On the accurate prediction of the optical absorption energy of F-centers in MgO from explicitly correlated ab initio cluster model calculations

Abstract

A systematic study of the different computational requirements that affect the accuracy of the ab initio prediction of excitation energies of F and F+ centers on cluster models of MgO is reported. It is found that rather limited basis sets are enough to predict excitation energies of the F and F+ centers that are near to each other as experimentally observed. However, the absolute value of the excitation energy is in error by ∼1 eV or ∼20%. Increasing the basis set reduces the calculated excitation energy for the allowed transition, reaching a value of 5.44 eV for the F center, only 9% in error with respect to experiment. Improving the basis set does not result in a better value of the excitation energy of the charged F+ center. Attempts to improve the calculated result by geometry optimization of the region near the oxygen vacancy, enlarging the cluster model, improving the primitive Gaussian set, or enlarging the auxiliary basis set centered on the vacancy failed to further reduce the error. It is concluded that much larger basis sets are required to predict excitation energies of electrons trapped at oxygen vacancies in ionic oxides with accuracy of or better than 0.4 eV.