cond-mat0310235/prb.tex
1: \documentclass[prb,aps,twocolumn,home,floats,pdflatex]{revtex4}
2: %\documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
3: 
4: \usepackage{graphics}
5: %\usepackage{pdflatex}
6: \usepackage{amsmath}
7: \bibliographystyle{apsrev}
8: \usepackage{thumbpdf}
9: %\usepackage[colorlinks,backref,pagebackref]{hyperref}
10: \usepackage{citesort}
11: %\usepackage{multicol}
12: \usepackage{dcolumn}
13: 
14: \begin{document}
15: 
16: \title{{\rm\small\hfill (submitted to Phys. Rev. B)}\\
17: On the thermodynamic stability of PdO surfaces}
18: 
19: \author{Jutta Rogal, Karsten Reuter and Matthias Scheffler}
20: 
21: \address{Fritz-Haber-Institut der Max-Planck-Gesellschaft, Faradayweg
22: 4-6, D-14195 Berlin, Germany}
23: 
24: \date{Received: 9 October 2003}
25: 
26: \begin{abstract}
27: As a first step towards understanding the morphology of PdO crystals
28: we performed a systematic full-potential density-functional theory 
29: study of all possible ($1 \times 1$) terminations of the low-index 
30: surfaces of tetragonal PdO. Applying the concept of
31: \emph{first-principles atomistic thermodynamics} we analyze the 
32: composition, structure and stability of these PdO orientations in 
33: equilibrium with an arbitrary oxygen environment. Within the studied 
34: subset of ($1 \times 1$) geometries the polar PdO-terminated PdO(100) 
35: orientation turns out to be surprisingly stable over the whole range 
36: of experimentally accessible gas phase conditions. Setting up a 
37: constrained \emph{Wulff construction} within the compiled data set, 
38: this PdO(100)-PdO facet correspondingly dominates the obtained
39: polyhedron by far. The real PdO crystallite shape will however
40: likely be affected by surface reconstructions, which are not
41: covered by the present study.
42: \end{abstract}
43: 
44: \pacs{68.47.Gh, 71.15.Mb, 82.65.+r}
45: 
46: %68.47.Gh: Oxide surfaces
47: %71.15.Mb: Density functional theory, local density approximation,
48: %          gradient and other corrections
49: %82.65.+r: Surface and interface chemistry; heterogeneous catalysis at surfaces
50: 
51: 
52: \maketitle
53: 
54: \section{Introduction}
55: 
56: Metal oxides are compounds of widespread technological interest, and 
57: one field of application is catalysis, where they can act as the
58: active material or the (often not that passive) support 
59: \cite{henrich94,noguera96}. To obtain a microscopic understanding of 
60: the function of these compounds in such applications it is necessary 
61: to know their surface atomic structure, which is also 
62: influenced by temperature and partial pressures in the environment. 
63: This can be particularly important for oxygen containing environments, 
64: where the stability of different surface terminations of varying 
65: stoichiometry may well be anticipated as a function of oxygen in the 
66: surrounding gas phase. Considering the technological importance of 
67: oxides the scarcity of such atomic-level information even for 
68: well-ordered single-crystal surfaces is surprising. The little that is 
69: known stems almost exclusively from ultra-high vacuum (UHV)
70: experiments, and is furthermore largely concentrated on some specific 
71: oxides like the vanadium oxides, rutile (TiO${}_2$) or corundum
72: (Al${}_2$O${}_3$) \cite{henrich94,noguera96}. For other 
73: oxides often not even the low-energy surface orientations are firmly 
74: established, and this also applies for the case of palladium oxide (PdO).
75: 
76: Although PdO is renowned for its high activity in the catalytic 
77: combustion of methane \cite{burch95,mccarty95,ciuparu01,ciuparu02},
78: and the involvement of oxidic structures in high-pressure CO oxidation 
79: reactions at Pd surfaces is now being discussed 
80: \cite{hendriksen03a,hendriksen03b}, virtually no information about the 
81: electronic and geometric structure of PdO surfaces is presently 
82: available, neither from experiment nor from theory. As a first step we 
83: therefore investigate the surface structure and composition of all 
84: low-index surfaces of tetragonal PdO in equilibrium with an arbitrary 
85: oxygen environment using the concept of {\em first-principles
86: atomistic thermodynamics} \cite{weinert86,scheffler87,kaxiras87,qian88}
87: based on density-functional theory (DFT) calculations (Section II). 
88: Lacking any experimental data on surface reconstructions we first
89: focus on all possible $(1 \times 1)$ terminations and set up a 
90: constrained Wulff construction for this limited set for the whole
91: range of experimentally accessible gas phase conditions (Section
92: IIIB+C). This provides a first data base against which future models of 
93: reconstructed surfaces may be compared, in particular by how much they 
94: would have to reduce the surface free energy in order to have
95: corresponding facets contribute significantly to the real PdO crystal shape.
96: In addition, relatively high surface free energies and work functions
97: obtained for $(1 \times 1)$ terminations might point at likely
98: candidates for surface reconstructions. Interestingly, within the studied 
99: subset of $(1 \times 1)$ geometries, one termination (PdO(100)-PdO) 
100: turns out to be much more stable than all others, and correspondingly 
101: dominates our constrained Wulff polyhedron by far. And this 
102: although it represents a so-called polar termination, which are traditionally 
103: dismissed on electrostatic grounds \cite{tasker79,noguera00} (Section IIID). 
