1: %\documentclass[twocolumn,showkeys,preprintnumbers,amsmath,amssymb]{revtex4}
2: %\documentclass[prb,twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
3: %\documentclass[preprint,showkeys,preprintnumbers,amsmath,amssymb]{revtex4}
4: \documentclass[prb,preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
5:
6:
7: % Some other (several out of many) possibilities
8: %\documentclass[preprint,aps]{revtex4}
9: %\documentclass[preprint,aps,draft]{revtex4}
10: %\documentclass[prb]{revtex4}% Physical Review B
11: %\documentclass[twocolumn,aps,prl,showpacs]{revtex4}
12:
13: \usepackage{graphicx}% Include figure files
14: \usepackage{dcolumn}% Align table columns on decimal point
15: \usepackage{bm}% bold math
16:
17: %\nofiles
18:
19: \begin{document}
20:
21: %\preprint{APS/123-QED}
22:
23: \title{Effect of c(2x2)-CO overlayer on the phonons of Cu(001): a first principles study}% Force line breaks with \\
24:
25: \author{Marisol Alc{\'a}ntara Ortigoza}
26: \email{alcantar@physics.ucf.edu}
27: \affiliation{Department of Physics, University of Central Florida\\
28: Orlando, Florida 32816, USA\\
29: %}%
30: %dec11
31: and\\
32: Forschungszentrum Karlsruhe, Institut f\"ur Festk\"orperphysik\\
33: D-76021 Karlsruhe, Germany
34: }%
35:
36: \author{Rolf Heid}
37: \email{heid@ifp.fzk.de}
38: \author{Klaus-Peter Bohnen}%
39: \email{bohnen@ifp.fzk.de}
40: \affiliation{%
41: Forschungszentrum Karlsruhe, Institut f\"ur Festk\"orperphysik\\
42: D-76021 Karlsruhe, Germany
43: }%
44:
45: \author{Talat S. Rahman}
46: \email{talat@physics.ucf.edu}
47: \affiliation{Department of Physics, University of Central Florida\\
48: Orlando, Florida 32816, USA
49: }%
50:
51:
52:
53:
54: \date{\today}% It is always \today, today,
55: % but any date may be explicitly specified
56:
57: \begin{abstract}
58: We have examined the effect of a c(2x2) overlayer of CO on the surface phonons of the substrate, Cu(001), by applying the density functional perturbation theory with both the local (LDA) and the generalized-gradient (GGA) density approximations, through the Hedin-Lundqvist and the Perdew-Burke-Ernzerhof functionals, respectively. Our results (GGA) trace the Rayleigh wave softening detected by helium atom scattering (HAS) experiments to changes in the force constants between the substrate surface atoms brought about by CO chemisorption, resolving an ongoing debate on the subject. The calculated surface phonon dispersion curves document the changes in the polarization of some modes and show those of the modes originally along the $\overline{Y}$ direction of the clean surface Brillouin zone (SBZ) which are back-folded along the $\overline{\Delta}$ direction of the chemisorbed SBZ, to be particularly consequential. The vertical and shear horizontal section of $S_1$ in the SBZ of
59: the clean surface, for example, is back-folded as a longitudinal-vertical mode, indicating thereby that $S_1$ $-$ predicted a long time back along $\overline{\Delta}$ for the clean surface $-$ may be indirectly assessed at $\overline{X}$ upon CO adsorption by standard planar scattering techniques. These findings further suggest that some of the energy losses detected by HAS along $\overline{\Delta}$, which were associated to multiphonon excitations of the adlayer frustrated translation mode, may actually correspond to the back-folded substrate surface modes.
60: \end{abstract}
61:
62: \pacs{68.35.Ja, 63.20.D-, 68.43.-h, 63.22.-m}% PACS, the Physics and Astronomy
63: % Classification Scheme. Use showpacs class option if pacs is desired
64: %\keywords{c(2x2)-CO-Cu(001), Surface phonon dispersion, CO internal stretch mode, CO vibrational modes,
65: % chemisorption, CO overlayer, density functional theory, linear response theory.}
66: \maketitle
67:
68: \section{\label{sec:level1}introduction}
69:
70: Although it is understandable that a large part of theoretical investigations in catalytic surface science is dedicated to developing an understanding of the surface electronic and geometric structure and
71: energetics of processes such as chemisorption and thermal activation, as a function of catalyst element and surface geometry, two decisive aspects are often overlooked: attestation of the dynamical stability of the model system\cite{bohnen} and understanding of the vibrational dynamics of the reactant-catalyst complex. Clearly, analyses lacking considerations of system dynamics disregard the fact that reaction paths and the so-called \emph{attempt frequencies} or rate preexponential factors pertaining to dynamical processes are themselves determined by the displacement patterns and frequencies of the phonons of the system under consideration, respectively.
72: Moreover, according to the harmonic transition-state theory, the entire spectrum of phonons
73: (at the equilibrium and the transition states) of the composite system is required to
74: determine the attempt frequency of any given process, and not just those vibrational modes in which reactants are primarily involved.
75: Regardless, it is frequently assumed that the changes in the free energy, which govern the rates of adsorbate processes, involve little
76: contribution from the substrate because of relatively low chemisorption energies,
77: as in the case of carbon-monoxide (CO) adsorbed on noble metal surfaces.
78: The CO binding energy on Cu(001), for example, is 20 times smaller than the carbon-oxygen binding energy and
79: about three times smaller than the chemisorption energy of CO on Ni(111).\cite{r43, r12} Nevertheless, examination of ultra-violet and X-ray photoelectron spectra signalize a significant interaction between the molecular and the Cu orbitals.\cite{allyn, mariani, umbach, isa, sandell, bjorneholm} Namely, the valence levels of CO molecules adsorbed on Cu are rearranged, broadened, and shifted with respect to those of molecules in the gas phase, in a similar fashion to that observed for Ni(001) $-$ albeit, of course, to a lesser degree.
80: In each of these cases, molecular adsorption also impacts the substrate phonons. One such example is the case of the hydrogen overlayer on Pt(111), \cite{sampyo,HAS} in which the frequency of the substrate Rayleigh wave at the zone boundaries is modified substantially from its value on clean Pt(111).
81: Furthermore, even in the case of Cu self-diffusion,\cite{kong} in which the substrate may be expected to play a less important role due to the short-range interaction among coinage metal atoms, it has been shown that the contribution of the substrate to free energy changes in the course of adatom hopping is non negligible. Should the Cu-CO interaction have a range comparable to or longer than the Cu-Cu interaction, for instance, grasping the microscopic details of CO-related reactions facilitated by metallic surfaces will demand taking into consideration all inter-atomic couplings within the system.
82:
83:
84: We have recently applied the density-functional perturbation theory\cite{r1,r62} (DFPT) to examine the vibrational modes of the c(2x2)-CO adlayer on the Cu(001) surface. Since all pertinent interatomic interactions are automatically included in our approach, we were able to calculate the phonon frequencies at
85: arbitrary propagation directions, i.e., in the entire surface Brillouin zone (SBZ).\cite{ours3} Analysis of the displacement vectors showed that the
86: CO-modes are influenced by molecule-substrate and molecule-molecule interactions.\cite{ours3} Interactions among neighboring CO molecules, separated by $\sim$~3.6{\AA}, were such as to disperse and/or split the C-O stretch, the frustrated rotation (FR), and the frustrated translation (FT) modes. Interestingly, we also found that omission of the dynamics of the substrate in the calculations lowers the frequency of the Cu-CO stretch mode by $\sim$8 meV with respect
87: to the value obtained from DFPT calculations. The frequencies of two adlayer modes, the Cu-CO stretch and the FT modes,
88: were found to depend strongly on whether the local density (LDA) or the generalized gradient (GGA) approximation, as formulated by
89: Perdew, Burke, and Ernzerhof (PBE), was used.
90: In fact, the results are a testimony to the unsuitability of LDA to describe the CO adlayer since, contrary to experimental assessments,\cite{allyn, r2}
91: not only does it render the top adsorption site as a shallow local minimum,\cite{r43} but also implies that the FT mode of the adsorbed
92: CO molecules is unstable almost everywhere along the $\overline{\Delta}$ and $\overline{\Sigma}$ directions of the c(2x2) SBZ (Fig.~\ref{fig:2a} (b)).
93: Such dynamical instabilities are indicators of the inability of LDA to describe the Cu-CO interaction which consequently leads to errors in the predicted CO adsorption site on the surface. The problem, in turn, originates from the expression for the exchange-correlation energy, which is inherently approximated in the
94: Kohn-Sham formulation and thus leads to a non-exact cancelation of the Coulomb self-interaction.~\cite{kg}
95: GGA-PBE, though not systematically self-interaction free, introduces an enlargement of the energy gap between the highest occupied orbital (HOMO) $-$ $5\sigma$ $-$ and the lowest unoccupied orbital (LUMO) $-$ $2\pi$* $-$ of CO, thus reducing the hybridization between the $2\pi$* orbital and the metallic $d$-states,~\cite{kre,rap} removing the discrepancy between theory and experiment regarding the preferred adsorption site of CO on Cu(001),~\cite{r43} and accurately reproducing the dispersion of the FT mode as measured by helium atom scattering (HAS) experiments.~\cite{ours3,r2}
96:
97:
98: In the present work, we turn again to the real-space force constants and the phonon dispersion of c(2x2)-CO/Cu(001) to analyze the effect of CO molecules on the dynamics of the Cu(001) substrate. First of all, the different scenario exhibited by GGA-PBE regarding the C-Cu interaction calls for revisiting the effect of the CO adlayer on Cu(001). Ellis et al.\cite{r2} observed via HAS measurements that the Rayleigh wave (RW) of Cu(001) softens upon CO adsorption with respect to the clean surface. The later effect was only partially explained by the mass overloading of the CO-\emph{covered} Cu atoms since the softening obtained by simply increasing the mass of such atoms while keeping intact the force constants of the clean surface underestimates the observed softening.
99: Since no significant changes in the force constants of the Cu substrate was found in the "frozen-phonon" LDA DFT calculations of
100: Lewis and Rappe\cite{r44} who obtained reasonable agreement with the experiment for the frequency of the \emph{back-folded} RW at $\overline{\Gamma}$, the mass overloading effect was accepted as the main reason for the RW softening. In reality, it is not straightforward to interpret or classify the softening as due to either factor. On the one hand, while only one of the two Cu surface atoms in the unit cell adsorbs and carries CO, the RW refers (mainly) to the vibration of the first layer which is represented by both atoms in the surface unit cell. On the other hand, \emph{covered} and \emph{bare} atoms relax in opposite directions upon CO chemisorption. In other words, \emph{covered} atoms not only support a CO molecule but also relax outwards, whereas bare atoms relax further inwards with respect to bulk interlayer spacing (GGA-PBE). Developing a rationale for the softening of the RW becomes even more complex given that despite the disagreement between LDA and GGA-PBE in
101: the predicted structural features and changes in the force constants of the substrate ~\cite{thesis}, both reproduce reasonably well the HAS measurements of the RW at the zone center.
102:
103: In this work we also focus on the substrate modes which exist along the $\overline{Y}$ direction of the clean SBZ and are now back-folded along the $\overline{\Delta}$ direction of the chemisorbed SBZ with changed polarization. Our calculations in fact suggest that some of the energy losses detected by HAS in this direction $-$ and ascribed to the multiphonon excitations of the adlayer FT mode {REF} $-$ may actually correspond to surface back-folded modes. Such backfoldings and polarization changes may be the key for the experimental detection of modes that are otherwise inaccessible to planar scattering techniques. Of particular relevance is the vertical (V - vibration perpendicular to the surface) and shear horizontal (SH - vibration perpendicular to the propagation direction) section of $S_1$ predicted for the clean surface, which may unfold $S_1$ at $\overline{X}$ to planar scattering detection. Specifically, such a mode is back-folded as a longitudinal (L - vibration pa
104: rallel to the propagation direction) and vertical mode for the chemisorbed surface and is degenerate with the pure shear-horizontal mode, $S_1$, at the zone boundary.
105:
106: The rest of this work is organized as follows: Section II contains the computational details. Section III is a summary of results concerning the structure of bulk Cu, the clean Cu(001) surface, and the c(2x2)-CO/Cu(100) chemisorbed surface. In Section IV, we present our results and discussion of the dynamics of all three systems. Finally, Section V contains concluding remarks of this study.
107:
108: \section{\label{sec:level2}COMPUTATIONAL DETAILS}
109:
110: Periodic super-cell calculations are performed on the basis of the DFT
111: formalism and the norm-conserving pseudopotential approach.~\cite{r1} The present results are
112: derived from the mixed basis (MB) technique.\cite{r52}
113: Results using both LDA and GGA are
114: obtained. The former is applied through the Hedin-Lundqvist~\cite{r54}
115: parameterization of the exchange-correlation functional, whereas GGA is
116: introduced via the PBE expression.~\cite{r56}
117:
118: %The radius around Cu sites, at which the \emph{d}-type local functions are smoothly
119: %cut off, $r_{cutoff}$, is 2.3 au for both LDA and GGA-PBE. The $r_{cutoff}$ of \emph{s}- and \emph{p}-
120: %type local functions is 1.2 and 1.08 au, respectively, for C and O in both LDA and GGA-PBE. In
121: %the present GGA-PBE calculations, \emph{d}-type basis functions are included in the
122: %description of the valence states of C atoms.
123: %The maximum kinetic energy,
124: %$E_{max}$, of the plane waves used to describe valence states has been set to 20 Ry
125: %for LDA and increased up to 33 Ry for GGA-PBE to reach convergence of the
126: %CO bond length and the atomic forces in c(2x2)-CO/Cu(001). The energy
127: %at which the Fourier expansion of the charge density is truncated, $G_{max}$, has
128: %been set to 50 Ry for both LDA and GGA-PBE.
129:
130: %MODIFY
131: The clean and the CO-chemisorbed Cu(001) surfaces are simulated with symmetric slabs
132: inside a tetragonal unit cell containing either 9 (for LDA) or 7 (for GGA-PBE)
133: layers of Cu. On the chemisorbed surface, CO molecules are
134: symmetrically located on each side of
135: the slab. Periodically repeated slabs are separated by a distance equivalent to
136: 11 and 9 vacuum layers, correspondingly.
137: Integrations inside the Brillouin zone are performed over a discrete mesh of 8x8x8 k-points for bulk Cu and of
138: 8x8x1, and 6x6x1 k-points
139: for Cu(001) and CO-c(2x2)-Cu(001), respectively.
140: %MODIFY
141: %Integrations up to the Fermi surface at the irreducible set of k-points are obtained using
142: %the broadening technique for the level occupation,\cite{r59} where the Gaussian smearing
143: %parameter, $\sigma$, is set to 0.2 eV.
144:
145: %Minimization of the slab total energy as a function of the atomic
146: %positions is based upon the reduction of Hellmann-Feynman forces\cite{r60} below
147: %$10^{-3}$ Ry/au, using the Broyden-Fletcher-Goldfarb-Shanno (BFGS)
148: %algorithm.\cite{r61}
149:
150: The calculation of the lattice dynamical matrices at specific q-points of the SBZ is
151: based on the linear response theory embodied within
152: DFPT,\cite{r1,r62}
153: as implemented in the MB scheme in Ref.\onlinecite{r58}.
154: % ADD
155: %Studies of the dynamics of CO/Cu(001) have also been performed
156: %using the plane-wave technique and ultra-soft pseudopotentials
157: %with the Quantum ESPRESSO
158: %package,~\cite{r57} which essentially provides the same
159: %results when LDA is used, but overestimates the frequency
160: %of the FT mode when GGA-PBE is used.\cite{thesis}
161: %A comparison of the corresponding results is presented
162: %elsewhere.\cite{thesis}
163: % ADD
164: The dynamical matrices for bulk Cu, Cu(001), and c(2x2)-CO/Cu(001) are
165: calculated at the q-points of a 4x4x4, a 4x4x1, and a 2x2x1 mesh, respectively.
166: The real-space force constants in these systems
167: are obtained by the standard Fourier transform of the
168: corresponding dynamical matrices.\cite{refagcu}
169: The force constants of both surfaces are then
170: matched with those of bulk Cu to model a clean and a
171: chemisorbed asymmetric slab of 50 layers
172: and used to obtain the frequencies at
173: arbitrary q-points.
174: Surface modes on the clean surface have been identified as those
175: whose amplitude weight in the two outermost layers is larger than 20$\%$.
176: On the chemisorbed slab, surface modes and resonances have been identified
177: as those whose amplitude weight
178: in the 6 outermost atoms (including C and O) is larger than 20 and 5$\%$, respectively.
179: For further details of the computational methods, we refer the reader to Ref.\onlinecite{thesis}.
180:
181: % MODIFY
182:
183:
184: \section{\label{sec:level3} RESULTS AND DISCUSSION OF STRUCTURAL PROPERTIES}
185:
186:
187: Our results for the bulk Cu lattice parameter (3.57 {\AA} [LDA] and 3.68 {\AA} [GGA-PBE]) are in good
188: agreement with all-electron\cite{r67} (AE) and pseudopotential\cite{r43,r44} calculations. Nevertheless, our calculated bulk modulus, $B$ (170 GPa [LDA] and 128 GPa [GGA-PBE]), falls
189: below that provided by AE
190: calculations.\cite{r67} Discrepancies in this respect are, in fact,
191: expected since $B$ involves the second derivative of the energy
192: with respect to the volume, being thus more susceptible
193: to the differences between AE and PP calculations than the lattice parameter.
194: As for agreement with
195: experiment,\cite{r69,r70,r71} LDA underestimates the lattice
196: constant and yields larger bulk modulus. GGA-PBE overcorrects
197: LDA, though it reproduces better the experimental bulk modulus
198: than does LDA. Further details and comparisons can be found in Ref.\onlinecite{thesis}.
199:
200: A schematic top view of the structure of Cu(001) and the c(2x2)-CO adlayer on Cu(001) is shown
201: in Fig.~\ref{fig:2a}(a).
202: The relaxation of the
203: interlayer distances normal to the surface of the clean Cu(001) has been
204: extensively studied theoretically.\cite{r1,r7,r43,r44,r48}
205: Nevertheless, before conducting our
206: study on the CO-chemisorbed surface, it is essential to test the
207: applied methodology on the well characterized clean Cu(001) surface; which will also
208: serve as a reference to appraise
209: the extent to which CO chemisorption affects it.
210: Notice in Table~\ref{tab:tab2} that both LDA and GGA-PBE produce
211: an inwards relaxation of the surface layer of $\sim 3\%$.
212: These results are in agreement with earlier pseudopotential
213: calculations\cite{r1,r7,r43,r44,r48} and with
214: surface structure measurements via the
215: medium ion energy scattering (MEIS) technique (see Table~\ref{tab:tab2}).\cite{r73}
216:
217:
218: As illustrated in Fig.~\ref{fig:2a}(a), the primitive super-cell of the CO-covered surface contains two Cu atoms per
219: layer, which are non-equivalent in odd numbered layers since CO
220: sits directly above only one of them. Accordingly, atoms in the first layer
221: are referred either as \emph{covered} or \emph{bare} atoms. To our knowledge, there is no experimental characterization of the substrate
222: geometry after CO adsorption.
223: Nevertheless, in agreement with previous calculations,\cite{r43,r44} our results\cite{ours3,thesis}
224: indicate that the first and third
225: layers rumple, while the second layer atoms do not.
226: Quantitative differences, however, arise between LDA and the GGA-PBE results
227: regarding the interlayer relaxations of the Cu(001) surface upon CO
228: adsorption.
229: We refer the reader to Table 1 in Ref.\onlinecite{ours3} for the details of the structure of c(2x2)-CO/Cu(001) and their comparison
230: with available experimental data and other calculations.\cite{r43,r44}
231: Here, we only stress that, although both LDA and GGA agree on the fact that CO raises the original inward contraction
232: of \emph{covered} atoms and even makes them relax slightly outwards with respect to
233: the bulk situation, such outward
234: relaxation is two times larger within the GGA-PBE than what it is within the LDA.
235: In addition, according to LDA, the inward relaxation of bare atoms is slightly decreased by CO, whereas GGA-PBE predicts that bare atoms undergo an
236: inward relaxation which is even larger ($\sim$ -4.0$\%$) than that of the topmost atoms of the clean surface. Note in passing that for the
237: Cu-C and C-O bond lengths, for which experimental data is available,\cite{r29}
238: GGA-PBE gives slightly better agreement to the experimental results than LDA does.\cite{ours3,thesis}
239: Likewise, our GGA-PBE calculation produces a
240: chemisorption energy of 0.68 eV at $\theta = 0.5$ ML, to be compared to
241: the experimental value\cite{r31} of 0.57 eV. LDA, on the other hand, gives a much higher value of 4.00 eV.
242:
243:
244: \section{RESULTS AND DISCUSSION OF THE LATTICE DYNAMICS}
245:
246:
247: Just for reference, we mention that our calculated phonon dispersion of
248: bulk Cu (see Ref.~\onlinecite{thesis}) is in reasonable agreement with neutron inelastic-scattering
249: measurements (NIS).\cite{r77,r78}
250: As for the dispersion of the surface phonons of Cu(001), they have been studied at length recently by DFPT methods
251: using both LDA and GGA-PBE.\cite{r1,r80} Nevertheless, we have repeated such calculations for consistency in comparison with results for the chemisorbed surface, which is the main subject of this work.
252: The notation used in this paper for the surface phonons is in accordance with that introduced by Allen et al.\cite{r84}
253: The (1x1) and c(2x2) SBZ corresponding to the clean and the chemisorbed surfaces, respectively, are shown in Fig.~\ref{fig:2a} (b).
254: We reiterate that the (1x1) SBZ is defined so that its zone boundaries along the [100] and [110] directions correspond to the $\overline{M}$ and $\overline{X}$ points, respectively. In addition, the $\overline{\Gamma}$-$\overline{M}$, $\overline{\Gamma}$-$\overline{X}$, and $\overline{X}$-$\overline{M}$ segments are denoted as the $\overline{\Sigma}$, $\overline{\Delta}$ and $\overline{Y}$ directions, respectively.
255: Before proceeding with our analysis, it is important to notice that the q-space of (1x1) SBZ which is outside the c(2x2) SBZ is \emph{back-folded} within the c(2x2) SBZ.
256:
257: We now turn
258: briefly to the highlights of the results of our calculations of the phonons of Cu(001). Full details of our results can be found in Ref.\onlinecite{thesis}.
259: Note that although a number of surface modes and resonances are found $-$ especially along $\overline{Y}$ $-$,\cite{r84} we describe below only those modes which are
260: of interest in the discussion of the chemisorbed surface (see Figs.~\ref{fig:5} and ~\ref{fig:7}).
261: Table~\ref{tab:tab5} summarizes the Cu(001) phonon frequencies at high symmetry q-points ($\overline{X}$ and $\overline{M}$) and compares with those reported in Ref.~\onlinecite{r1} and experiment.\cite{r82,r83}
262: We should point out that the sharp longitudinal resonance detected in HAS measurements,\cite{r85} is not reproduced in DFPT calculations. An explanation as to why theory and experiment differ on this particular issue remains an open question that deserves further investigation. Nevertheless, for the purpose of this work, we shall not pursue the origin and/or modifications of such longitudinal resonances since they completely vanishe after CO chemisorption.\cite{r2} Notice that the issue of the longitudinal resonance persists for several metal surfaces.\cite{r1} Chis \emph{et al}. have recently addressed the case of Al(001)\cite{chis} and Cu(111).\cite{r87}
263:
264: With regard to the c(2x2)-CO/Cu(001) system, as discussed above, LDA indicates that the chemisorption of
265: CO has little impact on the force constants mediating the
266: interaction between \emph{bare} atoms and their first NN in the
267: second layer, whereas those mediating the interaction between
268: \emph{covered} atoms and their first NN in the second layer, YY,
269: YZ, ZY, and ZZ, are mildly softened by 12, 12, 24 and 6$\%$,
270: respectively, as compared to the clean surface.\cite{thesis} GGA-PBE, in contrast,
271: finds that CO chemisorption modifies the force constants of not
272: only \emph{covered} atoms but also of \emph{bare} ones.
273: Naturally, the major effect occurs on \emph{covered} atoms,
274: whose corresponding force constants are softened by 40, 20, 38,
275: and 14$\%$, respectively, while those of \emph{bare} atoms are
276: stiffened by 14, 14, 14, and 9$\%$, respectively. Unlike the
277: \emph{interlayer} force constants, \emph{intralayer} force
278: constants between \emph{bare} atoms and their first NN
279: \emph{covered} atoms are barely altered by CO chemisorption.
280: Namely, they are $\sim$5$\%$(GGA-PBE) or $\sim$8$\%$(LDA) stiffer
281: than would be if CO were not present.
