s u r f a c e science

ELSEVIER Surface Science 343 (1995) 211-220

Electronic factors determining the reactivity of metal surfaces

B. H a m m e r a.b.,, J.K. N o r s k o v ~
a Center for Atomic-Scale Materials Physics and Physics Department, Technical University ofDenmark, DK-2800 Lyngby, Denmark
b Joint Research Centerfor Atom Technology, 1-1-4 Higashi, Tsukuba, lbaraki 305, Japan
Received 8 Juno 1995; accepted for publication 21 August 1995

Abstract

Based on density functional theory calculations of H2 dissociation on AI( 111 ), Cu( I I 1), Pt(l 11) and CusPt(l 11) we present a
consistent picture of some key physical properties determining the reactivity of metal and alloy surfaces. The four metal surfaces are
chosen to represent metals with no d-bands, with filled d-bands and with d-states at the Fermi level. We show that electronic states
in the entire valence band of the metal surface are responsible for the reactivity, which consequently cannot be understood solely in
terms of the density of states at the Fermi level nor in terms of the empty d-states above it. Rather we suggest that trends in
reaetivities can be understood in terms of the hybridization energy between the bonding and anti-bonding adsorbate states and the
metal d-bands (when present), and we demonstrate that a simple frozen potential based estimate of the hybridization energy correlates
well with the calculated variation of the barrier height for the different metal surfaces.

Keyword~: Alloys; Chemisorption; Copper; Density functional calculations; Hydrogen; Models of surface chemical reactions; Platinum;
Surface chemical reaction

1. Introduction [5,6] have recently been developed [7], which

give the two simple measures, LDOS(EF) and Nh,
It is one of the long term goals of surface science as limiting cases of a more general dependence of
to establish an understanding of the important the reactivity on the surface local density of
physical properties of a surface determining its states [ 8].
chemical reactivity. Such an understanding could In the present paper, we present calculations of
be the basis for an "atomistic engineering" of the energy along the reaction path for a set of
surfaces with desired catalytic activity or selectivity. systems selected to test out these concepts. The
Two simple properties of the surface which have calculations are based on the density functional
been successfully correlated with its reactivity are theory (DFT) total energy method. We choose the
the local density of one-electron states at thc Fermi dissociation of H2 as a test reaction and consider
level LDOS(EF) [1-3] and the number of holes four metal surfaces AI( 111 ), Cu( 111 ), Pt( 111 ) and
Nh in the d-bands [4]. More general formulations Cu3Pt( 111 ) chosen to illustrate metal surfaces with
of surface reactivity measures based on concepts and without d-bands, with l~rge and small Fermi
from the theory of gas phase chemical reactivity level surface densky of states LDOS(EF) and with
and without d-holes. We conclude that neither of
* Corresponding author. Fax: +,15 45 93 23 99; E-mail! the two simple reactivity measures are sufficient to
hammer@fysik.dm.dk. understand all the systems. The ~-'a3Pt(ll 1) surface

0039-6028/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved

SSDI 0039-6028 (95)00000-0
212 R Hammer and J. K Norskov/Surface Science 343 (1995) 211-220

