We present ab initio density functional studies of the stability and the formation of oxygen vacancies on three low-index surfaces of ${\text{V}}_2{\text{O}}_5$. In agreement with the experimental results on the morphology of ${\text{V}}_2{\text{O}}_5$ crystallites we find that the surface energies of the (100) and (001) surfaces are considerably higher than that of the (010) surface such that in equilibrium 85% of the surface area is occupied by (010) surfaces. However, we also find that the energies required for the formation of oxygen vacancies are considerably lower on the energetically less favorable surfaces. As reported in earlier theoretical studies we find that on the (010) surface, the elimination of an oxygen atom from the vanadyl group [denoted O(1)] requires considerably less energy ($E_f=4.7\text{eV}$ relative to atomic oxygen) than the formation of a vacancy by desorption of twofold or threefold coordinated O(2) and O(3) oxygen atoms $\left(E_f=6.5\text{eV}\right)$. In addition, the thermodynamic analysis shows that under conditions required for vacancy creation (very low value of the oxygen partial pressure) the ${\text{V}}_2{\text{O}}_5$ phase is unstable and may transform into ${\text{VO}}_2$ by releasing molecular oxygen. Due to extensive framework relaxation, the formation of vacancies at the O(1) and O(2) sites of the (001) and (100) surfaces requires 1.0–1.5 eV less energy than on the (010) surface. A thermodynamic analysis demonstrates that on the (010) surfaces only O(1) vacancies are marginally stable against reoxidation under strongly reducing conditions, while on the (100) and (001) surfaces all types of vacancies are stable even under much higher partial pressures of oxygen. In addition, the adsorption of atomic hydrogen and the formation of hydroxyl groups had been studied on all three surfaces. Hydrogen adsorption is an exothermic process. Adsorption energies on vanadyl O(1) atoms are larger by about 1.1 and 1.6 eV on the (001) and (100) surfaces than on the (010) surface and adsorption energies on O(2) sites are larger by 0.8–1.2 eV. Vacancy formation by elimination of a hydroxyl group requires less energy than abstraction of an oxygen atom. On the (010) surface the vacancy formation energies are reduced by 1.6–1.7 eV; on the (001) and (100) surfaces the reduction varies between 0.1 and 0.9 eV. However, the lowest vacancy formation energies are still 3.1 eV on the (010) and 2.9 and 2.3 eV on the (001) and (100) surfaces, respectively. The lower vacancy formation energies mean that although in equilibrium only about 15.5% of the surface area of a crystallite consists of (001) and (100) facets, these surfaces will make a considerable contribution to the activity of ${\text{V}}_2{\text{O}}_5$ as an oxidation catalyst.

Ab initio density functional calculations of the structural and electronic properties of V2O5 bulk and its low-index surfaces are presented. For the bulk oxide and the (010) surface (the natural cleavage plane) a good agreement with experiment and with earlier ab initio calculations is found. For the first time, the investigations are extended to other low-index surfaces: (001) and (100). On both surfaces, termination conserving a bulk-like stoichiometry is preferred, but—in contrast to the (010) surface—a strong structural relaxation takes place. Relaxation reduces the surface energy from 1.16 to 0.48 J m−2 for the (001) and from 0.61 to 0.55 J m−2 for the (100) surface. Although the relaxed surface energies are still one order of magnitude higher than calculated for the (010) surface (0.047 J m−2), the Wulff construction demonstrates that (001) and (100) surfaces contribute about 15% of the total surface area of a V2O5 crystallite, indicating a non-negligible role in the catalytic activity of V2O5.