AIRSS data for the prediction of the FCC high-pressure phase of Na3P

The ab initio random structure searching (AIRSS) techniqueis used to search for candidate structures. AIRSS is a stochastic method which generates structures randomly with a given number of formula units. A minimum atom-atom separation is specified for the generated structures. These separations are chosen based on short AIRSS runs. Symmetry constraints are imposed on our generated structures. This strategy tends to speed up thesearches because low-symmetry structures are unlikely to have low energies according to Pauling’s principle 30,31, although we allocate part of our searching time to check low symmetry structures, for completeness. The generated structures are relaxed using first-principlesquantum mechanical density functional theory (DFT)methods as implemented in the CASTEP code. This procedure is carried out in parallel to ensure a high throughput of results. AIRSS has a proven track record of predicting structures in a diverse variety of systems that have subsequently been verified by experiment, such as in compressed silane, aluminium hydride, high-pressure hydrogen sulfide, and xenon oxides. We limit our searches to a maximum of eight formula units (24 atoms) per cell.