant was decanted from the suspension as much as possible without loss of solids and was discarded. The remaining kaolinite solid suspension was evenly partitioned into four 700-mL centrifugation vessels, the masses of the vessels were equalized by distilled water addition, and the vessels were centrifuged at high speed until all the solid matter was deposited. The supernatants were discarded, 300 mL of distilled water was added to the vessels, the solids were re-dispersed, and the centrifugation procedure was repeated. This distilled-water washing and centrifugation procedure was repeated two more times, the final isolated solids in the four vessels were combined together, and the solids were oven dried at 80ºC for 24 h.

Dimethyl sulfoxide -kaolinite intercalate (DMSO-K).
The method used to prepare DMSO-K was similar to the one described by Olejnik et al. (1968). A 10-g sample of KGa-1b <2mm was dispersed in a mixture of 60 mL of DMSO and 0.1 mL of deionized water at 80ºC and allowed to equilibrate for three weeks. The resulting solid product was isolated by centrifuging at high speed until all the solid matter was deposited and the supernatant was discarded. The solid was dispersed in 200 mL of isopropanol to wash out excess DMSO and centrifuged using the same procedure described above. This procedure was repeated three times and the final solid material was oven dried at 80ºC overnight.
Deuterated kaolinite. A sample of DMSO-K (100 mg) was added to 5 mL of D 2 O in a glass vial, the vial was capped, and the resulting suspension was stirred using a magnetic spin bar at room temperature for three d. The DMSO-K solid matter was isolated by centrifuging the suspension at high speed until all the solid matter was deposited and the supernatant was discarded. The solids were then transferred to an evaporation dish using 2À3 mL of acetone and oven dried at 70ºC for one h. The dried solid was then stored in a desiccator under N 2 prior to analysis.

Methods
Solid-state 1 H nuclear magnetic resonance. The 1 H MAS NMR experiments were carried out at 899.69 MHz on a 21.1 T Bruker AVANCE II NMR spectrometer using a 4 mm H/X MAS Bruker probe (Bruker BioSpin GmbH, Rheinstetten, Germany). All experiments used an 18 kHz MAS spinning rate with a modified 90-90 rotor-sync quad-echo pulse sequence for background signal suppression. A recycle delay time (d1) of 5 s was used, which was sufficient for complete relaxation and allowed quantitative analysis of peak integrals. The chemical shift scale was referenced to the tetramethylsilane (TMS) peak at 0.0 ppm.
Solid-state 27 Al nuclear magnetic resonance. Lower field solid-state 27 Al MAS NMR spectra were obtained at 52.1 MHz on a 4.7 T Bruker AVANCE III NMR spectrometer (Bruker BioSpin GmbH, Rheinstetten, Germany) equipped with a 7 mm triple resonance MAS probe. A d1 time of 2 s was used and the solid samples were packed into a 7 mm zirconia rotor and spun at a frequency of 5 kHz. High-field 27 Al MAS NMR spectra were measured at 234.5 MHz on a 21.1 T Tesla Bruker AVANCE II NMR spectrometer (Bruker BioSpin GmbH, Rheinstetten, Germany) equipped with a 4 mm H/X MAS Bruker probe, using a MAS spinning rate of 18 kHz. A d1 time of 2 s was used and was sufficient for complete relaxation. In both cases, the chemical shift scale was referenced to the 27 Al(NO 3 ) 3 peak in water at 0.0 ppm (0.5 M at 21.1 T; 0.1 M at 4.7 T). Twodimensional 27 Al triple-quantum (3Q) MAS NMR experiments were performed at 21.1 T under 18 kHz MAS. All experiments were rotor-synchronized in the f1 dimension (SWH1=18 kHz, t1 increment of 55.56 ms) and employed a three-pulse z-filter 3QMAS pulse sequence. The spectral width in the direct F2 dimension was 100 kHz. For each t1 increment, 48 transients were accumulated using a 1 s relaxation delay. The total number of t1 increments was 128. The excitation and conversion pulses optimized using 60 kHz r.f. power for Al sites with small C Q ( 27 Al) values were 5.75 ms and 1.75 ms, respectively. The 27 Al selective 90º pulse width was 9 ms. The chemical shift referencing in the indirect dimension was done according to the method of Millot and Man (2002).
