Structure prediction based on ab initio simulated annealing for boron nitride

Possible crystalline modifications of chemical compounds at low temperatures correspond to local minima of the energy landscape. Determining these minima via simulated annealing is one method for the prediction of crystal structures, where the number of atoms per unit cell is the only information used. It is demonstrated that this method can be applied to covalent systems, at the example of boron nitride, using ab initio energies in all stages of the optimization, i.e. both during the global search and the subsequent local optimization. Ten low lying structure candidates are presented, including both layered structures and 3d-network structures such as the wurtzite and zinc blende types, as well as a structure corresponding to the beta-BeO type.