which we find to have essentially no d-holes just paper, we use the DFT to describe H2 dissociation
above the Fermi level and a low LDOS(EF) is on a number of metal surfaces, which we choose
found to be as reactive as the P t ( l l l ) surface, to describe in a (repeated) slab geometry. An
where there are both d-holes and a large obvious advantage of this choice is that the metal
LDOS(EF). Instead we discuss our results in terms states are inherently extended Bloch states. The
of the d-hybridization picture originally proposed main drawback of the slab geometry is that a
in Refs. [9] and [10] and recently explicitly complete array of molecules rather than just one
demonstrated to be of relevance in studies o[ H2 single H2 is reacting with the surface. The error
dissociation on NiAl(110) and on the late trans- introduced by this can, however, be controlled by
ition and noble metal surfaces [11,12]. We show systematically increasing the surface cell size.
that the d-hybridization picture which associates The elemental metals AI, Cu, and Pt have fcc
the trends in reactivity with the extra hybridizction structure, while Cu3Pt has the fee-like Cu3Au struc-
between the metal d-states and adsorbate states ture. The CusPt(111) surface is known experimen-
that are renormalized due to the interaction with tally to show a (2 × 2) LEED pattern indicative of
the metal sp states can be used to understand all the surface being bulk stoichiometric with 25% Pt
the effects observed. We establish this by studying and 75% Cu in the first surface layer [21]. In the
in detail the differences of the densities of states of present theoretical work, the surfaces are all real-
the molecule outside the surfaces. We show that to ized with super cells having a (2 × 2) surface unit
understand the trends in the reactivities of all the cell, comprising six (111) layers of substrate atoms
surfaces it is important to include the interaction and five "empty" layers (i.e. > 10 ,~) of vacuum.
between the metal d-states and both the bonding Considering the small relaxations observed for the
and the anti-bonding molecular states. In particu- close-packed (111) fcc metal surfaces [22] we
lar, we show that whether or not a barrier for the choose to use the truncated bulk positions neglect-
dissociation exists is determined by the relative
ing the relaxations. H2 is adsorbed on one slab
position in energy of the renormalized adsorbate
surface and the difference in workfunctions of the
states and metal d.bands, the molecule~surface s - d
two slab surfaces is compensated for by a dipole
coupling matrix element, and the degree of filling
layer in the middle of the vacuum [23]. Using the
of the molecule=surface anti-bonding states. Based
methods of Payne, Teter and co-workers [24] and
on the effective medium theory [ 13] we formulate
an approximate quantitative measure of the role of Gillan [25], the Kohn-Sham equations are
of the d-hybridization on the barrier height and solved in a basis of plane waves of kinetic energy
we demonstrate how this measure correlates with up to 50 Ry at 6 or 15 Bloch wave vectors (k-
the barrier heights found in the full, self-consistent points) in the C3v and C,v irreducible Brillouin
calculations. zones respectively. The vectors are chosen accord-
ing to the method of Chadi and Cohen [26] and
correspond to a total of 54 vectors in the k:--0
plane of the entire first Brillouin zone. Fermi
2. Method distributed occupation numbers at kT,-0.1 eV are
used to stabilize the iterative procedures and all
Density functional theory with the local density total energies are extrapolated to 0 K [25]. The
approximation (LDA) augmented by non-local proton is described with the Coulomb potential,
terms (GGA) for the description of exchange- Al with a Bachelet-Hamann-SchltRer [27]
correlation effects [ 14,15] has been shown recently pseudopotential, while for Cu and Pt the soft,
to describe adsorption and dissociation of simple scalar-relativistic pseudopotentials of Troullier and
molecules on metal surfaces semi-quantitatively Martins [28 ] are used. The charge density and the
[ 16-19]. For H2 dissociation over Cu(l 11), for density of states are found self-consistently within
instance, the experimentally determined barrier is the LDA, as are the theoretical lattice constants,
reproduced to within 0.1 eV [ 16,20]. In the present aAi = 3.96 ~, ao. = 3.57 ~, apt = 3.93 ~,, and acu3pt=
 R Hammer and J. K Norskov/Surface Science 343 (1995) 211-220 213

3.68 ,g, used throughout this work. All total energies

reported are GGA values derived from the LDA
densities.

1.5
A
3. Reaction energetics

Fig. 1A summarizes the calculated energetics of 1.0

H2 outside the (111) surfaces of AI, Cu, Pt, and
Cu3Pt. In order to facilitate the qualitative and ~2
0.5
quantitative discussion of the trends in the He q2
dissociation over the different metal surfaces, we
choose to study the H, in the same reaction 0.0
geometry and for one fixed set of (b, Z) coordinates
over all the surfaces. The atop dissociation with
the two hydrogen atoms going into the third -0.5
nearest neighbor three-fold sites is chosen as this
simplifies the dis,:ussion of the molecule--surface
interaction, by mainly involving one surface atom -1.0
3.0 2.5 2.o i.5 ].0 0.5
(cf. Figs. IB-1D).
z(~,)
A large barrier is seen to build up towards
dissociation over the AI( 111 ) and Cu( 111 ) surfaces Fig. 1. (A) The potential energy along the unrelaxed atop
[16] and over the Cu sites in Cu3Pt(lll). The reaction path. The chosen sets of H-H bond length, b, and H2
P t ( l l l ) and the Pt sites in Cu3Pt(lll), on the height, Z, above the plane of the surface atomic position are:
(b, Z)=(0.78 A, 3.0 A), (0.8 A, 2.0 A), (0.9 A, 1.75 A), (1.2 A,
other hand, dissociate the Hz without a barrier. 1.5 A), (2.0 A, 1.25 A), (2x~a, 0.9/~), (2x/~a, 0.75 A), where a
For the Cu(l 11) surface where the adsorption is is the theoretical lattice parameter of AI, Cu, Pt and Cu3Pt
activated, we have checked the influence of the respectively. The energy zero is the sum of the total energies of
chosen reaction path by also allowing the molecule the separate super cell calculations of the clean slab and the
to lind the minimum energy path [16]. This is isolated He (b-0.78 A) systems respectively. For Cu, the dotted
line shows the energy along the path where the mdecule has
shown by the dotted line in Fig. 1A. While the been allowed to move parallel to the surface and to rotate to
barrier is lower than for the fixed path, the inclu- find the minimum energy path which turns out to be ibr the
sion of relaxations does not alter the trends seen molecule dissociating perpendicular to the Cu=Cu bridge into
for the unrelaxed calculations [29]. For the non- nearest neighbor three-fold sites [16"1. (B) The geometry over
AI(111), Cu(l 11) and Pt(111). Large open circles represent the
activated cases even lower energy paths might
substrate atoms, small solid circles (@) the H atoms in the Hz
exist, but this would again not change the trends. molecule with b= 1.2 A. The (2 x 2) cell i,q indicated. (C) and
We note that all features of the calculated poten- (D) The geometry for He over the Cu (light gray) and Pt (dark
tials are in agreement with experiment. Cu and AI gray) sites in Cu3Pt(ll I ).
surfaces show a large barrier for dissociation
1"30,31] while Hz dissociation occurs spontane- of states (DOS) projected on the d-orbitals of the
ously on Pt and Cu3Pt surfaces [21,32]. Cu(111) and the Pt(111) surface atoms and the Pt
and Cu atoms in the CihPt(111) surfsce. It can be
seen that the d-bands of C u 3 P t represent some
4. Electronic factors intermediate between those of pure Cu and pure
Pt; the main qualitative difference being that for
As a starting point for our discussion of the the alloy, the Fermi level intersects right at the top
surface properties determining the reactivity of a of the d-bands. The d-bands of Cu3Pt are therefore
metal surface, consider in Fig. 2 the local density practically filled (i.e. there are essentially no
214 R Hammer and J.K. Norskov/ Surface Science 343 (1995) 211-220