NMR Spectral deconvolution and curve fitting. The experimental 1 H MAS NMR spectra were fitted with the help of the peak fitting software FityK (Wojdyr, 2010), the peak fitting was done using Voigt curves, and the rotational sidebands were not included in this analysis. This still gave an accurate fit since the vast majority of the signal intensity was in the central transition and the relative signal intensities of each peak was the same for the sidebands as it was for the central transition.
Because 27 Al is a quadrupolar nucleus, it produces NMR line shapes that cannot be described by classical curves and requires specialized software to simulate and fit the NMR curves. The experimental 27 Al MAS NMR spectra were fitted using Dmfit (Massiot et al., 2002), which is an NMR software package with several options for simulating the NMR spectra of quadrupolar nuclei.
Fitting of the NMR spectra was done using the built in 2QUAD function, which allows the computation of spectra at static or finite MAS rates and takes into account 2 nd order quadrupolar effects and chemical shift anisotropy. The central (i.e. CT) and satellite (i.e. ST1, ST2) transitions were simulated using MAS rates of 5 kHz for the 4.7 T spectra and 18 kHz for the 21.1 T spectra and 10 and 2 ppm of Gaussian line broadening, respectively.
Attenuated total reflectance Fourier-transform infrared analysis (ATR-FTIR). The ATR-FTIR spectra were obtained using a Nicolet 6700 FTIR (Thermo Scientific, 81 Wyman Street, Waltham, Massachusetts, USA) in attenuated total reflection mode using the SMART diamond ATR attachment. A background spectrum from 4500 to 650 cm -1 was obtained before each sample was analyzed using an average of 128 scans with a resolution of 4 cm -1 and the spectra were obtained with the aid of the OMNIC software package. Sample analyses were done by evenly spreading out~2À5 mg of each solid sample and firmly pressing it to the surface of the ATR crystal. Spectra were measured using the same number of scans and spectral resolution as the background.
Powder X-ray diffraction (XRD). XRD traces were obtained using a Philips PW 3710 diffractometer (PANanalytical, Lelyweg 1, 7602 EA Almelo, The Netherlands) with a Cu Ka radiation source (l = 1.5406 Å ), operated at 45 kV generator voltage and 40 mA current. Sample analyses were accomplished by pressing the solid materials into cylindrical pellets with a diameter of~0.5 cm and thickness of~2 mm and mounting the sample pellets on disk shaped sample holders.
Thermal gravimetric analysis (TGA). TGA mass loss curves were obtained using a Q500 Thermal Gravimetric Analyzer (TA Instruments, New Castle, Delaware, USA). Sample solids (2À5 mg) were loaded into a 50 mL tared high temperature Pt pan and analyzed using a steady temperature ramp program of 10ºC min À1 under a N 2 flow rate of 25 mL min À1 . A one min isothermal temperature hold was done at the beginning of each experiment to allow the movements of the sample pan to stabilize after it was loaded inside the furnace.
CHNS Elemental analysis. Elemental analyses were performed at the G.G. Hatch Stable Isotope Laboratory of the University of Ottawa. The standard protocol employed for elemental analysis has a routine analytical precision of Ô0.1% and is summarized as follows: (1) samples and standards were weighed into Sn capsules and loaded into an Elementar Isotope Cube elemental analyzer (Elementar, Langenselbold, Germany); (2) the Sn sample capsules were flash combusted in the analyzer at about 1800ºC in O 2 carried by a He flow through columns of reducing/oxidizing chemicals to produce N 2 , CO 2 , and SO 2 ; and (3) the N 2 , CO 2 , and SO 2 gases were separated in a trap and purge system of adsorption columns to separately measure each gas using the thermal conductivity detector.

