The cohesive energetics of three phases of solid cubic rubidium chloride, the zinc blende structured 4:4 phase, the 6:6 sodium chloride polymorph and the 8:8 phase with the cesium chloride structure, are computed using a non-empirical fully ionic model. The rearrangement energies needed to convert free anions to their optimal states in-crystal, two-body inter-ionic potentials, plus the further contributions arising from electron correlation, are reported. The 'optimal' anion–anion potentials, computed by using at each geometry the optimal wavefunction, are compared with the 'frozen' potential using the same wavefunction at all geometries.

The lattice energy of the 4:4 structure is predicted to be some 40 kJ mol⁻¹ smaller than that of either the 6:6 or the 8:8 phases. Introduction of the Axilrod–Teller triple dipole dispersion interactions and the vibrational zero point energy predicts the 8:8 phase to lie 3.2 kJ mol⁻¹ lower in energy than the 6:6 structure. This is both consistent with radius ratio arguments and supported by two separate experiments that strongly suggest that the 8:8 phase is favoured over the 6:6 structure at low temperatures even though the latter is more stable at ambient temperatures.

A shell model description is presented for the ion-induced dipole interactions that arise both in small clusters and in crystals encapsulated in nanotubes. The elastic constants and entropy at 300 K predicted for the 6:6 phase from this model by using the GULP program agree well with experiment. A smaller entropy is predicted for the 8:8 structure.