The quality of the extraction of electron density distributions by means of
a multipole refinement method is investigated. Structure factors of the ur
ea crystal have been obtained from an electron density distribution (EDD) r
esulting from a density function calculation with the CRYSTAL95 package. To
account for the thermal motion of the atoms, the stockholder-partioned den
sities of the atoms have been convoluted with thermal smearing functions, w
hich were obtained from a neutron diffraction experiment. A POP) multipole
refinement yielded a good fit, R = 0.6%. This disagreement factor is based
on magnitudes only. Comparison with the original structure factors gave a d
isagreement of 0.8% owing to differences in magnitude and phase. The fitted
EDD still showed all the characteristics of the interaction density After
random errors corresponding to the experimental situation were added to the
structure factors, the refinement was repeated. The lit was R = 1.1%, This
time the resulting interaction density was heavily deformed. Repetition wi
th another set of random errors from the same distribution yielded a widely
different interaction density distribution, The conclusion is that interac
tion densities cannot be obtained from X-ray diffraction data on non-centro
symmetric crystals.