Cu(l I 1) Cu3Pt(I 11) It(Ill) can "freeze" the density and potential of the rest
Cu site Pt site
5 of space and let it be independent of the presence
of the adsorbate. In that case it can be shown that
0 the difference in interaction energy of the adsorbate
in the two surroundings can be written [ 10,33]
-5
~Ead s - ' ~ enads(8 ) d8 -I- ~Ees, ( 1)
-10 • A ° • . . D"
where the first term is the difference in the sum of
Projected DOS (arb. units)
one electron energies associated with the adsorbate
Fig. 2. The density of states (DOS) for (A) Cu, (B-C) CuaPt and the second is the difference in the electrostatic
and (D) Pt. In (A) and (B) the solid curves are the DOS energy of the adsorbate in the two surroundings.
projected onto the five d-orbitals of atomic Cu positioned in Due to the variational principle the errors in this
the surface, while onto the d-orbitals of atomic Pt in (C) and
(D). The total DOS (on a different scale) for bulk Cu, Pt, and
energy difference are second order in the errors
Cu3Pt are shown by the dotted curves in the respective figures. that the "freezing" has produced in the density
No truncation of the atomic orbitals is used hence also some s and potential ['13,34]. If we are comparing two
character on neighboring atomic sites is measured by the solid metals that are not too different this should be a
curves and causes the background outside the energy range of reasonable procedure.
the d bands.
Eq. (1) shows that in cases where there is no
large charge transfer and the electrostatic energy
difference is of a minor importance, the changes in
d-holes!) and the DOS at the Fermi level is low the barrier can be estimated from the one electron
compared to what it is for Pt. These properties energy sum. In the following we will use this
render Pt in Cu3Pt(lll) the ideal test case for principle in two ways. First, we shall use it in
reactivity theories. It is immediately clear that conjunction with the self-consistently calculated
simple indicators of reactivity like large Fermi level one electron spectra to gain a qualitative under-
DOS or localized d-holes just above the Fermi standing of the differences between the metals.
level cannot explain the chemistry of the Pt sites Later we will also apply it in a more quantitative
in Cu3Pt(111). analysis using the frozen potential approximation
To relate the barrier for dissociation to the directly.
electronic structure of the unperturbed surface it We apply Eq. (1) to build a qualitative picture
is very tempting to build the theory on the changes of the differences in activation barriers for dissoci-
in the Kohn-Sham one electron levels of density ation in the following way. We consider as a
functional theory. The problem is that the total common starting point an Hz molecule at a fixed
energy entering into the determination of the bar- position (corresponding e.g. to the top of the
rier is not related in a simple way to the sum of barrier over the Cu(111) surface) interacting with
the one electron energies, A possible solution has a metal surface with only s and p electrons (AI or
been devised b a s h on the effective medium theory. jellium). This interaction gives rise to a shift in the
If we are not interested in the absolute magnitude Hz bonding (%) and anti-bonding (¢*) states as
of the barrier but only the variation from one illustrated in Figs. 3A-3B 1"9,35]. The new % and
system to the next, we can use the variational ¢* states show up as resonances in agreement with
properties of the total energy functional to arrive the Newns-Anderson "weak chemisorption" pic-
at a very simple expression for total energy differ- ture [38]. The anti-bonding state ends up just
encos. If we compare the energy of an adsorbate above the Fermi level while the bonding state is
in two different surroundings we can "freeze" the about 7 eV below. This is for instance illustrated
electron density and one-electron potential close by the self-consistent results for the DOS at the
to the adsorbate and use the same density and transition state in Fig. 4. The energy curve for the
potential in the two different systems. Likewise, we AI(I 11) surface in Fig. 1 shows that such a free
 R Hammer and J.K. Norskov/Surface Science 343 (1995) 211-220 215

Schematic (renormalized) molecular levels and the center of

HJjellium HdCu(I ! i ) Cu( 111
a 2
the d-bands. As was demonstrated in Ref. [ 11 ] by
projections onto multi-center orbita]s, these DOS
... 0 features represent orbitals of molecule-surface
>,
bondi.,:g and anti-bonding character• The inter-
m~ -5 action between the molecular resonance and the
metal d-bands reflects the simplest two level inter-
-10 D" action problem of basic chemistry, as indicated
schematically by the lines between the panels in
Projected DOS (arb. units) Fig. 3. This is a consequence of the narrow band
width of the metal d-bands and can be viewed as
Fig. 3. Schematic drawing of the interaction between the
molecular levels and the metallic states. The Ha bonding, 0.t, the strong coupling limit of the Newns-Anderson
and anti-bonding, 0.*, level (A) are broadened into resonances model [38]. Fig. 1 shows that the effect of includ-
and shifted down due to the interaction with the metallic s ,and ing the coupling to the d-states is to lower the
p states (B). Interaction with the metal d-states (D) further barrier in all cases.
splits the % and 0.* projected DOS in molecule-surface bonding
The interaction between the ° adserbate states
and anti-bonding contributions (C). The % (0.*) projected DOS
is shown with solid (dotted) curves and the occupied part i~ and the metal d-bands gives rise to a down-shift
marked with light (dark) gray shading. of the lowest of the two states and an up-shift
of the highest state which for the parameters
electron like surface gives rise to a considerable o[ interest he e i~ given by the perturbation
barrier for dissociation [35-37]. expression:
For the other metals we now have to include
V2
the effect of the metal d-states, and according to At~ . . . . . V<<[~d- e~[. (2)
Eq. (1) we can estimate the extra energy contribu.-
tion from this hybridization just from the one-
electron energy differences. Fig. 3B-D shows the The effect of this interaction is then determined by
effect of introducing the d-states outside Cu( 111 ). three factors:
For both % and a*, the molecule-metal d inter- (i) The position of the mole::ular bonding and
action results in the formation of two DOS features anti-bonding state~: relative to the d-bands
- one below and one above the center of the (1~:,,-%1 and ]~:d--r~.,!;I),
AI(III) Cu(Ill) Cu~Pt(l 11 )
Cu site Pt site
Pt(ll I)

5E ~ (°:-d)"1 anti-
L,-":" ....

%
.. i
0
1| ) q,----

t....

-10
":N" V o.,
I ' . . lC2 ,,,.ll..ll,l . b . l l h .

Projected DOS (art). units)

Fig. 4. The density of states (DOS) for H2 over the atop sites of CA) AI(II1), (B) Cu(lll), (C) Cu in Cu3Pt(IIlL° (D) Pt in
Cu3Pt(111), and (E) Pt(111). A fixed configuration of the He, with an H-H separation of 1.2 A, and an H2 height of 1.5 A above the
plane of the surface atoms, is used. The DOS is projected onto the 1-12 bonding, 0"8, (solid curves) and anti-bonding, 0"*, (dotted
curves) linear combinations of the atomic H Is orbitals (truncated at 2 bohr). The arrows inside panels (B) through (E) indicate the
0"fd anti-bonding states. The nature of the wavefunction at the differ,nt peaks in the DOS is depicted by the schematics at the right
of the figure (the o"s - represented by two spheres - are shown atop the d's with the sign of the wavefunction indicated by the shading).
216 R Hammer and J. K Norskov / Surface Science 343 (1995) 211-220

(ii) The coupling matrix element V between the All in all the Pt sites show a considerably smaller
molecular orbitals and the metal d-states. (negative) barrier than the Cu sites.
(iii) The filling, f, of the molecule-surface anti- Clearly, in this picture there is only a small
bonding states given by the position of the Fermi difference between e.g. the Pt sites in CuaPt(lll)
level, Ev. and in Pt(111) even though there is a large differ-
These three factors can systematically explain ence in the local state densities just around the
the behavior of the total energies given in Fig. 1. Fermi level for the two sites. The reason is that
For the Cu example in Fig. 3, the o*-d interaction the molecule perturbs the surface so strongly (V is
must be attractive, since the anti-bonding state is of the order eV) that it is the availability of states
almost empty. The old interaction on the other in a window of several eV around the Fermi level
hand must be repulsive because here the anti- that is important for the interactions.
bonding state is filled causing a Pauli type repulsive
energy term arising from the orthogonalization of
the overlapping molecular and metal states. The
5. Approximate reactivity measure
total energy calculations shown in Fig. 1 show that
the net result of the molecule-metal d interaction
We can make the arguments more quantitative
is lowering the energy barrier over Cu(l 11) relative
by introducing as an approximate reactivity mea-
to over AI( 111 ).
sure, ~Et,, the energy difference due to the coupling
Moving now from Cu(l I 1) to the other metal
to the d-bands estimated directly in the frozen
surfaces we have to consider the changes in the
potential approximation. For the molecule at the
interaction between the % and o* and the d-states
transition state we write:
separately. The o* interaction is always attractive,
since the anti-bonding ~r*-d state will always be Vz Vz
above the Fermi level. The size of this interaction ~Et,= - 2 2(I - f ) ~ + ~Vz. (3)
must bv about the same outside Cu(111) and the
Cu sites in Cu3Pt(l 11) since both the coupling Here the first term describes the energy gain due
strength and the position of the d-bands is about to the hybridization between the o'* and the d-
the same. Outside the Pt site in the alloy and the states (the factor of 2 is for spin). The second term
P t ( l l l ) surface, the center of the d-bands is not is the corresponding old interaction. Here there
very different from the case of Cu (cf. Fig. 2), but is an extra factor of (1 - f ) because the degree of
the coupling matrix element must be considerably filling f of the anti-bonding state varies from one
larger because the 5d states are much more metal to the next. The last term is the repulsion
extended than are the 3d states. This leads to a due to the orthogonalization of both o* and os
larger attraction. Fig. 4 supports this view. The states to the metal d-states. In both cases the
figure shows the DOS for the molecule held at orthogonalization energy will be proportional to
1.5 A above the respective surfaces and sites with ItS where S is the overlap matrix element. In Eq.
an H-H separation of 1.2 A, corresponding approx- (3) we have assumed that S is proportional to V
imately; to the transition state geometry over pure and lumped the proportionality constant into the
Cu(lll). Clearly the shifts in the bonding and factor 0c. The present reactivity measure, ~Ets is
anti-bonding states are considerably larger outside clearly approximate and can certainly be refined.
Pt atoms than outside Cu atoms. However, the basic idea of Eq. (3), to explicitly
The o s interaction also changes considerably as include the direct covalent interaction between the
the surface changes. The anti-bonding %-d peak renormali~,ed molecular states and the metal d
(indicated by the arrows in Fig. 4) just above the states is what makes this reactivity measure power-
metal d-bands ~]~ieh is filled for Cu is shifted ful compared to reactivity theories that only rely
gradually above tile Fermi level as the d-bands are on the propert;es of the unperturbed metal surfaces
shifted up for the Cu3Pt and the Pt surfaces. The [ 1-7]. Hoffmann has also pointed out the impor-
old interaction thus changes sign from Cu to Pt. tance of the interaction between the molecular
 R Hammer and Z K Norsko v~Surface Science 343 (1995) 211-220 217

states and the metal d states [8]. One important all, we take ~%= - 7 eV and ~,. = 1 eV relative to
difference between his approach and the present is the Fermi level for all the metals. Tile ~ appearing
that rather than using the free molecular states we in the orthogoaalization term, we use as a single
use the ones that have been renormalized to fitting parameter independent of the metal consid-
stronger binding energies through the interaction ered. Finally, but most importantly, we calculate
with the metal sp states [9,10]. This enables us to the matrix dement V within the LMTO-ASA
obtain a semi-quantitative, yet simple, estimate of approximation as [41 ]:
the role of the d-states for the reaction energetics.
In Eq. (3) we assume that the renormalized o'* MHMd
level has its center above the Fermi level. As the V=r/ ra (5)
dissociation completes, 8,, and e,. both converge
to the position oI the H adsorbate resonance, ~u, where Mn and Md are given by the potential
well below the Fermi level. The frozen potential around the hydrogen and metal atoms respectively,
estimate of the chemisorption energy difference, r is the H-metal separation and r/ is a constant
~Echm, for one H due to the coupling to the d- independent of the metal [42]. In the frozen poten-
bands therefore takes the form: tial approximation Md is independent of the pres-
ence of the hydrogen and MH does not depend on
V2 the metal. Mu is given by the independent estimate
6Echem '~ - 2 ( 1 --f) ~ " t " I~¢hemV2. (4) in Ref. [41] and the Md's (or the potential parame-
8d--~ H
ters) are tabulated for all the metals in Ref. [34].
The ( l - f ) dependence of the first term in this With all the terms in Eq. (3) thus accounted for,
expression is able to account for the main trend in we can now get the scaling of JEt~ from one metal
the ehemisorption energies of adsorbates like H all to the next. In Table 1 we show the three terms in
through the transition metal series [39]. The Eq. (3) and in Fig. 5 we plot the calculated barriers
second term in Eq. (4) is the only one active for for adsorption from Fig. 1 as a function of the new
the noble metals. Gold is the noble metal with the reactivity measure ¢~Ets. Clearly there is an excellent
largest 1/2 and will therefo~e according to t~q. (4) correlation between the two. In the table and the
have the largest adsorbate s-metal d repulsion. figure we have included results for Ni(lll),
This explains why gold in terms of its surface NiAI(110), and Au(l 11) for completeness
reactivity - is the noblest metal [ 12]. [11,12,43].
In order to use Eq. (3) in the discussion of the Having shown the strong correlation between
relative reactivity of the metals, we need to evaluate the calculated energy barriers and tile here pro-
8d, 8,,., ~%, f, 0e and V2 for the different systems. posed reactivity measure, JEt,, we are now in
The center ot the d-bands, ~d, we readily take from position to discuss - on a quantitative level -
Fig. 2, while the filling, f, of the anti-bonding %--d which electronic factors are the most important in
state we approximate by the local filling of the determining the reactivity of metal surface. Table 1
surface d-states [40]. The positions, ~% and ~,,, of reveals that among the two attractive terms,
the bonding and anti-bonding H2 resonances after --2(V2/~,.--Sd) and --2(1--f)(V2/~d--e%), the
hybridization with the metal sp electrons, we expect former is by far the most prominent; i.e. we identify
to vary only little from metal to metal because the the a*u-d interaction as the dominant attractive
s-s coupling matrix element varies little for the term, which can be traced to the fact that a* is
metals considered [34]. We further neglect any initially empty. Another important observation can
influence of the workfunction change from one be made in Table 1. This is that due to the similar
metal to the next on the renormalized H2 levels, scaling with V of the attractive --2(VZ/t;cr~*--t~d)
and measure the positions relative to the Fermi term and the repulsive 0cV2 term which represents
level. For the transition state confi~,uration consid- the orthogonalization of both occupied and unoc-
ered this is a reasonable approximation due to the cupied adsorbate states with the metal d states, the
close proximity of the H, and the surfaces. All in two terms largely cancel each other. On Au, where
218 B. Hammer and J. K. Norskov / Surface Sck, nce 3,43 (1995) 211-220

Table 1
The center of the d-bands, ~a, measured relative to the Fermi level; the coupling strength, V2, and the three contributions to the
activation energy difference lEts in Eq. 13); all energies are in eV; the position of the anti-bonding H2 resonance after hybridization
with the metal sp electrons, ~,,., is assumed 1 eV above the Fermi level for all the metals; likewise the position of the bonding
resonance, e,,,, is assumed 7 eV below the Fermi level; the filling, f, of the anti-bonding ag--d state is taken to be 1 over Cu and Au
sites and to be 0.9 over Pt and Ni sites (corresponding to the degree of filling of the d-states in all cases) [40]; the constant a is used
as a fitting parameter and takes upon the value of 0.38

Metal ~a V" V2 V2 ~V2 lets

2 - - -2(1-/)

Cu - 2.67 2.42 - 1.32 0 1.02 -0.30

Cu: CuaPt - 2.35 2,42 - 1.44 0 1.02 - 0.42
Pt : CuaPt - 2.55 9.44 - 5.32 -0.42 3.96 -- 1.78
Pt - 2.75 9.44 - 5.03 - 0.44 3.96 - 1.51
Ni - 1.48 2.81 -2.27 -0.10 1.18 - 1.19
Ni. NiA! - 1.91 2.81 - 1.93 -0.11 1.18 -0.86
Au -3.91 8.10 -3.30 0 3.40 0.10

1.5

> 1.0 Cu:Cu3Pt

o

Ni:NiAI / o
~ 0.5 o Cu
<

i 0.0
Pt:Cu3Pt/_ 0 Ni
Pt
-0.5
-2.0 -I.5 -1.0 -0.5 0.0
BE,,,,
(eV)
Fig, 5, Correlation between the reactivity measure 6Et, and the calculated activation barrier for adsorption (the total energy at (b,
Z)=( !,2 A, 1.5 A) inFig, 1 [43]). :c in Eq, (3) has been used as a fitting parameter.