RESULTS AND DISCUSSION
The 1 H MAS NMR spectrum of the <2 mm KGa-1b sample was obtained at 900 MHz ( Figure 1) and the spectrum was consistent with previously reported results (Wang et al., 2002;Zhu et al., 2016). Two overlapping, but distinctly different, signals in the 1 to 3 ppm range with a broad shoulder at higher chemical shifts were observed. The spectrum was deconvoluted using Voigt curves. The Voigt function is a combination of Cauchy (i.e. Lorentzian) and Gaussian functions and is commonly used for curve fitting in spectroscopy and crystallography due to the wide variety of line shapes it can simulate (Armstrong, 1967;Langford, 1978). Whereas the Fourier transformation of the free induction decay plot in NMR is expected to give a purely Lorentzian line shape, the line broadening effect from dipolar interactions will produce a Gaussian convolution in the solid state (Van Vleck, 1948;Kittel and Abrahams, 1953). Consequently, a function which combines both line shapes is commonly used to obtain an accurate deconvolution of the observed signal. The deconvolution result is also given in Figure 1. Two major signals at 1.70 and 2.68 ppm were measured with relative intensities of 1.00 and 3.09. This result is consistent with the expected proton site ratio in the clay mineral, where 1 out of every 4 hydroxyls in the idealized Al 2 Si 2 O 5 (OH) 4 kaolinite structure is an inner hydroxyl and the other 3 hydroxyls are inner surface hydroxyls. As a result, the ratio of the integrated signal of the inner surface hydroxyl protons to the integrated signal of inner hydroxyl protons should be 3:1, which is in close agreement with the experimental ratios. Consequently, the 1.70 ppm signal was attributed to the inner hydroxyl, while the 2.68 ppm signal was attributed to the inner surface hydroxyls. One should also note that the best fits to the experimental spectra were obtained when the inner surface hydroxyl peak at 2.68 ppm was broader and had a more Gaussian character than the inner hydroxyl peak signal at 1.70 ppm. This further supports the signal assignments. The inner hydroxyls are segregated from the other hydroxyl groups, but the inner surface hydroxyls are close to other hydroxyl groups and experience more 1 H-1 H dipolar interactions. This should result in a more Gaussian line shape. In addition, one should expect a larger chemical shift dispersion from the inner surface hydroxyls than from the inner hydroxyls. These observed 1 H chemical shift values are in good agreement with reported chemical shifts of~0.5 ppm for trioctahedral smectite internal hydroxyl groups and~2.0 ppm for dioctahedral smectites (Alba et al., 2000). In dioctahedral sheets, the internal OH bonds are almost parallel to the layers and allow an H-bond with the apical oxygen atoms of the tetrahedral sheet, which results in decreased chemical shielding. In trioctahedral sheets, the internal OH bonds are perpendicular to the layers and point toward the hexagonal cavities, which is not favorable for strong H-bonds. Wang et al. (2002) reported chemical shift values of 2.4 and À1.3 ppm for one kaolinite and 3.0 and À0.9 ppm for another kaolinite. These values were attributed to inner surface (2.4 and 3.0 ppm) and inner (À1.3 and À0.9 ppm) hydroxyls. The negative chemical shifts were surprising and are closer to what was expected for a trioctahedral sheet than for a dioctahedral sheet. In addition, Wang et al. (2002) reported relative peak intensities for the inner surface and inner hydroxyls of 62% and 38% for one kaolinite and 27% and 73% for the other kaolinite. This significantly deviated from the expected 3:1 relative intensity ratio. Two 1 H NMR peaks were observed and deconvoluted by Zhu et al. (2016), with chemical shifts in the 2.8À3.1 ppm and 1.7À1.9 ppm ranges that were attributed to inner surface and inner hydroxyls, respectively. The Zhu et al. (2016) relative peak intensities also did not have the expected 3:1 ratio and the chemical shift values were close to those observed in the present study. In only one case was the measured relative hydroxyl population ratio (3.3:1) fairly close to the expected 3:1 structural value and the ratios were 2:3 in the other two cases. The discrepancy between measured and expected hydroxyl populations was attributed to different degrees of bond rupture on kaolinite platelet edges where the inner hydroxyls are exposed to give a similar environment to the inner surface hydroxyls. This argument was difficult to follow using the published FESEM images, which indicated particle dimensions in the 0.5À1 mm range and consequently a very small percentage of edge hydroxyls (Zhu et al., 2016). To account for the broad shoulder in the 1 H MAS NMR spectrum at higher frequencies in the present work (Figure 1), the deconvolution included a third weak and broad signal centered at 4.92 ppm. This resonance was attributed to surface water present in the material.
TGA weight loss analysis of the sample indicated that 0.46 % (w/w) water was present in the kaolinite under study. This quantity of water would give a proton abundance of 0.13 relative to the inner hydroxyls. Using this abundance, a best fit to the experimental spectrum was obtained when this peak was centered at 4.92 ppm. This value was expected to depend greatly on the amount of water adsorbed on the external surfaces (Li et al., 2010;Zhu et al., 2016).
