Mixed phase TiO₂ powders are very effective in photo-catalysis. It is assumed that this efficiency is connected to the separation of photogenerated mobile electrons and holes between rutile and anatase, controlled by the offsets between their valence and conduction bands. The actual alignment is, however, debated. Our high level electronic structure calculations suggest that, at typical rutile/anatase interfaces, the band edges of rutile lie higher in energy than those of anatase. This conclusion is reached by taking into account the generic alignment of the bulk band structures and the specific structure of the interface. The latter has been obtained by simulated annealing, using approximate quantum mechanical molecular dynamics on models of the rutile(100)/anatase(100) and of the rutile(110)/anatase(101) interface. Our results are corroborated by photoelectron spectroscopy of actual rutile/anatase interfaces in the literature. The predicted band offsets would lead to accumulation of mobile electrons in anatase and mobile holes in rutile, with the process being also supported by electron self-trapping in rutile and hole self-trapping in anatase. We show, however, that interface defects, like the oxygen vacancy, may provide a recombination channel. Such effects may explain the variation in the suggested direction of charge transfer in many experiments.

The rutile(100)/anatase(100) interface. Titanium atoms are drawn with gray and oxygen with red circles. The black arrows show the preferred direction in the charge transfer of photogenerated electrons and holes.