At the heart of photocatalysis lies the recognition that chemical reactions are catalyzed on specific surfaces of catalyst particles. Because of the specific structure of such particles and the coordination of exposed surface atoms, the reactivity of different surfaces and their mechanisms for bonding to adsorbed molecules can vary dramatically. [1,2] Information about the size and shape of titanium dioxide (TiO 2 ) nanocrystals and, in particular, about the abundance and relative surface areas of specific crystal facets is highly desirable for the development of sophisticated photocatalytic experiments that are based on powdery materials. Unfortunately, for real catalysts, the kind of crystal facets that are exposed and their relative abundances are often not known. Profound knowledge about the relationship between processing conditions and the abundance of specific crystal facets would allow for tailoring improved catalysts that show predominantly those surfaces that are catalytically most active for specific reactions. The discovery of the decomposition of water on TiO 2 electrodes [3] stimulated many further investigations and opened up the research area of photocatalysis. Today, titanium dioxide is by far the most important photocatalyst material. It is commercially available as powders in two crystalline modifications, anatase and rutile, with surface areas ranging from 10 to more than 300 m 2 g
À1. There is some evidence that the surface energy densities of anatase are generally lower than those of rutile. [4][5][6] However, theoretical predictions of the relative stability of perfect rutile and anatase surfaces do not give a coherent picture. Depending on the quantum-chemical method, the calculated rutile (110) surface energy density is found to be smaller [7] or larger [8,9] than those of all anatase low-index surfaces. Calculations based on classical pair potentials [6,10] disagree as to the relative stability of different anatase surfaces. All of these theoretical calculations are for perfect clean TiO 2 surfaces. A study for water-covered surfaces predicts anatase (101) and anatase (100) to be more stable than rutile (110), depending on the coverage with hydrogen or oxygen. [5] In another theoretical study of rutile and anatase nanoparticles of 2-6 nm diameter, [11] it was concluded that anatase particles have smaller surface energy densities than rutile up to a diameter of 2.5 nm. Consequently, for small crystals that show high surface-tovolume ratios, it is advantageous to adopt the anatase modification, and large titania crystals are thermodynamically favored if they have the rutile structure. The rutile-to-anatase transition occurs at particle sizes of a few tens of nanometers, and it depends on temperature, grain size, nanocrystal morphology, impurity content, reaction atmosphere, and synthesis conditions. [5,12]