Conceptually, the governing principle of any energy-minimisation technique of computer modelling of crystal structure and properties is self-evident. At 0 K the atomic array corresponding to the minimum cohesion energy, which represented as the sum over all interatomic interaction energies, is thought of as the most stable structure. However, this general principle occurs to be somewhat ambiguous, when the calculated minimum energy should be compared with experimental data. Indeed, a reference state of the cohesion energy for crystals of various types could be defined in a different way. For instance, the most natural way is to compare the calculated cohesion energy of covalent and metallic crystals with the experimental atomisation energy, i.e. the energy that gains in formation of such a crystal from a gas of isolated (noninteracting) atoms.
The cohesion energy of molecular crystals (either organic or inorganic) could be most appropriately defined as the sublimation energy, i.e. in this case the reference state is a gas of isolated and noninteracting molecules.
On the other hand, the cohesion energy of an inorganic crystal or mineral is most often regarded as the lattice energy which is the energy gain in formation of a crystal from an infinitely diluted gas of isolated ions (cations and anions). It implies that such ions do exist as stable particles in the gaseous state. It holds true for all cations and for the univalent anions like H–, F–, Cl–, Br–, I–, OH–, but that is not the case for all multivalent anions like O2–, S2–, Se2–, N3– etc.