1: \documentstyle[aps,prb,floats,epsf,epsfig]{revtex}
2: \def\btt#1{{\tt$\backslash$#1}}
3: \def\baselinestrech{1.4}
4:
5: \unitlength1cm
6:
7: \begin{document}
8: \input epsf
9: \draft
10:
11: \renewcommand{\floatpagefraction}{1.00}
12: \renewcommand{\topfraction}{1.00}
13: \renewcommand{\textfraction}{0.00}
14: \renewcommand{\bottomfraction}{1.00}
15:
16: \twocolumn[\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse%
17: \endcsname
18: \tightenlines
19: \draft
20:
21: \title {{\rm\small\hfill to appear in Phys. Rev. B}\\
22: Stability of sub-surface oxygen at Rh(111)}
23:
24: \author{M.Veronica Ganduglia-Pirovano \cite{humboldt}, Karsten Reuter, and Matthias Scheffler}
25:
26: \address{Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg
27: 4-6, D-14195 Berlin, Germany}
28:
29: \date{\today}
30: \maketitle
31:
32: \begin{abstract}
33: Using density-functional theory (DFT) we investigate the incorporation of
34: oxygen directly below the Rh(111) surface. We show that oxygen incorporation
35: will only commence after {\it nearly} completion of a dense O adlayer
36: ($\theta_{\rm tot}\approx 1.0$ monolayer) with O in the fcc on-surface sites.
37: The experimentally suggested octahedral sub-surface site occupancy, inducing
38: a site-switch of the on-surface species from fcc to hcp sites, is indeed
39: found to be a rather low energy structure. Our results indicate that at even
40: higher coverages oxygen incorporation is followed by oxygen agglomeration in
41: two-dimensional sub-surface islands directly below the first metal layer.
42: Inside these islands, the metastable hcp/octahedral (on-surface/sub-surface)
43: site combination will undergo a barrierless displacement, introducing a
44: stacking fault of the first metal layer with respect to the underlying
45: substrate and leading to a stable fcc/tetrahedral site occupation. We suggest
46: that these elementary steps, namely, oxygen incorporation, aggregation into
47: sub-surface islands and destabilization of the metal surface may be more
48: general and precede the formation of a surface oxide at close-packed late
49: transition metal surfaces.
50: \end{abstract}
51: \pacs{PACS numbers: 81.65.Mq, 68.43.Bc, 82.65.My}
52: ]
53:
54: \narrowtext
55: \vskip2pc
56:
57: \section{Introduction}
58: Exposing a metal surface to oxygen can result in simple adsorbate covered
59: surfaces, O sub-surface penetration or oxide formation, depending markedly on
60: the oxygen partial pressure, substrate temperature, the crystallographic
61: orientation of the surface, and the time of exposure. As the chemical activity
62: of the metal surface in these different states can vary significantly, knowledge
63: of the interaction of oxygen with metal surfaces is critical for understanding
64: technologically important processes like e.g. oxidation catalysis. The formation
65: of surface oxides on metal surfaces can in this respect both be beneficial, as
66: well as detrimental. In particular, for applications such as the CO oxidation on
67: transition metal based catalysts, oxide formation at the catalyst's surface in
68: the reactive environment was primarily viewed as leading to an inactive surface
69: oxide poisoning the catalytic reaction, as e.g. in the case of Rh
70: surfaces,\cite{oh83,kellogg85,peden88} while recent experiments showed that the
71: high catalytic activity of the Ru(0001) surface with respect to the CO oxidation
72: relates in fact to the existence of RuO${}_2$(110) oxide patches that form under
73: oxidizing conditions. \cite{boettcher97,over00,kim01} Thus, apparently,
74: unreactive surface oxides are more likely to form on Rh than on Ru surfaces.
75: These findings call for an atomistic understanding of the oxidation of these
76: transition-metal surfaces and of the structure of their oxide surfaces.
77:
78: There is general agreement that the mechanistic steps leading to oxide formation
79: involve dissociative oxygen chemisorption on the metal surface, lattice
80: penetration of atomic oxygen, and the crystallization and growth of the
81: stoichiometric oxide phases. The sequence of these events may be complex, and
82: rather than successively, they may occur simultaneously, depending on the
83: temperature and pressure. The first of these steps, i.e. the chemisorption
84: of oxygen on Rh single crystal surfaces, has been the subject of intensive
85: experimental and theoretical research,
86: \cite{comelli98,loffreda98,pirovano99,pirovano01} but only few experimental
87: studies have addressed the ensuing formation of sub-surface oxygen,
88: \cite{thiel79,rebholz92,peterlinz95,gibson95,janssen96,wider99,gibson99}.
89: In one of these studies it could recently be
90: shown \cite{wider99} that exposure of the Rh(111) surface to O$_2$ at
91: moderatly elevated temperatures ($\sim 470$ K) leads to the formation of
92: sub-surface oxygen species, and there is evidence that at least $\sim 0.9$
93: monolayer (ML) of O is adsorbed on the surface before sub-surface sites are
94: occupied. Interestingly, the analysis of the X-ray photoelectron diffraction
95: (XPD) data of the corresponding study suggested that the sub-surface species
96: ($\sim 0.1$ ML) occupies octahedral sites between the first and second Rh
97: outermost layers, which lie just underneath the fcc on-surface adsorption
98: sites, while the neighboring on-surface oxygen has {\it switched} from
99: its normal fcc to the hcp sites.
100:
101: On theory side, {\em ab initio} studies of O sub-surface species on
102: transition metal surfaces are even more scarce
103: \cite{kiejna00,reuter02a,reuter02b} and as yet lacking for Rh surfaces.
104: The present theoretical study therefore specifically examines
105: the incorporation of O into the close-packed Rh(111) surface using
106: density-functional theory. Extending preceding work focusing on ordered
107: oxygen phases {\em on} this surface \cite{pirovano99,pirovano01} we now
108: discuss the stability of oxygen {\em below} the surface, and address
109: questions concerning the minimum oxygen coverage at the onset of oxygen
110: penetration (section III.A), the stability of various available sub-surface
111: sites (section III.B), as well as the effect of a continued oxygen
112: incorporation with increasing coverage up to 2.0 ML (section III.C). The
113: present results for sub-surface oxygen on Rh(111) are discussed and compared
114: with corresponding results for the Ru(0001) surface \cite{reuter02a,reuter02b}
115: as a first step towards the understanding of the elementary steps governing
116: the onset of surface oxide formation. Questions remain as to the nature of
117: the distinct chemical activity of oxidized Ru(0001) and Rh(111) surfaces.
118:
119:
120: \section{Calculational details}\label{sectionII}
121:
122: All calculations have been performed using
123: density-functional theory (DFT) and the generalized
124: gradient approximation (GGA) of Perdew {\em et al.}
125: \cite{perdew92} for the exchange-correlation functional
126: as implemented in the full-potential linear augmented
127: plane wave method (FP-LAPW) \cite{blaha99,kohler96,petersen00}.
128: The Rh(111) surface is modeled in the supercell approach,
129: employing a 7-layer (111) Rh slab with a vacuum region
130: corresponding to 6 interlayer spacings ($\approx 13$\,{\AA}).
