The number of protein structures re®ned at a resolution higher than 1.0 A Ê is continuously increasing. Subatomic structures may deserve a more sophisticated model than the spherical atomic electron density. In very high resolution structural studies (d < 0.5 A Ê ) of small peptides, a multipolar atom model is used to describe the valence electron density. This allows a much more accurate determination of the anisotropic thermal displacement parameters and the estimate of atomic charges. This information is of paramount importance in the understanding of biological processes involving enzymes and metalloproteins. The structure of the scorpion Androctonus australis Hector toxin II has been re®ned at 0.96 A Ê resolution using synchrotron diffraction data collected at room temperature. Re®nement with a multipolar electron-density model in which the multipole populations are transferred from previous peptide studies led to the observation of valence electrons on covalent bonds of the most ordered residues. The re®ned net charges of the peptide-bond atoms were of the correct sign but were underestimated. Such protein-structure re®nements against higher resolution data collected at cryogenic temperature will enable the calculation of experimental atomic charges and properties such as electrostatic potentials.

The experimental electron density distribution in glycyl-L-threonine dihydrate has been investigated using single-crystal X-ray diffraction data at 110 K to a resolution of (sin theta/lambda) = 1.2 A(-1). Multipolar pseudo-atom refinement was carried out against 5417 observed data and the molecular electron density was analyzed using topological methods. The experimental electrostatic potential around the molecule is discussed in terms of molecular interactions. Crystal data: C6H12N2O4. 2H2O, Mr = 212.2, orthorhombic, P2(1)2(1)2(1), Z = 4, F(000) = 456 e, T = 110 K, a = 9.572 (3), b = 10.039 (3), c = 10.548 (2) A, V = 1013.6 (4) A3, Dx = 1.3 g cm(-3), mu = 1.2 cm(-1) for lambdaMo = 0.7107 A.