104: 
105: 
106: \section{Theory}
107: 
108: \subsection{Atomistic thermodynamics}
109: 
110: In order to describe the thermodynamic stability of PdO surfaces in an 
111: oxygen environment we use DFT total-energy calculations as input to 
112: atomistic thermodynamics considerations 
113: \cite{weinert86,scheffler87,kaxiras87,qian88,reuter01,reuter03},
114: which treat the effect of the surrounding gas phase via the contact
115: with a corresponding reservoir. In equilibrium with such a reservoir
116: the most stable surface structure in the constant temperature and pressure 
117: ($T$,$p$)-ensemble minimizes the surface free energy, which is defined as
118: \begin{equation}
119: \label{equ:gamma}
120: \gamma(T,\{p_i\})= \frac{1}{A} \left[ \; G - \sum_{i}N_i\mu_i(T,p_i) \; \right] \qquad .
121: \end{equation}
122: 
123: \noindent
124: Here $G$ is the Gibbs free energy of the solid with the surface of
125: area $A$, to which in a supercell calculation with symmetric slabs
126: both the top and bottom surface contribute equally. $\mu_i(T,p_i)$ 
127: is the chemical potential of the various species $i$ present in the 
128: system, i.e. in this case $i =$ Pd, O. $N_i$ gives the number 
129: of atoms of the $i$th component in the solid. For not too low
130: temperatures and sufficiently large particles bulk PdO may further 
131: be assumed as a second thermodynamic reservoir with which the surface 
132: is equilibrated. This constrains the chemical potentials of O and 
133: Pd to the Gibbs free energy of bulk PdO, 
134: $g_{\textrm{PdO}}^{\textrm{bulk}}$ (where the small $g$ denotes the 
135: Gibbs free energy per formula unit), and allows to eliminate $\mu_{\rm
136: Pd}$ from eq. (\ref{equ:gamma}). The remaining quantities to be
137: determined for the calculation of the surface free energy are then the 
138: chemical potential of the oxygen gas phase, as well as the difference in 
139: Gibbs free energies of slab and bulk PdO.
140: 
141: The computation of the prior is straight-forward, as $\mu_{\rm O}$ is
142: completely fixed by the surrounding gas phase reservoir, which may
143: well be approximated as an ideal gas. Ideal gas laws then relate
144: the chemical potential to temperature and pressure
145: \cite{reuter01,janaf}, and we will convert the dependence of the 
146: surface free energy on $\mu_{\rm O}(T,p)$ in all figures also into the 
147: more intuitive pressure scales at $T=300$\,K and $T=600$\,K. The
148: second input to $\gamma(T,p)$, i.e. the Gibbs free energy difference 
149: of the bulk phase and the slab, receives contributions from changes
150: in the vibrational and configurational degrees of freedom at the
151: surface, as well as from the $pV$-term and as leading contribution from
152: the difference in total energies. From a dimensional analysis, the
153: $pV$-term may well be neglected \cite{reuter01}. The configurational 
154: contribution for a system like PdO can further be estimated as below 
155: 5\,meV/{\AA}${}^2$, \cite{reuter03} again negligible for a study that
156: aims at a first, rather coarse comparison of different $(1 \times 1)$ 
157: surface terminations. 
158: 
159: The vibrational contribution can be obtained from first-principles 
160: using the computed phonon density of states (DOS) at
161: the surface and in the bulk, cf. e.g. ref. \onlinecite{heid00}.
162: Alternatively, the Einstein approximation to the phonon DOS
163: can be employed to get an order of magnitude estimate \cite{reuter01}.
164: Allowing a 50\% variation of the characteristic frequency 
165: \cite{mcbride91} of each atom type at the surface, the vibrational
166: contribution to the surface free energy at all considered PdO
167: terminations stays in this model always within a range of about
168: 10-20\,meV/{\AA}${}^2$ for all temperatures up to $T=600$\,K.
169: Although this is certainly not a small contribution in general 
170: anymore, we will nevertheless neglect it in this particular study.
171: Being interested in a coarse, first screening of the stability of
172: various PdO terminations, it will become apparent below that even 
173: a 10-20\,meV/{\AA}$^2$ contribution will not affect the physical
174: conclusions drawn. If in other studies a higher accuracy is
175: required, the vibrational term may either be taken into account
176: by e.g. simplified treatments of the most relevant vibrational
177: modes \cite{sun03}, or eventually by a full DFT calculation of
178: the phonon DOS. Here, we will neglect it however, and may with
179: all other approximations discussed in this section then replace
180: the difference of bulk and slab Gibbs free energies entering into
181: the computation of $\gamma(T,p)$ simply by the corresponding
182: difference of total energies.
183: 
184: \subsection{DFT basis set and convergence}
185: 
186: The DFT total energies that are thus needed are obtained using a
187: mixture of the full-potential augmented plane wave + local orbitals (APW+lo) and 
188: the linear augmented plane wave (LAPW) method together with the 
189: generalized gradient approximation (GGA) for the exchange-correlation 
190: functional \cite{perdew96} as implemented in the \textsf{WIEN2k}
191: code \cite{wien2k,sjoestedt00,madsen01}. To simulate the different 
192: PdO surfaces we use supercells containing symmetric slabs with 7-11 
193: layers and 12-15\,{\AA} vacuum between subsequent slabs. The 
194: outermost 2-4 layers are fully relaxed for all surfaces. For all
195: orientations we have also performed test calculations with thicker
196: slabs and relaxing more layers, without obtaining any significant
197: changes to the chosen setup ($\leq$ 3\,meV/\AA$^2$ in the surface 
198: energy).