the I¢a-~.l is larger than on any other metal tive to the repulsive term and the two therefore
surface considered, the attractive term. through almost completely cancel each other leaving the
large in absolute numbers, is therefore small rela- Au surface with a large barrier despite the presence
 I3. Hammer mid J.K. Norskov/Surface Science 343 (1995) 21 !-220 219

of d-states actively involved in the chemical [4] J. Harris and S. Andersson, Phys. Rev. Lett. 55 (1985)
1583.
reaction.
[5] K. Fukui, T. Yonezawa and H. Shingu, J. Chem. Phys.
2011952) 722.
[6] R.G. Parr and W. Yang, Density Flmctional Theory of
6. Conclusions Atoms and Molecules tOxford Univ. Press, New York,
1989).
[7] M.H. Cohen, M . V . Ganduglia-Pirovano and
We conclude that to understand the trends in J. Kudrnovsky, Phys. Rev. Lett. 72 (1994) 3222;
reactivity of H2 outside transition and noble metals M.H. Cohen, M . V . Gandu~flia-Pirovano and
and their alloys, one has to consider the interaction J. Kudrnovsky, J. Chem. Phys. 101 1994) 8988.
between the metal d-states and the bonding and [8] R. Hoffmann, Rev. Mod. Phys. 60 (1988) 601.
anti-bonding molecular states after they have been [9] B.I. Lundqvist, O. Gunnarsson, H. Hjeimberg and J.K.
Norskov, Surf. Sci. 89 (i979) 196.
shifted due to the interaction with the free-electron- [10] S. Holloway, B.I. Lundqvist and J.K. Norskov, Proc. Int.
like metallic electrons. It then turns out that the Congress on Catalysis, Berlin, Vol. 4 (1984), p. 85;
important surface parameters determining the reac- J.K. Norskov, Rep. Prog. Phys. 53 11990) 1253.
tivity are the position of the whole d-bands relative [11] B. Hammer and M. Schemer, Phys. Rev. Left. 74
{ 1995) 3487.
to the Fermi level, because this determines both
[12] B. Hammer and J.K. Norskov, Nature 376 (1995) 238.
the size of the bonding and anti-bonding energy [133 K.W. Jacobsen, J.K. Norskov and M.J. Puska, Phys.
shifts and the degree of filling of the anti-bonding Rev. B 35 (1987) 7423.
states as well as the size of the coupling matrix [14] D.M. Ceperley and B.J. Alder, Phys. Rev. Lett 45 11980)
elements which also enter into the size of the 566;
energy shifts. We expect that this picture can be J.P. Perdew and A. Zunger, Phys. Rev. B 23 11981 ) 5048.
[15] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson,
transferred directly to other adsorbing molecules M.R. Pederson, D.J. Singh and C. Fiolhais, Plays. Rev. B
and surfaces, but more calculations are required 46 {1992) 6671.
to test this out in detail. [16] B. Hammer, M. Scheffler, K.W. Jacobsen and J.K.
Norskov, Phys. Rev. Lett. 73 11994) 141)0.
[t7] J.A. White, D.M. Bird, M.C. Payne and I. Stich, Phys.
Rev. Lett. 7311994) 1404.
Acknowledgements [18] P.H.T. Philipsen, G. te Veide and E.J. Baerends, Chem.
Phys. Lett. 226 (1994) 583.
These calculations have only been possible due [19] P. Hu, D.A. King, S. Crampin, M.-H. Lee and M.('.
Payne, Cheln. Phys. I,eit. 231)(1994) 501.
to the computational resources of the JRCAT
[20] A. Gross, B. Hammer, M. ScheMer and W Brenig, Phys.
supercomputmg system, which is supported by the Rev. Lett. 73 11994) 3121.
New Energy and Industrial Technology [21] R. Linke, U. Schneider, H. Busse, C. Becket, U. Schri~der,
Development Organization (NEDO) of Japan. G.R. Castro and K. Wandelt. Surf. Sci. 307309 (1994)
Center for Atomic-scale Materials Physics 407.
(CAMP) is sponsored by the Danish National [22] F. Jona and P.M. Marcus, in: The Structure of Surfaces
II, Eds. J.F. van der Veen and M.A. Van ttove (Springer,
Research Foundation. This work was supported in Berlin, 1988) p. 90.
part by the Danish Research Councils through the [23] J. Neugebauer and M. Schefl]er, Phys. Rev. B 46
Center for Surface Reactivity. (1992) 16067.
[24] M.C. Payne, M.P. Teter, D.C. Allan, T.A. Arias and J.D.
Joannopoulos, Rev. Mod. Phys. 64 (1992) 1045.
[25] M.J. Gillan, J. Phys.: Cortdens. Matter I (1989) 689.
References [26] D.J. Chadi and M.L Cohen, Phys. Rev. B 8 (1973) Y747:
S.L. Cunningham, Phys. Rev. B I0 (1974) 4988.
[1] P.J. Feibelman and D. Hamann, Phys. Rev. Lett. 52 [27] G.B. Bachelet, D.R. Hamann and M. Schi0ter, Phys. Rev.
(1984) 61; Surf. Sci. 149 (1985) 48. B 26 (1982) 4199.
[2] L.M. Falicov and G.A. Somorjai, Proc. Nat. Acad. Sci. [28] N. Trouilier and J.L. Martins, Phys. Rev. B 43 11991)
USA 82 (1985) 2207. 1993; For the Pt s and d potentials r:=2.26 bohr and a
[3] W. Yang and R.G. Parr, Proe. Nat. Acad. Sci. USA 82 6st5d 9 atomic configuration are used, while for the p
11985) 6723. potential r~ = 2.50 bohr and 6s°n56p°255d 8 are used.
220 R Hammer and £ &: Norskov / Surface Science 343 (1995) 211-220

