Revisiting the Structure of the p 4 4 Surface Oxide on Ag ( 111 )

0031-9007= Scanning tunneling microscopy (STM) and density-functional theory are used to reexamine the structure of the renowned p 4 4 -O=Ag 111 surface oxide. The accepted structural model [C. I. Carlisle et al., Phys. Rev. Lett. 84, 3899 (2000)] is incompatible with the enhanced resolution of the current STM measurements. An ‘‘Ag6 model’’ is proposed that is more stable than its predecessor and accounts for the coexistence of the p 4 4 and a novel c 3 5 3 p rect phase. This coexistence is an indication of the dynamic complexity of the system that until now has not been appreciated.

Under ambient and oxygen-rich conditions, ultrathin oxide films may form on the surfaces of transition metals [1][2][3][4][5][6][7].Such nanoscale surface oxides are not only relevant to corrosion and corrosion resistant materials, but they may also be the active phases of transition metal oxidation catalysts, as has been demonstrated by recent theoretical and experimental studies [1][2][3][4][5][6][7].The surface oxide that has perhaps received the most attention is a p4 4 phase that forms on Ag(111) [6].This is in some respects the ''original'' surface oxide since it was the first to be identified in the 1970s [8,9], and indeed ever since it has been a key challenge of surface science to achieve an atomic-scale characterization of its structure [6].In fact, for some years it was believed that the structure of this surface oxide had been ''solved'' by means of the scanning tunneling microscopy (STM) experiments of Carlisle et al. [10,11].The unique structural model that emerged from this work and which was supported by subsequent density-functional theory (DFT) studies [12 -14] was an Ag 2 O-like oxide adlayer with a stoichiometry of Ag 1:83 O. Through some of the first ab initio (T; p) phase diagrams, the Ag 1:83 O overlayer was predicted to be the thermodynamically most stable phase of those identified at ambient oxygen pressures, and was then credited with being of importance to one of the top ten industrial catalytic processes, namely, the Ag catalyzed partial oxidation of ethene to ethene epoxide [12 -15].However, this interpretation and the correctness of the Ag 1:83 O model was challenged recently by a literature review supplemented with a new DFT study [6] and by a surface x-ray diffraction investigation [16].
Here we show through a combination of STM and DFT that the existing model for the p4 4 overlayer is indeed incorrect and propose an alternative comprised of Ag 6 motifs, which bears little or no resemblance to its predecessor.Of broader significance, however, is our characterization of a second phase, a c3 5 3 p rect overlayer, built from the same Ag 6 motifs as the p4 4, and the realization that the O=Ag111 system is much more complex than hitherto anticipated.With DFT a wealth of structures, all with comparable energetic stability, are identified, and with STM we see that the oxide overlayer(s) which form depend sensitively upon the preparation conditions of temperature, pressure, and oxidant.More often than not, the substrate is shrouded in a coexistence of many phases rather than a single well-formed p4 4 overlayer.A clear implication of the present results is that at the temperatures and pressures characteristic of industrial epoxidation catalysis this Ag surface is likely to comprise a complex coexistence or mix of oxide overlayers.
Figure 1(a) shows a typical image obtained with our home-built Aarhus STM [17] after oxidizing the Ag(111) surface by exposing it to atomic oxygen [18].Although there are well-ordered domains with p4 4 symmetry on the surface, they coexist with equally well-ordered domains of different symmetry.Apart from a p4 5 3 p rect phase discussed already by Carlisle et al. [11], there is a c3 5 3 p rect phase identified here for the first time [19].Notably, the bright features exhibited by this and the p4 4 phase are exceedingly similar.Both the line scan taken along the dashed line in Fig. 1(a) and the superposition of peaks from each phase, left and right insets, respectively, show that the apparent heights and diameters of the bright features match extremely well.This suggests that the two phases share a common structural element or ''building block,'' with the difference between the phases apparently being how these building blocks are arranged.