199: 
200: The parameters for the mixed APW+lo and LAPW (L/APW+lo) basis set 
201: are:  $R_{\textrm{MT}}^{\textrm{Pd}} = 1.8$\,bohr, 
202: $R_{\textrm{MT}}^{\textrm{O}} = 1.3$\,bohr, wave function expansion 
203: inside the muffin tins up to $l_{\textrm{max}}^{\textrm{wf}}=12$, 
204: potential expansion up to $l_{\textrm{max}}^{\textrm{pot}}=6$. The
205: Brillouin zone (BZ) integration is performed using Monkhorst-Pack 
206: (MP) grids with 6, 8 and 16 \textbf{k}-points in the irreducible
207: part of the BZ for the (110)/(111), the (100)/(010)/(101)/(011) and
208: the (001) directions, respectively. The energy cutoff for the plane 
209: wave representation in the interstitial region is 
210: $E_{\textrm{max}}^{\textrm{wf}} = 17$\,Ry for the wave function and 
211: $E_{\textrm{max}}^{\textrm{pot}} = 196$\,Ry for the potential. With 
212: these basis sets the surface energies of the different PdO surfaces are 
213: converged within 1-2\,meV/{\AA}$^2$ regarding the \textbf{k}-points and 
214: 3-4\,meV/{\AA}$^2$ regarding $E_{\textrm{max}}^{\textrm{wf}}$. Errors
215: introduced by the fundamental approximation in the DFT approach, i.e.
216: our use of a GGA as exchange-correlation functional, will be commented
217: on below.
218: 
219: 
220: \section{Results}
221: 
222: Since there is virtually no atomically-resolved information about 
223: the structure and composition of crystalline PdO surfaces available 
224: from the experimental side, we start our theoretical investigation 
225: from a very basic point of view. That is, to get a first idea about 
226: the geometric and electronic properties of different PdO surfaces we 
227: focus here on a rather coarse comparison of the subset of all 
228: possible ($1\times1$) terminations of the low-index surfaces
229: of tetragonal PdO.
230: 
231: \subsection{Geometric bulk and surface structure}
232: 
233: \begin{figure}
234: \scalebox{0.35}{\includegraphics{pics/fig1}}
235: \caption{\label{fig:PdOeinheit}
236: The tetragonal bulk unit cell of PdO.  Small dark spheres indicate
237: oxygen atoms, large light ones Pd atoms.}
238: \end{figure}
239: 
240: PdO crystallizes in a tetragonal structure within the space group 
241: $D_{\textrm{4h}}^9$ \cite{rogers71}. There are two formula units 
242: of PdO in the tetragonal unit cell with Pd atoms at all corners 
243: and in the centre, and O atoms at (0, 1/2, 1/4), (0, 1/2, 3/4), 
244: (1, 1/2, 1/4) and (1, 1/2, 3/4), as shown in Fig. \ref{fig:PdOeinheit}.
245: All Pd resp. O atoms are equivalent, with each Pd atom planar 
246: coordinated by 4 oxygen atoms, and each O atom tetrahedrally 
247: surrounded by 4 Pd atoms. Within our DFT-GGA approach the optimized 
248: lattice constants of the PdO unit cell are obtained as $a =
249: 3.051$\,{\AA} and $c = 5.495$\,{\AA}, which is in reasonably
250: good agreement with the experimental lattice constants,
251: $a_{\rm exp} = 3.043$\,{\AA} and $c_{\rm exp} = 5.336$\,{\AA}
252: \cite{rogers71}.
253: 
254: Due to the tetragonal structure of the PdO unit cell, there are 
255: 5 inequivalent low-index orientations, each with 2-3 different 
256: $(1 \times 1)$ surface terminations. The PdO(100) surface 
257: (parallel to the $yz$-plane) is equivalent to the (010) surface 
258: (parallel to the $xz$-plane). For this surface direction there 
259: are two different $(1 \times 1)$ surface terminations, one 
260: containing only Pd atoms in the topmost layer (PdO(100)-Pd),
261: and the other Pd as well as O atoms (PdO(100)-PdO). Schematic
262: drawings of these and all following surface geometries are
263: shown as insets in Fig. \ref{fig:stability} below. The PdO(001) 
264: surface is parallel to the $xy$-plane, and there are again two 
265: $(1 \times 1)$ terminations, one with only Pd atoms (PdO(001)-Pd) 
266: and one with only O atoms (PdO(001)-O) in the topmost layer.
267: 
268: For the PdO(101) (equivalent to PdO(011)) surface there exist 
269: three different terminations. One termination is stoichiometric
270: cutting the stacking of O-Pd${}_2$-O trilayers along the (101) 
271: direction just between consecutive trilayers (PdO(101)). The other 
272: two terminate either after the Pd layer (the O-deficient PdO(101)-Pd) 
273: or have two O layers at the top (the O-rich PdO(101)-O). The 
274: remaining PdO(110) and (111) directions are on the other hand
275: characterized by alternating layers of Pd and O atoms along the
276: surface normal, exhibiting therefore either a completely Pd 
277: (PdO(110)-Pd, resp. PdO(111)-Pd) or a completely O (PdO(110)-O, 
278: resp. PdO(111)-O) terminated surface.
279: 
280: Looking at these 11 different $(1 \times 1)$ terminations we 
281: notice that only one of them is stoichiometric. The other 10
282: exhibit either an excess of oxygen or palladium atoms, and 
283: belong thus to the class of so-called \emph{polar} surfaces
284: \cite{tasker79,noguera00}, the stability issue of which
285: will be discussed in more detail in Section IIID.
286: 
287: \subsection{Surface free energies}
288: 
289: As mentioned above we want to analyze the stability of these
290: different PdO surfaces when in contact with an oxygen gas
291: phase characterized by a given O chemical potential. This 
292: $\mu_{\textrm{O}}(T,p)$ can experimentally (and assuming that 
293: thermodynamic equilibrium applies) only be varied within certain 
294: boundaries. The lower boundary, which will be called the 
295: \emph{O-poor limit}, is defined by the decomposition of the 
296: oxide into palladium metal and oxygen. A reasonable upper boundary 
297: for $\mu_{\textrm{O}}$ on the other hand ({\em O-rich limit}) is 
298: given by gas phase conditions that are so oxygen-rich, that O 
299: condensation will start on the sample at low enough temperatures. 
