The Gram-Charlier temperature factor formalism has been applied to a set of accurate low-temperature data on bis(pyridine)(meso-tetraphenylporphinato)iron(II), and to a theoretical set of static structure factors on the hexaaquairon(lI) ion. The refinements are compared with the multipole treatment for atomic asphericity due to chemical bonding. In a treatment of the experimental data in which only the iron atom asphericity is considered, the 'thermal motion' formalism is as efficient as the multipole formalism in accounting for the observations. It is slightly less efficient when applied to the static theoretical data, though model maps based on the two treatments are remarkably similar. A high-order Gram-Charlier refinement of the porphyrin data, followed by a multipole refinement of all data with the Gram-Charlier parameters initially fixed, and later varied, shows that simultaneous refinement of anharmonic and aspherical effects is possible, though the resulting separation may not be accurate. A combined Gram-Charlier multipole refinement on the static data, however, leads to non-significant thermal parameters. It is concluded that the statistical GramCharlier formalism is remarkably successful in representing bonding effects in the valence charge density if these are not specifically accounted for in the scattering formalism. Statistical anharmonic thermal motion formalisms should only be used for X-ray data analysis in combination with a formalism accounting for the effect of bonding on the atomic charge density. * Permanent address: Department of Chemistry, The University, Glasgow G12 8QQ, Scotland.t Permanent address: Central Research Institute for Chemistry, Hungarian Academy of Sciences, Budapest 11, Pusztaszeri UT 59-67, H-1525 Budapest, POB 17, Hungary. ~t To whom all correspondence should be addressed.0108-7673/88/030336-07503.00
IntroductionIt is commonly assumed in crystallographic studies that thermal motion can be adequately described by a formalism based on a harmonic force field. The Fourier transform of the Gaussian probability distribution of harmonically vibrating atoms is the temperature factor component of the time-averaged atomic form factor first introduced by Cruickshank (1956). It has also been realized, however, that there exist not uncommon phenomena in which anharmonic motion might be a significant effect, for example pseudorotation (Cremer & Pople, 1975), ring oscillations in biphenyl-type compounds (Rietveld, Maslen & Clews, 1970), anharmonic vibrations in alloys (Kontio & Stevens, 1982) and semiconductors (McIntyre, Moss & Barnea, 1980) and many others. Internal vibrations of bonds in molecules have a less dominant effect in crystallographic work, but evidence for their anharmonicity is abundantly present in spectroscopic data. Indeed, inclusion of anharmonic covariant tensor coefficients has been reported to be essential in some structure determinations (e.g. Marsh & Abrahams, 1987;Zucker & Schulz, 1982;Johnson, 1969; Willis, 1969). Such anharmonicity leads to devia...