The characteristic honeycomb arrangement of the bright features in the p4 4 phase, apparent in Fig. 1(a), was also observed in the original STM study by Carlisle et al. [10].Interpreting each bright feature as a single Ag adatom led to the Ag 1:83 O model shown in Fig. 1(b), which finds its roots in the Ag 2 O epitaxial overlayer models originally proposed by Rovida et al. [9] and Campbell [20].However, higher resolution images of the p4 4 obtained with the current STM setup, and at a lower bias than the image in (a), contradict this interpretation.An example of one such higher resolution image of the p4 4 phase is shown in Fig. 1(c).Clearly the p4 4 looks significantly different from the region of p4 4 in (a), and its appear-ance in (c) indicates that each of the bright features in (a) contains a detailed substructure.Indeed, the somewhat triangular appearance in (c), with clearly separable corner atoms and a bright protrusion in the middle of each triangle, is clear evidence that the prevailing Ag 1:83 O model is incorrect.As shown in Fig. 1(e), we obtain an even better resolved image of the related c3 5 3 p rect phase.Apart from the already identified corner atoms, which we anticipate are Ag atoms since O atoms tend to image as depressions when adsorbed on Ag(111) [11], this image suggests that the bright protrusion in the middle of the triangles is composed of another three Ag atoms.This is also confirmed by the image in the inset of Fig. 1(e), where an even higher resolution could be obtained.We therefore arrive at the conclusion that the common building block of the p4 4 and the c3 5 3 p rect phases is a triangular arrangement of six Ag atoms, not just a single Ag atom as in the Ag 1:83 O model.This provides the foundation of the alternative ''Ag 6 model'' that we now propose for the p4 4 phase, in which there are two six-atom Ag triangles facing each other as shown in Fig. 1(d).
Taking this embryonic model provided by STM, we turn to DFT to determine if a stable p4 4 model consistent with the new STM data can be identified.And if so, how many oxygen atoms does it contain and what is the registry of this oxide with the substrate?To this end we performed an extensive series of DFT calculations [21].The p4 4 model that emerges from these calculations is shown in Fig. 2(a).In addition to the 12 Ag atoms of the two Ag 6 triangles, there are six O atoms, two located in each ''trough'' between the Ag 6 triangles.The 12 overlayer Ag atoms are located approximately above the threefold sites of the underlying Ag(111) substrate, six close to hcp sites and six close to fcc sites [22].While the stoichiometry of this new model is formally Ag 2 O, it does not resemble the former Ag 2 O-derived models, of which the Ag 1:83 O model was one.These models resemble the structure of bulk Ag 2 O, being comprised of a trilayer of Ag 2 O cut along the (111) plane and epitaxed upon Ag(111) [9,20].In contrast, the new Ag 6 model is not related to the bulk oxide.Instead it is better understood in terms of an arrangement of O atoms embedded between separate six-atom .Thus, whether it remains appropriate to describe the overlayer models proposed here as surface oxides is somewhat unclear.However, since the physical differences between overlayers variously described as ''surface oxides,'' ''oxygen adsorbate phases,'' or ''oxygen actuated reconstructions'' are often difficult, if not impossible, to define, this issue is little more than one of semantics.This is particularly true for the models proposed here which exhibit elements characteristic of a local oxidic structure and of an adsorbate induced metal-surface reconstruction.
Within the current computational setup, the new p4 4 and the c3 5 3 p rect models have comparable energetic stability.Furthermore, they are both more stable than the prevailing Ag 1:83 O model for the p4 4, by circa 0:09 eV=O and circa 0:05 eV=O, respectively.Additional support for the models proposed here comes from the Tersoff-Hamann STM simulations shown in Figs.2(c) and 2(d) [23].Clearly the agreement with the experimental images is good.In fact, by slightly modifying the tip height and bias conditions in the STM simulations we are able to reproduce not only the triangular appearance of the building blocks shown in Figs.2(c) and 2(d), but also the more spherical appearance of the bright features observed in Fig. 1(a) [23].Furthermore, we verified that the simulated O 1s x-ray photoemission spectrum for the p4 4 agrees with the published experimental data [20,24].
