The prediction of high pressure phases of a chemical system is realized as
a two-step process: identification of structure candidates through the glob
al exploration of their energy landscapes over a range of different pressur
es, followed by a limited local optimization using ab initio methods. The a
pplication of this recipe is presented for several systems, Li3N, Na3N, Lix
Na(6 - x)N2 (x = 1,...,5), Li2S and Na2S. We find that at standard pressure
, the optimal configurations for the binary end compositions of the alkali
nitrides, Li3N and Na3N, exhibit the Li3N and Li3P structures, respectively
. Among the ternary compounds, the compositions with x = 2 and x = 4 are pr
eferred, the optimal structures being ternary variations of the Li3N struct
ure type. At moderately high pressures, phase transitions from Li3N to Li3B
i related structures are predicted for all ternary compounds, while Na3N an
d Li3N exhibit transitions from Li3P to Li3Bi and Li3N via Li3P to Li3Bi, r
espectively. The analogous study of the landscapes of Li2S and Na2S shows t
hat a sequence of phase transitions from the CaF2 structure via the PbCl2 s
tructure to the Ni2In structure is expected.