Oxygen triclusters in crystalline CaAl4O7 (grossite) and in calcium aluminosilicate glasses: 17O NMR

We present O MAS NMR data for crystalline calcium dialuminate (grossite), CaAl4O7 and monoaluminate, CaAl2O4. The first of these contains an oxygen tricluster site and serves as a model compound for sites of this type in aluminosilicate glasses. Tricluster site NMR parameters are distinct from those of bridging O atoms (Al-O-Al), allowing partial resolution in triple quantum MAS NMR spectra. Such spectra for calcium aluminosilicate glasses are consistent with the presence of a small fraction of tricluster sites. Observed chemical shifts for non-bridging oxygen (NBO) atoms in an impurity phase in the CaAl2O4 sample are distinct from those for NBO in Ca-aluminosilicate glasses, indicating that the latter are primarily bonded to Si, not Al. been presented for their existence, although speculation was made about assigning an “extra” feature in the O spectra to such groups (Stebbins and Xu 1997). Formation of triclusters was also suggested to explain changes in Raman spectra of CaAl2O4 melts at high temperature (Daniel et al. 1996). If triclusters are present, compositional effects on viscosity anomalies in the high silica part of the Na-aluminosilicate system suggested that O(AlSi2) groups predominate, but that O(Al2Si) and/or O(Al3) groups are most important in the Caaluminosilicate system (Toplis et al. 1997; Toplis and Dingwell 1998). Local charge balance would seem to favor O(Al3), as in crystalline mullite [Al2(Al2+xSi2–x)O10–x] (Angel et al. 1993). To accurately interpret NMR results on glasses, data on representative crystalline model compounds are needed, although ab initio quantum mechanical calculations of NMR observables such as isotropic chemical shift (δiso) and quadrupolar coupling constant (CQ) and asymmetry parameter (η) are rapidly becoming accurate enough to be predictive (Vermillion et al. 1998; Xue and Kanzaki 1999). Calcium dialuminate (grossite, CaAl4O7), in which one of four O sites is a simple tricluster with three Al and no other neighbors, is an excellent potential model for O(Al3) sites in Ca-aluminosilicate glasses. The remaining O sites in CaAl4O7 all are “bridging O atoms” (AlO-Al), with one or two Ca neighbors for charge compensation (Goodwin and Lindop 1970; Ponomarev et al. 1971), and can serve as models for such sites in glasses as well. Here we present O NMR data on CaAl4O7. We suggest that these results for O(Al3) triclusters are consistent with their presence in Ca-aluminosilicate glasses, although future work will be needed to confirm this. New, higher resolution spectra for CaAl2O4 at a field of 18.8 Tesla also clarify previous results for Al-O-Al groups (Stebbins et al. 1999): this phase has a “stuffed silica” structure in which all O atoms are in such linkages (Hörkner and Müller-Buschbaum 1976). In addition, new results on NBO in Ca-aluminate crystals and glasses better refine the role of these types of O atoms in glasses. STEBBINS ET AL.: OXYGEN TRICLUSTERS 1308 EXPERIMENTAL METHODS CaAl4O7 enriched in O was synthesized from CaCO3 and isotopically-enriched Al2O3 (Stebbins et al. 1999). Three cycles of solid-state reaction for 2 h at 1600 °C under Ar, followed by grinding, were required to produce a monophasic (by XRD) sample, but apparently led to significant loss of O and a relatively low enrichment level (probably <10%). O NMR results on this material (see below) suggested that a previously described sample of O-CaAl2O4 (Stebbins et al. 1999) might contain a significant amount of CaAl4O7 in addition to the reported non-bridging oxygen-containing (glassy?) impurity noted at the time, apparently missed in the routine powder XRD scan used to characterize the material. Therefore, the CaAl2O4 sample was reacted with an additional 5 wt% CaO by partially melting under Ar at 1560 °C followed by crystallization for 2 h at 1340 °C in order to eliminate residual glass and CaAl4O7. Synthesis of the calcium aluminosilicate glass (“CAS”) was also described previously (Stebbins et al. 1999); its nominal composition lies on the CaAl2O4–SiO2 join with a Si/Al ratio of 0.5 (CaAl2SiO6). NMR spectra for O were collected at 54.2 MHz with a Varian 400 (9.4 Tesla field) and at 81.3 and 108.4 MHz with Varian Inova 600 and 800 spectrometers (14.1 and 18.8 T). At 9.4 T, a 5 mm MAS probe from Doty Scientific was used with sample spinning rates of about 15 kHz; at 14.1 and 18.8 T, 3.2 mm Varian/Chemagnetics probes were used with spinning rates of 18 to 20 kHz. Radio frequency tip angles of about 15 to 30° were used (0.2 to 0.8 μs); recycle delays of 1 to 5 s were chosen to optimize signal to noise ratios but were tested to ensure that there was no differential relaxation among different sites (no change in the shapes of the spectra with increased delay time); frequencies are reported relative to 20% O H2O. Spectra were fitted with the Varian software “STARS,” with manual optimization of the agreement between calculated and observed spectra at one or two magnetic fields as needed. The triplequantum magic angle spinning (3QMAS) method and the spectra for the CAS glass were previously described (Stebbins et al. 1999). Long spin-lattice relaxation times for the crystalline Ca-aluminates made collection of 3QMAS spectra on these samples impractical. The single-pulse Al spectrum for CaAl4O7 was acquired at 18.8 T (208.3 MHz), and was referenced to 1 M, acidified, aqueous Al(NO3)3.