While the specific Ag 6 models proposed here for the p4 4 and c3 5 3 p rect phases are consistent with the present STM and DFT data, the DFT calculations provide additional, more general insight that broadens the discussion to the relevance of these phases to oxidation catalysis.In the search for structural models with DFT, we identified a wealth of structures of similar energetic stability, some of which were already discussed in Ref. [6].
Considering the uncertainty (to a lesser extent) due to the finite basis set and (to a larger extent) due to the approximate exchange-correlation functional, it is not possible to give preference to one or the other of these models.Hunting for ''the most stable'' structure may not even be particularly useful, especially when taking the vibrational and configurational free energy contributions at the elevated temperatures of epoxidation catalysis into account.
In fact, within the current computational setup the energy difference between the new p4 4 model and some of those discussed in Ref. [6] is less than k B T per oxygen adatom.In light of these results, this system appears much more complex than previously recognized, and at the high temperatures of oxidation catalysis there is no reason why only the ordered p4 4 surface oxide should prevail.Indeed, from the present STM experiments it is clear that the overlayer which forms depends sensitively on the precise preparation conditions of temperature, pressure, and oxidant (O 2 , atomic oxygen, or NO 2 ).For any of the employed oxidants, conditions can be found under which a well-ordered p4 4 overlayer covers the entire surface.However, it was much more common to observe a coexistence of several structures such as those depicted in Fig. 1(a) [25].This is, in particular, valid for O 2 , which is a more common oxidant than atomic oxygen or NO 2 and which is used in industrial Ag catalysis, too.In summary, our combined STM and DFT study shows that the presently accepted structural model for the renowned p4 4 overlayer phase of O on Ag(111) [10] is incorrect.We propose an alternative Ag 6 model that has an O coverage within the experimentally estimated range [6,20], according to DFT is more stable than its predecessor, and yields simulated STM images that agree with experiment.However, before the case can be closed on this ''evolving oxide structure'' [26], further experimental support will be needed for a full and final structure determination.Aside from the atomic-scale characterization of this complex reconstructed surface, it is intriguing how sensitively the O overlayers that form depend on the precise preparation conditions of temperature, pressure, and oxidant.While previous studies focused almost exclusively on the p4 4 phase, we find here that the Ag(111) substrate is almost always cloaked in a complex coexistence of surface structures, among which the p4 4 and the novel c3

FIG. 1 (
FIG. 1 (color online).STM images and partial structural models of the p4 4 and c3 5 3 p rect overlayers.(a) A 200 A 2 patch of Ag(111) covered by the p4 4 and c3 5 3 p rect overlayers with inclusions of isolated p4 5 3 p rect units (box), measured at ÿ0:51 nA and ÿ131:5 mV.The inset displays a line scan along the dashed line (left) and the superposition of a p4 4 (green or light gray) with a c3 5 3 p rect (black) feature (right).(b) The Ag 1:83 O model for the p4 4 proposed by Carlisle et al. [10].Solid (open) large gray balls represent the overlayer (substrate) Ag atoms, small (red) balls the oxygen atoms.(c),(e) STM images of 35 A 2 patches of the p4 4 and c3 5 3 p rect overlayers, respectively.(c) is at ÿ0:42 nA and ÿ21:7 mV, and (e) is at ÿ0:40 nA and ÿ34:2 mV.The inset in (e) (dotted lines) displays the six-atom structural element of the c3 5 3 p rect phase measured with a tip state that allowed resolution of the central parts of the proposed Ag 6 triangles (ÿ0:42 nA, ÿ21:7 mV).(d),(f) The distribution of the overlayer Ag atoms in the p4 4 and c3 5 3 p rect phases, respectively, proposed here.

5 3 pFIG. 2 (
FIG. 2 (color online).Structural models and Tersoff-Hamann simulated STM images for (a),(c) the p4 4 and (b),(d) the c3 5 3 p rect phase.The dark (light) gray balls are Ag atoms in the overlayer (substrate), and the small dark (red) balls are O atoms.In all four images a circa 30 A 2 region of the surface is displayed.