To further confirm the signal attributions in the 1 H MAS NMR spectra, an attempt was made to deuterate the KGa-1b by washing the DMSO-K intercalate with deuterated water (see Materials and Methods). The D 2 O intercalation process should remove DMSO and leave unexpanded kaolinite with deuterated inner surface hydroxyls. This allows the inner surface hydroxyls to be differentiated from the inner hydroxyls, which are known to be much more resistant to deuteration (Ledoux and White, 1964). Deuterium exchange, thus, could be an effective way to identify the inner hydroxyl groups in kaolinite. In addition, this would also be expected to have a line narrowing effect due to a reduction in dipolar interactions. The attempt to deuterate directly KGa-1b by washing with D 2 O repeatedly at room temperature was, as expected, unsuccessful. This is clearly because hydroxyl groups in natural kaolinite exchange with deuterium only under high temperature and pressure (Romo, 1956). Even under high temperature and pressure conditions, however, the inner hydroxyls are exchanged with difficulty. FTIR spectra of KGa-1b and deuterated kaolinite in the region expected for the O-D stretching bonds are shown in Figure 2. The absorption frequency that corresponds to the KGa-1b inner hydroxyl was observed at 3620 cm À1 (Johnston et al., 1990). The deuterated KGa-1b inner hydroxyl was observed at 2676 cm À1 and had an intensity that was strongly reduced in comparison to the other character- istic absorption frequencies of kaolinite. This was because the inner surface hydroxyls are much more accessible than the inner hydroxyls, which resulted in selective deuteration. Similarly, Johansson et al. (1998) found that the vibrational frequency n H was not exactly 1.37 times n D as predicted by Hooke's law, but rather was 1.35 and likely resulted from an isotopic effect on the O-H/O-D bond strengths. Consequently, the 1 H MAS NMR spectrum of deuterated kaolinite, KGa-1b-D, should confirm the attribution of the 1.70 ppm signal to the inner hydroxyls because this peak should have become more intense in comparison to the 2.68 ppm signal that was attributed to the inner surface hydroxyls. This is exactly what was observed experimentally (Figure 1). The 1 H MAS NMR spectrum of deuterated KGa-1b-D (Figure 1) appeared very different from the initial KGa-1b sample. The deconvolution gave two main signals at 1.74 ppm and 2.58 ppm, which were similar to the spectrum of KGa-1b but with different ratios. The 1.74 and 2.58 ppm signals were, respectively, attributed to the inner hydroxyls and to the inner surface hydroxyls. In comparison to the KGa-1b spectrum, the 2.58 ppm signal population was reduced by almost a factor of two in the KGa-1b-D spectrum, which resulted from the more efficient deuteration of the inner surface hydroxyls. Additionally, the signal from surface adsorbed water disappeared. The new broad signal at 2.87 ppm was attributed to residual DMSO molecules and the DMSO contents determined from TGA, and CHNS elemental analysis were 0.022 and 0.025 moles/ 100 g, respectively. Deconvolution of the 1 H MAS NMR signal indicated a similar DMSO content of 0.032 moles/ 100 g of material. One should note that only one signal was sufficient to account for the small amounts of DMSO. This contrasts with the deconvolution of DMSO-K signals which was based on two signals of equal intensity for DMSO (see below). The use of two DMSO signals with a 1 H population of 0.49/2 each and chemical shifts provided by the DMSO-K spectrum did not improve the deconvolution. As mentioned above, several previous reports demonstrated that one of the two DMSO methyl groups in DMSO-K was keyed into the kaolinite siloxane ring of the tetrahedral sheet, while the other methyl group was parallel to the layer (Johnston et al., 1984, Thompson, 1985Hayashi, 1997). The 13 C MAS NMR spectrum of DMSO-K was recorded and two signals were observed at 44.2 and 43.1 ppm, which is in agreement with this structure. Consequently, to analyze the 1 H NMR spectrum of DMSO-K, a model to take this fact into consideration was used. The 1 H NMR spectrum of DMSO-K and a deconvolution of the spectrum into several components (Figure 1) included the two kaolinite hydroxyl signals with respective intensities of 1 and 3 and two DMSO methyl signals of equal intensity. A small amount of water confirmed by TGA had a molar fraction of 0.03 and was included in the deconvolution as a broad and low intensity band at 4.89 ppm. The best fit was obtained when the chemical shift of the keyed-in methyl group of DMSO was 4.23 ppm, while the other methyl group was observed at 2.86 ppm.