300: Appropriate and well-defined estimates for these limits are
301: given by \cite{reuter01}
302: \begin{equation}
303: \label{equ:Orange}
304: \Delta G_f(T = 0 \textrm{K}, p = 0) < \Delta\mu_{\textrm{O}}(T, p_{\textrm{O}_2}) < 0 \qquad ,
305: \end{equation}
306: 
307: \noindent
308: where the O chemical potential is referenced with respect to the 
309: total energy of an oxygen molecule, $\Delta\mu_{\textrm{O}} = \mu_{\textrm{O}} 
310: - (1/2)E_{\textrm{O}_2}^{\textrm{total}}$, and 
311: $\Delta G_f(T = 0 \textrm{K}, p = 0)$ is the low temperature
312: limit for the heat of formation of PdO. Within DFT-GGA we compute 
313: $-0.88$\,eV for this quantity, which compares well with the 
314: experimental value of $\Delta G_f^{\rm exp}(T \rightarrow  0 \textrm{K}, 
315: p = 1 ^{}\textrm{atm}) = -0.97$ eV \cite{CRC95}. To also consider 
316: the uncertainty in these theoretically well-defined, but approximate 
317: limits for $\Delta\mu_{\textrm{O}}$, we will always plot the
318: dependence of the surface free energies also some tenths of eV outside 
319: these boundaries. From this it will become apparent below that the 
320: uncertainty in the boundaries does not affect at all our physical 
321: conclusions drawn.
322: 
323: \begin{figure*}
324: \scalebox{0.32}{\includegraphics{pics/fig2}}
325: \caption{\label{fig:stability}
326: Surface free energies for the 5 inequivalent low-index PdO surfaces.
327: Solid lines indicate $(1 \times 1)$ terminations, and dashed lines
328: larger unit cell reconstructions discussed in Section IIID.
329: The vertical dotted lines specify the range of $\Delta\mu_{\textrm{O}}$
330: considered in this study (see text), while in the top two $x$ 
331: axes the dependence on the O chemical potential has been converted
332: into pressure scales at $T = 300$ K and $T = 600$ K. The insets
333: show the surface geometries of the corresponding $(1 \times 1)$ 
334: terminations (O small dark spheres, Pd large light spheres).}
335: \end{figure*}
336: 
337: We proceed to show all surface free energy plots
338: of the eleven discussed $(1 \times 1)$ terminations in Fig. 
339: \ref{fig:stability}. Terminations with an O excess (deficiency) 
340: show a negative (positive) slope, i.e. they will become the more
341: favorable (unfavorable) the more O-rich the surrounding
342: gas phase is. Comparing the five plots shown in
343: Fig. \ref{fig:stability} the very high stability of the PdO(100)-PdO
344: termination is easily recognized. The only other terminations
345: that exhibit comparably low surface energies are the PdO(101), 
346: and towards the O-rich limit the PdO(110)-O and PdO(111)-O 
347: surfaces. All other considered terminations are rather 
348: high in energy, in particular all those that feature Pd atoms
349: in their outermost layer. Looking at the energy scale in the
350: plots we also notice that this stability ordering will not be
351: affected by the afore discussed 10-20\,meV/{\AA}$^2$ uncertainty
352: in our surface free energies.
353: 
354: \begin{table}
355: \caption{\label{tab:gamma}
356: Surface free energies of all low-index PdO ($1\times1$) terminations
357: at the oxygen-poor limit, as calculated within the GGA or the 
358: LDA.  All energies are in meV/\AA$^2$, and the numbers in brackets 
359: denote the energetic difference with respect to the lowest-energy
360: PdO(100)-PdO termination.}
361: 
362: \begin{ruledtabular}
363: \begin{tabular}{l|rr|rr}
364: Surface 			&
365: \multicolumn{2}{c}{$\gamma_{\textrm{O-poor}}$}  	&
366: \multicolumn{2}{c}{$\gamma_{\textrm{O-poor}}$}  \\
367: termination			&	
368: \multicolumn{2}{c}{GGA}         &
369: \multicolumn{2}{c}{LDA}			\\ \hline
370: \\
371: PdO(100)-PdO		        &	33    &    (0)
372: & 59&  (0)	       	\\
373: PdO(100)-Pd			&	119   &  (+86)				& 170 &(+111)			\\
374: 				&						&				\\
375: PdO(001)-O			&	119  &	(+86)				& 162 &(+103)			\\
376: PdO(001)-Pd			&	156  &  (+123)				& 212 &(+153)	 		\\
377: 				&						&				\\
378: PdO(101)			&	57 &	(+24)				& 86  &(+27)      		\\
379: PdO(101)-O			&	134 &   (+101)				& 180 &(+121)			\\
380: PdO(101)-Pd			&	128&	 (+95)
381: &  173 &(+114)		\\
382: 				&						&				\\
383: PdO(110)-O			&	86&	 (+53)				& 119 &(+60)			\\
384: PdO(110)-Pd			&	137&	(+104)				& 173 &(+114)			\\
385: 				&						&				\\
386: PdO(111)-O            		&         72&     (+39)                		& 109 &(+50)        
387:      \\
388: PdO(111)-Pd     	        &        105 &    (+72)
389: & 143& (+84)            \\
390: \end{tabular}
391: \end{ruledtabular}
392: \end{table}
393: 
394: As already indicated above this uncertainty in the surface energies
395: does not yet include the error due to the choice of GGA as 
396: exchange-correlation functional. We have therefore calculated the surface 
397: free energies of all considered terminations also within the local 
398: density approximation (LDA) \cite{perdew92}. To set up the supercells 
399: for these LDA calculations we first optimized the lattice constants 
400: for PdO bulk, obtaining $a_{\rm LDA} = 2.990$\,{\AA} and $c_{\rm
401: LDA} = 5.292$\,{\AA}, i.e. values that are as expected slightly
402: smaller than the afore mentioned experimental lattice constants.
