Modeling doped and defective oxides in catalysis with density functional theory methods: Room for improvements

Due to the well-known problem of the self-interaction, standard density functional theory (DFT) methods tend to produce delocalized holes and electrons in defective oxide materials even when there is ample experimental evidence of a strong localization. For late transition metal compounds or rare earth oxides, this results in the incorrect description of the electronic structure of the system (e.g., magnetic insulators are predicted to be metallic). Practical ways to correct this deficiency are based on the use of hybrid functionals or of the DFT+U approach. In this way, most of the limitations related to the self-interaction are removed, and the electronic structure is properly described. What is less clear is to what extent hybrid functionals, DFT+U approaches, or standard DFT functionals can properly describe the strength of the chemical bonds at the surface of an oxide. This is a crucial question if one is interested in the catalytic properties of oxide surfaces. Oxidation reactions often involve oxygen detachment from the surface and incorporation into an organic substrate. Oxides are doped with heteroatoms to create defects and facilitate oxygen removal from the surface, with formation of oxygen vacancies. Do standard DFT calculations provide a good binding energy of the missing oxygen despite the failure in giving the right electronic structure? Can hybrid functionals or the DFT+U approach provide a simple yet reliable way to get accurate reaction enthalpies and energy barriers? In this essay, we discuss these problems by analyzing some case histories and the relatively scarce data existing in the literature. The conclusion is that while modern electronic structure methods accurately reproduce and predict a wide range of electronic, optical, and magnetic properties of oxides, the description of the strength of chemical bonds still needs considerable improvements. (C) 2008 American Institute of Physics.