The 27 Al MAS NMR spectra were recorded for KGa-1b at two different magnetic fields of 21.1 T and 4.7 T (Figure 3). The signal maxima were observed, respectively, at 6.4 ppm and À18.2 ppm and had noticeable differences in the line shapes. The quadrupolar effects were minimized at 21.1 T, but strongly affected the signal at 4.7 T. The signal observed at 6.4 ppm was consistent with a six-coordinated Al(III), which is expected to give a signal between À10 to +15 ppm (Mackenzie and Smith, 2002). The 6.4 ppm value at 21.1 T should be close to the true isotropic chemical shift, d iso , because the signal center of gravity is proportional to the inverse square of the applied magnetic field (MacKenzie and Smith, 2002;Wasylishen et al., 2012). An extrapolation to B 0 À2 = 0 for the spectra centers of gravity at 4.7 T and 21.1 T gave a d iso value of 8.3 ppm. Close examination of the 21.1 T spectrum (Figure 3) revealed that the central peak (CT) base is not symmetrical and the line shape is distorted at higher frequencies. This was attributed to the presence of the satellite transitions, ST1 and ST2, which respectively resulted from the spin transitions of Ô1/2 ? Ô3/2 and Ô3/2 ? Ô5/2. This is more obvious in the spinning side band structure (Figure 4). In the first spinning side band (ssb) pairs, the CT transition signal dropped dramatically, but the ST transition intensities dropped to a much lesser degree such that the transitions became clearly resolved from the CT. Trace amounts of tetrahedral IV Al(III) were also observed in the close-up of the spectrum (Figure 4). This peak was centered at 71.0 ppm, which is consistent with a four-coordinated Al-O species that was expected to give a signal between Figure 3. 27 Al MAS NMR spectra of KGa-1b obtained under magnetic fields of 21.1 T and 4.7 T. 50 and 80 ppm (Mackenzie and Smith, 2002) and plausibly resulted from Al substitution for Si in the tetrahedral sheet. If multiple sites for the Al(III) cation in the kaolinite structure existed, the sites could not be identified from the spectrum at very high field due to the very similar isotropic chemical shifts. On the other hand, if the quadrupolar coupling constant, C Q , and/or the asymmetry parameter, Z Q , were sufficiently different, a low magnetic field might be used to maximize the quadrupolar effects on the NMR spectra. This would result in the detection of separate and crystallographically different Al(III) cationic sites even if the chemical shifts were similar or even identical. The 27 Al MAS NMR spectrum of KGa-1b was recorded at 4.7 T and the central peak position was significantly shifted towards lower frequencies with a maximum at À18.2 ppm (Figure 3). A number of features were observed on both sides of the peaks, which resulted from the increased quadrupolar effects at lower field strengths. Line shapes were simulated using models for one and for two Al(III) sites ( Figure S2 in Supplemental Materials section, Figure 5). Using the d iso value found by extrapolation at infinite magnetic field strength, the best fit in the case of one site was found using C Q = 3.21 MHz and Z Q = 0.75 values when fitted to the 4.7 T experimental spectrum. This gave a reasonable fit overall, but the line shape did not match well with the experimental line shape on the lower frequency side ( Figure S2 in Supplemental Materials section), suggesting that at least another Al site was present. Simulation was then repeated by assuming two unique Al sites of equal intensity. Fitting this model to the 4.7 T experimental spectrum gave an improvement in comparison to the one-site model. This fit ( Figure 5) and the parameters extracted from it (Table 1) were then successfully used to fit the spectrum obtained at 21.1 T. The deconvolution values were very close to the values that Paris (2014) reported for kaolinite and were in agreement with other previously reported values (Hayashi et al., 1992a;Rocha and Pedrosa de Jesus, 1994;Crosson et al., 2006;Zhou et al., 2009) as discussed by Paris (2014). This described approach was consequently validated and could be applied to other systems.
The 27 Al MAS NMR spectrum of DMSO-K recorded at 21.1 T (Figure 6) was compared to the corresponding   (Paris, 2014) spectra of KGa-1b and KGa-1b-D at the same field. The KGa-1b and KGa-1b-D spectra were almost identical despite the small amount of DMSO that remained trapped in KGa-1b-D, which confirms the successful deuteration of the inner surface hydroxyls of kaolinite. The 27 Al MAS NMR spectrum of DMSO-K showed subtle line-shape changes compared to KGa-1b. A shift in the central peak maximum from 6.4 to 5.8 ppm was also observed, which occurred only on the lower frequency side of the peak. This may indicate that the shift was due to changes in the quadrupolar interaction (i.e. C Q ) rather than in the isotropic chemical shift (d iso ) of the Al species.
To verify this, the spectrum was recorded at 4.7 T in order to amplify the quadrupolar effects and compared to the KGa-1b spectrum (Figure 7). Similar to the 21.1 T case, a shift towards lower frequency was observed, but the shift was much more dramatic at low magnetic field Figure 5. 27 Al MAS NMR spectra of KGa-1b at 4.7 T fitted to a simulated spectrum using two unique Al sites. A close up of the side bands is shown as an inset. The chemical shift value and quadrupolar parameters of each site are summarized in Table 1. Figure 6. 27 Al MAS NMR spectra of DMSO-K compared to KGa-1b and KGa-1b-D. These spectra were obtained at 21.1 T using a MAS rate of 18 kHz. The removal of 88% of interlayer DMSO resulted in a line shape of KGa-1b-D almost identical to the one of KGa-1b. Figure 7. Superposition of the experimental 27 Al MAS NMR spectra of KGa-1b and DMSO-K recorded at 4.7 T. The broadening towards lower frequency in DMSO-K was likely due to higher values of C Q (see text). strength and the DMSO-K central peak maximum was observed at À26.4 ppm. The superposition of this spectrum with KGa-1b showed that this shift occurred on the lower frequency side of the signal, which is strong evidence that it resulted from changes in quadrupolar parameters rather than in d iso . This is because the 2 nd order quadrupolar interaction with the chemical shift is negative. At higher magnetic fields, these 2 nd order effects were reduced and gave a central transition that is fairly symmetric at a position close to d iso . Spectra obtained at lower field, on the other hand, exhibited much stronger 2 nd order effects, which caused the central peak to broaden towards more negative frequencies (Mackenzie and Smith, 2002). At a low magnetic field, the details of the central peak became much more apparent. From the shape of the signal, a superposition of the Al sites was apparent with some sites that had parameters identical to KGa-1b and some sites that had an increased C Q value. The 4.7 T spectrum was simulated by assuming that the DMSO molecule cover-age of internal kaolinite surfaces was complete in the 11Å XRD phase and accounted for the 88% intercalation ratio revealed by XRD. The 12% abundance of the unmodified kaolinite phase was taken into account and gave relative abundances of 0.44:0.44 and 0.06:0.06 for the two DMSO-K Al sites and the two KGa-1b Al sites, respectively ( Figure 8 and Table 2). One should note that a 3-site model with only one DMSO-K Al site gave an unsatisfactory fit to the experimental spectrum. The DMSO intercalation of KGa-1b resulted in a pronounced increase in the quadrupolar coupling constant, C Q , for both Al sites and reduced the asymmetry parameter, whereas the chemical shifts were mostly unchanged. These results pointed to a distortion in the Al octahedron that resulted from interactions of intercalated DMSO through the aluminol groups. The value of C Q ( 27 Al) depends on the Al-O bond length and on the O-Al-O bond angle (Ghose and Tsang, 1973;MacKenzie and Smith, 2002). Interactions between the interlayer DMSO sulfoxide groups and Al-OH groups in the clay mineral Figure 8. 27 Al MAS NMR spectra of DMSO-K recorded at 4.7 T fitted to a simulated spectrum using two unique Al sites each for the DMSO modified and unmodified kaolinite phases. A close up of the side bands is shown as an inset. The chemical shift value and quadrupolar parameters of each site are summarized in Table 2.  (Johnston et al., 1984;Frost et al., 1998). These results support the model of a more distorted Al octahedron as a result of interlayer hydrogen bonding, which produces a greater C Q value. Two-dimensional multiple-quantum (MQ) 27 Al MAS NMR spectra are often recorded to resolve overlapping Al sites and is aided by the fact that the quadrupolar broadening effects are attenuated in these spectra in the indirect (f1) multi-quantum dimension. Triple-quantum (3QMAS) 27 Al NMR spectra in KGa-1b and DMSO-K samples were acquired at 21.1 T (Figure 9). Because 27 Al NMR parameters of the two KGa-1b Al sites are very similar and overlap strongly at such high magnetic fields, the individual Al sites cannot be resolved by these 27 Al 3QMAS NMR spectra. Comparison of apparent differences, however, between the 27 Al 3QMAS NMR spectra of KGa-1b and DMSO-K (Figures 9a, 9b) clearly revealed that the effective (averaged over two Al sites) isotropic chemical shifts in the two samples are nearly identical within the experimental errors of +8.0 ppm in KGa-1b and +7.5 ppm in DMSO-K. At the same time, the second order quadrupolar effect (SOQE), which is a combined function of C Q and Z Q , is noticeably larger in DMSO-K (4.1 MHz) in comparison to  and this is in agreement with the analysis and results of the one-dimensional MAS experiments.

CONCLUSIONS
The 1 H MAS NMR chemical shifts observed in the highly crystalline, low defect, Georgia kaolinite, KGa-1b, were unambiguously attributed to the internal surface hydroxyls at 2.7 ppm and to the internal hydroxyls at 1.7 ppm. The ratio of the internal surface and the internal hydroxyl intensities was 3:1 as expected from the kaolinite structure. Kaolinite was partially deuterated by exchanging the hydroxyl protons in a kaolinite expanded with DMSO molecules with deuterated water. After deuterated water treatment, the 3:1 ratio dropped to 1.8:1 as the result of partial deuteration of the internal surface hydroxyls. This is in agreement with the signal attribution and with the observed O-D stretching region of the FTIR spectrum. The 1 H MAS NMR chemical shifts of the DMSO intercalate, DMSO-K, were obtained by deconvoluting the spectra into four major signals. Two signals with a 3:1 ratio were attributed to the kaolinite hydroxyls and two signals were attributed to the two non-equivalent DMSO methyl groups with identical intensities at 2.9 and 4.2 ppm. The 27 Al MAS NMR spectra of KGa-1b were obtained at 21.1 T and at 4.7 T. Most of the quadrupolar effects were highly reduced at 21.1 T where the signal shape mainly resulted from chemical shift dispersion effects and the low field spectra were dominated by quadrupolar effects. The two non-equivalent octahedral Al(III) sites were identified by simulating the low-field spectra. Because the isotropic chemical shifts were very similar, the shifts were differentiated by analyzing the line shape to determine the quadrupolar coupling constants and the asymmetry parameters. The quadrupolar coupling constants and asymmetry parameters, which are much more sensitive to local Figure 9. 27 Al 3QMAS NMR spectra of (a) KGa-1b and (b) DMSO-K obtained at 21.1 T using a MAS rate of 18 kHz. The peak positions in the two dimensions (f 2 , f 1 ) were identified using the automated MestReNova software peak picking tool. symmetry perturbations in comparison to the isotropic chemical shifts, and the 27 Al NMR isotropic chemical shifts of 7.7 and 7.8 ppm were generally in good agreement with previously reported values. The 27 Al NMR values obtained for kaolinite allowed the simulation of the DMSO intercalate 27 Al NMR spectrum using four signals, which used two sites for the 12% nonintercalated kaolinite and two sites for the 88% intercalated kaolinite. The quadrupolar constants notably increased and the asymmetry parameters decreased after DMSO intercalation and showed that the major perturbation effect on the local Al symmetry resulted from DMSO intercalation and the rigid structural effect on the interlayer space. Finally, multiple quantum 27 Al MAS NMR spectra of KGa-1b and DMSO-K were recorded at 21.1 T and indicated that line-shape variations between KGa-1b and DMSO-K observed in the 27 Al MAS NMR spectra resulted from changes in quadrupolar parameters rather than changes in the d iso values. Both 1 H and 27 Al MAS NMR studies at several different magnetic fields have the potential to reveal important information about the local environment of kaolinite hydroxyl groups and structural Al(III), which could prove an essential complement to other techniques.