Evidence for Mixed Mg Coordination Environments in Silicate Glasses: Results from 25Mg NMR Spectroscopy at 35.2 T

The Mg–O coordination environment of silicate glasses of composition CaMgSi2O6, Na2MgSi3O8, and K2MgSi5O12 is probed using ultrahigh-field (35.2 T) 25Mg magic angle spinning nuclear magnetic resonance (MAS NMR) and triple-quantum MAS NMR spectroscopy. These spectra clearly reveal the coexistence of 4-fold- (MgIV) and 6-fold- (MgVI) coordinated Mg in all glasses. The MgIV/MgVI ratio implies an average Mg–O coordination number of ∼5 for CaMgSi2O6 glass, bringing NMR results for the first time in good agreement with those reported in previous studies based on diffraction and X-ray absorption spectroscopy, thus resolving a decade-long controversy regarding Mg coordination in alkaline-earth silicate glasses. The Mg–O coordination number decreases to ∼4.5 in the alkali–Mg silicate glasses, indicating that Mg competes effectively with the low field strength alkali cations for the nonbridging oxygen in the structure to attain tetrahedral coordination. This work illustrates the promise of ultrahigh-field NMR spectroscopy in structural studies involving nuclides with low gyromagnetic ratio.


INTRODUCTION
Recent studies have indicated that Mg plays a key role as a constituent element in silicate glasses in improving their crack and corrosion resistance as well as certain aspects of their bioactivity. 1,2−5 These functional roles of Mg are critically dependent on its bonding and coordination environment in the structures of silicate glasses and liquids.On one hand, it is well-established that the addition of MgO to a silicate glass results in the formation of nonbridging oxygen atoms (NBOs).In that sense, it can be argued that MgO plays the role of a modifier oxide. 6On the other hand, increasing addition of MgO to SiO 2 beyond 50 mol % results in an increase in the glass transition temperature, which could be suggestive of the participation of strong Mg−O bonds in the integrity of the structural network. 7owever, despite a large number of structural studies utilizing a variety of spectroscopic and diffraction methods, the nature of the Mg−O coordination environment in silicate glasses has remained rather controversial in the literature.For example, previous X-ray absorption spectroscopic studies at the Mg K-edge indicated the presence of Mg IV (tetrahedral or 4fold coordination), Mg V (5-fold coordination), and Mg VI (octahedral or 6-fold coordination) environments in the structure of CaMgSi 2 O 6 glass. 8,9On the other hand, 25 Mg magic angle spinning nuclear magnetic resonance (MAS NMR) studies conducted at a magnetic field of 14.1 T by Kroeker and Stebbins indicated that, like crystalline CaMgSi 2 O 6 , Mg is present predominantly as Mg VI in the structure of the glass of the same composition. 10The same conclusion was reached by Sen et al. for glasses along the join Mg 2 SiO 4 −MgSiO 3 based on 25 Mg MAS NMR spectra acquired at a magnetic field of 21.8 T and by Shimoda et al. for MgSiO 3 , CaMgSi 2 O 6 , and a few other Ca, Mg silicate and aluminosilicate glasses on the basis of 25 Mg triple-quantum MAS (3QMAS) NMR studies at 16.4 and 21.8 T. 6,11,12 These results on the Mg coordination environment in MgSiO 3 and CaMgSi 2 O 6 glasses have recently been corroborated in a 25 Mg NMR study conducted at 14.1 T by Eckert and co-workers; 13 but this study also indicated that the presence of multiple Mg coordination environments in these glasses cannot be precluded solely on the basis of NMR data at these moderate magnetic fields.In contrast, isotope-substituted neutron diffraction studies by Cormier and Cuello and combined Xray and neutron diffraction and reverse Monte Carlo simulation by Wilding et al. indicated that Mg is present predominantly as Mg IV in MgSiO 3 and other MgO−SiO 2 glasses, with an average Mg−O coordination number ranging between 4.4 and 4.8. 7,14,15uch a significant discrepancy between diffraction and NMR results may be indicative of the challenges associated with each technique for the determination of the coordination environment of cations, such as Mg.While, for diffraction, a distorted nearest-neighbor environment may lead to an erroneous estimation of the coordination number of an atom, NMR spectroscopy of the 25 Mg nuclide presents its own formidable challenges owing to the low gyromagnetic ratio of this nuclide, combined with its low natural abundance (10%) and large quadrupolar broadening.These challenges with 25 Mg NMR can be circumvented by isotopically enriching the samples with 25   Mg and by performing MAS and multiple-quantum MAS (MQMAS) NMR spectroscopy at ultrahigh magnetic fields (>20 T) to minimize or eliminate the effects of quadrupolar broadening.The MQMAS experiment is a two-dimensional NMR spectroscopic technique that averages out the secondorder line-broadening effects from quadrupolar interactions for I > 1/2 nuclides such as 25 Mg (I = 5/2), by mixing single-and multiple-quantum coherences to generate a high-resolution isotropic spectrum in one dimension, which is correlated in the second dimension with the single-quantum MAS centraltransition spectrum. 16The isotropic spectrum is only broadened by the distributions of the chemical shift and the isotropic quadrupolar shift.Mg NMR study by Kroeker and Stebbins, 10 except the glass samples were all rejuvenated for this study before the NMR data collection by melting them at 1450 °C for 1 h in a platinum crucible followed by quenching in air.All samples were originally prepared by Kroeker and Stebbins using 96.75% 25 Mg-enriched MgO. 10 The crystalline compounds were prepared via solidstate synthesis, and their phase identification was carried out using powder X-ray diffraction.The glasses were prepared by using the typical melt-quenching route.Details of the sample synthesis can be found in that study.

