General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright
owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
You may not further distribute the material or use it for any profit-making activity or commercial gain
You may freely distribute the URL identifying the publication in the public portal
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Downloaded from orbit.dtu.dk on: Aug 09, 2022
CO Chemisorption at Metal Surfaces and Overlayers
Hammer, Bjørk; Morikawa, Y.; Nørskov, Jens Kehlet
Published in:
Physical Review Letters
Link to article, DOI:
10.1103/PhysRevLett.76.2141
Publication date:
1996
Document Version
Publisher's PDF, also known as Version of record
Link back to DTU Orbit
Citation (APA):
Hammer, B., Morikawa, Y., & Nørskov, J. K. (1996). CO Chemisorption at Metal Surfaces and Overlayers.
Physical Review Letters, 76(12), 2141-2144. https://doi.org/10.1103/PhysRevLett.76.2141
VOLUME 76, NUMBER 12 PHYSICAL REVIEW LETTERS 18MARCH 1996
CO Chemisorption at Metal Surfaces and Overlayers
B. Hammer,
1,2
Y. Morikawa,
2
and J. K. Nørskov
1
1
Center for Atomic-scale Materials Physics and Physics Department, Technical University of Denmark, DK-2800 Lyngby, Denmark
2
Joint Research Center for Atom Technology (JRCAT), 1-1-4 Higashi, Tsukuba, Ibaraki 305, Japan
(Received 31 August 1995)
A database of ab initio calculations of the chemisorption energy of CO over Ni(111), Cu(111),
Ru(0001), Pd(111), Ag(111), Pt(111), Au(111), Cu
3
Pt(111), and some metallic overlayer structures
is presented. The trends can be reproduced with a simple model describing the interaction between
the metal d states and the CO 2p
p
and 5s states, renormalized by the metal sp continuum. Our
model rationalizes the results by Rodriguez and Goodman [Science 257, 897 (1992)] showing a strong
correlation between the CO chemisorption energy and the surface core level shift.
PACS numbers: 73.20.At, 71.15.Mb, 73.61.At, 82.65.My
Over the past three decades the field of surface sci-
ence has produced a series of accurate spectroscopical
techniques that can provide detailed information about the
electronic structure at surfaces [1]. It would be extremely
useful if spectroscopic data could also be used directly to
give information about the chemical activity of the sur-
face. This would open up new possibilities in the future
search for, e.g., more efficient catalysts. Recently, Ro-
driguez and Goodman [2] have established a spectacu-
lar correlation between spectroscopical data (surface core
level shifts) and the chemisorption energy of CO on a se-
ries of metal surfaces and overlayers. If such an approach
can be generalized, we would have a means of predict-
ing the chemical activity of surfaces based on the surface
electronic properties alone.
In the present Letter, we discuss the physics of CO ad-
sorption over metal surfaces and overlayers by presenting
an extensive ab initio database of CO chemisorption en-
ergies calculated within density functional theory (DFT)
using the generalized gradient approximation (GGA). We
demonstrate that the trends in the database can be un-
derstood using a simple two-level model describing the
coupling of the CO 5s and 2p
p
states to the metal d va-
lence states. One key surface parameter determining the
strength of the bonding turns out to be the energy of the
center of the metal d band. This surface property can
be obtained from spectroscopical methods either directly
with photoemission (UPS) or indirectly through the sur-
face core level shifts [3,4]. Using this, we demonstrate
explicitly how our model of the CO chemisorption energy
can account for all of the experimental data of Rodriguez
and Goodman.