403: After a full relaxation of the 2-4 outermost layers, the relative
404: changes in the surface geometries with respect to these bulk values
405: are found to be very similar to the ones obtained within the GGA.
406: The resulting absolute surface free energies of the eleven
407: terminations in the oxygen-poor limit are listed in Table
408: \ref{tab:gamma} (together with the corresponding GGA values). Although
409: the absolute values of the LDA surface energies are
410: 30-50\,meV/{\AA}$^2$ higher than within the GGA, the relative
411: differences between them are within 10-30 meV/{\AA}$^2$. This is exemplified
412: in Table \ref{tab:gamma} by also indicating the relative energetic
413: difference with respect to the PdO(100)-PdO termination, which is the
414: lowest-energy surface in both the LDA and the GGA. As only this
415: energetic ordering among the considered $(1 \times 1)$ terminations 
416: enters into the targeted constrained Wulff construction, and we find this
417: ordering to be similar for calculations with two quite
418: different exchange-correlation functionals, we expect the DFT
419: accuracy to be rather high for this system.
420: 
421: 
422: \subsection{Constrained Wulff construction}
423: 
424: \begin{figure}
425: \scalebox{0.40}{\includegraphics{pics/fig3}}
426: \caption{\label{fig:wulff}
427: {\em Constrained} Wulff construction at the oxygen-poor 
428: (left) and oxygen-rich limit (right). The construction is 
429: constrained to the studied $(1 \times 1)$ terminations 
430: and reflects therefore only the relative energies of 
431: different PdO orientations, rather than the real PdO crystal
432: shape, which will most likely be affected by surface
433: reconstructions. The polyhedra are symmetric with respect 
434: to the $xy$-plane, and only the upper half is shown
435: correspondingly.}
436: \end{figure}
437: 
438: With the results obtained for the surface free energies of the
439: different $(1 \times 1)$ PdO terminations we set up a constrained 
440: Wulff construction \cite{wulff01} for a PdO single 
441: crystal. Since this construction is constrained to reflect
442: only the studied $(1 \times 1)$ terminations, its intention
443: is more to compare the relative energies of different 
444: surface orientations, rather than to really predict the
445: equilibrium PdO crystallite shape (which will most likely be 
446: affected by surface reconstructions on which we comment below). 
447: Since the surface free energies depend on the oxygen chemical
448: potential, the obtained construction will vary with the conditions 
449: in the surrounding gas phase. We therefore present in Fig. 
450: \ref{fig:wulff} the Wulff polyhedra for the two considered 
451: extremes, i.e. in the O-poor and the O-rich limit. Due to the 
452: tetragonal symmetry of PdO, the polyhedra are symmetric with 
453: respect to the $xy$-plane indicated in Fig. \ref{fig:PdOeinheit}, 
454: and only the upper half of each polyhedron is shown correspondingly. 
455: 
456: As one would already assume from the stability plots in Fig. 
457: \ref{fig:stability}, the low-energy PdO(100)-PdO surface forms the 
458: largely dominating facet (rectangular, dark gray facets) at both the 
459: O-poor and the O-rich limit of $\Delta\mu_{\textrm{O}}$. The other 
460: triangular, light gray facets are built by the stoichiometric
461: PdO(101) surface termination. All other investigated terminations 
462: do not contribute at all to the present construction, since the 
463: corresponding planes lie outside of the polyhedra and do not cross 
464: them at any point. The polyhedra in Fig. \ref{fig:wulff} are scaled 
465: in a way, that the absolute value of the area belonging to the 
466: PdO(101) termination is the same in the O-poor and O-rich limit,
467: since the surface free energy of this termination is also constant 
468: with respect to $\Delta\mu_{\textrm{O}}$. In turn, the remaining surface
469: area built up by the O-rich PdO(100)-PdO termination increases 
470: strongly in going from O-poor to O-rich limit. In
471: the prior limit, this termination forms already 72\% of the whole 
472: area of the polyhedron, while at the latter limit this fraction has
473: increased to even 94\%. As would already be expected from the
474: similar relative energy differences, these Wulff construction
475: properties are also almost equally obtained within the LDA: again the
476: PdO(100)-PdO and PdO(101) terminations are the only ones contributing
477: to the polyhedron, and the PdO(100)-PdO facet covers 66\% (82\%) of
478: the surface area in the {\rm O-poor} ({\rm O-rich}) limit. Although
479: such a comparison between two functionals forms no formal proof,
480: we would therefore assume the obtained shape to be quite
481: accurate with respect to this DFT approximation.
482: 
483: This does, however, not comprise the major limitation of our study,
484: given by the restriction to $(1 \times 1)$ terminations.
485: Surface reconstructions could considerably lower the surface free
486: energy of any of the PdO surfaces and correspondingly significantly
487: influence the overall shape of the Wulff polyhedron. Unfortunately, 
488: to the best of our knowledge no experimental information on such surface
489: reconstructions is presently available for crystalline PdO. Without any such 
490: information, not even on the surface periodicity, the phase space
491: of possible reconstructions is simply too huge to be assertively
492: screened by todays first-principles techniques alone \cite{lundgren02}. 