Materials. The crystalline
2.2.Ultrahigh-Field 25 Mg NMR. 25 Mg solid-state NMR experiments recorded at 35.2 T were performed on the seriesconnected hybrid (SCH) magnet ( 25 Mg Larmor frequency of 91.83 MHz) at the National High Magnetic Field Laboratory (NHMFL) in Tallahassee, USA. 17  Simulation parameters are listed in Table 1.
The Journal of Physical Chemistry B 14.9 kHz, recycle delay of 1 s, and 1024 averaged transients.For the CaMgSi 2 O 6 , K 2 MgSi 5 O 12 , and Na 2 MgSi 3 O 8 glass samples, the 1D NMR spectra were acquired using QCPMG 18 signal enhancement with π/2 and π pulses of 3 and 6 μs, 0.5 s recycle delay, 54 CPMG pulse acquisition loops with each loop lasting 20 rotor periods, and 3840, 5120, and 8192 averaged transients, respectively.A phase-modulated split-t 1 shifted-echo sequence 19 was used for acquisition of the 2D 3QMAS NMR spectra in combination with QCPMG for signal enhancement.Pulses of 9.9 and 3.3 μs with rf field of ∼44 kHz were used for 3Q excitation and conversion, along with soft pulses of 5 and 10 μs at 16.7 kHz rf field, 0.5 s recycle delay, and 68 CPMG pulse-acquisition loops, with each loop lasting 20 rotor periods.Acquisition of the t 1 evolution was performed by rotor synchronization of the delay between the 3Q excitation and conversion pulses; 8 complex t 1 points were acquired for each spectrum with a total of 2880, 8640, and 5760 and averaged transients per t 1 point for the CaMgSi 2 O6, K 2 MgSi 5 O 12 , and Na 2 MgSi 3 O 8 glass samples, respectively. 25Mg NMR spectra were externally referenced by recording the 17 O signal of D 2 O and using the 17 O and 25 Mg frequency ratios reported in the IUPAC recommendations. 20

RESULTS AND DISCUSSION
The ultrahigh-field 25 Mg MAS NMR spectra of the three crystalline compounds CaMgSi 2 O 6 (diopside), Ca 2 MgSi 2 O 7 (akermanite), and K 2 MgSi 5 O 12 are shown in Figure 1.The center bands of these central-transition line shapes were simulated using the ssNake software package to obtain the NMR parameters including the isotropic chemical shift δ iso , the quadrupolar coupling constant C Q , and the asymmetry parameter η of the electric field gradient tensor at the site of the nucleus. 21As shown in Table 1, the resulting NMR parameters are consistent with those obtained in a previous study by Kroeker and Stebbins from simulation of low-field (14.1 T) line shapes on the same samples. 10It is apparent from Figure 1 that while the central-transition MAS line shapes of crystalline CaMgSi 2 O 6 and Ca 2 MgSi 2 O 7 can be described well with a single set of δ iso , C Q , and η, that of crystalline K 2 MgSi 5 O 12 requires a distribution of quadrupolar parameters.This difference was explained by Kroeker and Stebbins 10 to be due to a disordered structure of the K 2 MgSi 5 O 12 sample, owing to its synthesis conditions.An extended Czjzek distribution of the quadrupolar parameters 21,22 was therefore employed within the ssNake program to simulate the 25 Mg NMR line shape for this crystal.The Mg atoms are known to be octahedrally coordinated in crystalline CaMgSi 2 O 6 , and tetrahedrally coordinated in K 2 MgSi 5 O 12 and Ca 2 MgSi 2 O 7 , consistent with their 25 Mg NMR δ iso values of ∼8, 39, and 47 ppm, respectively, as obtained from these simulations (Table 1).It may be noted here that in the present study, the significant lowering of the quadrupolar broadening in the ultrahigh-field 25 Mg MAS NMR spectra of these crystals reveals the presence of a small amount (∼4%) of a secondary "impurity" phase in the Ca 2 MgSi 2 O 7 sample in the form of a resonance centered near ∼5 ppm (Figure 1).The presence of this phase went undetected in the lower-field 25 Mg MAS NMR spectrum of this sample in the previous study by Kroeker and Stebbins. 10oreover, the small fraction of this phase must have remained undetected in their powder X-ray diffraction measurement as well.The spectral line shape of this resonance could indeed be simulated with a set of NMR δ iso , C Q , and η that is consistent with this secondary phase being crystalline CaMgSi 2 O 6 .
The ultrahigh-field 25 Mg MAS NMR spectra of the three glasses (CaMgSi 2 O 6 , K 2 MgSi 5 O 12 , and Na 2 MgSi 3 O 8 ) investigated in this study are shown in Figure 2.These spectral line shapes can be simulated well, with a single site having an extended Czjzek distribution of quadrupolar parameters (Table 2).A direct comparison between the 25 Mg NMR parameters (Tables 1 and 2) of the crystalline and glassy phases of CaMgSi 2 O 6 and K 2 MgSi 5 O 12 shows relatively small change in δ iso between the crystal and the glass, suggesting that Mg is predominantly 6-fold (4-fold)-coordinated in CaMgSi 2 O 6 (K 2 MgSi 5 O 12 ) glass.On the other hand, as expected, the average C Q for the 25 Mg MAS NMR line shape increases by nearly 4× in the CaMgSi 2 O 6 glass and by 2× in the K 2 MgSi 5 O 12 glass, with respect to their crystalline counterparts (Tables 1 and 2), which can be attributed to a corresponding increase in the structural disorder of the Mg coordination environment.Finally, a strong similarity between the 25 Mg NMR parameters of K 2 MgSi 5 O 12 and Na 2 MgSi 3 O 8 glass indicates a corresponding similarity in their Mg coordination environments (Table 2).These 25 Mg MAS NMR results and the general conclusions obtained at the ultrahigh magnetic field of 35.2 T are completely consistent with those obtained in previous studies carried out at 14.1 T, and besides validating each other they do not provide any new information. 10,13owever, as discussed below, a significantly different structural scenario emerges when the results of the ultrahigh-field 25 Mg 3QMAS NMR spectra of these glasses are considered.
The 25 Mg 3QMAS NMR spectra of the CaMgSi 2 O 6 , K 2 MgSi 5 O 12 , and Na 2 MgSi 3 O 8 glasses are shown in Figure 3.The total projection of these spectra along the isotropic dimension resolves the presence of at least two Mg resonances in all three cases.For a spin I = 5/2 nuclide such as 25 Mg, the δ iso and the quadrupolar product P Q = C Q (1 + η 2 /3) for the two resonances can be estimated from their isotropic peak position δ F1 and their center of gravity δ F2 CG along the MAS dimension using the following relations 23 17 27 10 27

P I I I I (
) 1000 680 2 (2 1) 81 4 ( 1) 3 where ν 0 is the resonance frequency (91.83 MHz) and I = 5/2 is the spin quantum number of the nuclide.Such calculations yield δ iso and P Q values ranging between 12 and 19 ppm and 3.3 and 5.5 MHz, respectively, for one and 30 and 40 ppm and 5.0 and 7.0 MHz, respectively, for the other 25 Mg NMR resonance (Table 3).Unfortunately, the structural assignment of these 25 Mg resonances on the basis of their δ iso values is not an entirely straightforward task due to the fact that highresolution 25 Mg NMR measurements on well-characterized Table 1. 25

The Journal of Physical Chemistry B
crystalline compounds remain rather limited in the literature to establish clear chemical shift ranges for specific Mg−O coordination environments in silicates.Previous 25 Mg NMR studies 11,24,25 indicated that the δ iso for Mg VI environments in some silicate minerals ranged between ∼1 and 15 ppm, while the results of the present study on crystalline silicates as listed in Table 1 indicate that the δ iso for Mg IV environments can at least be as low as ∼39 ppm.Therefore, the two resonances with low (12−19 ppm) and high (30−40 ppm) δ iso values in the 25 Mg 3QMAS NMR spectra of CaMgSi 2 O 6 , K 2 MgSi 5 O 12 , and Na 2 MgSi 3 O 8 glasses can be tentatively assigned to the Mg VI and Mg IV environments, respectively.However, the validity of this assignment hinges somewhat upon whether a fraction of Mg atoms in the structure of these glasses can also be present in 5-fold coordination with oxygen, i.e., as Mg V .Unfortunately, the characteristic δ iso for Mg V species remains unknown at present, though it is expected to lie between ∼20 and 30 ppm (see below).Nevertheless, the existence of Mg V species in these glasses cannot be precluded solely on the basis of the NMR data presented in this study.
It may be noted here that in the absence of any constraint on η, the P Q values obtained from the peak positions in the 3QMAS spectra bracket the estimate of C Q to within ±15%.A second estimate of δ iso and C Q for these two resonances can be obtained from simulations of the MAS projections of these 3QMAS spectra at the two isotropic peak positions using a Czjzek distribution.These simulations are shown in Figure 4, which yield average δ iso and C Q values for the Mg VI and Mg IV environments in these glasses that are consistent with the estimates obtained from the 3QMAS spectral peak positions.To ensure further consistency, these average δ iso and C Q values for the three glasses are used as initial guesses along with an extended Czjzek distribution of the quadrupolar parameters to simulate their 1D 25 Mg MAS NMR line shapes in Figure 2. It is clear from Figure 5 that such two-component simulations are indeed able to reproduce the 1D MAS NMR line shapes quite well.The peak areas obtained from these simulations provide the relative fraction of the Mg VI and Mg IV environments in these glasses (Table 3).This procedure yields a Mg VI /Mg IV ratio of 56:44 for the CaMgSi 2 O 6 glass.This ratio decreases significantly to 30:70 for the K 2 MgSi 5 O 12 glass and 25:75 for the Na 2 MgSi 3 O 8 glass.These ratios are found to be completely consistent with those estimated from a simulation of the total isotropic projection spectra with Gaussian components under the assumption of uniform triple-quantum excitation for both Mg environments.
The resulting average Mg−O coordination number of ∼5.1 for the CaMgSi 2 O 6 glass is consistent with that (∼5.0) reported on this composition by Ildefonse et al. 8 on the basis of a Mg K-edge extended X-ray absorption fine structure (EXAFS) spectroscopic study.Moreover, the simulation Table 2. 25  The Journal of Physical Chemistry B parameters in Table 3 yield an average 25 Mg δ iso of ∼22 ppm, in excellent agreement with the same value reported by George and Stebbins 26 for the CaMgSi 2 O 6 liquid using high-temperature 25 Mg NMR spectroscopy, lending further support to the Mg−O coordination number obtained in the present study.This Mg−O coordination number is also consistent with that (∼5.0) obtained in a recent study by Cormier and Cuello 27 for glass of composition Ca 1.5 Mg 0.5 Si 2 O 6 using combined X-ray and neutron diffraction and reverse Monte Carlo simulation.When taken together, this agreement between NMR, EXAFS, and diffraction studies resolves the long-standing inconsistency between the Mg−O coordination numbers for silicate glasses reported in previous studies and the related controversy in the literature, as noted earlier.It is expected that a similar situation will also hold for the MgSiO 3 glass as diffraction studies have indicated that the average Mg−O coordination number for this glass is lower (∼4.5)compared to its Ca-rich analogues, and this coordination number monotonically decreases with increasing Mg/Ca in glasses along the binary join CaSiO 3 − MgSiO 3 . 14,15,27,28n a previous study, Kroeker and Stebbins hypothesized that Mg competes more effectively with the low field strength alkali cations than with the higher field strength Ca for the NBOs in the silicate network to form more compact tetrahedral coordination environments.This hypothesis is corroborated by the 25 Mg 3QMAS results, which clearly show that compared to the CaMgSi 2 O 6 glass, the Mg VI /Mg IV ratio is significantly lowered in the alkali−Mg silicate glasses (Table 3).It may be noted that a lowering of the average Mg−O coordination number with increasing Na/Mg ratio was also reported in a previous study by Bisbrouck et al. 29 for Na−Mg boroaluminosilcate glasses and a predominantly tetrahedral Mg−O coordination was suggested by Shimoda et al. for Na−Mg and K−Mg silicate glass. 11The results of the present study indicate Table 3. 25 Mg NMR Parameters for Glasses Obtained from Simulation of 1D 25 3).The average 25 Mg δ iso of ∼31−32 ppm for these glasses (Table 3) is again in good agreement with those (∼27−30 ppm) reported by George and Stebbins 26 for Na−Mg and Na, K−Mg silicate liquids using high-temperature 25 Mg NMR spectroscopy.It may be noted here that the tetrahedral coordination of Mg does not necessarily mean that a part of the Mg ions play the role of network-forming cations in these glasses.−32

CONCLUSIONS
In summary, a combination of 25 Mg MAS and 3QMAS NMR at an ultrahigh magnetic field enables identification of mixed  Mg δ iso values for all glasses are found to be in good agreement with those reported for liquids of similar composition in previous high-temperature 25   The Journal of Physical Chemistry B A Bruker AVANCE NEO console and a single-resonance MAS probe designed and constructed at the NHMFL were used with 3.2 mm o.d.pencilstyle rotors spinning at a MAS frequency of 17.86 kHz.The 1D NMR spectra for the crystalline CaMgSi 2 O 6 (diopside), Ca 2 MgSi 2 O 7 (akermanite), and K 2 MgSi 5 O 12 samples were obtained using a single-pulse direct excitation of 2 μs, rf field of

A
Mg NMR Parameters for Glasses Obtained from Simulation of Spectra Acquired at 35.2 T Using a Single Site with an Extended Czjzek Distribution a Gaussian line-broadening of 10 ppm was used to account for a distribution of isotropic chemical shifts.The mean value of η for the distribution was kept constant for all simulations.b Mean value of the extended Czjzek distribution.

Figure 3 .
Figure 3. Contour plots of 25 Mg 3QMAS NMR spectra of (a) CaMgSi 2 O 6 , (b) K 2 MgSi 5 O 12 , and (c) Na 2 MgSi 3 O 8 glasses.Total projections on isotropic and MAS dimensions are shown on the left and top.The red straight line across each plot denotes a quadrupolar-induced shift.

A
Mg MAS Spectra Acquired at 35.2 T Using Two Sites with Extended Czjzek Distributions a composition sites δ iso (±2 ppm) average C Q (±0.1 MHz) b width of Czjzek distribution σ (±0.1 MHz) average η b relative fraction Gaussian line-broadening of 10 ppm was used to account for a distribution of isotropic chemical shifts.The mean value of η for the distribution was kept constant for all simulations.Values in parentheses are δ iso and P Q obtained from peak positions in F1 and center of gravity positions in F2 dimensions of the 3QMAS spectra using eqs 1 and 2. b Mean value of the extended Czjzek distribution.The Journal of Physical Chemistry B that the average Mg−O coordination number is ∼4.6 for the K 2 MgSi 5 O 12 glass and ∼4.5 for the Na 2 MgSi 3 O 8 glass (Table Mg−O coordination environments in silicate glass.The results are interpreted in terms of the coexistence of Mg IV and Mg VI environments in all glasses, although the possibility of the presence of some fraction of Mg in a Mg V environment cannot be ruled out.The relative ratios of Mg IV and Mg VI , thus obtained for CaMgSi 2 O 6 glass, yield an average Mg−O coordination number of ∼5, which is in good agreement with the results obtained in previous Mg K-edge EXAFS and diffraction measurements.This agreement resolves a longstanding inconsistency between Mg−O coordination numbers in silicate glasses derived via NMR spectroscopy vis-a-vis X-ray absorption spectroscopy and diffraction techniques.When taken together, the25 Mg δ iso values for various silicate crystals, glasses, and liquids obtained in this study and in previous studies in the literature allow for the establishment of25 Mg chemical shift ranges of ∼5−17, 20−30, and 30−50 ppm, for Mg IV , Mg V , and Mg VI environments, respectively.Mg shows a stronger preference for tetrahedral coordination in alkali−Mg

Figure 5 .
Figure 5. Two-component simulation of experimental 1D 25 Mg MAS NMR line shapes (black solid line) for (a) CaMgSi 2 O 6 , (b) K 2 MgSi 5 O 12 , and (c) Na 2 MgSi 3 O 8 glasses.Simulation is shown as a green dotted line and individual components are shown with red and blue dot-dashed lines.Asterisks denote spinning sidebands.Note different ppm scales for the spectra.