“…Furthermore, the phase problem, which is mainly important when non-centrosymmetric space groups are involved, is another implicit source of inaccuracy. Even if the multipolar models of &(r) permit a much better description of phases than the spherical-atom model ( Souhassou et al, 1991), the results have to be analysed critically (El Haouzi et al, 1996;Spackman & Byrom, 1997; Pe  re Á s et al, 1999): atomic constraints must be applied in some cases and several re®nement strategies must be tested to drive the convergence to the most realistic minima. Finally, it is noteworthy that in the intermolecular regions on which this study is focused features in &(r) are of much lower magnitude than those around atomic positions, and thus they can be more sensitive to small variations induced by different experiments, methods, models and strategies.…”