131: Oxygen atoms are adsorbed on both sides of the slab. Their
132: positions as well as those of all Rh atoms in the two
133: outermost substrate layers are allowed to relax while
134: the central three layers of the slab are fixed in their
135: calculated bulk positions. In a preceding publication
136: we have detailed the geometrical properties and stability
137: of ordered adlayers of O on Rh(111) \cite{pirovano99}.
138: Exactly the same calculational setup is used here, such
139: that the results of our prior study can be used for
140: comparison. Further technical details of the calculations
141: can correspondingly be found in Ref. \onlinecite{pirovano99}.
142:
143: We address the stability of O/Rh(111) structures with respect
144: to adsorption of $\rm O_2$ by calculating the average
145: adsorption energy per O adatom,
146:
147: \begin{equation}
148: \label{eq1}
149: E_{\rm ad}(\theta_{\rm tot}) = E_{\rm b}(\theta_{\rm tot}) - D/2\ ,
150: \end{equation}
151:
152: \noindent
153: where $D$ is the dissociation energy of the $\rm O_2$ molecule
154: and $E_{\rm b}(\theta_{\rm tot})$ the average binding energy
155: per oxygen atom as a function of the total coverage (i.e., on- + sub-surface)
156: $\theta_{\rm tot}$. In turn, $E_{\rm b}(\theta_{\rm tot})$ is defined as
157:
158: \begin{eqnarray}
159: \label{eq2}
160: E_{\rm b}(\theta_{\rm tot}) &=& - \frac{1}{N_{\rm tot}}
161: [\; E_{\rm O/Rh(111)}^{\rm slab}(\theta_{\rm tot}) - \\ \nonumber
162: &&(E_{\rm Rh(111)}^{\rm slab} + N_{\rm tot} E_{\rm O}^{\rm atom}) \;]\ ,
163: \end{eqnarray}
164:
165: \noindent
166: where $N_{\rm tot}$ is the total number of oxygen atoms in the unit-cell, and
167: $E_{\rm O/Rh(111)}^{\rm slab}$, $E_{\rm Rh(111)}^{\rm slab}$, and $E_{\rm
168: O}^{\rm atom}$ are the total energies of the O/Rh(111) adsorbate system, of the
169: clean Rh(111) surface, and of the free oxygen atom, respectively. A positive
170: value of the average adsorption energy indicates that the dissociative
171: adsorption of O$_2$ is exothermic. That is, the binding energy per O adatom on
172: Rh(111) is larger than that which the O atoms have in $\rm O_2$(gas), i.e.,
173: $D/2$. The total energies of the adsorbate system and clean surface are
174: calculated using the same supercell. Details on the calculations of the isolated
175: O atom and the free $\rm O_2$ molecule are given in Ref.
176: \onlinecite{pirovano99}.
177:
178: To test the accuracy of the calculated binding energies, $E_{\rm b}(\theta_{\rm
179: tot})$, on numerical approximations due to the finite FP-LAPW basis set and the
180: finite slab and vacuum thicknesses in the supercell approach, selected
181: calculations were repeated with higher accuracy. The self-consistent
182: calculations of $E_{\rm b}$ were routinely conducted for a 7-layer slab, a 16 Ry
183: plane-wave cutoff in the interstitial region, and a $(12\times 12\times 1)$
184: Monkhorst-Pack grid for the $(1\times 1)$ unit-cell with 19 {\bf k}-points in
185: the irreducible wedge \cite{pirovano99}. Changing to denser {\bf k}-meshes up to
186: a $(18\times 18\times 1)$ grid with 37 {\bf k}-points in the irreducible wedge
187: resulted in negligible variations of $E_{\rm b}$ within $\sim 10$ meV/O atom.
188: Similarly, extending the Rh(111) slab from 7 to 9 and 11 layers led only to
189: changes in the calculated binding energies up to $\sim 10$ meV/O atom. The only
190: really notable effect on the computed $E_{\rm b}$ is caused by variations of the
191: finite plane-wave cutoff in the interstitial region: Increasing this cutoff from
192: 16 Ry to 24 Ry decreases the binding energies by $\approx 100$ meV/O atom. Yet,
193: this decrease can be largely attributed to an improved description of both the
194: free O atom and the chemisorbed species and thus affects all structures that
195: contain the same amount of O alike. Consequently, the calculation of
196: relative stabilities of different phases, i.e. the difference between two
197: binding energies, has a significantly smaller error than this variation in the
198: absolute value of $E_{\rm b}$. Combining all these tests, we give a conservative
199: estimate of this latter error of $\pm 30$ meV/O atom, which in turn does not
200: affect any of the conclusions made in this work.
201:
202:
203: \section{Stability of sub-surface O}
204:
205: \subsection{Minimum coverage for O incorporation}
206:
207: The initial oxidation of the Rh(111) surface proceeds via the dissociative
208: chemisorption of oxygen in the threefold fcc hollow sites of the basal surface.
209: \cite{comelli98,loffreda98,pirovano99} O${}_2$ exposure under UHV conditions
210: leads to two well ordered superstructures, namely a $(2 \times 2)$-O and a
211: $(2\times 1)$-O phase at coverages of 0.25\,ML and 0.5\,ML, respectively. The
212: latter coverage was for a long time believed to correspond to a saturated
213: surface. \cite{comelli98} Yet, after theoretical prediction,
214: \cite{loffreda98,pirovano99} a $(1\times 1)$-O phase
215: ($\theta_{\rm tot} = 1.0$\,ML) could be stabilized experimentally
216: under UHV conditions using atomic O as oxidant, thus demonstrating that the
217: apparent saturation results from a kinetic energy barrier to O${}_2$
218: dissociation. \cite{gibson99} This barrier can also be overcome by
219: enhanced O${}_2$ dosage at moderately elevated temperatures
220: \cite{peterlinz95,wider99,castner80} or by exposure to more oxidizing carrier
221: gases like e.g. NO${}_2$. \cite{peterlinz95,gibson99} In all cases, sub-surface
222: O incorporation or bulk diffusion was reported for higher exposures at
223: temperatures above $\sim 400$ K, and as already mentioned a recent experimental
224: study brought evidence that at least $\sim 0.9$\,ML of oxygen is adsorbed on the
225: surface, before sub-surface sites become occupied. \cite{wider99}
226:
227: To address this initial incorporation theoretically we study the stability and
228: properties of sub-surface oxygen species as function of the total coverage,
229: $\theta_{\rm tot}$ ($0.25\leq\theta_{\rm tot}\leq 1.0$ ML). We employ
230: $(2 \times 2)$ unit-cells and calculate the average adsorption energy
231: of fully relaxed structures containing from
232: $N_{\rm tot} = 1$ ($\theta_{\rm tot} = 0.25$\,ML) up to
233: $N_{\rm tot} = 4$ ($\theta_{\rm tot} = 1$\,ML) oxygen atoms (see Eq.
234: (\ref{eq2})). The occupation of sub-surface sites will commence, when a
235: structure with one of these O-atoms located below the surface becomes
236: energetically more favorable than the most stable one with all oxygens on
237: the surface at the same total coverage. That is, we compare the stability of
238: $(2 \times 2)-((N_{\rm tot}-1) {\rm O}_{\rm on} + {\rm O}_{\rm sub})$/Rh(111)
239: mixed phases relative to the $(2 \times 2)-(N_{\rm tot} {\rm O}_{\rm on})$/Rh(111)
240: structure with O${}_{\rm on}$ atoms in fcc hollow sites for
241: $N_{\rm tot} = 1,\ldots 4$.
242:
243: Between the first and second outermost metal layers, there are three different
244: high-symmetry interstitial sites available for O incorporation. The octahedral
245: site (henceforth octa) lies just underneath the fcc on-surface site, and one
246: tetrahedral site (tetra-I) lies below the hcp on-surface site. A second
247: tetrahedral site (tetra-II) is located directly below a first layer metal
248: atom. Considering both three-fold hollow sites (fcc and hcp) as possible
249: adsorption sites on the surface, leads then to a many-fold of possible
250: structural combinations of how to place the $N_{\rm tot}$
251: oxygen atoms into the unit-cell, in particular as at most coverages several
252: symmetry inequivalent possibilities for the same on- and sub-surface site
253: combination exist.
254:
255: \begin{figure}
256: \begin{center}
257: \psfig{file=pics/fig1.ps,width=1.0\columnwidth,angle=0}
258: \end{center}
259: \caption{Top view of all possible on-surface/sub-surface
260: site combinations at $\theta_{\rm tot} = 1.0$\,ML with one oxygen atom located
261: below the surface (see text for the explanation of the different sites). Rh =
262: big spheres (white = surface layer, grey = 2nd layer), O = small spheres (black
263: = on-surface, grey = sub-surface). Oxygens in tetra-II sites below the first
264: layer Rh atoms are invisible in this plot and are schematically shown as small
265: white circles. Symmetry inequivalent occupation of the same kind of on- and
266: sub-surface sites are denoted with (a) and (b), respectively.}
267: \label{fig1}
268: \end{figure}
269:
270: As will be shown below, the stability of structures involving on- and
271: sub-surface sites becomes only comparable to that of the pure chemisorbed
272: on-surface phase at $\theta_{\rm tot} = 1.0$\,ML. Thus, only at that coverage
273: we will address the complete set of possible structural combinations, shown
274: in Fig. \ref{fig1}, while at the lower coverages we only exemplify the
275: stability trends by computing three likely combinations that will become
276: relevant at a later stage of our discussion. Namely, these are geometries
277: where on-surface oxygen is in fcc sites and the sub-surface oxygen in either
278: the octa or the tetra-I sites, fcc/octa and fcc/tetra-I respectively.
279: Thirdly, we included the experimentally suggested possibility \cite{wider99}
280: that oxygen in octahedral sub-surface sites induces a site-switch of the
281: nearby on-surface oxygens from fcc to hcp, viz hcp/octa. The thus resulting
282: set of considered structures is shown in Fig. \ref{fig2} for
283: $\theta_{\rm tot} = 0.5$\,ML and $\theta_{\rm tot} = 0.75$\,ML. At the
284: latter coverage we also tested the possibility of a simultaneous occupation
285: of hcp and fcc on-surface sites and octahedral sub-surface sites, leading to
286: the fcc hcp/octa structure. Figs. \ref{fig1} and \ref{fig2} also show the
287: considered symmetry inequivalent structures with the same kind of on- and
288: sub-surface sites occupation, e.g. hcp/octa (a) and hcp/octa (b).
289:
290: \begin{figure}
291: \begin{center}
292: \psfig{file=pics/fig2.ps,width=1.0\columnwidth,angle=0}
293: \end{center}
294: \caption{Top view of selected on-surface/sub-surface
295: site combinations at $\theta_{\rm tot} = 0.5$\,ML and
296: $\theta_{\rm tot} = 0.75$\,ML with one oxygen atom located
297: below the surface (see text). Rh = big spheres
298: (white = surface layer, grey = 2nd layer), O = small spheres
299: (black = on-surface, grey = sub-surface).}
300: \label{fig2}
301: \end{figure}
302:
303: \begin{table}
304: \caption{\label{tableI}
305: Average adsorption energies (in eV/O atom) of Rh(111) geometries
306: containing O in
307: on-surface and/or sub-surface sites. The label on/sub indicates the site type
308: (fcc,hcp for on-surface; octa/tetra-I/tetra-II for sub-surface, see text).
309: Calculated values for the most stable on-surface adlayer with O occupying fcc
310: sites (fcc/---) are from Ref. {\protect\onlinecite{pirovano99}}.
311: For $\theta_{\rm tot} \leq 1.25$\,ML, 0.25\,ML of O is contained
312: below the surface, for $\theta_{\rm tot} = 2.0$\,ML, 1.0\,ML of
313: O is contained below the surface.}
314: \begin{tabular}{c|rccccc}
315: Sites &\multicolumn{6}{c}{$\theta_{\rm tot}$} \\
316: on/sub & 0.25& 0.50 & 0.75 & 1.00 & 1.25 & 2.00 \\ \hline
317: fcc/--- & 2.24 & 1.95& 1.66 & 1.40 & & \\
318: ---/octa &$-1.25$ & & & & & \\
319: ---/tetraI &$-0.78$ & & & & & \\
320: fcc/octa & & 0.36 & 0.75 & 0.81 & 0.69 & 0.56 \\
321: fcc hcp/octa & & 0.68 & & & & \\
322: hcp/octa (a) & & 0.81 & 1.02 & 0.92 & 0.81 & 0.89 \\
323: hcp/octa (b) & & 0.34 & 0.68 & 0.74 & & \\
324: fcc/tetraI (a) & & 1.02 & 1.19 & 1.07 & 0.93 & 1.13 \\
325: fcc/tetraI (b) & & & & 0.74 & & \\
326: hcp/tetraI (a) & & & & 0.79 & & \\
327: fcc/tetraII (a)& & & & 0.91 & 0.77 & 0.83 \\
328: fcc/tetraII (b)& & & & 0.91 & & \\
329: hcp/tetraII (a)& & & & 0.80 & 0.68 & 0.86 \\
330: hcp/tetraII (b)& & & & 0.85 & & \\
331: \end{tabular}
332: \end{table}
333:
334: The calculated average adsorption energies of all investigated structures
335: are compiled in Table \ref{tableI}, while Fig. \ref{fig3} additionally
336: visualizes the trends for three selected on-surface/sub-surface combinations
337: as described above. Concentrating first on the lowest tested coverage,
338: $\theta_{\rm tot} = 0.25$\,ML, we find that the occupation of just
339: sub-surface sites without any on-surface oxygen is not even exothermic
340: and by $\approx 3$\,eV/O atom less favorable than adsorption into the
341: on-surface fcc hollow sites. We have recently shown that this is generally
342: the case for the closed-packed late $4d$ transition metal surfaces,
343: which is largely largely due to the significant local expansion of
344: the metal lattice induced by occupation of sub-surface sites.
345: \cite{todorova02} The corresponding cost of distorting the metal lattice
346: and breaking metal bonds renders sub-surface sites initially always less
347: stable than on-surface chemisorption. Yet, upon increasing the
348: on-surface coverage, repulsive interactions between the
349: adsorbates drive the adsorption energy down, cf. Fig. \ref{fig3},
350: until eventually occupation of sub-surface sites might
351: become more favorable compared to a continued filling
352: of on-surface sites.
353:
354: \begin{figure}
355: \begin{center}
356: \psfig{file=pics/fig3.ps,width=0.8\columnwidth,angle=-90}
357: \end{center}
358: \caption{Average adsorption energies (in eV/O atom) of
359: selected Rh(111) geometries containing O in on-surface
360: and/or sub-surface sites, see text and Table \ref{tableI}.}
361: \label{fig3}
362: \end{figure}
363:
364: Turning therefore to higher coverages up to 1.0 ML, where we continuously
365: increase the number of adsorbed on-surface oxygen atoms per unit cell, we
366: find the average adsorption energies of all mixed structures now to be
367: positive, i.e. they should be able to form. Still, in the whole sub-monolayer
368: regime the values of $E_{\rm ad}(\theta_{\rm tot})$ for the mixed phases
369: are significantly lower compared to that of the pure on-surface adsorption
370: at the same total coverage, reflecting the above described fact, that the
371: occupation of sub-surface sites is energetically considerably less favorable.
372: Only at $\theta_{\rm tot} = 1.0$\,ML does the adsorption energy of the
373: selected mixed phases approach that of the pure on-surface fcc phase to
374: within $\approx 0.3$\,eV/O atom (cf. Fig. \ref{fig3}). To make sure that
375: no other structural on-surface/sub-surface combination would be even slightly
376: more stable and then eventually more favorable than the pure on-surface
377: phase, we tested all of the possible structures shown in Fig. \ref{fig1} at
378: this particular coverage. As can be seen from Table \ref{tableI} neither of
379: these configurations leads to a more favorable binding than the pure fcc phase.
380: Therefore, O incorporation into the Rh(111) surface can only occur at just
381: about the completion of the full monolayer coverage on the surface, i.e.
382: $\theta_{\rm tot}\approx 1.0$\,ML.
383:
384: The calculated energy difference of $\approx 0.3$\,eV/O atom between the
385: pure fcc and the mixed phases for coverages $\theta_{\rm tot} \approx 1.0$\,ML
386: indicates that sub-surface O penetration into Rh(111) is an energetically
387: activated process and that a small, but finite concentration of sub-surface
388: oxygen can be expected already at on-surface coverages slightly below 1.0\,ML
389: at elevated temperatures. Wider {\it et al.} have found that extended exposure
390: of the Rh(111) surface to oxygen at 470 K led to the formation of a sub-surface
391: species, though the on-surface coverage remained slightly below 1.0\,ML.
392: \cite{wider99} In comparison, on Ru(0001) we find the energy difference
393: between the pure on-surface and the mixed phases at $\theta_{\rm tot} = 1.0$\,ML
394: to be $\approx 0.8$\,eV/ O atom, \cite{ruthenium} that is about a factor
395: of three larger than for Rh(111). Correspondingly, oxygen penetration has at
396: the prior surface only been reported to occur {\em after} the $(1 \times 1)$-O
397: phase on the surface has been completed. \cite{stampfl96,boettcher99}
398:
399:
400: \subsection{Site preferences and site-switch}
401:
402: Trying to understand the oxygen site preferences in the various
403: on-surface/sub-surface phases at coverages up to 1.0 ML, we note
404: first that the low stability of a number of structural combinations
405: can be understood in terms of electrostatic repulsion between
406: the oxygens, i.e. whenever the electronegative oxygens come too
407: close to each other. This holds for mixed fcc hcp combinations
408: (cf. Table \ref{tableI}), as well as for the less stable of the
409: two symmetry inequivalent possibilities of the same
410: on-surface/sub-surface site combination (rendering e.g. the hcp/octa
411: (b) and fcc/tetra (b) geometries unfavorable, cf. Figs. \ref{fig1}
412: and \ref{fig2} and Table \ref{tableI}).
413:
414: Moreover, we find the stability of all other combinations at
415: $\theta_{\rm tot}=1.0$ ML to be rather similar, i.e. within
416: $\approx 0.2$\,eV/O atom, cf. Table I. This suggests that at
417: elevated temperatures oxygen incorporation could start initially
418: in all available sites, and that kinetic factors (like penetration
419: barriers) more than energetic factors determine which of the
420: possible metastable sites get populated first in an experiment
421: with controlled oxygen dosage. As noted in the introduction,
422: a recent analysis of XPD data indicated the presence of a
423: sub-surface oxygen species ($\sim 0.1$\,ML) in octahedral sites
424: between the first and second metal layer at an almost
425: oxygen-covered ($\sim 0.9$\,ML) Rh(111) surface after $10^5$
426: Langmuir O${}_2$ exposure at 470 K. The analysis of the data
427: yields to the suggested simultaneous occupation of neighboring
428: on-surface hcp sites, which is different from the otherwise
429: preferred fcc sites on Rh(111) before oxygen incorporation.
430: \cite{wider99}
431:
432: \begin{figure}
433: \begin{center}
434: \psfig{file=pics/fig4a.ps,width=0.6\columnwidth,angle=-90}
435: \end{center}
436: \caption{Calculated work function change for the
437: fcc/octa and hcp/octa (a) structures at $\theta_{\rm tot}= 1.0$\,ML
438: (see Fig.~\protect\ref{fig1}). Calculated values for the most
439: stable adlayers with O occupying only on-surface fcc sites
440: are from Ref.~\protect\onlinecite{pirovano99}.}
441: \label{fig4}
442: \end{figure}
443:
444: \begin{figure}
445: \begin{center}
446: \psfig{file=pics/fig4b.ps,width=0.3\columnwidth,angle=-90}
447: \end{center}
448: \caption{
449: Calculated initial-state on-surface O $1s$ core-level shifts for
450: the fcc/octa and hcp/octa (a) phases at $\theta_{\rm tot}=1.0$
451: relative to the O $1s$ level for the adatoms of the O-$(1\times 1)$/Rh(111)
452: structure at $\theta_{\rm on}=1.0$ ML, with O occupying on-surface
453: fcc sites (fcc/---). The label on/sub indicates the on- and
454: sub-surface adsorption sites, respectively. There are two symmetry
455: inequivalent O adatoms in the hcp/octa phase, which is why two levels
456: are shown for this structure.}
457: \label{fig4b}
458: \end{figure}
459:
460: This intriguing finding is in line with our calculations, which
461: indeed indicate, that {\em if} sub-surface oxygen was to occupy
462: octahedral sites, then the site-switch combination hcp/octa
463: would be considerably more favorable compared to the fcc/octa
464: combination with the on-surface oxygens in their normal adsorption
465: sites, cf. Fig. \ref{fig3}.
466:
467: In seeking for a qualitative explanation for the stability of the
468: hcp on-surface sites upon occupation of octahedral sites,
469: and consistent with the discussion on the preferred fcc site
470: adsorption begun in earlier papers, \cite{pirovano99,pirovano01}
471: we examine the calculated changes in the work-function, $\Delta\Phi$,
472: as well as the difference in O $1s$ binding energy of on-surface
473: O-atoms, $\Delta{\rm O}_{1s}$, of the mixed fcc/octa and hcp/octa
474: phases at $\theta_{\rm tot}=1.0$ ML relative to the values for the
475: stable $(1\times 1)$-O/Rh(111) structure, with O just occupying fcc
476: sites. The calculated $\Delta\Phi$ and $\Delta{\rm O}_{1s}$ values
477: are shown in Figs. \ref{fig4} and \ref{fig4b}, respectively. Initially,
478: the work function rises as a function of the coverage, while O remains
479: in the surface fcc sites; at $\theta_{\rm tot}\approx 0.75$ ML it
480: reaches a saturation value. Such an increase in the work-function
481: upon O chemisorption reflects the high electronegativity of the adspecies
482: that results in an induced inward dipole moment, i.e., with the negative
483: charge at the vacuum side of the surface. The saturation at higher
484: coverages is then a consequence of the dipole-dipole interaction giving
485: rise to a depolarization with decreasing O-O distance.
486:
487: The initial incorporation of oxygen in the fcc/octa and hcp/octa phases
488: leads to a decrease in the work function compared to the saturation value
489: at 1.0 ML with O just occupying on-surface fcc sites. This decrease is
490: considerably smaller for the hcp/octa structure, cf. Fig. \ref{fig4};
491: where it should also be noticed that the average bond length,
492: O$_{\rm on}$-Rh$_{1}$ and interlayer spacing, $\bar d_{\rm O_{\rm on}-Rh_1}$,
493: between chemisorbed O and Rh atoms at the surface are practically
494: independent of the occupation of the sites at a given coverage, cf.
495: Tables \ref{tableII} and \ref{tableIII}. Moreover, the calculated
496: initial-state O $1s$ shifts schematically shown in Fig. \ref{fig4b},
497: indicate that the O $1s$ levels of adatoms in the hcp/octa geometry
498: are $\sim 0.5$ eV {\it less} bound compared to corresponding levels
499: of oxygen atoms in the fcc/octa geometry. This reflects a larger
500: electrostatic repulsion at the hcp sites, where a somewhat more
501: negatively charged O would be adsorbed, which in turn correlates with
502: the smaller decrease in the work function for the preferred site-switch
503: phase.
504:
505: On comparing the values of $\Delta\Phi$ and $\Delta{\rm O}_{1s}$ for the
506: hcp/octa and fcc/octa phases with those for the pure fcc adsorption in
507: Figs. \ref{fig4} and \ref{fig4b} it can thus be suggested that a similarly
508: charged O would sit in hcp sites (rather than in fcc sites) upon
509: octahedral sub-surface occupation compared to the oxygens adsorbed in fcc
510: sites before O penetration. Hence, similarly to the pure on-surface fcc
511: adsorption, \cite{pirovano99,pirovano01} we suggest that a stronger
512: {\it ionic} bonding favors the hcp sites in the site-switch phases at the
513: onset of O penetration ($\theta_{\rm tot}\approx 1.0$ ML).
514:
515: Although we can thus rationalize the higher stability of the hcp/octa
516: site-switch phase suggested by Wider {\em et al.} \cite{wider99} compared
517: to the fcc/octa phase, we nevertheless stress that the occupation of
518: octahedral sub-surface sites is according to our calculations only of a
519: metastable character, as the fcc/tetra-I structural combination
520: is consistently energetically more favorable over the complete
521: tested coverage range, cf. Fig. \ref{fig3}.
522:
523:
524: \subsection{Continued oxidation: island formation and trilayer shift}
525:
526: We investigate the continued oxidation after sub-surface O
527: has been initially incorporated in the coverage region
528: $1.25\leq\theta_{\rm tot}\leq 2.0$ ML by calculating the
529: average adsorption energy at both 1.25\,ML ($\theta_{\rm sub}=0.25$ ML)
530: and 2.0\,ML ($\theta_{\rm sub}=1.0$ ML) coverages. Again,
531: we find that the site combinations hcp/octa and fcc/tetra-I
532: give the highest adsorption energies, while the occupation
533: of tetra-II sites leads to energetically slightly less favorable
534: phases, cf. Table I. In the following we will restrict our discussion
535: to the two most stable phases, i.e. hcp/octa and fcc/tetra-I.
536:
537: \begin{table}[ht]
538: \caption{\label{tableII}
539: Calculated structural parameters in {\AA} for the hcp/octa and
540: fcc/tetra-I phases at $\theta_{\rm tot}=1.25$ ML. O atoms and all
541: Rh atoms in the two outermost layers were allowed to relax. For
542: the interlayer distances, the center of mass of each layer is
543: used. Numbers in parenthesis correspond to bulk values, which
544: were fixed. O$_{\rm on,sub}$-Rh$_{1,2}$ indicate (averaged) bond
545: lengths to first and second layer Rh atoms, and $\Delta z_{{\rm Rh}_1,2}$
546: and $\Delta z_{{\rm O}_{\rm on}}$ the magnitude of the buckling
547: of the outermost Rh layers and of the on-surface O adlayer,
548: respectively. The lateral displacements, radially away from the
549: ideal lattice positions, are denoted by $\Delta r_{{\rm Rh}_{1,2}}$,
550: and $\Delta r_{{\rm O}_{\rm on}}$, respectively.}
551: \begin{tabular}{lccc}
552: & hcp/octa & fcc/tetra-I\\ \hline
553: O$_{\rm on}$-Rh$_{1}$ & 1.94 & 1.95 \\
554: O$_{\rm sub}$-Rh$_{1}$ & 2.15 & 2.00 \\
555: O$_{\rm sub}$-Rh$_{2}$ & 2.21 & 1.94 \\
556: $\bar d_{\rm O_{\rm on}-Rh_1}$ & 1.14 & 1.15 \\
557: $\bar d_{12}$ & 2.79 & 2.78 \\
558: $\bar d_{23}$ & 2.30 & 2.32 \\
559: $\bar d_{34}$ & (2.213) & (2.213) \\
560: $\Delta z_{{\rm Rh}_1}$ & 0.30 & 0.23 \\
561: $\Delta z_{{\rm Rh}_2}$ & 0.26 & 0.30 \\
562: $\Delta z_{{\rm O}_{\rm on}}$ & 0.18 & 0.18 \\
563: $\Delta r_{{\rm Rh}_1}$ & 0.06 & 0.08 \\
564: $\Delta r_{{\rm Rh}_2}$ & 0.03 & 0.00 \\
565: $\Delta r_{{\rm O}_{\rm on}}$ & 0.07 & 0.07 \\
566: \end{tabular}
567: \end{table}
568:
569: \begin{table}[ht]
570: \caption{\label{tableIII}
571: Calculated structural parameters in {\AA} for the hcp/octa,
572: fcc/tetra-I, and fcc/tetra-I$_{\rm fault}$ phases at
573: $\theta_{\rm tot}=2.0$ ML. O atoms and all Rh atoms in the
574: two outermost layers were allowed to relax. Numbers in
575: parenthesis correspond to bulk values, which were fixed.
576: O$_{\rm on,sub}$-Rh$_{1,2}$ indicate bond lengths to first
577: and second layer Rh atoms, and $\bar d_{ij}$ interlayer
578: spacings.}
579: \begin{tabular}{lccc}
580: & hcp/octa & fcc/tetra-I & fcc/tetraI$_{\rm fault}$\\\hline
581: O$_{\rm on}$-Rh$_{1}$ & 1.96 & 1.97 & 1.97 \\
582: O$_{\rm sub}$-Rh$_{1}$ & 2.10 & 2.05 & 2.06 \\
583: O$_{\rm sub}$-Rh$_{2}$ & 2.51 & 2.01 & 2.02 \\
584: $\bar d_{\rm O_{\rm on}-Rh_1}$& 1.19 & 1.19 & 1.19 \\
585: $\bar d_{12}$ & 3.36 & 3.34 & 3.35 \\
586: $\bar d_{23}$ & 2.14 & 2.17 & 2.18 \\
587: $\bar d_{34}$ & (2.213) & (2.213) & (2.213) \\
588: \end{tabular}
589: \end{table}
590:
591: Tables \ref{tableII} ($\theta_{\rm tot} = 1.25$\,ML) and
592: \ref{tableIII} ($\theta_{\rm tot} = 2.0$\,ML) show that
593: the geometrical changes in response to sub-surface penetration
594: of oxygen are significant and it is therefore not accurate to
595: assume that the Rh(111) lattice will remain essentially
596: undisturbed upon the occupation of sub-surface sites.
597: For both hcp/octa and fcc/tetra-I phases at 1.25 ML (0.25 ML
598: below the surface), the mean outermost substrate interlayer
599: spacing, $\bar d_{12}$, expands by about $\sim 22\%$, and
600: $\sim 26\%$, relative to the corresponding value for the
601: bulk-terminated surface, respectively (cf. Table \ref{tableII}).
602: Prior to the present work, we have calculated an increase of
603: $\bar d_{12}$ from 2.21\,{\AA} for the clean surface to 2.37{\AA}
604: for the $(1\times 1)$-O/Rh(111) structure with oxygens in fcc
605: sites, \cite{pirovano99} which means that an additional
606: $\sim 15-19\%$ expansion is induced by the 0.25 ML of sub-surface
607: oxygen.
608:
609: \begin{figure}
610: \begin{center}
611: \psfig{file=pics/fig5a.ps,width=1.0\columnwidth,angle=0}
612: \psfig{file=pics/fig5b.ps,width=1.0\columnwidth,angle=0}
613: \end{center}
614: \caption{Top and side view of the fcc/tetra-I geometry at
615: $\theta_{\rm tot} = 1.25$\,ML. Small and large spheres
616: represent oxygen and Rh atoms respectively, where those
617: lying in the same plane and equivalent under the threefold
618: rotation symmetry have the same color. The arrows indicate
619: the direction of reference for the atomic in-plane displacements
620: $\Delta r_{{\rm Rh}_{1,2}}$, and $\Delta r_{{\rm O}_{\rm on}}$
621: of Table~\protect\ref{tableII}. Distances are in {\AA}.}
622: \label{fig5}
623: \end{figure}
624:
625: For the $(2 \times 2)$-($4\rm O_{\rm on}+\rm O_{\rm sub}$)/Rh(111)
626: phases at 1.25\,ML, symmetry allows for (in part considerable)
627: buckling of the outermost Rh and on-surface O layers, i.e. quite
628: distinct local contractions and expansions together with lateral
629: shifts (cf. Table \ref{tableII}), which can most often be understood
630: as a local expansion of the metal lattice around the occupied
631: sub-surface site. \cite{todorova02} In view of keeping the paper
632: within a limited length, we show only one example. For instance,
633: in the fcc/tetra-I geometry, cf. Fig. \ref{fig5}, the first layer
634: Rh atoms above the occupied sub-surface site, are pulled out of the
635: surface by $\sim 0.23$\,{\AA}, while the neighboring Rh atoms in
636: the second substrate layer are pushed inwards by $\sim 0.30$\,{\AA}.
637: Also, there are local vertical displacements of the O adatoms
638: ($\sim 0.18$\,{\AA}), and considerable lateral shifts radially away
639: from the ideal lattice positions for both substrate and adsorbate
640: atoms.
641:
642: \begin{figure}
643: \begin{center}
644: \psfig{file=pics/fig6.ps,width=0.6\columnwidth,angle=-90}
645: \end{center}
646: \caption{Calculated average adsorption energy during the
647: registry shift of the O${}_{\rm on}$-Rh-O${}_{\rm sub}$
648: trilayer along the $[\bar 2 1 1]$ direction over the Rh(111)
649: substrate. The initial hcp/octa phase ends in a fcc/tetra-I
650: configuration with a fault in the stacking sequence of the
651: surface Rh layer after shifting the trilayer by
652: ${\frac{a_0}{3}}\sqrt{3/2}$ ($a_0=3.83$ {\AA}).}
653: \label{fig7}
654: \end{figure}
655:
656: The expansion of the first Rh layer distance becomes with $\approx 51$\%
657: even more pronounced at $\theta_{\rm tot} = 2.0$\,ML (1\,ML O below the
658: surface), cf. Table \ref{tableIII}. Despite this significant distortion
659: of the metal lattice, we find increased adsorption energies for both
660: the hcp/octa and fcc/tetra-I phases compared to the low sub-surface O
661: coverage at $\theta_{\rm tot} = 1.25$\,ML, cf. Fig. \ref{fig3}. This
662: reflects an attractive interaction between the sub-surface species, which
663: implies that the latter have a tendency to form {\em sub-surface islands}
664: (i.e. nucleate) with a local $(1 \times 1)$ periodicity. For both the hcp/octa and
665: fcc/tetra-I phase, the structure of these islands can be viewed as an
666: O${}_{\rm on}$-Rh-O${}_{\rm sub}$ trilayer on top of a Rh(111) substrate,
667: cf. Fig. \ref{fig6}. Interestingly, the internal geometry of the strongly
668: bound trilayer is almost identical for both phases, cf. Table \ref{tableIII}
669: and Fig. \ref{fig6}; it is only the coordination to the underlying
670: substrate which is different in both cases. In fact, we even find that the
671: higher energy hcp/octa configuration is unstable against a registry shift of
672: this whole trilayer along the $[\bar 2 1 1]$ direction. The calculated
673: average adsorption energy along this {\it sliding} of the trilayer over
674: the Rh(111) surface is shown in Fig. \ref{fig7}. We have fully optimized
675: the structures at the calculated points along the {\em barrierless}
676: displacement, starting from the hcp/octa structure. At the end of the shift
677: by $\frac{a_0}{3}{\sqrt{3/2}}$ ($a_0=3.83$ {\AA}, Rh lattice constant
678: \cite{pirovano99}), the on-surface oxygen atoms are located in fcc and the
679: sub-surface oxygen atoms in tetra-I sites, but the surface Rh layer inside
680: the trilayer does not continue the fcc lattice stacking, but is now located
681: in a stacking fault position. The calculated structural parameters
682: (interlayer spacings and bond lengths) between these two fcc/tetra-I$_{\rm fault}$
683: and fcc/tetra-I geometries have remained virtually unchanged (cf. Table
684: \ref{tableIII}), and the average binding energies differ by only $\sim 3$
685: meV/O atom, i.e. they are degenerate within our calculational uncertainty,
686: which is plausible as the two structures differ in fact only by a 60${}^o$
687: rotation of the trilayer with respect to the underlying substrate. We
688: have similarly compared the geometries and average binding energies of both,
689: fcc/tetraI and fcc/tetraI$_{\rm fault}$ structures at 1.25 ML and also found no
690: significant difference; we find a difference of $\sim 1$ meV/atom in the
691: calculated binding energies.
692:
693: The reason behind the $\sim 0.2$\,eV/O atom ($\sim 0.4$\,eV/unit cell)
694: preference for the fcc/tetra-I
695: structure compared to the hcp/octa phase can be found when analyzing
696: the coupling of the formed trilayer to the underlying Rh(111) substrate.
697: Whereas we consistently find O-Rh bondlengths of $\sim 2.0$\,{\AA}
698: for both on- and sub-surface oxygens in the majority of tested structures,
699: the O${}_{\rm sub}$-Rh${}_2$ bondlength of the sub-surface oxygen to
700: the second layer Rh atom in the hcp/octa phase at $\theta_{\rm tot} = 2.0$\,ML
701: is significantly larger (2.51\,{\AA}). This points to a weak coupling
702: of the trilayer to the Rh(111) substrate in this geometry, which can also
703: be seen in the density plots shown in Fig. \ref{fig6}, where
704: the incorporation of sub-surface O in octahedral sites is found
705: to induce only very small changes on the valence charge of the second
706: layer Rh atoms. These changes are considerably larger for the fcc/tetra-I
707: configuration, suggesting that the energetic preference of this structure
708: is primarily due to an improved coupling of the formed trilayer to the
709: underlying substrate.
710:
711: Hence, even if sub-surface oxygen was initially incorporated into the
712: metastable octahedral sites as suggested by experiment, \cite{wider99}
713: it would due to this instability transform into the fcc/tetra-I
714: configuration upon continued oxygen penetration into the Rh(111) surface.
715: It is interesting to notice that on the Ru(0001) surface \cite{reuter02a,reuter02b}
716: a plausible oxidation pathway has been identified, in which after the
717: formation of such trilayers a phase transformation into the rutile
718: RuO${}_2$(110) structure was obtained for local oxygen coverages exceeding
719: 5\,ML. The local formation of such precursing structures via the
720: agglomeration of sub-surface oxygen atoms below the surface that follow
721: the initial oxygen incorporation, could therefore be a more general
722: phenomenon in the oxidation of transition metal surfaces. Once the local
723: oxygen coverage exceeds a critical value, these precursor configurations
724: will then undergo a structural change and actuate the formation of bulk
725: oxide phases. In this respect we notice that already in the preferred
726: fcc/tetra-I configuration the Rh first layer atoms are sixfold coordinated
727: to oxygens and the oxygen atoms in the tetrahedral site fourfold coordinated
728: to metal atoms, i.e. they exhibit identical local coordinations as in the
729: most stable, corundum-structured Rh${}_2$O${}_3$ bulk oxide.
730:
731:
732: \section{Summary}
733:
734: In conclusion we have presented a density-functional theory
735: study addressing the initial penetration of oxygen into the
736: Rh(111) surface. Due to the large local expansion of the metal
737: lattice induced by the occupation of sub-surface sites, O
738: chemisorption on the surface is initially significantly more
739: favorable. In agreement with recent experimental findings,
740: we therefore find oxygen penetration to occur only after
741: the adsorption of {\it almost} a full monolayer on the surface.
742: A particularly interesting result is that the calculated
743: energy difference of $\sim 0.3$ eV/atom between the pure
744: on-surface and the mixed on/sub-surface phases at
745: $\theta_{\rm tot} = 1.0$\,ML is about a factor of three
746: smaller than that for Ru(0001). Oxygen penetration has at the
747: Ru surface only been reported to occur {\em after} the
748: $(1 \times 1)$-O phase on the surface has been completed.
749:
750: The experimentally suggested occupation of octahedral sites
751: in connection with a site-switch of the on-surface oxygens
752: from fcc to hcp sites is indeed found to be initially possible
753: as a metastable configuration. Increased sub-surface O
754: incorporation will then lead to the agglomeration of
755: sub-surface species. The hcp/octa phase is unstable against a
756: registry shift of the O${}_{\rm on}$-Rh-O${}_{\rm sub}$
757: outermost layers by which the initial hcp/octa phase ends
758: in a fcc/tetra-I configuration with a fault in the stacking
759: sequence of the surface Rh layer. In this most stable phase,
760: the on- and sub-surface oxygens occupy fcc and tetra-I sites,
761: respectively, and metal atoms are sixfold coordinated to oxygens
762: while the sub-surface oxygen atoms in tetrahedral sites are
763: fourfold coordinated to metal atoms.
764:
765: The similarities between these results and those of a simultaneous
766: theoretical study of the oxidation of the Ru(0001) surface provide
767: a basis for the interpretation of the role of oxygen incorporation,
768: and nucleation as precursors of the final bulk oxide structure.
769: Addressing the phase transformation to the Rh${}_2$O${}_3$ bulk
770: oxide will thus be of considerable future interest, completing the
771: atomistic pathway of oxide formation at the Rh(111) surface.
772:
773:
774: \section{Acknowledgements}
775:
776: We thank T. Greber, J. Osterwalder, A. Seitsonen, C. Stampfl,
777: F. Illas, and M.E. Grillo for helpful discussions.
778:
779:
780: \begin{references}
781:
782:
783: \bibitem[*]{humboldt}
784: Present address: Institut f\"ur Chemie, Humboldt Universit\"at zu Berlin,
785: Unter den Linden 6, D-10099 Berlin (Germany).
786:
787: \bibitem{oh83}
788: S.H. Oh and J.E. Carpenter, J. Catal. {\bf 80}, 472 (1983).
789:
790: \bibitem{kellogg85}
791: G.L. Kellogg, Phys. Rev. Lett. {\bf 54}, 82 (1985);
792: Surf. Sci. {\bf 171}, 359 (1986).
793:
794: \bibitem{peden88}
795: C.H.F. Peden, D.W. Goodman, D.S. Blair, P.J. Berlowitz, G.B. Fisher,
796: and S.H. Oh, J. Phys. Chem. {\bf 92}, 1563 (1988).
797:
798: \bibitem{boettcher97}
799: A. B\"ottcher, H. Niehus, S. Schwegmann, H. Over, and G. Ertl,
800: J. Phys. Chem. B {\bf 101}, 11185 (1997).
801:
802: \bibitem{over00}
803: H. Over, Y.D. Kim, A.P. Seitsonen, S. Wendt, E. Lundgren, M. Schmid,
804: P. Varga, A. Morgante, and G. Ertl, Science {\bf 287}, 1474 (2000).
805:
806: \bibitem{kim01}
807: Y.D. Kim, H. Over, G. Krabbes, and G. Ertl, Topics in Catal. {\bf 14},
808: 95 (2001).
809:
810: \bibitem{comelli98}
811: G. Comelli, V.R. Dhanak, M. Kiskinova, K.C. Prince, and R. Rosei,
812: Surf. Sci. Rep. {\bf 32}, 165 (1998); and references therein.
813:
814: \bibitem{loffreda98}
815: D. Loffreda, D. Simon, and P. Sautet, J. Chem. Phys. {\bf 108}, 6447 (1998).
816:
817: \bibitem{pirovano99}
818: M.V. Ganduglia-Pirovano and M. Scheffler, Phys. Rev. B {\bf 59},
819: 15533 (1999).
820:
821: \bibitem{pirovano01}
822: M.V. Ganduglia-Pirovano, M. Scheffler, A. Baraldi, S. Lizzit, G. Comelli, G.
823: Paolucci, and R. Rosei, Phys. Rev. B {\bf 63}, 205415 (2001).
824:
825: \bibitem{thiel79}
826: P. A. Thiel, J. T. Yates Jr., and W. H. Weinberg,
827: Surf. Sci. {\bf 82}, 22 (1979).
828:
829: \bibitem{rebholz92}
830: M. Rebholz, R. Prins, and N. Kruse, Surf. Sci. {\bf 269/270}, 293 (1992).
831:
832: \bibitem{peterlinz95}
833: K.A. Peterlinz and S.J. Sibener, J. Phys. Chem. {\bf 99}, 2817 (1995).
834:
835: \bibitem{gibson95}
836: K. D. Gibson, J. I. Colonell, and S. J. Sibener, Surf. Sci. {\bf 343},
837: L1155 (1995).
838:
839: \bibitem{janssen96}
840: N. M. H. Janssen, A. Schaak, B. E. Nieuwenhuys, and R. Imbihl, Surf. Sci.
841: {\bf 364}, L555 (1996).
842:
843: \bibitem{wider99}
844: J. Wider, T. Greber, E. Wetli, T.J. Kreutz, P. Schwaller,
845: and J. Osterwalder, Surf. Sci. {\bf 417}, 301 (1998); {\em ibid.}
846: {\bf 432}, 170 (1999).
847:
848: \bibitem{gibson99}
849: K.D. Gibson, M. Viste, E.C. Sanchez, and S.J. Sibener, J. Chem.
850: Phys. {\bf 110}, 2757 (1999); {\em ibid.} {\bf 112}, 2470 (2000).
851:
852: \bibitem{kiejna00}
853: A. Kiejna and B. I. Lundqvist, Phys. Rev. B {\bf 63}, 085405 (2001); {\em ibid.}
854: Phys. Rev. B {\bf 64}, 049901 (2001).
855:
856: \bibitem{reuter02a}
857: K. Reuter, C. Stampfl, M.V. Ganduglia-Pirovano, and M. Scheffler,
858: Chem. Phys. Lett. {\bf 352}, 311 (2002).
859:
860: \bibitem{reuter02b}
861: K. Reuter, M.V. Ganduglia-Pirovano, C. Stampfl, and M. Scheffler,
862: Phys. Rev. B ({\em in press}).
863:
864: \bibitem{perdew92}
865: J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson,
866: D.J. Singh, and C. Fiolhais, Phys. Rev. B {\bf 46}, 6671 (1992).
867:
868: \bibitem{blaha99}
869: P. Blaha, K. Schwarz, and J. Luitz, {\bf WIEN97}, {\em A Full
870: Potential Linearized Augmented Plane Wave Package for Calculating
871: Crystal Properties}, Karlheinz Schwarz, Techn. Universit\"at Wien,
872: Austria, (1999). ISBN 3-9501031-0-4.
873:
874: \bibitem{kohler96}
875: B. Kohler, S. Wilke, M. Scheffler, R. Kouba, and C. Ambrosch-Draxl,
876: Comp. Phys. Commun. {\bf 94}, 31 (1996).
877:
878: \bibitem{petersen00}
879: M. Petersen, F. Wagner, L. Hufnagel, M. Scheffler, P. Blaha, and K. Schwarz,
880: Comp. Phys. Commun. {\bf 126}, 294 (2000).
881:
882: \bibitem{castner80}
883: D.G. Castner and G.A. Somorjai, Appl. Surf. Sci. {\bf 6}, 29 (1980).
884:
885: \bibitem{todorova02}
886: M. Todorova, W.X. Li, M.V. Ganduglia-Pirovano, C. Stampfl, K. Reuter,
887: and M. Scheffler, Phys. Rev. Lett. ({\em submitted}).\
888:
889: \bibitem{ruthenium}
890: The value of $\approx 0.8$ eV/O atom was calculated by comparing the
891: average adsorption energy of
892: $(2\times2)$-($3{\rm O}_{\rm on}+{\rm O}_{\rm sub}$)/Ru(0001)
893: ordered structures with that of a full adlayer on the surface
894: [$(1\times1)$-O/Ru(0001), with O on hcp sites]. The coadsorbate
895: structures considered on Ru(0001) are analogous to the fcc/tetraI
896: (a)-(b), and the hcp/oct (a)-(b) of Fig. {\protect\ref{fig1}}.
897: Calculations were performed within the same computational scheme
898: as employed in the present study, modelling the metal surface by a
899: six layer slab. Details of the used FP-LAPW basis set are those
900: described in Refs. {\protect\onlinecite{reuter02b,lizzit01}}.
901:
902: \bibitem{lizzit01}
903: S. Lizzit, A. Baraldi, A. Groso, K. Reuter, M. V. Ganduglia-Pirovano,
904: C. Stampfl, M. Scheffler, M. Stichler, C. Keller, W. Wurth, and
905: D. Menzel, Phys. Rev. B {\bf 63}, 205419 (2001).
906:
907: \bibitem{stampfl96}
908: C. Stampfl, S. Schwegmann, H. Over, M. Scheffler, and G. Ertl,
909: Phys. Rev. Lett. {\bf 77}, 3371 (1996).
910:
911: \bibitem{boettcher99}
912: A. B\"ottcher and H. Niehus, J. Chem. Phys. {\bf 110}, 3186 (1999).
913:
914: \end{references}
915:
916: \onecolumn
917:
918: \begin{figure}
919: \begin{center}
920: \psfig{file=pics/fig7.ps,width=0.8\columnwidth,angle=90}
921: \end{center}
922: \caption{Left: Contour plots of constant electron density in a
923: $[\bar 2 1 1]$ plane perpendicular to the (111) surface of
924: $(2\times 2)-(4 \rm O_{\rm on}+ 4 \rm O_{\rm sub}$)/Rh(111) structures
925: with oxygens in hcp/octa (top) and fcc/tetra-I (bottom) sites.
926: Right: Plot of the difference charge density of the same
927: plane, i.e. where the charge density of the clean Rh(111)
928: surface (with distances between Rh atoms as in the chemisorbed system)
929: and those of the isolated oxygens have been subtracted from the
930: electron density shown to the left. The distance between contours
931: is 0.3 $e/${\AA}${}^3$ (left) and 0.06 $e/${\AA}${}^3$ (right). In
932: both cases, the improved bonding of the trilayer to the
933: underlying Rh(111) substrate can be seen as an increase in
934: the bonding charge density to the second layer Rh atoms.}
935: \label{fig6}
936: \end{figure}
937:
938: \twocolumn
939:
940: \end{document}
941: