Surface reconstruction and band alignment of non-metallic spinel oxides from first principles

A systematic prediction of the reconstructed surfaces of non-metallic normal spinel oxides is conducted using a global structure search scheme that combines Bayesian optimisation and the evolutionary algorithm, adopting first principles calculations for local optimisations. Their (100) surfaces with macroscopically stoichiometric and nonpolar terminations are investigated, yielding various representative reconstructed surface geometries for $A^{2+}{B}_2^{3+}{\mathrm{O}}_4$$A^{2+}{B}_2^{3+}{\mathrm{O}}_4$, $A^{4+}{B}_2^{2+}{\mathrm{O}}_4$$A^{4+}{B}_2^{2+}{\mathrm{O}}_4$, and $A^{6+}{B}_2^{+}{\mathrm{O}}_4$$A^{6+}{B}_2^{+}{\mathrm{O}}_4$ systems. The competition between A cations on tetrahedral sites and B cations on octahedral sites, in the covalency of their bonds with O ions, emerges as a primary factor that accounts for the observed chemical trends in the formation of distinct reconstructed geometries and their associated surface energies. This covalency competition identifies the key cations that govern the surface reconstruction patterns, with the coordination numbers of surface cations acting as a secondary factor influencing the surface energy. Additionally, the ionic-size effect plays a supplementary role in surface reconstruction to covalency competition, particularly when there is a significant ionic-size difference between A and B cations. The band alignment of these surfaces, determined by calculating their ionisation potentials and electron affinities, is also presented and discussed in relation to the reconstructed geometries. This study offers useful insights into the exposure of cation/anion sites on the reconstructed surfaces of normal spinel oxides that is often relevant to desired surface functionalities, such as those for electrocatalysts and photocatalysts, and provides a framework that can be reasonably generalised to the surface reconstruction of other spinel compounds.