A number of recent studies indicate that, under the oxygen rich conditions of oxidation catalysis, some transition metal catalysts may be covered by thin oxide overlayers. Moreover, it has been suggested that such “surface-oxide” layers are catalytically active, possibly more active than the pure metal surfaces as was traditionally assumed. This contemporary picture can be traced back to Ag catalysis, where over ago it was suggested that the top layer of Ag(111) reconstructed to an epitaxial like overlayer upon exposure to oxygen [Rovida et al, Surf. Sci. 43, 230 (1974)]. Extensive experimental work, including scanning tunneling microscopy studies in which the oxide was apparently imaged with atomic resolution, as well as density-functional theory calculations, largely confirmed this interpretation. However, a review of published experimental data and new density-functional theory results presented here indicate that previous conclusions are significantly incomplete and that the structure of this original surface oxide must be reconsidered.
This estimate was made through a somewhat circuitous route, namely, by comparing the ratio of O to Ag Auger peak heights for O on Ag(111) to similar data for a phase of known coverage of O on Ag(100).
Although a slow heating rate of was used, the double peak nature of the desorption spectra was not resolved. Instead the main peak was rather asymmetric with a pronounced high temperature tail. It is likely that this high temperature tail, which extended up to , would have concealed a second peak, if one were present.
The isolated depressions on the Ag(111) triangular terraces are believed to be O atoms. See Refs. 20 and 21 for more details.
These Ag adatoms correspond to Ag atoms in bulk that would be bonded to O atoms in adjoining trilayers.
The latter differs from the former merely by the surface energy of the clean reference surface [here this is Ag(111)].
These calculations were done in unit cells similar to the cell used for the adsorption calculations. The structure of a free-standing trilayer film was optimized as a function of unit cell area by varying together both unit cell vectors from to of their value in the adsorption cell.
The O–Ag distances within the triangular building blocks of each overlayer are all . The Ag–Ag distances differ by as much as , specifically the Ag–Ag distances representing the base of each pyramid are . This reflects, however, large variations in Ag–Ag distances within individual overlayers, in particular within the like models, rather than an inherent difference between the various overlayer structures themselves.
We notice that the coordination of the basal oxygen atoms to Ag atoms in the overlayer also affects the registry of the oxide-like overlayer with the Ag(111) substrate. Stable structures with onefold coordinated O atoms invariably have their Ag–O bonds of these “undercoordinated” basal O atoms directed toward atop sites of the underlying substrate. Structures with twofold coordinated O atoms tend to have these oxygens located at bridge sites on the underlying Ag(111) substrate. For the structures in which all oxygen atoms are threefold coordinated, i.e., the like overlayers, a strong preference for the location of the basal oxygen atoms has not been observed.
We add a (somewhat obvious) word of caution concerning the interpretation of Fig. 5 and indeed plots like this in general. Only phases which we have calculated will appear on it. Thus there may always exist a lower energy structure which we have not yet identified. Of course, however, nature does not always find the lowest energy structures either, diamonds and nanotubes are metastable yet long lived forms of carbon!
Some valence band photoemission experiments have also been reported for the overlayer (Refs. 18 and 30). However, they do not agree on the location of the O adsorbate induced resonances. Moreover, when we calculate the total and partial densities of states for the various low energy structures under consideration here we find them to be essentially indistinguishable.