Before presenting the DFT-GGA database we first dis-
cuss our simple model of the trends in CO chemisorption
energies. When an atom or a molecule is adsorbed on sim-
ple metal surfaces like Na, Mg, or Al without d states, the
electronic states of the adsorbate are broadened into reso-
nances and shifted down in energy through the interaction
with the broad continuum of metal sp states [5,6]. This is
illustrated in Fig. 1. While the adsorbate states may con-
sist of closed shells the metallic sp states have formed open
bands, which enable energy gain through hybridization of
adsorbate and metal electronic states. These metal sur-
faces, thereby, build up bonds to many adsorbates, includ-
ing CO [7]. The surfaces of transition metals and noble
metals also have open sp bands and therefore also form
bonds to adsorbates. However, the presence of the d states
in these metals enables a further bonding interaction be-
tween the metal d states and the adsorbate related states
(that are already renormalized through the interaction with
the metal sp states) [8–13] as illustrated in Fig. 1. It can
easily be shown in a tight-binding framework that the total
energy change caused by this interaction takes the form of
a hybridization energy gain and an orthogonalization en-
ergy cost. In the limit of a small overlap S between the
adsorbate states and the metal d states and of a small cou-
pling matrix element V compared to the energy separation,
FIG. 1. The self-consistent electronic density of states (DOS)
projected onto the 5s and 2p
p
orbitals of CO: in vacuum and
over Al(111) and Pt(111) surfaces. Also shown is the DOS
from the d bands in the Pt(111) surface. The sharp states of
CO in vacuum are seen to broaden into resonances and shift
down in energy over the simple metal surface (mixing with
the 4s state causes additional structure in the 5s resonance).
Over the transition metal surfaces the CO resonances further
hybridize with the metal d states. This leads to shifts in the
5s and 2p
p
levels and to antibonding 5s-d states at the top
of the d bands and bonding 2p
p
-d states at the bottom. These
states have low weight in the 5s and 2p
p
projections shown.
0031-9007y96y76(12)y2141(4)$10.00 © 1996 The American Physical Society 2141
VOLUME 76, NUMBER 12 PHYSICAL REVIEW LETTERS 18MARCH 1996
De between the two states, the hybridization energy gain
becomes proportional to V
2
yjDej while the orthogonaliza-
tion energy cost scales with SV (i.e., roughly as V
2
).
For CO, adsorption experiments [14] as well as many
theoretical studies [10,11,15,16] suggest that the filled 5s
and the doubly degenerate, empty 2p
p
electronic states
are the ones mainly responsible for the bonding to metal
surfaces. We therefore write the following simple model
expression for the d contribution to the CO chemisorption
energy over transition metal surfaces:
E
d-hyb
. 24
"
f
V
2
p
e
2p
2e
d
1fS
p
V
p
#
22
"
s1 2 fd
V
2
s
e
d
2e
5s
1s11fdS
s
V
s
#
, (1)
where 2 is for spin, f is the fractional filling of the d
bands, e
2p
and e
5s
are the positions in energy of the
(renormalized) adsorbate states, and e
d
is the center of
the metal d bands. V and S are labeled according to the
symmetry of the orbitals they describe.
The fractional filling factors, the coupling matrix ele-
ments, and the overlaps in Eq. (1) we take to be dependent
only on the atomic number of the metal atom to which the
CO bonds, i.e., independent on the environment of this
metal atom. The environment will manifest itself through
the position of the center of the d states on the metal atom
in the surface before the CO adsorption. Values for the
center, determined by DFT calculations, are included in
Table I. We approximate f with the idealized fractional
filling factor sy21dy10 where y is the valence of the
metal atom. As we will be concerned with the variation
of the E
d-hyb
from one metal to the next it suffices to es-
timate V
p
and V
s
in an LMTO (linear muffin tin orbital)
framework [17], where they factorize as products of terms
dependent only on the adsorbate and the substrate prop-
erties, respectively. This means that the present coupling
matrix elements must scale precisely as the LMTO based
V
sd
used in Ref. [13] for a different adsorbate H
2
inter-
acting with the transition metal surfaces. Introducing a
and b as adjustable parameters common to all the met-
als, we write V
2
p
. bV
2
sd
and S
p
. 2aV
p
. From the
DFT orbitals of CO and the various metals we find that
S
s
yS
p
. 1.3 is a good approximation and we therefore
write V
2
s
.s1.3d
2
bV
2
sd
and S
s
. 2aV
s
.
As the transition metal surfaces considered have very
similar half filled s bands, the renormalization of the
CO states by the delocalized metal states will be very
alike. Guided by the results of DFT calculations for
CO adsorption on Al(111) (Fig. 1) and on the transition
metal surfaces with small coupling matrix elements, we
use 12.5 and 27 eV (with respect to the Fermi level) for
the renormalized e
2p
and e
5s
positions, respectively.
In Fig. 2 we present the scaling of the chemisorption
energy for CO within the model as compared to the set
of ab initio DFT calculations given in Table I. Adjusting
only a and b in the model, we obtain the excellent
correlation in Fig. 2. a and b values of 0.063 eV
21
TABLE I. Parameters and results for CO chemisorption atop
a metal atom (first column) in metal surfaces and overlayers
[second column: M
1
yM
2
means a monolayer of M
1
on a M
2
substrate. “Ni@Cu(111)” refers to a Cu(111) substrate with
every fourth surface Cu substituted by a Ni]. The center e
d
of the local d bands at the metal atom measured relative to
the Fermi level and the fractional filling f of the these bands.
The coupling matrix element V
2
sd
(normalized to 1.0 for
Cu) and the chemisorption energy E
chem
from the DFT-
GGA calculations. The last column gives the experimental
chemisorption energies for CO on Ni(111) [27], Cu(111) [28],
Ru(0001) [29], Pd(111) [30], Ag(111) [31], Pt(111) [32], and
Al(111) [7]. All energies are in eV.
Atop Surface e
d
f
V
2
sd
E
chem
E
exp
Ni Ni(111) 21.48 0.9 1.16 21.36 21.26
Ni
NiyRu(1000)
21.27 0.9 1.16 21.51
Ni Ni@Cu(111) 21.18 0.9 1.16 21.56
Cu Cu(111) 22.67 1.0 1.00 20.62 20.52
Cu
CuyPt(111)
21.88 1.0 1.00 20.94
Cu Ni@Cu(111) 22.56 1.0 1.00 20.61
Cu Cu
3
Pt(111) 22.35 1.0 1.00 20.53
Ru Ru(0001) 21.41 0.7 3.87 21.80 21.66
Pd Pd(111) 22.16 0.9 2.78 21.30 21.47
Pd
PdyRu(0001)
22.86 0.9 2.78 20.98
Ag Ag(111) 24.28 1.0 2.26 0.09 20.28
Pt Pt(111) 22.75 0.9 3.90 21.45 21.50
Pt Cu
3
Pt(111) 22.55 0.9 3.90 21.51
Au Au(111) 23.91 1.0 3.35 20.04
Al Al(111) 20.49 20.21
and 1.5 eV
2
, respectively, have been used, both of which
are of the right order of magnitude compared to DFT
estimates [17] of 0.09 eV
21
and 2 eV
2
. We note that in
Fig. 2 the slope of the least square fitted curve guiding the
eye is close to one, which means that the adsorbate-metal
d interactions described by our model can account for the
main trends in CO bonding from one surface to the next.
The curve is offset by , 0.5 eV on the vertical axis for
E
d-hyb
0, which fits with CO forming a bond of this
strength on simple metal surfaces (see Table I). Outside
the scale in Fig. 2 is the result for COyRu(0001). Here
the simple model [Eq. (1)] estimates a d contribution of
22.08 eV to the chemisorption energy. Both the model
and the full calculation thus show that the CO-Ru bond is
very strong. However, in this case (as for most of the
metals to the left in the transition metal series) where
e
2p
2e
d
is small and the d band width is large, the
neglect of the latter in the two level model of Eq. (1) leads
to a largely overestimated E
d-hyb
.
As an important property of the model, we further note
that it captures the shifts in the CO chemisorption energy
from the single crystal surfaces to the overlayer structures.
This is apparent in Fig. 2 as seen by the dashed lines. For
Cu
3
Pt, which is a very stable covalently bonded alloy, the
Cu sites become less reactive than predicted by the model,
while the Pt sites behave roughly as expected. As the
model only describes the coupling of the metal d states
to the renormalized CO 2p
p
and 5s states we conclude
2142
VOLUME 76, NUMBER 12 PHYSICAL REVIEW LETTERS 18MARCH 1996
FIG. 2. Comparison of the model and the full DFT-GGA
chemisorption energies for a number of metal systems.
that it is this interaction which is responsible not only for
the gross trends in CO chemisorption energies over the
wide range of the late transition metal surfaces considered
but also for the details for metallic overlayers and alloy
surfaces. We return to this below.
The ab initio DFT calculations [18] are performed
using the local density approximation (LDA) [19] for
finding self-consistent charge densities and densities of
states (DOS) while using for the exchange-correlation
energy in all reported total energy differences the GGA
[20]. A quarter monolayer of CO is adsorbed on one
side of slabs having six fcc (111)—for Ru: hcp (0001)—
layers of metal atoms. Ionic cores are described with
pseudopotentials [21]. The Kohn-Sham equations are
solved in a basis of plane waves of kinetic energy up to
40 Ry (for Ni and Cu: up to 50 Ry) at 6 and 15 k points in
the C
3y
and C
2y
irreducible Brillouin zones, respectively.
For the present purposes, we choose to consider the
DFT-GGA calculations as a computer experiment of CO
adsorption on a number of metal surfaces with all ionic
degrees of freedom kept fixed. Hereby we can concen-
trate on the ability of the model [Eq. (1)] to capture the
trends caused by the electronic factors of the CO-metal
bonding. Substrate relaxations are therefore not consid-
ered, but rather the truncated bulk geometries are used
[22]. Further, for all surfaces, CO is put at the top posi-
tion with a fixed metal-carbon distance of 1.94 Å and a
CO bond length of 1.14 Å as reported from calculations
for COyPd(110) [23]. The use of fixed CO coordinates
is actually a good approximation. For CO on Pt(111)
we find that the relaxed values are r
MC
1.88 Å and
r
CO
1.15 Å and that this relaxation influences E
chem
by less than 0.05 eV. That the DFT-GGA is capa-
ble of describing the CO itself and the CO-metal inter-
action well is suggested both by previous calculations
for Pd and Cu surfaces [23,24] and by our calculations:
For CO in vacuum we get a bond strength of 10.88 eV
and a vibrational frequency of 2162 cm
21
both of which
compare well with the experimental values of 11.09 eV
and 2169.8 cm
21
[25]. For CO adsorbed atop Pt(111)
we calculate a downshift of the CO stretch frequency to
2120 cm
21
in agreement with an experimentally observed
downshift to 2104 cm
21
[26]. Finally, considering the
use of fixed site and coordinates, the chemisorption en-
ergies in our database agree well overall with the exper-
imentally determined heats of CO adsorption included in
Table I. In particular, for Ag(111) and Au(111) where the
orthogonalization terms dominate we expect an outward
relaxation of the CO to influence the chemisorption ener-
gies bringing theory and experiment in better agreement.
Our present model of the CO bonding is in complete
agreement with the theoretical interpretations developed
by Blyholder [15], Bagus [11], and others. The language
of electron donation from the CO 5s to the metal and
backdonation from the metal to the CO 2p
p
describes
the concerted action of the coupling of the CO levels to
the metal sp states and the d states. With the present
division of the donation and backdonation into separate
metal sp and d steps which follows the reasoning of
Bagus and Pacchioni [11] we obtain a simple picture and
a quantitative model of the electronic reason for the trends
in the CO chemisorption energies over metal surfaces and
overlayers.
We now return to the experimental observation by
Rodriguez and Goodman of a strong correlation between
the surface core level shift of different overlayers and
the CO chemisorption energy [2]. Our analysis goes in
two steps. First, we build on the extensive theoretical
insight into the origin of the surface core level shifts by
Weinert and Watson (WW) [3] and Hennig, Ganduglia-
Pirovano, and Scheffler (HGS) [4]. WW show that the
variation in surface core level shifts for metal overlayers
is accompanied by a similar shift in the center of gravity
of the d bands—at least towards the right in the transition
metal series, while charge transfer effects are inadequate
for explaining the shifts. The latter is confirmed by
HGS, who also show that the trends in variations in the
surface core level shifts for different overlayers are given
by the initial state shift, that is, by the changes in the
electronic structure of the unperturbed surface. From this
we conclude that we can view the variation in the surface
core level shifts as a measure of the variation in the d
band center.
The second step in our analysis is then to use our
model [Eq. (1)] to establish the relationship between vari-
ations in the d band center and the chemisorption en-
ergy. We note that in Eq. (1) the hybridization energy
term related to the 2p
p
dominates the expression. It
therefore also dominates the differential change in E
d-hyb
for a change de
d
in the position of the d-band center,
which may be caused by changes in the surroundings
of the metal atom at which the CO bonds. We have
2143
VOLUME 76, NUMBER 12 PHYSICAL REVIEW LETTERS 18MARCH 1996
FIG. 3. The experimental data of Rodriguez and Goodman [2]
for CO adsorption on Ni, Cu, and Pd overlayers plotted as the
renormalized TPD peak shifts
s
DTyg
d
vs the measured shifts in
surface core level positions. The reference systems for the CO
TPD peaks and the surface core level positions for the overlayer
structures are the respective single crystal surfaces.
dE
d-hyb
.f24fV
2
p
yse
2p
2e
d
d
2
gde
d
. This suggests that
different metal overlayers share the same direct propor-
tionality relation between de
d
and dE
d-hyb
yg, where g
is proportional to fV
2
sp
yse
2p
2e
d
d
2
, which only depends
on the overlayer metal and not on the substrate. For Cu,
Ni, and Pd overlayers considered by Rodriguez and Good-
man [2] g has values of 1, 1.76, and 3.08 relative to Cu
(cf. Table I). In Fig. 3 we examine if this analysis can
be used on the original experimental data [2]. The fig-
ure strongly supports our thesis. Identifying the shift in
the d-band center with (minus) the core level shifts and
assuming the CO chemisorption energy shift proportional
to (minus) the temperature-programmed desorption (TPD)
peak shift (DTyg), a linear relation between the renormal-
ized TPD peak shift and the core level shift is expected.
Such a relationship is clearly present in Fig. 3, however,
slightly offset from the origin. The offset might reflect a
systematic offset in the core level shifts compared to the
shifts of d-band centers.
The present work was in part financed by the Danish
Research Councils through the Center for Surface Reac-
tivity and Grant No. 9501775. Center for Atomic-scale
Materials Physics is sponsored by the Danish National
Research Foundation. JRCAT is supported by the New
Energy and Industrial Technology Development Organi-
zation (NEDO) of Japan.
[1] D. P. Woodruff and T.A. Delchar, Modern Techniques of
Surface Science (Cambridge University Press, Cambridge,
1986).
[2] J. A. Rodriguez and D. W. Goodman, Science 257, 897
(1992).
[3] M. Weinert and R.E. Watson, Phys. Rev. B 51, 17 168
(1995).
[4] D. Hennig, M.V. Ganduglia-Pirovano, and M. Scheffler,
Phys. Rev. B (to be published).
[5] D. M. Newns, Phys. Rev. 178, 1123 (1969).
[6] J. K. Nørskov, Rep. Prog. Phys. 53, 1253 (1990).
[7] T.-C. Chiang, G. Kaindl, and D.E. Eastman, Solid State
Commun. 36, 25 (1980).
[8] B. I. Lundqvist, O. Gunnarsson, H. Hjelmberg, and J. K.
Nørskov, Surf. Sci. 89, 196 (1979).
[9] S. Holloway, B. I. Lundqvist, and J. K. Nørskov, in
Proceedings of the International Congress on Catalysis,
Berlin, 1984 (Verlag Chemie, Weinheim, Germany, 1984),
Vol. 4, pp. 85–95.
[10] R. Hoffmann, Rev. Mod. Phys. 60, 601 (1988).
[11] P.S. Bagus and G. Pacchioni, Surf. Sci. 278, 427 (1992).
[12] B. Hammer and M. Scheffler, Phys. Rev. Lett. 74, 3487
(1995).
[13] B. Hammer and J. K. Nørskov, Nature (London) 376, 238
(1995); Surf. Sci. 343, 211 (1995).
[14] J.C. Campuzano, in The Chemical Physics of Solid
Surfaces and Heterogeneous Catalysis, edited by D.A.
King and D. P. Woodruff (Elsevier, New York, 1990),
Vol. 3A, p. 389.
[15] G. Blyholder, J. Phys. Chem. 68, 2772 (1964).
[16] M.A. van Daelen, Y. S. Li, J. M. Newsam, and R. A. van
Santen, Chem. Phys. Lett. 226, 100 (1994).
[17] O.K. Andersen, O. Jepsen, and D. Glötzel, in Highlights
of Condensed Matter Theory, Proceedings of the Interna-
tional School of Physics “Enrico Fermi,” Course LXXXIX
(North-Holland, Amsterdam, 1985), p. 59; J. K. Nørskov,
J. Chem. Phys. 90, 7461 (1989).
[18] M.C. Payne, M.P. Teter, D.C. Allan, T.A. Arias, and
J. D. Joannopoulos, Rev. Mod. Phys. 64, 1045 (1992).
[19] D.M. Ceperley and B. J. Alder, Phys. Rev. Lett. 45, 566
(1980); J. P. Perdew and A. Zunger, Phys. Rev. B 23, 5048
(1981).
[20] J.P. Perdew, J. A. Chevary, S.H. Vosko, K. A. Jackson,
M. R. Pederson, D. J. Singh, and C. Fiolhais, Phys. Rev. B
46, 6671 (1992).
[21] G.B. Bachelet, D.R. Hamann, and M. Schlüter, Phys.
Rev. B 26, 4199 (1982); N. Troullier and J. L. Martins,
ibid. 43, 1993 (1991).
[22] When allowing for relaxations of the uppermost two layers
of the clean CuyPt(111) surface the DFT-GGA values for
the CO chemisorption change by less than 0.01 eV.
[23] P. Hu, D. A. King, S. Crampin, M.-H. Lee, and M. C.
Payne, Chem. Phys. Lett. 230, 501 (1994).
[24] P.H.T. Philipsen, G. te Velde, and E. J. Baerends, Chem.
Phys. Lett. 226, 583 (1994).
[25] K.P. Huber and G. Herzberg, Constants of Diatomic
Molecules, Molecular Spectra and Molecular Structure
Vol. IV (Van Nostrand Reinhold, New York, 1979).
[26] E. Schweizer, B. N.J. Persson, M. Tüshaus, D. Hoge, and
A. M. Bradshaw, Surf. Sci. 213, 49 (1989).
[27] J.B. Miller, H.R. Siddiqui, S.M. Gates, J. N. Russell, Jr.,
and J. T. Yates, Jr., J. Chem. Phys. 87, 6725 (1987).
[28] P. Hollins and J. Pritchard, Surf. Sci. 89, 486 (1979).
[29] H. Pfnür, P. Feulner, H.A. Engelhardt, and D. Menzel,
Chem. Phys. Lett. 59, 481 (1978).
[30] H. Conrad, G. Ertl, J. Koch, and E. E. Latta, Surf. Sci. 43,
462 (1974).
[31] G. McElhiney, H. Papp, and J. Pritchard, Surf. Sci. 54,
617 (1976).
[32] H. Steininger, S. Lehwald, and H. Ibach, Surf. Sci. 123,
264 (1982).
2144