282: Tables~\ref{tab:tab6} and
283: \ref{tab:tab7} summarize the frequencies obtained at $\overline{\Gamma}$ and $\overline{X}$,
284: respectively, and compare them with those found in experiments, when available,
285: and former theoretical studies. In the sections below, we discuss in detail the
286: characterization of the
287: vibrational modes displayed by the Cu(001) substrate and shown in Fig.~\ref{fig:7}. Note, however, that because of the failures of LDA outlined in
288: Section I and for the sake of clarity, our discussion will be focused on the results obtained via GGA-PBE. LDA results are mentioned only for cases of substantial discrepancy with GGA-PBE.
289:
290:
291: \subsubsection{ Substrate modes along $\overline{\Sigma}$}
292:
293: In the following analysis, one must bear in mind
294: that modes proper to the clean Cu(001) surface (i.e. within the (1x1) SBZ)
295: along the $\overline{\Sigma}$ direction (from the $\overline{\Gamma}$ point
296: to the $\overline{M}$ point) which appear in the latter half of the same
297: are now accessible in the first half, which lies inside the c(2x2) SBZ (see Fig.~\ref{fig:2a}).
298:
299: \emph{(a) The $S_1$ mode}. The Rayleigh wave, known as $S_1$,\cite{r84} is the surface mode with the lowest frequency
300: along $\overline{\Sigma}$ direction. On the clean surface, at $\overline{M}$,
301: its polarization is mainly vertical (V) and localized in the first-layer, acquiring an additional longitudinal (L) polarization
302: as it approaches the $\overline{\Gamma}$-point.
303: On the chemisorbed surface, $S_1$ increases its frequency along
304: $\overline{\Sigma}$, until it reaches the zone boundary and matches, except
305: for a 1 meV gap, the \emph{back-folded} $S_1$, which has its
306: maximum at $\overline{\Gamma}$ and decreases its frequency along
307: $\overline{\Sigma}$.
308: Regarding the gap between the RW branches at the zone boundary,
309: we see that the higher (lower) branch
310: corresponds to a mode whose amplitude weight is primarily
311: vertical in \emph{bare} atoms (\emph{covered} atoms dragging the
312: molecule to some extent), involving also a small contribution
313: from the L-polarized vibration of \emph{covered} (\emph{bare}) atoms.
314: In fact, the higher branch matches the RW of the clean surface at the zone boundary, suggesting
315: that \emph{bare} atoms are not affected by the presence of CO.
316: Nevertheless, we will see that this is not always the case. At $\overline{\Gamma}$, where the softening with respect to the clean
317: surface is maximal, the \emph{back-folded} $S_1$ mode
318: corresponds to an \emph{out-of-phase} vibration between
319: \emph{covered} and \emph{bare} atoms. In this case, the
320: contribution of \emph{bare} atoms is 50$\%$ larger than that of
321: \emph{covered} atoms and CO molecules are dragged parallel to
322: the latter.
323: Close to $\overline{\Gamma}$, $S_1$ is broadened and appears as a
324: finite-width resonance whose maximum amplitude weight can be
325: as low as 6$\%$ in the first two layers. It softens by $\sim$16$\%$ at $\overline{\Gamma}$,
326: overestimating in fact HAS measurements ($\sim$10$\%$). We observe that although LDA and GGA-PBE predict different effects of
327: CO chemisorption on the force constants of the first layer, in both
328: cases $S_1$ softens by the same percentage.
329:
330:
331: \emph{(b) The $S_2$ mode}. This mode is mostly V-polarized and localized in the second layer at the zone boundary. On the clean surface, this
332: mode soon forms part of a band of bulk resonances along $\overline{\Sigma}$
333: whose maximum
334: amplitude weight comes from either the second and first layers (with V- and L- polarization,
335: respectively) when close to the SBZ boundary, or first
336: layer (with mixed V-L polarization) when close to $\overline{\Gamma}$. GGA-PBE and LDA yield slightly different results as to its prevalence along $\overline{\Sigma}$ and its degree of localization (see Ref.\onlinecite{thesis}).
337:
338: On the chemisorbed surface, however, $S_2$ only appears \emph{back-folded}
339: from $\overline{M}$ to $\overline{\Gamma}$. It steeply disappears as
340: soon as it immerses into the bulk band. At $\overline{\Gamma}$, $S_2$ is
341: well inside the bulk band, even though, no coupling to bulk modes
342: is observed. On the contrary, it is more localized on the
343: chemisorbed surface than on the clean surface (at $\overline{M}$). $S_2$, in addition, stiffens by $\sim$~3 meV on the chemisorbed surface.
344:
345: \emph{(c) The $L_1$ mode and the corresponding shear-horizontal (SH) branch}.
346: $L_1$ refers to a L-polarized mode mostly localized in the first layer of the clean surface at the zone boundary of the SBZ. It rapidly becomes a resonance as it approaches $\overline{\Gamma}$. As a consequence, on the
347: CO-covered surface, it appears mainly as a \emph{back-folded} mode.
348: At $\overline{\Gamma}$, the \emph{back-folded} $L_1$ and its degenerate SH
349: pair stiffen very slightly ($\sim$~1$\%$) due to the chemisorption. Neither $L_1$ nor its degenerate SH
350: pair are totally localized in the first or second
351: layers (as in the clean surface) but exhibit a significant contribution of $\sim$60$\%$ to the amplitude
352: weight from deeper layers. Outside $\overline{\Gamma}$, the \emph{back-folded}
353: $L_1$ rapidly disappears, while the \emph{back-folded} SH mode
354: disperses and becomes more localized towards the zone boundary, where it matches the branch that originally
355: appears on the clean surface but vanishes rapidly back to
356: $\overline{\Gamma}$.
357:
358: \subsubsection{Substrate modes along $\overline{\Delta}$}
359:
360: Here, one should notice that, like the \emph{back-folding} along $\Sigma$, modes proper to the clean surface along $\overline{Y}$ are
361: \emph{\emph{back-folded}} to $\overline{\Delta}$ in the c(2x2) SBZ (see Fig.~\ref{fig:2a} (b)).
362:
363: \emph{(a) $S_1$ and the corresponding L-branch}.
364: $S_1$ is the mode of lowest energy of Cu(001) and is totally localized in the first layer at the $\overline{X}$ point.
