Efforts to compute accurate all-electron density-functional energies f
or large molecules and clusters using Gaussian basis sets are reviewed
and their use in fullerene science described. The foundation of this
effort, variational fitting, is described first. When discovered exper
imentally, C-60 was naturally assumed to be particularly stable, but l
ocal-density-functional calculations showed that C-60 is quite unstabl
e relative to the higher fullerenes and graphene (a single sheet of gr
aphite). In addition to raising questions about the relative abundance
of the various fullerenes, this work conflicted with the then state-o
f-the-art density-functional calculations on crystalline graphene. Now
high accuracy molecular and band structure calculations are in fairly
good agreement with each other and experiment. These calculations cle
arly demonstrate that each of the 12 pentagons, which are necessary to
close a fullerene, is best viewed as a rather high-energy, more than
2 eV, defect in a graphene sheet. The effect of the heptagon, the seco
nd most common defect in fullerene materials, is described. Most recen
tly, we have developed accurate, variational gradient-corrected forces
for use in geometry optimization of clusters and in molecular-dynamic
s simulations of friction. The gradient-corrected optimized geometry o
f C-60 is given. (C) 1997 John Wiley & Sons, Inc.