A systematic theoretical study of stoichiometric clusters (N aI) n up to n = 15 is performed using the ab initio Perturbed-Ion (PI) model. The structures obtained are compared to previous pair potential results, and observed differences between (N aI) n clusters and previous ab initio results for other alkali halide clusters are discussed. (N aI) n clusters with n up to 15 do not show yet a marked preference for geometries which are fragments of the bulk lattice. Instead, stacks of hexagonal rings or more open structures are obtained as ground structures in clusters with n = 3, 6, 7, 9, 10, 12, 13 and 15, indicating that convergence to bulk structure is not achieved yet at this size range. Low lying isomers which are fragments of the crystal lattice exist, nevertheless, for those cases. The binding energies show that clusters with n = (4), 6, 9 and 12 molecules are specially stable. The binding energy has been decomposed in contributions which allow for an intuitive interpretation. Some electronic properties like ionization potentials and electronic energy levels are also studied.

The melting-like transition in sodium clusters NaN , with N=55, 92, and 142 is studied by using constant-energy molecular dynamics simulations. An orbital-free version of the Car-Parrinello technique is used which scales linearly with system size allowing investigation of the thermal behaviour of large clusters. The ground state isomer of Na142 (an uncomplete three-shell icosahedron) melts in two steps: the first one (at ≈ 240 K) is characterized by the high mobility of the atoms located on the cluster surface; the second, homogeneous melting (at ≈ 270 K), involves diffusive motion of all the atoms across the cluster. For the case of Na92, the icosahedral structure has a larger number of surface vacancies, and melts in two well separated steps, surface melting at ≈ 130 K and homogeneous melting at ≈ 240 K. Na55, a complete two-shell icosahedron, melts in a single stage at ≈ 190 K. Our results on homogeneous melting for Na142 and Na92 are in excellent agreement with recent experimental determinations of melting temperatures and latent heats. However, the experimentally observed enhancement of the melting temperature around N=55 is not reproduced by the calculations.