High-resolution 27A1 MAS NMR spectra of natural leucite recorded at Ho = 11.7T contain three resolvable resonances at 27A1 c5~ =69.2, 64.7, and 61.0+0.5ppm. These three resonances are assigned to the three inequivalent framework positions of leucite: T3, T2, and T1, respectively. Fitting the observed spectra yields a Si,A1 distribution for leucite in which approximately one-half of the A1 is in T1 and one-quarter in each of T2 and T3. This Si,A1 distribution differs substantially from those obtained by previous workers using 29Si NMR spectroscopy and X-ray diffraction. New 29Si NMR spectra and revision of previously reported 29Si NMR peak assignments, however, make the 27A1 and 29Si NMR results consistent. The 27A1 6~ correlate linearly with the mean T-O-T' bond angles of the average structure, which allows the peak assignments to be made. However, this correlation lies distinctly toward higher frequency and larger bond angles than correlations for Si,A1 ordered aluminosilicates, suggesting that the mean T(A1)-O-T'(Si) bond angle for each site in leucite is smaller than the mean bond angle of the average structure, which is averaged over T(A1)-O-T'(Si) and T(Si)-O-T'(Si,A1) angles. framework of leucite are not evenly distributed among the three crystallographically inequivalent framework positions.
The 29Si NMR spectrum of leucite is complicated, consisting of eight readily discernible peaks or shoulders (Fig. 1 ; Fig. 1 of Brown et al. 1987; Fig. 4 of Murdoch et al. 1988). However, there are 15 different local Si-environments in leucite: 5 chemically distinct environments (0-4 A1 next-nearest framework cations) for each of 3 crystallographically different sites. The spectral features must, therefore, be some combination of the 15 peaks corresponding to these Si-environments. Brown et al. (1987) and Murdoch et al. (1988) obtained values for the distribution of Si and A1 among the three crystallographic positions by assigning the 29Si NMR peaks and calculating spectra based on statistical models for the extent of short-range Si,A1 ordering. (Throughout this paper, we use the term "short-range Si,A1 ordering" to denote how the Si on, for example, T1, are distributed among TI(4A1), TI(3A1), TI(2A1), TI(1A1), and TI(0A1) environments.) By varying the fraction A1 on each of the three crystallographic sites they were able to fit the observed spectra relatively well. These two studies differ in the assumed Si,A1 ordering used in the models; the spec-