The same Bragg reflection in TiO2 from 12 different (CBED)
patterns (from different crystals, orientations, and thicknesses) are
analyzed quantitatively to evaluate the consistency of the quantitative
CBED method for bond-charge mapping. The standard deviation in the
resulting distribution of derived X-ray structure factors is found to
be an order of magnitude smaller than that in conventional X-ray work,
and the standard error (0.026% for FX(110)) is slightly
better than obtained by the X-ray Pendellösung method applied to
silicon. This is sufficiently accurate to distinguish between atomic,
covalent, and ionic models of bonding. We describe the importance of
extracting experimental parameters from CCD camera characterization,
and of surface oxidation and crystal shape. The current experiments
show that the QCBED method is now a robust and powerful tool for
low-order structure factor measurement, which does not suffer from the
large extinction (multiple scattering) errors that occur in inorganic
X-ray crystallography, and may be applied to nanocrystals. Our results
will be used to understand the role of d-electrons in the chemical
bonding of TiO2.