365: It cannot be detected by standard planar scattering techniques on the clean surface
366: since its polarization is SH all along $\overline{\Delta}$. $S_1$
367: changes rapidly to V- polarization along $Y$ but, to our knowledge, no experimental
368: data along $\overline{Y}$ is yet available.
369:
370: On the chemisorbed surface, $S_1$ is totally localized in the first layer and the CO overlayer.
371: The latter vibrates \emph{in-phase} with the
372: \emph{covered} Cu atoms but with much smaller amplitude.
373: $S_1$ softens at $\overline{X}$ by $\sim$8$\%$ (see Tables~\ref{tab:tab5} and ~\ref{tab:tab7}). Interestingly, the calculations reveal a L-polarized mode that is degenerate with $S_1$ at $\overline{X}$, albeit the degeneracy is broken outside $\overline{X}$. Such mode
374: corresponds to the section of $S_1$ along $\overline{Y}$ in the SBZ of the clean surface, which is
375: \emph{back-folded} along $\overline{\Delta}$ on the
376: chemisorbed SBZ and changes polarization from SH to L.
377: The \emph{back-folded} $S_1$ remains L-polarized as it goes across
378: $\overline{\Delta}$ towards $\overline{\Gamma}$ up to the crossing point with the RW,
379: where it becomes V-polarized. It is slightly softened around the zone
380: boundary, just as much as $S_1$ at $\overline{X}$; nonetheless, the softening
381: becomes stronger $-$ similar to that of the RW $-$ right after crossing
382: the RW and the transitioning to V-polarization.
383: Note that Ellis et al. observed some
384: HAS peaks precisely at the region where \emph{back-folded}
385: $S_1$ is V-polarized.\cite{r2} Those peaks were at the time said to be associated with
386: multi-phonon excitation bands. The
387: excellent fit of their measured dispersion to our \emph{back-folded}
388: $S_1$ ($\overline{Y}$ is
389: \emph{back-folded} onto $\overline{\Delta}$), however, suggests that this latter mode is observed rather than multi-phonon excitation bands.
390:
391:
392: \emph{(b) The $S_4$ mode}. This is the RW along $\overline{\Delta}$.
393: On the clean surface, it is essentially V-polarized, although a
394: L-contribution is also present.
395: Although its amplitude weight is greatest at the first layer,
396: it decays slowly as a function of the layer depth.
397: On the chemisorbed surface, it is also
398: mostly localized ($\sim$60-70$\%$) in the first layer and, to
399: lesser degree, in the molecule. $S_4$ also displays a small splitting at $\overline{X}$ (see Table~\ref{tab:tab7}), yet in this case both branches
400: are softened with respect to that of the clean surface by 9.6 and
401: 15.0$\%$ at $\overline{X}$, while HAS measurements\cite{r2} find
402: a softening of $\sim$8.2$\%$. From the first
403: layer, only the \emph{covered} (\emph{bare}) atoms contribute to the
404: mode with lower (higher) energy.
405: The \emph{back-folded} $S_4$, on the other hand,
406: originally arises along $\overline{Y}$ on the clean surface and has V-polarization
407: at the zone boundary and a predominantly SH-character as it crosses $\overline{Y}$. On the chemisorbed surface, this mode is also
408: V-polarized close to the zone boundary. In GGA-PBE, it broadens and
409: becomes a resonance as soon as it immerses into the bulk band,
410: reappearing as a surface mode close to $\overline{\Gamma}$
411: with mixed L- and SH-polarization. In LDA, however, $S_4$ remains
412: highly localized on the surface and
413: changes smoothly to L-polarization.~\cite{thesis}
414:
415:
416: \emph{(c) The $S_5$ mode}. The polarization of this mode is SH and it is predominantly localized in the second layer on the clean surface. On the chemisorbed surface, it is also strongly localized in the second layer
417: but with mixed SH- and L-polarization. It
418: rapidly becomes a resonance along $\overline{\Delta}$ towards $\overline{\Gamma}$.
419: At $\overline{X}$, $S_5$ splits (see Table~\ref{tab:tab7}). In GGA-PBE
420: both branches are totally localized in the second layer. One of these
421: softens by 8.4$\%$ and the other stiffens by 3.5$\%$.
422: LDA shows that the lower branch
423: softens while the other - bearing $\sim$25$\%$
424: contribution from deeper layers - does not change at all.
425:
426:
427: \emph{(d) The $S_6$ mode}. This mode developes inside the spectrum gap, close to the zone boundary with a frequency of 23.6 meV (see Table~\ref{tab:tab7}). On the clean surface, it has a predominant L-polarization in the first layer.
428: On the chemisorbed surface, however, it is found to be degenerate with a
429: SH pair at $\overline{X}$.
430: Note in Fig.~\ref{fig:7} that
431: the degeneracy is broken inside the SBZ.
432: Our calculations find that $S_6$ - and the SH-branch - remain nearly
433: dispersionless and involve
434: an additional V polarized second layer vibration, as found on the clean
435: surface.
436: These modes are, incidentally, more localized
437: in the first layer of the chemisorbed surface than in that of clean Cu(001).
438: At the right end of the bulk band gap (see Fig.~\ref{fig:7}), $S_6$ slightly softens and becomes a
439: resonance while it penetrates the bulk band. The SH branch, in contrast,
440: extends well inside the bulk band and becomes
441: more localized at the top two layers. According to GGA-PBE, $S_6$ softens at $\overline{X}$ by 2.0$\%$,
442: while LDA predicts no softening (see Table~\ref{tab:tab7}).
443:
444: \section{SUMMARY}
445:
446:
447: A first-principles study of the dynamics of a c(2x2)CO overlayer chemisorbed on Cu(100) has been presented. Our calculations show that LDA displays a mild effect of CO chemisorption on the relaxations
448: and force constants of the surface. The LDA predicted adsorption site is also not in agreement with experimental results.
449: Nevertherless, LDA is able to give good agreement with the HAS data for the frequency at the zone center of
450: the \emph{back-folded}
451: RW. This feature is rather unexpected in
452: consideration of the poor ability of LDA to describe the Cu-CO
453: interaction and its less successful description of the acoustic
454: modes of bulk Cu and the RW in Cu(001).\cite{thesis} GGA-PBE, on the other
455: hand, finds that CO chemisorption significantly perturbs the
456: structure and the first NN force constants of the surface layer
457: atoms. Softening of the RW is well reproduced; only slightly
458: overestimated by $\sim$1 meV at $\overline{\Gamma}$. It is surprising that,
459: while LDA and GGA-PBE differ considerably in the response of the substrate
460: to CO chemisorption, the actual percentage softening of
461: the \emph{back-folded} RW with respect to the clean surface is
462: comparable.
463: We find that softening of the RW along $\overline{\Sigma}$ and $\overline{\Delta}$ is
464: not necessarily connected with the vibration of the \emph{covered}
465: atoms, which indicates that mass overloading alone cannot account for it.
466: If that were to be the case, all surface modes involving the first layer atoms
467: would be softer. On the contrary, $L_1$, for example, does not
468: soften, notwithstanding the leading involvement of first layer
469: atoms. In fact, it slightly stiffens, consistent with the
470: hardening of intralayer force constants of the first layer. Moreover,
471: softening/stiffening of the various modes cannot be simply dictated by the
472: atomic layer, propagation direction, and polarization characteristics,
473: indicative of complexity in the modifications in Cu force constants.
474: For example, $S_2$, the V-mode in the second layer, stiffens.
475: In turn, while the $S_1$ mode (along $\overline{\Delta}$)
476: slightly softens in the region where it is L-polarized and totally
477: localized in the first layer, it
478: undergoes a stronger softening in the region where it is
479: V-polarized and involves contributions from deeper layers.
480: Softening of the RW seems thus due
481: not only to the mass of CO but also to
482: interactions of longer range, i.e. beyond first NN and involving
483: deeper layers, that subdue differences in the local bonding of
484: surface atoms and result in an overall softening of the RW, which is
485: sometimes independent on whether \emph{covered} or
486: \emph{bare} atoms are primarily involved.
487:
488: Our results call attention to the importance of the folding of $\overline{Y}$ (of the 1x1
489: SBZ) onto $\overline{\Delta}$ (of the c(2x2) SBZ), which displays
490: \emph{back-folded} modes along the latter. For example,
491: \emph{back-folded} $S_1$ and \emph{back-folded} $S_4$ are
492: found along $\overline{\Delta}$ with changed polarization that may make
493: them observable by standard scattering techniques. In particular, $S_1$ acquires V-polarization
494: close to $\overline{\Gamma}$ and L-polarization close to $\overline{X}$.
495: The good agreement for the dispersion of this latter mode
496: HAS measurements \cite{r2} is convincing evidence that these peaks arise from the \emph{back-folded} $S_1$ mode, rather than from multi-phonon excitation bands.
497: Perhaps even more importantly, \emph{back-folded} $S_1$ becomes as well
498: discernible to planar scattering spectroscopy techniques close to the zone
499: boundary since its polarization changes from SH- (along
500: $\overline{Y}$ of the (1x1) SBZ) to L- polarization (along
501: $\overline{\Delta}$ of the c(2x2) SBZ) in that region. The above thus implies
502: that the frequency of $S_1$ (SH-polarized in Cu(001)) can indirectly be
503: measured at $\overline{X}$ via \emph{back-folded} $S_1$. We believe, in fact, that
504: some of the peaks observed and assigned to \emph{2T overtones}\cite{r2} peaks by Ellis \emph{et al}.
505: along $\overline{\Delta}$ may instead correspond to \emph{back-folded}
506: $S_1$, in the region where is L-polarized. Our results here thus call for a new interpretation of the HAS data.
507: %The fact that
508: % $S_1$ has been predicted (along $\overline{\Delta}$) by theory
509: % since early studies of short-range interacting fcc
510: % structures,\cite{r84} and remained inaccessible to experiments
511: %due to its polarization, renders its detection and confirmation of the
512: % predicted frequency as an additional worthy verification of the
513: % success of DFT, the linear response theory, and the slab method in
514: % describing the fcc(001) surfaces.
515:
516:
517:
518: \begin{acknowledgments}
519: This work was supported in part by grant CHE-0741423 from NSF-USA. Computations
520: were performed at the Institut f\"ur Festk\"orperphysik, Forschungszentrum Karlsruhe, Germany.
521: Marisol Alc{\'a}ntara Ortigoza is thankful to the Forschungszentrum Karlsruhe for
522: financial support during her stays in Karlsruhe.
523:
524: \end{acknowledgments}
525:
526: %\newpage %Just because of unusual number of tables stacked at end
527: \bibliography{prb}% Produces the bibliography via BibTeX.
528:
529: \newpage
530:
531: \begin{table}
532: \caption{\label{tab:tab2}
533: Percentage change of the interlayer spacing of the outermost layers of Cu(100) compared to the bulk situation.
534: }
535: \begin{ruledtabular}
536: \begin{tabular}{|c|c|c|c|c|}
537: \hline
538: &\multicolumn{2}{c|}{Theory\footnotemark[1]} & \multicolumn{2}{c|}{Experiment}\\
539: \hline
540: &LDA & GGA-PBE&SPLEED\footnotemark[2] &MEIS\footnotemark[3] \\
541: \hline
542: $\Delta d_{12}$&-2.57&-2.82&-1.2&-2.4 \\
543: \hline
544: $\Delta d_{23}$&+0.55&+0.58&+0.9&+1.0\\
545: \hline
546: $\Delta d_{34}$&+0.30&fixed&-&-\\
547: \hline
548: \end{tabular}
549: \end{ruledtabular}
550: \footnotemark[1]{ This work }
551: \footnotemark[2]{ Ref.~\onlinecite{r72} }
552: \footnotemark[3]{ Ref.~\onlinecite{r73} }
553: \end{table}
554:
555:
556: \begin{table*}
557: \caption{\label{tab:tab5}
558: Frequencies (in meV) of the surface modes of Cu(100)
559: at the high symmetry points $\overline{X}$ and $\overline{M}$ (see Fig.~\ref{fig:2a} (b)).
560: The main polarization (SH, L, or V) is denoted in parenthesis and
561: the superscript indicates the layer showing largest amplitude weight.
562: }
563: \begin{ruledtabular}
564: \begin{tabular}{|c|c|c|c|c|c|c|}
565: \hline
566: & &\multicolumn{3}{c|}{Theory}& \multicolumn{2}{c|}{Experiment} \\
567: \hline
568: & &\multicolumn{2}{c|}{LDA} &GGA-PBE& EELS & EELS\\
569: \hline
570: & &This work&Ref.~\onlinecite{r1}&This work&Ref.~\onlinecite{r82}&Ref.~\onlinecite{r83} \\
571: \hline
572: $\overline{X}$&$S_1 (SH^1)$&9.7&9.9&9.1&-&-\\
573: \hline
574: &$S_4 (V^1)$&14.6&14.0&13.5&13.2&-\\
575: \hline
576: &$S_5 (SH^2)$&15.6 &15.0 &14.3 &- &- \\
577: \hline
578: &$S_6 (L^1)$&27.0 &26.1 &24.1 &- &25.2 \\
579: \hline
580: $\overline{M}$&$S_1 (V^1)$&18.7 &17.9 &16.9 &17.0 &- \\
581: \hline
582: &$S_2 (V^2)$&22.3&- &20.0 &- &20.3 \\
583: \hline
584: &$(SH^{1,2})$&22.3 &21.1-21.7 &20.1 &- &- \\
585: \hline
586: &$L_1(L^{1,2})$&22.3 &- &20.1 &- &~20.4 \\
587: \hline
588: \end{tabular}
589: \end{ruledtabular}
590: \end{table*}
591:
592:
593: \begin{table}
594: \caption{\label{tab:tab6}
595: Frequencies (in meV) at $\overline{\Gamma}$ of the surface vibrational modes of c(2x2)-CO/Cu(001).
596: }
597: \begin{ruledtabular}
598: \begin{tabular}{|c|c|c|c|c|}
599: \hline
600: &\multicolumn{3}{c|}{Theory} & Exp.\\
601: \hline
602: &\multicolumn{2}{c|}{LDA} &\multicolumn{1}{c|}{GGA} &HAS \\
603: \hline
604: &This work\footnotemark[1]&Ref.~\onlinecite{r44}\footnotemark[2]&This work\footnotemark[1]&Ref.~\onlinecite{r2}\\
605: \hline
606: $S_1$& 15.8 & 16.0 &14.2& 15.2\\
607: \hline
608: $L_1$& 22.6 & - & 20.2 & -\\
609: \hline
610: $S_2$& 23.1 & 23.2& 20.4& -\\
611: \hline
612: \end{tabular}
613: \end{ruledtabular}
614: \footnotemark[1]{DFPT.}
615: \footnotemark[2]{DFT-FD.}
616: \end{table}
617:
618: \newpage
619:
620:
621: \begin{table}
622: \caption{\label{tab:tab7}
623: Frequencies (in meV) at $\overline{X}$ of the surface vibrational modes of c(2x2)-CO/Cu(001).
624: }
625: \begin{ruledtabular}
626: \begin{tabular}{|c|c|c|c|}
627: \hline
628: &\multicolumn{2}{c|}{Theory} &Experiment\\
629: \hline
630: &LDA & GGA & HAS \\
631: \hline
632: &This work&This work&Ref.~\onlinecite{r2}\\
633: \hline
634: $S_1$&9.3&8.3&-\\
635: \hline
636: $S_4$&13.1&11.6&12.3\\
637: &13.3&12.2& \\
638: \hline
639: $S_5$&15.2&13.1&-\\
640: &15.6&14.8& \\
641: \hline
642: $S_6$&27.1&23.6&-\\
643: \hline
644: \end{tabular}
645: \end{ruledtabular}
646: \end{table}
647:
648:
649: \clearpage
650:
651: FIG.~\ref{fig:2a}: (a) The top view of the surface shows CO (grey circles),
652: and first (filled circles) and second (open circles) layer atoms of Cu(100).
653: The 1x1 (dashed line) and the c(2x2) (solid line) surface unit cells are underlined.
654: (b) The corresponding (1x1) (dotted line) and c(2x2) surface Brillouin zones (solid line)
655: showing the $\overline{\Gamma}$, $\overline{X}$, and $\overline{M}$ points; and the $\overline{\Delta}$, $\overline{\Sigma}$, and $\overline{Y}$ directions.
656:
657:
658: FIG.~\ref{fig:5}: GGA-PBE phonon dispersion of Cu(100), modelled by a 50-layer slab.
659: Theoretical surface modes (filled circles) are compared with HAS (open
660: triangles) and EELS (open circles) measurements taken from
661: Refs.~\onlinecite{r81,r82,r83}.
662:
663:
664: FIG.~\ref{fig:7} : GGA-PBE phonon dispersion of c(2x2)-CO/Cu(100), modelled by
665: a Cu 50-layer slab. Note that the dispersion of the high-lying modes, which correspond to the C-O stretch and the Cu-CO stretch modes, are omitted in this figure since the emphasis here is on the modes of the substrate. Filled circles denote theoretical surface modes.
666: Experimental data are taken from Ref.~\onlinecite{r2}: Filled circles and triangles
667: were associated with multiphonon processes. Open circles correspond to the substrate Rayleigh wave.
668: Squares were associated with the FT mode of CO on the perfect c(2x2)
669: structure (filled) and on defects in the adlayer at lower coverage (open).
670:
671:
672: \clearpage
673:
674:
675: \begin{figure}
676: {\huge (a)}
677: \vskip4pt\includegraphics[width=0.4\textwidth]{ARHBfig_1}{\tiny . \\ \huge (b) \\}
678: \vskip 0.29in minus -0.5in
679: \includegraphics[angle=0, width=0.4\textwidth]{ARHBfig_2}
680: \caption{\label{fig:2a}
681: }
682: \end{figure}
683:
684: \begin{figure}
685: \includegraphics[angle=0, width=0.467\textwidth]{ARHBfig_3}
686: \caption{\label{fig:5}
687: }
688: \end{figure}
689:
690: \begin{figure}
691: %\vskip -1.0in minus -0.1in
692: %\hskip -0.14in minus -0.1in
693: %\includegraphics[angle=0, width=0.467\textwidth]{fig12c}
694: %\vskip -0.2in minus 0.01in
695: %\includegraphics[angle=270, width=0.5\textwidth]{fig12b}
696: %\vskip -1.42in minus -3.9in
697: %\hskip -0.16in minus -0.1in
698: \includegraphics[angle=0, width=0.448\textwidth]{ARHBfig_4}
699: %\vskip 0.29in minus -0.5in
700: %\hskip -0.28in minus -0.1in
701: %\includegraphics[angle=0, width=0.435\textwidth]{fig_5}
702: \caption{\label{fig:7}
703: }
704: \end{figure}
705:
706:
707: \end{document}
708: %
709: ****** End of file prb.tex ******