[29] Unlike the barrier height, we expect the heats of [391 P. Nordlander, S. Holloway and J.K. Norskov, Surf. Sci.
dissociative H, chemisorption (i.e. the minima in 136 (1984) 59.
Fig. IA) to be quite sensitive to relaxations in the [40] Note here the difference between the filling of the
H coordinates parallel to the surface - in particular for d-bands and the local filling of the d-states. While the
Al(l 11) and CuaPt. For H on Al(110), for instance the d-bands of Cu3Pt are filled, we find a filling of the
lowest energy site is on top and the chemisorption energy d-projected density of states of 0,94, close to the value
is +0.1 eV [361 rather than +0.5 eV for the tht'eehold for the Pt surface of 0.92. For simplicity we therefore use
site in Fig. 1A. a value of 0.9 (the "ideal" filling) for all Pt (and Ni) sites.
[3o] H.A. Michelsen, C.T. Rettner and D.J. Auerbach, Phys. [41] J.K. Nerskov, J. Chem. Phys. 90 (1989) 7461.
Rev. Lett. 69 (1992) 2678; [42] q contains the geometry and the anguiar momentum
H.A. Michelsen, C.T. Rettner, DJ. Auerbach and R.N. dependent Slater-Koster terms between the two H IS
Zare, J. Chem, Phys. 98 (1993) 8294. orbitals and the relevant d-orbitals. Composing the trg
[31] H.F. Berger and K.D. Rendulic, Surf. Sci. 253 ( 1991 ) 325. and o'* as LCAOs of the H IS we obtain:
[32] K. Chfistmann, Surf. Sci. Rep. 9 (1988) 1;
J.K. Brown, A. Luntz and P.A. Schuitz, J. Chem. Phys. [ 2 :~(l'-.?)... t=z_, b/2
95 ( 1991 ) 3767.
[33] For a detailed derivation see e.g.: J.K. Norskov, in:
V/- r r

Physics and Chemistry of Alkali Metal Adsorption, Eds.

H.P. Bonzel, A,M. Bradshaw and G. Ertl (Elsevier, ~ 1 - 3n 2 b/2
%-d.'-= " 2,,/3, n=---,
Amsterdam, 1989) p. 253, 1 4 Sun 2 r
[341 O.K. Andersen, O, Jepsen and D. Gl6tzel, Highlights of
Condensed Matter Theory LXXXIX Corso Soc. Italiana where S , . is the overlap between two H Is at the
di Fisica, Bologna (1985) p. 59, separation b. Depending on the value of S,., q~,*-dx:
[35] J.K. Norskov, A. Houmoller, P.K. Johansson and B.I. and ~l,,-d=: are quite similar. As q is independent of the
Lundqvist, Phys. Rev. Lett. 46 (1981) 257. metal, it serves to consider one common q and thereby
[36] B, Hammer, K.W. Jacobsen and J.K. Norskov, Phys. only one V for the purpose of investigating the scaling of
Rev. Lett. 70 (1993) 3971; 6E, as a function of the metal. Using the average of
K, Oundersen, K,W. Jacobsen, J.K. Norskov and ~l~.-dx= and q,-d:~ and neglecting S . . the V becomes:
B, Hammer, Surf, Sci, 304 (1994) 131. 0.246 bohr-3M.Md at (h, Z)--(1.2 A, 1.5 A).
[373 D,M, Bird, L J, Clarke, M,C, Payne and i. Stich, Chem. [433 For the He on NiAI(II0) we use the calculated value
Phys. Lett, 212 (1993) 518. from Ref. [ I I ] where the (b, Z) is ( 1.5 A, 1.33 A) including
[3s] D,M. Newns, Phys, Rev. 178 (1969) 1123. in Z the -0.14 A relaxation of the surface Ni atom.