Ab initio molecular orbital calculations (Hartree-Fock, KF and density func
tional theories, DFTs) have been carried out for SiO2 polymorphs coesite, l
ow cristobalite, and alpha-quartz, in order to investigate the reliability
of this method for predicting Si-29 and O-17 nuclear magnetic resonance (NM
R) properties of silicates. Oxygen- and silicon-centered clusters consistin
g of one (1T) to three tetrahedral (3T) shells (one to four atomic shells),
taken from real crystal structure, have been investigated. It is found tha
t for reasonable predication of both the Si-29 and O-17 chemical shifts (de
lta(i)(Si) and delta(i)(O)), the minimum cluster is one that gives the corr
ect second neighbors to the nucleus of interest. Both the delta(i)(Si) and
delta(i)(O) have reached convergence with respect to cluster size at the OH
-terminated two tetrahedral (2T) shell (three atomic shells around Si and f
our atomic shells around O) model. At convergence, the calculated delta(i)(
Si) values agree well (within +/- 1 ppm) with experimental data. The calcul
ated O-17 electric field gradient (EFG)-related parameters also agree with
experimental data within experimental uncertainties. The calculation also r
eproduces small differences in delta(i)(O) for O sites with similar tetrahe
dral connectivities, but shows deviations up to about 10 ppm in relative di
fference for O sites with different tetrahedral connectivities. The poor pe
rformance for the latter is mainly due to the approximations of the HF meth
od. Our study thus suggests that the ab initio calculation method is a reli
able mean for predicting Si-29 and O-17 NMR parameters for silicates. Such
an approach should find application not only to well-ordered crystalline ph
ases, but also to disordered materials, by combining with other techniques,
such as the molecular dynamics simulation method. (C) 2000 Elsevier Scienc
e B.V. All rights reserved.