493: Until such information becomes available from experiment, the best we can do
494: at the moment, is to check by how much such surface
495: reconstructions would have to lower the surface free energy in order
496: to give rise to significant changes in the real PdO crystal shape
497: compared to the constrained construction presented in this study. 
498: This check is done for the three orientations presently
499: not contributing to the exterior of the polyhedron, i.e. the (001), 
500: (110) and (111) facets, and since the surface reconstructions are of
501: unknown stoichiometry their effect could be different in the
502: O-poor and in the O-rich limit.
503: 
504: \begin{table}
505: \caption{\label{tab:reduction}
506: Minimum energy by which surface reconstructions at the various
507: facets would have to lower the surface free energy, in order
508: for the facets to touch the presently obtained constrained
509: Wulff polyhedron (touching). Additionally, the corresponding 
510: lowering required for the facet to cover approximately 10\% of 
511: the total surface area of the polyhedron is listed (10\% area). 
512: All energies in meV/{\AA}$^2$ (and percent changes) are given 
513: with respect to the lowest-energy $(1 \times 1)$ termination of 
514: the corresponding orientation.}
515: 
516: \begin{ruledtabular}
517: \begin{tabular}{lrr}
518: Orientation & Touching & 10\% area \\ \hline
519: \\
520: Oxygen-poor:\\
521: (001) & -54 (-45\%) & -67 (-56\%) \\
522: (110) & -40 (-47\%) & -44 (-51\%) \\
523: (111) & -7  (-10\%) & -16 (-22\%) \\
524: \\
525: Oxygen-rich:\\
526: (001) &  -7 (-10\%) & -42 (-58\%) \\
527: (110) & -39 (-80\%) & -41 (-84\%) \\
528: (111) &  -1 (-2\%)  & -20 (-37\%) \\
529: \end{tabular}
530: \end{ruledtabular}
531: \end{table}
532: 
533: Table \ref{tab:reduction} lists the corresponding values for
534: each facet, of how much the surface free energy would have to be 
535: lowered by a reconstruction (with respect to the lowest-energy
536: $(1 \times 1)$ termination presently considered in our constrained
537: Wulff construction), such that the facet just touches the current
538: polyhedron. Additionally given is the corresponding value, required 
539: to have the reconstruction really contribute significantly
540: to the total Wulff shape, which we consider to be the case
541: when the facet covers approximately 10\% of the total surface area. 
542: From the compiled data in Table \ref{tab:reduction} it seems
543: that rather massive reconstructions would be necessary (both
544: in the O-poor and in the O-rich limit) to have 
545: appreciable (001) and (110) facets in the PdO equilibrium 
546: crystal shape. A much smaller energy reduction would on the
547: other hand be required to have a (111) facet contribute. 
548: 
549: \subsection{Stability of PdO surfaces}
550: 
551: \begin{figure}
552: \scalebox{0.40}{\includegraphics{pics/fig4}}
553: \caption{\label{fig:recon}
554: Unit cells for stoichiometric reconstructions of the PdO(001)-O 
555: and PdO(110)-O terminations, achieved by simply removing
556: the oxygen atoms marked with crosses. The two left figures 
557: show configurations, where the oxygen atoms are taken out 
558: along a row; the right ones, where the oxygen forms a 
559: checkerboard pattern.}
560: \end{figure}
561: 
562: As already mentioned above only one of the 11 possible 
563: $(1 \times 1)$ terminations, namely the PdO(101), is 
564: stoichiometric, whereas all others are so-called polar 
565: surfaces, which are not expected to be stable on electrostatic
566: grounds \cite{tasker79,noguera00}. For the three different 
567: surface terminations in the (101) direction this ionic
568: model certainly complies with our results, since the 
569: stoichiometric PdO(101) termination turns out much more 
570: stable than the polar PdO(101)-O and PdO(101)-Pd 
571: terminations. For the other orientations it is on the other
572: hand not possible at all to truncate the tetragonal PdO 
573: structure in a $(1 \times 1)$ cell and achieve charge
574: neutrality; and one might wonder whether this is the
575: reason why e.g. the considered (001) and (110) $(1 \times 1)$
576: terminations exhibit such high surface energies? As most obvious 
577: guess we therefore performed fully relaxed calculations in
578: larger unit cells, where we simply removed half the O
579: atoms in the top layer in order to achieve stoichiometric
580: terminations. As shown in Fig. \ref{fig:recon} there
581: are two possibilities for both the (001) and the (110)
582: orientation to remove these oxygen atoms, either all
583: along a row or in a checkerboard pattern. The corresponding
584: surface free energies are drawn as dashed lines in Fig.
585: \ref{fig:stability}, and are not at all lower than the
586: corresponding polar $(1 \times 1)$ O-rich terminations.
587: These results are therefore at variance with the suggestion made
588: by Ciuparu {\em et al.}, that such a simple removal of O atoms should
589: lead to charge compensated and thus stable PdO(001) and PdO(110)
590: surfaces \cite{ciuparu01}.
591: 
592: Apparently, charge neutrality is not the dominant
593: feature determining the energetic stability of PdO
594: surfaces, as is also directly reflected in the very 
595: low surface free energy of the polar PdO(100)-PdO
596: termination. This points at the most obvious shortcoming
597: of the electrostatic model, namely the assumption that
598: all atoms of one species are identical and in the same
599: charge state, irrespective of whether they are at the
600: surface or in the bulk. As we had shown before
601: \cite{reuter01,wang98}, structural and electronic
602: relaxation at the surface allows for appreciable
603: deviations from this picture, such that other factors
604: (like an appropriate excess stoichiometry at O-rich
605: conditions) might well overrule the polarity issue.
606: 
607: \begin{table}
608: \caption{\label{tab:workf}
609: Work functions for the different $(1 \times 1)$ PdO 
610: surface terminations in eV.}
611: \begin{ruledtabular}
612: \begin{tabular}{llll}
613: O-terminated  		&	$\Phi$ (eV)		&	Pd-terminated	&	$\Phi$ (eV)	\\ \hline
614: \\
615: 				&				&	PdO(100)-Pd		&	4.0		\\
616: PdO(001)-O			&	7.9			&	PdO(001)-Pd		&	4.8		\\
617: PdO(101)-O			&	7.7			&	PdO(101)-Pd		&	4.5		\\
618: PdO(110)-O			&	7.2			&	PdO(110)-Pd		&	4.4		\\
619: PdO(111)-O			&	5.9			&
620: 				PdO(111)-Pd		&	4.7 \\ 
621: \\ \hline
622: PdO-terminated		&	$\Phi$ (eV)		&	stoichiometric	&	$\Phi$ (eV)	\\ \hline
623: \\
624: PdO(100)-PdO		&	6.4			&	PdO(101)		&	5.4		\\
625: \end{tabular}
626: \end{ruledtabular}
627: \end{table}
628: 
629: Still, that there is a different degree of polarity associated
630: with the different terminations, is nicely reflected in the
631: corresponding work functions, cf. Table \ref{tab:workf}. While
632: the work function of the stoichiometric PdO(101) termination
633: is with 5.4\,eV in a medium range, the work functions of all
634: Pd-terminated surfaces are about 1.0 - 1.5 eV lower. The ones
635: of the O-terminated surfaces are on the other hand about
636: 1.0 - 2.5\,eV higher, just as expected from the ionic model.
637: Even the comparably low work function of the O-terminated
638: PdO(111)-O surface fits into this picture, as there the layer 
639: distance to the topmost oxygen atoms is with 0.51\,{\AA} significantly
640: smaller than for the other orientations, and leads therefore
641: to a smaller dipole moment. 
642: 
643: \begin{figure}
644: \scalebox{0.34}{\includegraphics{pics/fig5}}
645: \caption{\label{fig:binding}
646: Binding energies of the topmost oxygen atoms at the most
647: stable $(1 \times 1)$ termination of the various PdO 
648: surfaces. The various orientations are sorted along the
649: $x$ axes with higher surface free energies to the right.
650: The binding energies are given with respect to a free 
651: O${}_2$ molecule.}
652: \end{figure}
653: 
654: Discarding the polarity as a major factor governing the
655: surface stability, we proceed to correlate the latter with the binding
656: energy of surface oxygen atoms, as interestingly each most stable 
657: $(1 \times 1)$ termination of each orientation features oxygen atoms in
658: the topmost layer, cf. Fig. \ref{fig:stability}. The corresponding 
659: binding energies with respect to molecular oxygen are shown in 
660: Fig. \ref{fig:binding}, with positive values indicating that 
661: O${}_2$ dissociation would be exothermic. The various orientations
662: are sorted along the $x$ axes with higher surface energies to
663: the right, revealing a clear correlation. Moreover, there is
664: also a clear correlation with the coordination of the surface
665: atoms: The two most stable terminations, namely the ones contributing 
666: to our constrained Wulff polyhedron, feature three-fold coordinated
667: surface O atoms and with $\sim 2.5$\,eV rather strong binding
668: energies. This is followed by the other three orientations,
669: that do not contribute to the present Wulff shape, exhibiting
670: only two-fold coordinated oxygen atoms and somewhat lower
671: binding energies around 1.5\,eV. We therefore conclude that
672: the stability of the studied subset of $(1 \times 1)$ terminations
673: seems primarily governed by the openness of the surface 
674: orientation, i.e. whether its geometric structure offers 
675: highly-coordinated O binding sites.
676: 
677: 
678: \section{Summary}
679: 
680: In conclusion we have calculated the surface free energies of all 
681: low-index ($1\times1$) PdO terminations in equilibrium with an oxygen 
682: environment using the concept of \emph{atomistic thermodynamics}.
683: The PdO(100)-PdO termination exhibits an extraordinarily low surface
684: energy over the entire range of experimentally accessible gas phase
685: conditions. Correspondingly, this facet dominates the Wulff polyhedron
686: constrained to the studied $(1 \times 1)$ terminations by far,
687: with only the PdO(101) orientation also covering a smaller surface
688: area. The high stability of these two terminations is largely
689: connected to their closed geometric structure, allowing a strong 
690: oxygen binding in highly-coordinated surface sites, while the polarity 
691: of the non-stoichiometric PdO(100)-PdO termination plays apparently 
692: only a minor role. The equilibrium shape of a real PdO crystal is 
693: likely to deviate from the presently obtained constrained Wulff 
694: polyhedron due to surface reconstructions. Lacking experimental 
695: information on such reconstructions, only the minimum energy 
696: lowering required to have corresponding facets contribute to the 
697: overall crystal shape have been presented. 
698: 
699: 
700: \section{Acknowledgements}
701: Stimulating discussions with M. Todorova are gratefully
702: acknowledged.
703: 
704: 
705: \begin{thebibliography}{99}
706: 
707: \bibitem{henrich94}
708: V.E. Henrich and P.A. Cox, \emph{The Surface Science of Metal Oxides}, 
709: Cambridge Univ. Press, Cambridge (1994).
710: 
711: \bibitem{noguera96}
712: C. Noguera, \emph{Physics and Chemistry at Oxide Surfaces}, 
713: Cambridge Univ. Press, Cambridge (1996).
714: 
715: \bibitem{burch95}
716: R. Burch and M.J. Hayes, J. Mol. Catal. A \textbf{100}, 13 (1995).
717: 
718: \bibitem{mccarty95}
719: J.G. McCarty, Catal. Today \textbf{26}, 283 (1995).
720: 
721: \bibitem{ciuparu01}
722: D. Ciuparu, E. Altman and L. Pfefferle, J. Catal. \textbf{203}, 64 (2001).
723: 
724: \bibitem{ciuparu02}
725: D. Ciuparu and L. Pfefferle, Catal. Today \textbf{77}, 167 (2002).
726: 
727: \bibitem{hendriksen03a}
728: B.L.M. Hendriksen, {\em Model Catalysts in Action: High-Pressure Scanning
729: Tunneling Microscopy}, Ph.D. thesis, Universiteit Leiden (2003);
730: http://www.physics.leidenuniv.nl/sections/cm/ip/group/
731: theses.htm\#hendriksen
732: 
733: \bibitem{hendriksen03b}
734: B.L.M. Hendriksen, M.D. Ackermann, and J.W.M. Frenken, 
735: ({\em private communication}).
736: 
737: \bibitem{weinert86}
738: C.M. Weinert and M. Scheffler, In: \emph{Defects in Semiconductors}, 
739: H.J. von Bardeleben (Ed.), Mat. Sci. Forum \textbf{10-12}, 25 (1986).
740: 
741: \bibitem{scheffler87}
742: M. Scheffler, In: \emph{Physics of Solid Surfaces -- 1987}, J. Koukal 
743: (Ed.), Elsevier, Amsterdam (1988). M. Scheffler and J. Dabrowski, 
744: Phil. Mag. A \textbf{58}, 107 (1988).
745: 
746: \bibitem{kaxiras87}
747: E. Kaxiras, Y. Bar-Yam, J.D. Joannopoulos, and K.C. Pandey, 
748: Phys. Rev. B \textbf{35}, 9625 (1987).
749: 
750: \bibitem{qian88}
751: G.-X. Qian, R.M. Martin, and D.J. Chadi, 
752: Phys. Rev. B \textbf{38}, 7649 (1988).
753: 
754: \bibitem{tasker79}
755: P.W. Tasker, J. Phys. C {\bf 12}, 4977 (1979).
756: 
757: \bibitem{noguera00}
758: C. Noguera, J. Phys.: Condens. Matter {\bf 12}, R367 (2000).
759: 
760: \bibitem{reuter01}
761: K. Reuter and M. Scheffler, Phys. Rev. B \textbf{65}, 035406 (2002).
762: 
763: \bibitem{reuter03}
764: K. Reuter and M. Scheffler, Phys. Rev. Lett. \textbf{90}, 046103
765: (2003); Phys. Rev. B \textbf{68}, 045407 (2003).
766: 
767: \bibitem{janaf}
768: D.R. Stull and H. Prophet, \emph{JANAF Thermochemical Tables}, 2nd 
769: ed., U.S. National Bureau of Standards, Washington DC (1971).
770: 
771: \bibitem{heid00}
772: R. Heid, L. Pintschovius, W. Reinhardt, and K.-P. Bohnen,
773: Phys. Rev. B {\bf 61}, 12059 (2000).
774: 
775: \bibitem{mcbride91}
776: J. McBride, K. Hass, and W. Weber, Phys. Rev. B \textbf{44}, 5016 (1991).
777: 
778: \bibitem{sun03}
779: Q. Sun, K. Reuter, and M. Scheffler, Phys. Rev. B \textbf{67}, 205424 (2003).
780: 
781: \bibitem{perdew96}
782: J.P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. \textbf{77}, 
783: 3865 (1996).
784: 
785: \bibitem{wien2k}
786: P. Blaha, K. Schwarz, G.K. Madsen, D. Kvasnicka, and J. Luitz, 
787: \textbf{WIEN2k}, \emph{An Augmented Plane Wave + Local Orbitals
788: Program for Calculating Crystal Properties}, Karlheinz Schwarz, 
789: Techn. Universit\"at Wien, Austria (2001). ISBN 3-9501031-1-2.
790: 
791: \bibitem{sjoestedt00}
792: E. Sj\"ostedt, L. Nordstr\"om, and D.J. Singh, Solid State Commun. 
793: \textbf{114}, 15 (2000).
794: 
795: \bibitem{madsen01}
796: G.K.H. Madsen, P. Blaha, K. Schwarz, E. Sj\"ostedt, and 
797: L. Nordstr\"om, Phys. Rev. B \textbf{64}, 195134 (2001).
798: 
799: \bibitem{rogers71}
800: D. Rogers, R. Shannon, and J. Gillson, J. Solid State Chem. 
801: \textbf {3}, 314 (1971).
802: 
803: \bibitem{CRC95}
804: \emph{CRC Handbook of Chemistry and Physics}, CRC press, Boca Raton FL
805: (1995).
806: 
807: \bibitem{perdew92}
808: J.P. Perdew and Y. Wang, Phys. Rev. B \textbf{45}, 13244 (1992).
809: 
810: \bibitem{wulff01}
811: G. Wulff, Z. Kristallogr. \textbf{34}, 449 (1901).
812: 
813: \bibitem{lundgren02}
814: E. Lundgren, G. Kresse, C. Klein, M. Borg, J.N. Andersen, M. De Santis,
815: Y. Gauthier, C. Konvicka, M. Schmid, and P. Varga, Phys. Rev. Lett.
816: {\bf 88}, 246103 (2002).
817: 
818: \bibitem{wang98}
819: X.-G. Wang, W. Weiss, Sh.K. Shaikhutdinov, M. Ritter, M. Petersen,
820: F. Wagner, R. Schl\"ogl, and M. Scheffler, Phys. Rev. Lett. {\bf 81},
821: 1038 (1998).
822: 
823: \end{thebibliography}
824: 
825: \end{document}
826: 
827: 
828: