Handbook of Heterogeneous Catalysis
Edited by,G. Ertl, H. Knozinger, J. Weitkamc
5 Elementary Steps and Mechanisms
5.1 Chemisorption
5.2 Microkinetics
5.3 Factors Influencing Catalytic Action
5.4 Organic Reaction Mechanisms
5.5 Computer Simulation
5.1 Chemisorption
5.1.1 Principles of Chemisorption
H.-J. FREUND
5.1.1.1 Introduction
The term chemisorption was coined in order to classify
the interaction between a particle in the gas phase and
a solid surface, i.e. the result of the adsorption process
[l]. If the interaction leads to the formation of a
chemical bond the adsorbate formed is called a chem-
isorbate. Where chemical bond formation is not im-
portant the process is classified as physisorption. There
are several conceptual problems with such a differ-
entiation which we briefly address in the following, and
which indicate that a more detailed look at the entire
process of adsorbate formation is needed before a reli-
able classification may be carried out. In fact, as it
turns out, for a conclusive classification one would need
the full theoretical and experimental understanding of
the system under investigation. Such an approach must
include the static aspects, i.e. the energies involved, as
well as the dynamic aspects, i.e. the processes involved
in the formation of the adsorptive interactions.
Irving Langmuir in 1916 introduced and investigated
the idea that there can exist strong, short-range forces
between adsorbates and a substrate [2, 31. Previously, it
Copyright 0VCH Verlagsgesellschaft mbH.1997
was generally believed that more or less unspecified
long-range forces - we would today call this phys-
isorption - draw gases towards a solid. Langmuir,
shortly after the introduction of the concept of an or-
dered lattice for the arrangement of the atomic con-
stituents of a bulk solid by von Laue [4], considered an
arrangement of atoms at the surface, a surface lattice,
that defines a specific density of adsorption sites. Atoms
from a gas phase, for example, striking the surface may
either bounce back into the gas phase or establish a
bond to one of these sites. This process is equivalent to
the formation of a surface chemical bond and was
termed chemisorption [5-71. Chemisorption lowers the
free energy of the closed system containing the un-
covered, “clean” surface and atoms or molecules from
the gas phase. This lowering in energy can be measured
via calorimetry or - less well defined - by a Clausius-
Clapeyron analysis of isostere data. It was therefore
tempting to differentiate chemisorption with respect to
physisorption via the energy that is deliberated in bond
formation [l]. Such a definition involves a limiting en-
ergy which separates chemisorption and physisorption
regimes. It was put in the neighborhood of 40 kJ mol-’
El]. Obviously, such a definition is rather artificial, and
today one generally disregards this kind of differen-
tiation solely on the basis of the enthalpy of formation.
The accepted definition of chemisorption today is in-
dependent of thermochemical data and rests on the
concept of a short-range chemical bond, which only
forms when there is direct intermingling of the sub-
strate and the adsorbate charge densities. In order
then to differentiate between chemisorption and phys-
isorption one has to understand the electronic structure
of the system [6, 81. Experimentally, this means that we
cannot rest the definition on a single measurement of
the heat of adsorption but rather on an as complete as
possible spectroscopical characterization of the surface
interacting with the adsorbate in comparison with the
same measurements of the separated systems.
The interaction between say a gas phase, containing
molecules AB, and a surface is discussed by consider-
ing various aspects of the process of AB-surface bond
formation. We cover the dynamic aspects connected
with the sticking of AB such as its dependence on the
References see page 938
912 5 Elementary Steps and Mechanisms
population of internal and external degrees of freedom
of AB in the collision, mobility on the surface, i.e. sur-
face diffusion etc., and the energetics, which will be the
starting point, all as a function of the surface coverage.
We shall discuss associative versus dissociative chem-
isorption and its dependences on surface structure.
Consideration of coadsorption and cooperativity in
the adsorption process is as important as invoking the
structure of the adsorbate, as well as the restructuring
of the surface as it interacts with an adsorbate.
5.1.1.2 Thermodynamics and Energetics
As this point it is important to differentiate between
macroscopic and microscopic surface phenomena. Sur-
face phenomena can be treated macroscopically by
chemical thermodynamics, in which atomic concepts
are not neccessary. Accordingly, the thermodynamic
relationships can be derived on the basis of pressure,
volume, surface area, composition, and temperature,
which can be measured in a straightforward manner.
Historically, therefore, the thermodynamic approach
was pursued first. Before discussing the atomic aspects
of the energy content of an adsorbate phase we shall
briefly summarize the important thermodynamic as-
pects noting, however, that this cannot be a com-
prehensive treatment. For the latter we refer to the
literature [l, 7, 9-12].
Consider an adsorbate phase consisting of n, moles
of a nonvolatile adsorbent (surface) and n, moles of an
adsorbate (gas phase). They are assigned internal en-
ergy U, entropy S and volume V. The surface A of the
adsorbent is assumed to be proportional to the ad-
sorbent volume. The Gibbs fundamental equation for
the full system is then
dG=-SdT+ VdP+padna+p,dns (1)
For the pure adsorbent,
dG" = -Sa d T + Vad P + dna (2)
Consequently, for the interacting adsorbate-adsorbent
system, the difference dGS= dG - dG" gives
dGS= -Ss d T + Vs d P + @ dna + pSdns (3)
where S, = S - Sa, V, = V - Va, and @ = pa - pz.
Using the above mentioned proportionality between
adsorbent volume and adsorbent surface,
@dna=f@dA (4)
where f is a proportionality factor, the surface tension
v, is given by
which has the dimension of a two-dimensional pres-
sure. This leads to the final equation
dGS= -S, d T + Vsd P + VdA + 11, dn,. (6)
This equation refers to a system where the adsorbate
resides on a truly inert substrate. In other words, eq 6
can only rigorously be applied to weakly interacting
physisorbate systems. For chemisorbates this equation
is not strictly applicable because the thermodynamic
parameters of adsorbate and adsorbent cannot be
separated.
Bearing this restriction in mind, and assuming that
the adsorbate phase is in equilibrium with the gas
phase, a Clausius-Clapeyron analysis yields
where we have changed to molar quantities s, u, and
the enthalpy h. The slope of a semilogarithmic plot of
the equilibrium pressure versus the inverse temperature
at constant v, yields the adsorption enthalpy, Ahads, re-
leased upon adsorption of one mole of gas. The prob-
lem, of course, arises because the surface tension is
hard to determine in general. However, the problem
may be circumvented by considering the so-called sur-
face coverage 0 instead:
and resorting to partial molar quantities, e.g.
(")
6, = (9)
T,P,n,
Then, a similar Clausius-Clapeyron analysis leads to
where qst is the isosteric heat of adsorption. This
quantity can be measured quite easily because constant
coverage is not too difficult to establish experimentally.
However, qst represents the difference between molar
enthaply in the gas phase and partial molar enthalpy in
the adsorbed phase, a quantity not easily connected to
microscopic considerations.
In order to directly see how the isosteric heat of ad-
sorption is measured, eq 10 is written in the form
-
dP = ( 2 ) d T
P
and integrated for a reasonably small pressure and
temperature interval so that we can assume the isos-
teric heat to be constant. This yields

161
105 1’10
-
lo2
[K-’]
iis ‘b ’ 0:2 ’ O’L
Xe coverage 0
’ 0’6
Figure 1. Determination of the isosteric heat of adsorpton from
the measurement of adsorption isotherms for the system Xe/
Ni(100) [13].
for two pairs of temperatures and pressures that pro-
duce the same surface coverage. For true equilibrium
conditions, a straight line with negative slope should be
obtained for the semilogarithmic plot, which in turn
yields the isosteric heat of adsorption. Repeating this
procedure for various coverages allows the coverage
dependence of the heat of adsorption to be determined.
It is obvious that the isosteric heat is a differential
quantity, in contrast to the equilibrium heat of ad-
sorption. Both are, of course, isothermal quantities.
From the differential heats of adsorption the integral
heat of adsorption can be obtained as
e1ntegr = 1
ns
qst dns
In the following we present an example of isosteric
heat determination [13]. Figure 1 shows, in the upper
panel, a set of four isotherms for the physisorption
system Xe/Ni( 100). The second panel contains the data
set in the upper panel as a plot of lnppc, versus recip-
rocal temperature for various values of Ox,. From the
shape of the individual straight lines the isosteric heat
has been determined and plotted in the third panel as a
function of Oxe.From the plot we see that the isosteric
heat slowly varies with temperature, the decrease in-
dicating repulsive interactions.
As mentioned above, the application of an isosteric
heat analysis to a chemisorption system is rather prob-
lematic, because inherently the analysis starts from the
5.1 Chemisorption 9 13
18
Figure 2. Schematic drawing of the spherical calorimeter [ 191.
assumption of truly separable subsystems. Therefore,
for such systems it is more appropriate to resort to a
direct calorimetric measurement of the adsorption en-
thalpy. Until very recently, it has not been possible to
undertake such measurements for thin-film systems [l,
14-19]. The reason was that the increase in temper-
ature of the calorimeter depends on the heat capacity
of the system and the absolute number of adsorbed
particles. The most complete set of data for such thin-
film systems has been reported by Wedler and co-
(13)
workers [l, 201. They used the so-called “spherical cal-
orimeter” shown in Fig. 2 [17]. Briefly, the calorimeter
sphere is located inside an ultrahigh-vacuum glass re-
cipient and temperatures change of less than lop5K are
registered with a thermometer connected to the calo-
rimeter sphere. The metal film is evaporated onto the
sphere and gas is admitted. The change in temperature
measured by the thermometer surrounding the calo-
rimeter sphere upon gas exposure is plotted versus time
in Fig. 3. The heat of adsorption is determined from
the integral of the T versus r curve. The example here is
the adsorption of H2 onto a Pd film [21]. Converting
the temperature-time curve into a heat of adsorption
yields 88 kJmol-I [21].
References see page 938
9 14 5 Elementary Steps and Mechanisms
3- H,/Pd lT.273 K)
c
Y corresponding equilibrium gas pressure. Table 2 col-
-a
1 2-
52
l-
Y
1-
O M
0 1 2 3 i 5 6 7
I t / minl
Figure 3. Temperature-time curve of a calorimetric measure-
ment for the system HZ/Pd [21].
In Table 1 are assembled the heats of adsorption for
various absorbate systems on different substrates de-
termined via isosteric heat measurement as well as cal-
orimetric measurements. In some cases the heat of ad-
sorption for one system has been measured using
different techniques. This allows an estimation of the
error involved in using those values based on different
measurements. It is clear that the value for CO ad-
sorption on Fe, for example, is considerably higher
than values for other CO-adsorbate systems. In order
to judge this, it must be understood that at room tem-
perature CO partly dissociates on polycrystalline Fe
which contributes to the observed calorimetric value.
This is a useful reminder that consecutive processes
have to be considered in calorimetric measurements.
King and co-workers have recently extended calori-
metric measurements to single-crystal surfaces by ap-
plying molecular beam techniques in combination with
IR radiation emission measurements (Fig. 4) [23, 38,
46-48]. There are three important parts of the experi-
ment. First, there is a molecular beam source to pro-
vide accurate determination of coverage. Secondly, the
sample consists of a unsupported single-crystal thin
film to reduce the thermal mass to a minimum. Finally,
an infrared detector is used to sense the heat radiated
as the gas adsorbs. In order to reduce white noise
experiments are performed using a pulsed molecular
beam, which must be capable of producing a significant
enthalpy change per pulse upon adsorption. A set of
results is included in Table 1 where it is compared with
results from other thermodynamic measurements. It is
interesting to note that the results for the single crystal
surfaces are situated in the region of those determined
for the polycrystalline films, indicating that the latter
consist of many crystallites exposing low-index planes.
The free enthalpy of adsorption is determined by
the enthalpy as well as the entropy of adsorption. The
entropy of adsorption, which is the quotient of the
reversibly exchanged heat and the temperature, may be
calculated from the equilibrium heat of adsorption, if
the surface tension is known, or from the isosteric heat
of adsorption. Prerequisite is the knowledge of the
lects typical values for the entropy of adsorption of
several adsorbate systems.
The number of values available is much smaller than
for the enthalpies of adsorption. The interpretation of
entropies is considerably more involved than the inter-
pretation of enthalpies. Often the observed values and,
in particular, the coverage dependences, cannot be rec-
0 onciled on the basis of theoretical predictions. It ap-
pears that the predictions as to how the various degrees
of freedom of the adsorbate contribute are not accurate
enough to date. However, in most cases large entropy
values are found when the mobility of the adsorbate
was expected or known (from other methods) to be
large.
However, the most popular method by which to de-
duce the heat of adsorption of an adsorbate system is
thermal desorption spectrometry (TDS) [53-571. A
schematic setup for a TDS measurement is shown in
Fig. 5 [58]. The sample is heated resistively and the
temperature is monitored by a thermocouple. If the
sample is a single crystal it responds rather rapidly to
heating so that relatively high heating rates may be
used. The concentration of desorbing species is mea-
sured with a quadrupole mass spectrometer (QMS).
Pumping capacity is important in thermal desorption,
because only if it is high enough, readsorption of the
desorbing species back onto the surface is excluded. If
the pumping speed is infinitely high we can ignore
readsorption and the change in adsorbate coverage per
unit time; a measure for the desorption rate ( r d e s ) , is
given by the Wigner-Polanyi equation [7, 591:
= v(0)O”exp{ - w} (14)
+
where T = To Pt. This is the basis for the analysis of
thermal desorption spectra.
Figure 6 schematically shows a set of T D spectra for
various initial coverages 0, and a given heating rate p
[7, 581. The first step is the integration of the spectra
starting from highest temperature, i.e. coverage 0 = 0,
to a given coverage O’, say 0.15. This yields a pair
of (r, T ) values for each initial coverage larger than
0’ = 0.15. A plot of l n r versus 1 / T yields Edes(@’)
from the slope and v(0’) from the intercept, which
is given by n l n 0 ’ + Inv(@’), if the order n of the de-
sorption reaction is known. However, for coverages
above 0.1, the second term is much larger than the first,
 5.1 Chemisorption 9 15
~~ ~

Table 1. Enthalpies of adsorption.

~~

Adsorbate Substrate q (kJ mol-') Remarks References

co Ni(ll1) 111 (i5) WF '1 22

130 Microcalorimetry 23
Ni( 100) 125 (0 5) WF 24
115 TDS~) 25
119 TDS 26
138 TDS 27
109 isosteric &d 28
123 isosteric &d 24
130 isosteric Ead 29
134 isosteric Ead 30
123 Microcalorimetry 23
Ni(ll0) 133 Microcalorimetry 23
Pd( 100) 150 (+5) WF 31
161 (+8) WF, TDS, LEED 32
Pd(ll1) 142 ( i 3 ) WF 33
Ru(0001) 160 (&lo) WF 34
Ru( 1OLO) 157 (*lo) Contact-pot., TDS 35
Cu( 100) 58 (*lo) WF 36
Fe(ll1) 325 273 K (dissociative) 37
Fe( 11 1) 200 195 K (partially diss.) 37
Fe(ll1) 100 (not dissociative) 20
CO/K Ni(100) 190 Microcalorimetry 38
co2 Fe 300 195 K (dissociative) 37
H2 Ni( 100) 96.3 (k5) WF 39
Ni(ll0) 90.0 (i5) WF 39
Ni(ll1) 96.3 ( i 5 ) WF 39
85 ( i 5 ) 40
Ni 7 5 , ..I76 1
Pd(ll1) 88 (i5) WF 41
Pd( 110) 103 (k5) WF 42
Pd( 100) 102 ( k 5 ) WF 41
Rh( 110) 92 (k5) WF. TDS 43
Ru( 1010) 80 ( i 5 ) 44
Co(1010) 80 ( i 5 ) 45
Ta 188.1 1
W 188.1 1
Cr 188.1 1
Fe 133.8 1
Fe 100 dissociative (273 K) 20
Fe 97 ( 2 3 ) 37
Pd 80/96 21
Na W 133.8 1
cs 267.5 1
0 Ni( 100) 532 ( I S ) IR (300 K)
x432 IR (100 K)
532 (is) 23
Ni(ll1) 470 ( 5 1 5 ) 23
Ni( 1 10) 498 ( 5 5 ) 23
02/CO Fe(ll1) 490 273 K 20

'1 WF, work function; 2, TDS, thermal desorption spectroscopy

so the latter may be neglected without large errors. It
should be noted that there are methods to determine
the order rigorously. This analysis, called the "com-
plete analysis" was first proposed by King in 1975 [55].
The preexponential factor v ( 0 ) can be regarded as
representing the frequency of attempts of the adsorbed
particle to escape the chemisorptive potential. The
values determined vary by at least four orders of
magnitude, from lo'* to 10'6s-1 [7]. If we adopt the
simplest picture and set v ( 0 ) equal to the frequency
of vibration of the adsorbed particle, values near
1013s-l are expected. The problems become even more
involved if we consider the number of successful at-
tempts, i.e. after multiplication of v ( 0 ) by the ex-
ponential in eq 14. Here, the activation energy for
desorption Edes(@)comes into play; both v ( 0 ) and
References see page 938
916 5 Elementary Steps and Mechanisms

a/c pre -
amp ld ier P e lens .-.. a / c amplifier

I
i
I‘ . G e generator
quadrupole mass
spectrometer
.
t
I
stagnation detector I

beamstopper II
1

Figure 4. Schematic drawing of the setup for microcalorimetric measurements on single crystals [46].

Table 2. Entropies of adsorption. It results in reliable values only for first-order desorp-
tion and provided that a reliable value for v is avail-
Adsorbate Substrate AS:! (J K-’ mol-’ ) References able. The Redhead equation can be directly derived
Xe Ni(100) 056 13, 49 from the Wigner-Polanyi equation by determining the
Pd(100) 258 50 temperature derivative of the rate, and realizing that it
H Pd(100) 0263 49 must vanish at the peak maximum temperature [7].
N2 Ni(100) 150 51 Additional procedures are given in the literature [58,
Ni(l10) 112 (i.5) 52
60-631. In connection with the initial question con-
cerning the heat of adsorption, it must be realized that
the desorption energy may be directly related to the
heat of adsorption if adsorption is a nonactivated pro-
cess. In other words the adsorption process is, ener-
getically, continuously “downhill”. A detailed under-
standing, however, necessitates an understanding of the
dynamics of adsorption.
A connection exists between the phenomenological
view of the energetics from the standpoint of thermo-
dynamics, and the microscopic view of adsorbate en-
ergetics. In this context the question as to whether a
process is activated or nonactivated may already have
been answered.
Figure 5. Schematic drawing of the experimental setup for a This approach goes back to Lennard-Jones who dis-
thermal desorption expriment. cussed adsorption energetics in a landmark paper in
1932 applying a quasi-one-dimensional approach [64].
Edes(@)depend on coverage. These coverage depend- Neglect for the moment all problems connected with
ences partly compensate each other for certain systems the question as to how a gas-phase particle is actually
in the sense that high values of v ( 0 ) are associated with trapped in a bound state at the surface of a solid, and
large values of Edes [7]. This has to be considered when simply consider the interaction potential between the
dealing with predictions and interpretations of de- gas-phase particle and a surface. Figure 7(a) shows the
sorption rates. It is therefore important to resort to a well known Lennard-Jones potential energy diagram.
complete or close to complete analysis of desorption It represents the superposition of attractive (longer
data. Simplified analyses were published much earlier. range) and repulsive (short range) forces according to
The most popular one is the so-called Redhead analy- E ( z )= -AzP6 + Bz-12
(16)
sis, based on the peak maximum temperature observed
in a thermal desorption spectrum [54]: where A and B are empirical constants and I’ is the
distance between the adsorbed particle and the surface.
To describe this interaction on the basis of ab initio
 5.1 Chemisorption 917

Figure 6. Determination of the desorption energy Edes from a model-independent analysis of thermal desorption data. The analysis is
carried out for an artificially chosen coverage of 0 = 0.15: (a) TDS data; (b) integration of the TDS data to give 0 versus T plots; (c) plot
of In dO/dt versus 1/T to determine Edes according to eq 14. The basis for the diagram are data for Ag/Ru(0001) [58].

t-
quantum mechanical calculations it would be necessary
to consider a semiinfinite solid interacting with an atom
or a molecule. This can be done in favorable, simple
cases using various approaches [65-681. The most
a prominent one, at least for metal surfaces, is the den-
sity functional approach with which one can come
~~~~ , ads close to the exact solution [65, 671. Another approach is
the so-called embedded cluster ab initio approach
where the solid surface is represented by a cluster of
atoms, augmented by an embedding scheme to repre-

k-
sent more accurately the infinite extension of the two-
dimensional system [68]. Assume for the moment that

b \ , this problem has been solved. Then, the potential en-

ergy curve in Fig. 7(a) represents the case where the
particle incident from the gas phase “sees” a con-
m z&@ tinuously “downhill” energy change until it reaches the
Eods
2Eads equilibrium position at 2,. (Note that, for the present,
the dynamics of the trapping process is being re-

I f
.0--.-.- -.-.- gg glected.) In such a case the desorption energy, as de-
termined from thermal desorption data, is equivalent
to the heat of adsorption. It is this situation that is
l i dissociation often considered for associative molecular adsorption.
C energy However, the situation becomes more difficult if either
a molecule which is associatively adsorbed may assume

dEods mEads . =@ different adsorption geometries on the surface, or the

Figure 7. Schematic potential energy diagrams for a molecule AB
approaching a surface: (a) associative chemisorption (&&); (b)
associative chemisorption (’&&) with precursor (’ &&); (c) dis-
sociative chemisorption (dEads) with molecular precursor (mEads).
molecule may dissociate upon adsorption and - to in-
crease complexity - may do so via a molecularly ad-
sorbed precursor state. Figures 7(b) and 7(c) schemati-
cally show the corresponding quasi-one-dimensional
References see page 938
9 18 5 Elementary Steps and Mechanisms
unsuccessful
I Y
C X 4
Figure 8. Two-dimensional potential energy surfaces (schematic) for (a) early and (b) late barrier (B) of dissociation of H2 on a transition
metal surface.
potential energy diagrams. In Fig. 7(b) there is a sec-
ond minimum in the potential energy diagram repre-
senting the two possible adsorption geometries. It
is already obvious that, in this case, the use of such
a quasi-one-dimensional diagram becomes very prob-
lematic because only a single spatial coordinate is used
to represent the molecule-surface interaction. There-
fore, such a situation calls for a multidimensional po-
tential energy diagram, and we shall come back to this
more general requirement later on. For the moment,
however, Fig. 7(b) already allows us to visualize the
transformation between the two inequivalent molecular
adsorption geometries as an activated process. It is
immediately clear that a desorption experiment will
probe this more complicated potential energy curve,
and thus a simple interpretation of the measured de-
sorption energy as the heat of adsorption will not be
possible in general.
The situation becomes even more complicated if,
upon interaction with the surface, the molecule dis-
sociates. This is depicted in Fig. 7(c). In this case it is
necessary to consider two intersecting potential energy
curves which refer to two different zero-energy levels,
namely the diatomic molecule being infinitely sepa-
rated from the surface for the associative interaction,
as well as the two constituent atoms being infinitely
separated from the surface. The difference between the
reference levels, of course, represents the heat of for-
mation of the diatomic molecule in the gas phase. In
this case, the above-mentioned difficulty with the quasi-
one-dimensional representation becomes particularly
clear, in the sense that here the coordinate representing
the separation between the two constituent atoms has
not been considered at all. Nevertheless, it can be seen
that there may be a rather large activation energy be-
tween the molecularly adsorbed precursor and the dis-
sociatively adsorbed atoms, which is very crudely rep-
resented by the energy near the crossing point with
respect to the potential energy minimum of the mo-
lecular precursor. Clearly, the point of intersection may
be situated well above the reference level corresponding
to the infinitely separated molecule and surface, which
in turn has strong consequences for the ability to pop-
ulate the dissociative adsorbate. As will become clearer
considering multidimensional potential energy surfaces
in such a situation, the molecule has to have a certain
impact energy to be able to surmount the activation
barrier. Whether this impact energy should be repre-
sented by translational degrees of freedom or internal
(rotational or vibrational) degrees of freedom cannot
be concluded on the basis of the quasi-one-dimensional
potential energy surface. However, it is already fully
transparent that the shape of the potential energy sur-
faces will determine the kinetics as well as the dynamics
of the system, and thus the probability to chemisorb.
Experimentally, we measure (for example) the sticking
probability of a particle from the gas phase into a par-
ticular adsorbate channel by probing the number of
adsorbed species as a function of gas pressure and sur-
face temperature. In other words, a relatively complex
scenario is condensed into basically a single number.
As the next section shows, it is far from easy to resolve
the details.
Before tackling the problem of sticking consider, as
alluded to above, potential energy diagrams that allow
the incorporation of some essential additional features
such as simultaneous motion along several coordinates
(often normal coordinates). Clearly, the situation be-
comes very complicated as soon as many such coor-
dinates come into play. Consider therefore, for sim-
plicity the most simple case of a hydrogen molecule
interacting with a transition metal surface. In recent
years, this problem has been treated experimentally as
well as theoretically in great detail so that a clear pic-
ture of the factors influencing the activation process
has emerged. A good review for the case of H2/Cu can
be found in Ref. 69.
Figures 8(a) and 8(b) show potential energy dia-
grams for such a system [70]. The potential energy is
plotted as equipotential lines in a coordinate system

where the ordinate represents the surface-molecule
(center of mass) distance, and the abscissa the inter-
atomic distance of the diatomic molecule, i.e. the
hydrogen molecule in this case. Denoting the inter-
molecular distance in the molecule by x and the dis-
tance of the center of gravity of the bond to the surface
by y , small x values are found for large y values, in-
dicating the intact bond between the hydrogen atoms.
As the molecule gets closer to the surface, i.e. y de-
creases, x finally increases to large values that are
characteristic of the bond-breaking process. It is the
exact position of the barrier, indicated by the letter B at
the top of the saddle point in the potential energy dia-
gram, that now governs the dynamics of the process.
Two different situations are depicted. In Fig. S(a) the
activation barrier is located in the entrance channel. A
molecule entering the entrance channel with sufficiently
high translational energy can surmount the barrier, as
indicated by the trajectory. However, it may well move
up the wall before it can follow the bend (as if on a
“bobsleigh” course) and the system will consequently
come out the exit channel vibrationally excited (i.e. the
hydrogen surface modes are excited) as indicated by
the curved trajectory. In Fig. 8(b) the activation barrier
is located more towards the exit channel. Here a
vibrationally excited molecule has a better chance to
surmount the activation barrier as indicated by the full
trajectory. An unsuccessful attempt with a translation-
ally excited molecule is shown for comparison. Once
the vibrationally excited molecule has crossed the bar-
rier, the hydrogen atoms formed will move across the
surface with relatively high translational energy. The
whole problem outlined so far can be mapped almost
perfectly onto the so-called Polanyi rules [711, where-
after an exergonic reaction of type A + BC + AB + C
with an early barrier request translational energy,
whereas, if the reaction has a late barrier, it requires
vibrational excitation of the reactants.
Molecular beam studies [72] have been undertaken
in recent years to prepare selectively translationally or
vibrationally excited molecules before they were scat-
tered off the surface, and a great deal has been learned
about how the molecules stick to a metal surface, spe-
cifically for hydrogen-transition metal systems [65-731.
In the case of hydrogen absorption on Cu, the barrier
[74-771 is in an intermediate position, so that both
translational as well as vibrational excitation helps to
surmount the barrier. There are still a lot of open
questions as to which role rotational excitation plays
[69]. However, even with a full understanding of the
processes occurring on the potential energy surfaces
shown in Figs. S(a) and 8(b), there are still some im-
portant ingredients missing. This has particularly to do
with the fact that in the discussion so far the geometric
and electronic structure of the surface has not been
considered. It is known that chemisorption is structure
5.1 Chemisorption 9 19
sensitive, and therefore these aspects have to be con-
sidered. It should be pointed out that hydrogen
adsorption on Cu surfaces may not be typical for
interaction with transition metals in general [78]; in
particular, remember that H2 dissociates with almost
no barrier on metals such as Ni, Pd, etc. [73]. It is clear
that in order to understand the barrier heights elec-
tronic structure calculations must be resorted to [79,
SO]. However, the difference between Cu and Ni may
be argued on a qualitative basis [79, SO]. Cu has the
electronic configuration 3dI04s’, with the rather diffuse
4s orbitals occupied. If a closed-shell H2 molecule ap-
proaches the Cu surface it will be repelled by the diffuse
4s electrons so that it is hard for the H2 to come in
close to dissociate. Ni has the electronic configuration
3d94s’ in which the 4s orbital is occupied, which again
leads to Pauli repulsion with the H2 molecule. How-
ever, in Ni the 4s electron may be promoted into the
hole within the d shell, forming a 3dl04s0 config-
uration - this reduces the repulsion dramatically and
allows the H2 molecule to come in close and dissociate.
Therefore, the barrier for Ni is much lower than for Cu
where it is in the range of 1 eV [69].
Another aspect that is important in connection with
the discussion of adsorbate thermodynamics and en-
ergentics, so far neglected, is the aspect of interaction
between adsorbed species. In Langmuir’s picture of
adsorption [2, 31, mentioned in the introduction, the
adsorbed particles occupy the lattice points of a two-
dimensional substrate with equal probability and with
hard wall potentials between them, preventing double
or multiple occupancy of any particular site, and with
well defined adsorption energies typical of the site.
(Note that at this point structure sensitivity comes into
the picture; however, this aspect is deferred until later.)
As a result of this view of adsorption, saturation would
be characterized by complete coverage and the forma-
tion of a true l x l adsorbate layer. Obviously, the
formation of ordered layers with coverages far below
complete coverage are more the rule than the ex-
ception, and are a direct consequence of the existence
of interaction potentials. Such an interaction potential
is shown in Fig. 9 for the system CO/Pd(l00) reported
by Tracy and Palmberg in 1969, compared with a CO-
CO interaction potential in the gas phase [31]. Inter-
action potentials may be either attractive or repulsive
and may be classified into direct and indirect inter-
actions [81-83]. Direct interactions involve dipole-
dipole (multipole-multipole) and orbital-overlap inter-
actions, and are often repulsive. On the other hand,
indirect interactions mediated through the metal sur-
face may be either attractive or repulsive depending on
distance and surface sites, i.e. the kind of charge mod-
ification of the electronic structure of the substrate by
References see page 938
920 5 Elementary Steps and Mechanisms
gas phas
1.0 2.0 3.0 L.0 5.0
CO-co separation [ A 1
Figure 9. Intermolecular potential for CO in the gas phase and
CO adsorbed on a transition metal surface.
the adsorbate. The interplay of the interaction poten-
tial and the adsorption energy of the isolated particle
with the clean surface finally determines the observed
properties of the adsorbed layer. In other words, the
structure of the adsorbed layer depends on the heat of
adsorption as well as on the coverage [ 5 ] .
The situation again may be depicted in the form of a
potential energy diagram; however we have to include
the existence of different surface sites [84]. Figure 10(a)
[7] shows a one-dimensional potential energy diagram
where the spatial coordinate extends parallel to the
surface. It has been assumed that every surface site
provides identical binding conditions. All identical
a b
t'
attractive interaction
ground state
Figure 10. One-dimensional potential energy parallel to the surface: (a) empty surface with a single particle bound with adsorption energy
&&; (b) superposition of the potential energy in (a) with a painvise interaction potential of particles on the surface (E,,,,), which may be
either attractive or repulsive.
adsorption sites are separated by a small activation
energy if compared with the activation energy for
desorption, which gives rize to a sinusoidal energy
dependence across the surface. At low enough tem-
perature the adsorbed particles will reside within the
potential wells because their thermal energy is too
small to overcome the activation barrier for diffusion.
Correspondingly, for higher thermal energies, particles
will site exchange resulting in a mobile adsorbed layer
with short residence times in the individual wells. We
shall discuss this situation in more detail further below.
The potential energy diagram parallel to the surface
changes significantly if the interaction between ad-
sorbed particles is taken into account. This is sche-
matically depicted in Fig. 10(b) where we have added
an attractive as well as a repulsive potential to the one-
dimensional diagram of Fig. 1O(a). The consequences
are energetic heterogeneities, weakening the adsorbate
surface bond in the case of the repulsive interaction,
and strengthening the adsorbate surface interaction in
the case of attractive interaction potentials. As men-
tioned above, phenomenologically this leads to the
formation of ordered phases on surfaces. In fact, there
may be several different ordered structures depending
on both temperature and coverage, because surface
diffusion may act against the formation of ordered
structures, i.e. favoring disordered layers while, for ,
example, coverage increase locks in certain structures.
A way to represent the various structures is to plot a
so-called phase diagram [MI. An example is shown
in Fig. 11. For the system CO/Cu(lOO), two ordered
phases are found in the given temperature range [86].
These are denoted by I and II+, and they occur at
coverages 1/2 and 4/7. Phase I is a c(2 x 2 ) structure,
while phase 11, consists of stripes of the c(2 x 2)
structure of width n = 3 separated by domain walls.
The main part of the phase diagram is filled by a dis-
ordered phase. A very interesting and frequently stud-
ied aspect of such phase diagrams are the two-dimen-
sional phase transitions. In two dimensions, similar to
tE
 5.1 Chemisorption 921

1
7 CO/Pd( 1001

\
fl
A

0
\
E 120-
7
Y
Y
m
U
0
w 100-
coverage

80-

I I I

I 0 02 04 06
0
-_
Figure 12. Adsorution energy (E2,+ as a function of surface
coverage 0 [7]: (A\ CO/Pd(100) [32]; (v) CO/Ni(lll) [22]; ( 0 )
~ --I

II+ H/Ni(llO) [39].

Figure 11. Phase diagram for the system CO/Cu(lOO), also

showing the ordered structures [86].
I to the population of additional weakly bound species
three dimensions, phase transitions may be classified as
discontinuous first-order, and continuous higher order.
In general, phase transitions may be evaluated accord-
ing to the temperature dependence of the thermody-
namic functions. This subject is considered in greater
detail elsewhere [87-901.
More important with regard to the heat of adsorp-
tion are the particle-particle interactions. As stated
above, according to the Langmuir picture of adsorp-
tion [2, 31 we would expect constant adsorption energy
until saturation of the surface is reached. In reality, this
is never the case [85]. Rather, the adsorption energy
generally decreases at medium and high coverages due
to interactions between the adsorbed particles. It is
possible to estimate the interactions from the coverage
dependence of the isosteric heat of adsorption. In Fig.
12 [7] are several examples [22, 32, 391 where the work
function has been used as a measure for the coverage
(which may be sometimes dangerous). It is obvious
from Fig. 12 that in all cases the adsorption energy
sharply decreases as saturation is approached. At low
coverge, however, the isosteric heat turns out to be ei-
ther constant, decreasing, or increasing with coveage.
The observed changes are a consequence of the par-
ticle-particle interactions on the surface, in the sense
that increase means attractive interactions, as for ex-
ample in the case of hydrogen for low coverage [39],
decrease repulsive interaction, as in the case of CO on
Ni( 11 1) [22]. The step-like decrease of the heat of ad-
sorption close to saturation for CO/Pd(lOO) [32] is due
on the surface [91, 921. These weakly bound species
may be rather reactive. Owing to their small heat
of adsorption they may react rather easily with co-
adsorbed, neighboring functional groups. At high re-
actant pressures and not too elevated substrate tem-
peratures this kind of scenario may play a significant
role.
5.1.1.3 Sticking
This section considers the traditional description of the
process where a molecule approaches a solid surface
and eventually is trapped by the potential. A con-
venient way to gain access to this problem is through
the consideration of the rate of adsorption. In the most
simple case, the rate of adsorption is proportional to the
number of molecules impinging per unit time on the
surface, the so-called particle flux, and to the (dimen-
sionless) efficiency with which an impinging particle
actually sticks to the surface, i.e. the so called sticking
probability. The initial sticking coefficient so is the ratio
of the number of adsorbed particles oS and the num-
ber of impinging particles for the uncovered surface.
Therefore,
O<S,<l (17)
References see page 938
922 5 Elementary Steps and Mechanisms
In principle, determination of this quantity is straight-
forward. In an adsorption experiment a clean surface
held at temperature T is exposed to a well defined
pressure P for a given time t (exposure is measured in
Langmuir: 1 L = lop6 torr for 1 s), and the amount of
gas taken up by the surface (by a suitable surface sci-
ence technique) is compared with the total amount of
gas that has struck the surface. A method frequently
used is the one proposed by King and Wells [93, 941. In
this case a molecular beam strikes the surface and the
change in the background pressure of a given gas is
measured by a mass spectrometer. The procedure is
calibrated with respect to a gold sample that is known
not to adsorb any molecules in the considered temper-
ature range.
Knowing how s is measured experimentally, we can
turn to further conceptual considerations. The rate of
adsorption, i.e. the change of the number of adsorbed
particles with time is given by [7]
where the flux of impinging particles has been treated
according to the kinetic theory of ideal gases, and a
function f (g,) accounts for the loss of empty sites as
the adsorption process proceeds. The term so may be
written in terms of a preexponential O s and an activa-
tion energy adsEact as
so = Osexp( - +)
ads
There is a different adsorption probability depending
on whether the adsorption site is occupied or not.
From what has been said before, the sticking coefficient
must also depend on the population of internal and
external degrees of freedom of the impinging molecule.
This can be done in a closed form by assuming the
sticking probability s to be composed of terms for the
vibrational states involved, each weighted by a Boltz-
mann factor (FB) representing the population of the
corresponding vibrational state [69]:
s(ujEej T )= CFB(~,
V
T)sa(u,Ee) (19)
where v represents the vibrational quantum state under
consideration, which is populated according to the
Boltzmann factor depending on the temperature T of
the gas (effectively the nozzle temperature in a mo-
lecular beam experiment). The effective translation
energy Ee is given by [74]
E, = E i COS"($~) (20)
in whch E i is the translation energy of the incident
particle, and $i is the angle of incidence with respect to
the surface normal. If n = 2 then the effective trans-
lation energy shows a typical "normal energy scaling",
in other words an exclusive dependence of the sticking
probability on the normal component of the energy of
the incident particle, which has been found rather fre-
quently [95-1001, and in particular for hydrogen ad-
sorption on transition metals.
With this in mind we can go back to equation (18)
and analyse the rate of adsorption further. First, write
the rate in terms of coverage and not in terms of the
absolute number of particles:
In the case of the most simple treatment according
to Langmuir [2, 31, where it is assumed that each
adsorbed particle occupies only one surface site, the
adsorbed species does not interact with other adsorbed
particles present on the surface. It is further assumed
that the adsorption energy is completely exhausted as
soon as one monolayer has been formed; the function
f (0) reduces to (1 - 0).If the particle dissociates
upon adsorption - it occupies two sites - the function
f (0) becomes (1 - Remembering that under
equilibrium conditions the rate of adsorption must
equal the rate of desorption
Tad = rdes (22)
we arrive at the following condition for the coverage:
( 0 )= b(T)P
+
1 b(T)P
which is the famous Langmuir adsorption isotherm [2].
In its derivation we have employed eq 14 for the de-
sorption rate assuming a first-order process, and con-
secutively just solved for @. In addition, we have used
an abbreviation for a constant b ( T ) which only de-
pends explicitely on temperature once the adsorbate
parameters are known. b( T ) is given by
In this case the preexponential factor O s should not de-
pend on coverage because it has been assumed for the
derivation that there is no intermolecular interaction.
Many different adsorption isotherms may be derived
where all or some of the basic assumptions going into
the derivation have been released or relaxed [101- 1051.
It should be stated, however, that the general form of
the Langmuir isotherm, which is shown for two tem-
peratures in Fig. 13, may be used for a phenomeno-
logical description of many processes. It is clear, from
the adsorption isotherm, the sticking probability so
may also be determined given that all other parameters
are known [l, 71.
 5.1 Chemisorption 923

metal surfaces regardless of their crystallographic ori-

I/
II
" tant to consider the coverage dependence of the stick-
pressure p
Figure 13. Plot of the Langmuir isotherm for two temperatures
( T I and T2).
Table 3 collects a set of sticking probabilities deter-
mined for various absorbate systems. The values vary
between unity and lo-*, although the range is usually
between 0.15 and 1. Obviously, there is a clear trend
that sticking is higher for atomically rough surfaces as
compared with atomically smooth surfaces depending
on the nature of the gas. It seems that energy accom-
modation is particularly easy on the rough surfaces as
compared to the smooth ones. Carbon monoxide and
nitric oxide stick quite effectively on many transition
entation. On open surfaces they even tend to dissociate.
The tendency to dissociate increases when going from
the right to the left in the periodic table. Co is approx-
imately on the border line.
We note at this point that in addition to the surface
crystallography, surface defects (point defects as well
as steps) are important to accommodate chemically
active species [112, 127, 1281. Initial sticking proba-
I bilities are interesting, but for real systems it is impor-
ing coefficient. Of course, a model-free discussion of
this aspect is very difficult. It is therefore common
practice to assume a set of possible kinetic processes
which are important in connection with sticking to a
surface. A possible scenario is shown in Fig. 14 sepa-
rately for adsorption and desorption [7]. We introduce
precursor states which may be classified as either in-
trinsic or extrinsic precursor states [129, 1321. The for-
mer exist at empty surface sites and the latter at sites
already occupied. While trapped into such a precursor
state the particle is only weakly held to the surface.
Thus it can diffuse across the surface and be eventually
trapped into an empty surface site. Given a precursor
lifetime of lop6s, the molecule probes the surface for a
sufficiently long time to find an empty site, if the pre-

Table 3. Initial sticking coefficients

Adsorbate Substrate Sticking coefficient Remarks References

~~~

H Ni( 100) 0.06 106, 107

Ni(l11) 20.01 108
Ni(ll0) Z1 109
0.96 1 I0
Pt(ll1) 0.1 111
~0.0001 112, 113
Rh(ll0) ZI 43
Ru( 1010) Z l 44
Co( 1OLO) 0.75 (i20%) 45
W(100) 1 114
0 C U ( 100) 0.03 300 K 115
Ni( 100) 1 1 I6
Pt(l11) 0.2 1 I7
co Ni(lll) 1 22, 118
Ni( 1 10) 0.89 119
Pd( 100) 0.6 32
Pd(l11) 0.96 120
Ru( 1010) 1 35
Pt(l11) 1 121
N W( 100) 0.2-0.6 see Figure 94
W(110) 1-5 x P-N2 122, 123
0.22 Y-N~ 124
W(111) 0.08 93
N2 Fe( 100) 10-~-10-~ 125
Fe( 111) 10-~-10-~ 126
Fe( 11 1) > (100) > (1 10) 10-6- 10-8 126

References see page 938

924 5 Elementary Steps and Mechanisms

direct processes
0- Pch 0
0 pd

precursor
mediated
processes

-0 0.5 1.0

I ,
~ prec:sor
chemisorbed I gas phase
state
I
1
' - r - ; s to the surface
perpendicuiar
1 1
1
I
zch zp
Figure 14. Schematic representation of direct and precursor-
mediated processes on a surface [ 129, 1301. Processes occurring
along the surface normal are plotted along the abszissa. The pro-
cesses are correlated with the potential energy diagram of Fig.
7(b) (ex = extrinsic precursor, in = intrinsic precursor, n , = num-
ber of impinging particles from the gas phase, d and 2'' are frac-
tions of trapped molecules, p = probabilities, p k = migration
probability along the surface).
coverage is not too large. In order to set up a scheme
we have to define probabilities (p,) with which the var-
ious states at the surface are populated. On the basis
of this (Fig. 14), it is possible to arrive at equations
for the rate of adsorption and desorption. However, in
the present case, different from the situation discussed
above for direct sticking, the sticking probability s ( 0 )
will be dependent on the surface coverage. Kisliuk, as
one of the first, has proposed a coverage dependent
sticking coefficient based upon such considerations
[131, 1321:
The constant K is connected with the probabilities to
populate a chemisorptive state via the various pre-
cursor states or desorb from them, respectively (Fig.
14):
I
8
Figure 15. Relative sticking probabilities as a function of surface
coverage according to the Kisliuk model [131, 1321. For an ex-
planation of K see text.
K = - Pi (26)
Pih +PA
The Kisliuk model for a coverage-dependent sticking
coefficient contains the linear Langmuir behavior as
well as the coverage-independent sticking probability
as limiting cases. Clearly, as K = 0, s ( 0 ) = so. Also, as
J K = 1, s(O) = s o ( l - 0),i.e. the linear Langmuir be-
havior is retained. As K is always larger than zero, we
have to consider two cases, namely for K > 1, and for
0 < K < I . The result is a convex curve for the former,
and a concave curve for the latter case (Fig. 15) [131,
1321. Which behavior is actually encountered is largely
determined by the probability p i , i.e. the probability
for desorption out of the extrinsic precursor. It has to
become smaller than the sum of probabilities to desorb
out of the intrinsic precursor and the probability to
chemisorb out of the intrinsic precursor, in order to
achieve K < 1. Under ultrahigh-vacuum conditions,
the population of extrinsic precursors is only easy
to realize at low substrate temperatures. Therefore,
concave sticking probabilities are generally found, as
demonstrated for some examples in Fig. 16 [39, 120,
133, 1341. However, at higher pressures, the population
of weakly bound precursor states may be of im-
portance, so that the population of the chemisorbed
state through the precursor becomes rate limiting. In
such cases we may find a convex curve of the sticking
probabilities. Of course, additional complications may
arise if the structure of the surface changes upon
changes of coverage [ 1351. Then the dependences may
become very different altogether. Oscillatory surface
chemical reactions are connected with such behavior in
some cases [136].
To end the section on sticking we would like to de-
scribe a very interesting development that has recently
become more visible, namely the experimental in-
 5.1 Chemisorption 925

1
I
before colliding with the substrate. Depending on the
polarity of the electric field in front of the surface, two
different orientations can be achieved: preferential N-
m end and preferential O-end collisions. The rotational
temperature of the colliding molecules determines
0 0.25 0.5 '0 0.5 1.0 the degree of orientation of the molecules. Therefore
I e seeded pulsed nozzle beams are used to cool the par-
ticles before collision. The integral number of mole-
cules leaving the surface after scattering is detected
from the NO partial pressure with a quadrupole mass
spectrometer located behind the target and thus
CO/Pd(lll) shielded from the direct beam [93]. Figure 18 shows a
0' 0.25 0.5 0.25 0.5 typical result in terms of partial pressures (right panel)
for the scattering of NO from Pt(100) as a function of
field strength and orientation of the NO molecule [142].
Figure 16. Relative and absolute sticking probabilities for car-
bon monoxide as a function of surface coverage [39, 120, 133,
The observed asymmetry, which is plotted in the left
1341. panel, is very high. Note that the degree of orientation,
given as the averaged cosine (cos3) of the angle be-
tween molecular axis and external electric field 3, is
vestigation of the dependence of sticking on the ori- 30%. The result documents the strong preference for
entation of the particle, in particular a molecule, upon trapping in the chemisorptive potential if the molecules
surface impact. Kleyn and co-workers [137, 1381, as approach the surface with the N-end. At higher surface
well as Heinzmann and co-workers [139-1411 have temperatures the asymmetry decreases as expected,
shown that a molecule such as NO can be state selected because the number of molecules that do not stick
and focused by taking advantage of a hexapolar elec- increases for both orientations. It seems that for a
tric field, and subsequently oriented in a homogeneous detailed understanding of the temperature dependence
electric field, as schematically indicated in Fig. 17 [140], a kinetic model involving precursor states has to be

guiding and
orientation field
OMS
for determination
kV
+ 1_0 ot stickinq coefficient
hexapole - field
(with central beam stop1
velocity

Figure 17. Experimental setup to study sticking probabilities of oriented NO molecules [139].

7 3-

n N-end collisions
5 1-
._
.ad

n
b
0
- 0
6 8 10 12 1L
= 076 z

8
,
10
I I

12
3 I I

I
field strength [kVlcml field strength [kV/crnl

Figure 18. (Right) NO partial pressure after scattering from a Pt(100) surface as a function of field strength and NO orientation [141].
(Left) Corresponding orientation asymmetry of the partial pressure of NO [142].

References see page 938

926 5 Elementary Steps and Mechanisms

invoked. A fit to a Kisliuk model [131, 1321 (see ments, for example, via low energy electron diffraction
above) indicates that not only chemisorption is favored (LEED). Furthermore, surface diffusion helps to over-
for N-end oriented molecules but also trapping into a come lateral concentration gradients due to non-
presursor state. If we change the adsorbate system from equilibrium clustering phenomena often found at low
a chemisorptive system such as NO on Pt(100) to a temperatures. There is a large amount of information
more weakly interacting system such as NO/Ag(l 1 1) available on surface diffusion [129, 148, 1491, both on
[137] we realize that the observed symmetries are ac- the experimental methods to measure diffusion co-
tually much smaller even at low temperature, and in- efficients as well as on the theoretical aspects of the
deed, slightly favor trapping of NO molecules with the problem. We shall only give a brief, nonexhaustive
0-end approaching the surface even at lower coverage. overview of the situation [7].
Conceptually, the process is thought to occur as a
random walk where adparticles hop between adjacent
5.1.1.4 Surface Diffusion sites, i.e. from an occupied to an adjacent empty site.
The hopping frequency depends then exponentially on
The motion of adsorbed particles obviously plays an the temperature of the system which leads to the fol-
important role for adsorbates and for surfaces in gen- lowing form of the diffusion coefficient:
eral, because this process enables the system to achieve
its equilibrium structure. Particularly, at elevated tem- D = DOexp(--) AEdiff
(27)
peratures the atoms of the substrate material can move,
lowering the free energy content of the surface. The with the preexponential factor DO and the activation
process of diffusion of substrate atoms has been in- energy for diffusion A E d i f f . It is correlated with the
vestigated frequently in the past. Applying various height of the energy barrier in Fig. 10 parallel to the
methods such as scattering methods, field emission and surface. An expression for DO may be derived from
contact potential measurements Bonze1 and co-workers transition state theory and depends on the activation
[143-1451, Butz and Wagner [146, 1471, Ehrlich [148], entropy of the process. The important quantity for
and Holzl and co-workers [149, 1501 have contributed surface diffusion is the activation energy. Its magnitude
to this area. Due to the rather high activation energies is about a tenth of the adsorption energy for a typical
required for the substrate atom displacements, tem- chemisorbate such as CO/Pd, i.e. it amounts to ap-
peratures up to 1OOOK have to be employed in order proximately 15-20 kJ mol-' . For physisorbates it is
to obtain reasonable rates of diffusion of substrates probably considerably lower.
atoms. In connection with the discussion of chem- The diffusion coefficient may be measured via several
isorption, however, we are more concerned with a dif- experimental techniques. The most prominent ones
ferent type of surface diffusion, namely when diffusion at present are the direct observation of a diffusion
occurs within the adsorbate phase. Such processes may boundary in either a field electron microscope [159,
be separated from the motion of substrate atoms be- 1601 or a photoelectron emission microscope [158] or
cause much lower temperatures are needed to induce via laser desorption experiments [ 161, 1621. In the latter
diffusion. Typical diffusion coefficients are given in case a short laser pulse is used to heat the surface to
Table 4. momentarily desorb the adsorbate from a well defined
Diffusion within the adsorbed layer is instrumental region of the crystal. Subsequent laser pulses with well
to establish long-range order and to obtain optimal defined time delays with respect to the first one, and
experimental conditions to perform diffraction experi- measurement of the number of particles leaving the
surface, allow one to determine the rate of diffusion
into the depleted zone. Other methods to determine
Table 4. Diffusivities of adsorbates. surface diffusion are spectroscopicmeasurements which
cover the proper time window, for example magnetic
Adsorbate Substrate Do (cm2s-l) References
resonance-based methods [163, 1641. In favorable cases
cs W(110) 0.23 151 these methods may even be applied to single crystal
K ViP) 1o -~ 152 surfaces [165].
N W(110) 0.014 153 As mentioned above, the diffusion process is thought
0 W(110) 0.04-0.25 154
to be a random walk across the surface. Then the
H Ni( 100) 2.5 x 10-3 155
D Ni( 100) 8.5 x 10-3 155 mean-square displacement of the adparticles is related
H WiP) 1.8 x 10-5 156 to the diffusion coefficient via the relation
D Pt(ll1) 8 x lo-' 157
co Pt(ll1) 10-~-10-~ 157 (A?) = 4 Dt (28)
co Pt(iio) [iio] 2.1 x 10-9 158 where is it understood that the surface itself only con-
co Pt(ll0) [OOl] 0.8 x 10-9 I58
tains a very low concentration of adparticles which do

not interfer with each other. In other words, the model
so far is coverage independent. However, we know
from previous considerations that coverage dependence
has to be considered. For example, if a particle wants
to move to an empty site the probability to hop clearly
depends on the number of empty sites in the neigbor-
hood, or even on the concerted motion of adparticles.
Coverage dependences may be introduced by using the
general transport equations, or specifically Fick's law
[166]. The solution of Fick's law again yields an ex-
ponential dependence of the diffusion coefficient as in
equation (27):
(29)
where the coverage dependence of the process enters
through a coverage dependence of the activation energy:
( 30)
where B is the short-range order parameter, B =
1 - exp(Epair/RT),2 is the number of nearest neigh-
bor sites, and Epair is the nearest neighbor interaction
energy. Using this approach Fick's equation may be
solved numerically.
Table 4 contains a collection of diffusion coefficients
determined experimentally for a variety of adsorbate
systems. It shows that the values may vary consid-
erably, which is of course due to the specific bonding of
the adsorbate to the surface under consideration. Sur-
face diffusion plays a vital role in surface chemical re-
actions because it is one factor that determines the
rates of the reactions. Those reactions with diffusion as
the rate-determining step are called diffusion-limited
reactions. The above-mentioned photoelectron emis-
sion microscope is an interesting tool to effectively
study diffusion processes under reaction conditions
[158]. In the world of real catalysts, diffusion may be
vital because the porous structure of the catalyst par-
ticle may impose stringent conditions on molecular
diffusivities, which in turn leads to massive conse-
quences for reaction yields.
5.1.1.5 Structure Sensitivity
So far we have neglected the fact that the substrate has
a particular geometric structure which influences, as we
shall see further below, the adsorption behavior in a
very pronounced way. Furthermore, in practical cases
the macroscopic geometric structure is rather complex.
Consider, for example, a real catalyst used in hetero-
geneous reactions. It may consist of bimetallic precip-
itates, or of thin films supported on alumina, silica, or
5.1 Chemisorption 927
titania, or highly dispersed metals such as platinum
black. However, even these materials possess a regular
geometric structure on the microscopic scale. Often,
microscopically analyzed, these materials expose regu-
lar crystallographic planes, which may be characterized
via scattering methods or real-space imaging. In catal-
ysis, the correlation between surface geometric and
electronic structure, the geometric shape and electronic
structure of a molecule, and the observed macroscopic
reactivity represents a very important and long dis-
cussed, but not yet solved problem. One distinguishes
between structure-sensitive and structure-insensitive
reactions. Special site requirements have been discussed
in terms of the so-called ensemble effect [167-1691
whereafter a molecule can only adsorb if a certain
group of adjacent surface atoms is available. Studies
on bimetallic alloy surfaces have often been used as
examples for such ensemble effects [167, 1681.
The present section enters into the discussion of the
electronic and geometric structure by considering first
an example where we can vary the strength of inter-
action between a given adsorbate and various metal
and metal-oxide surfaces. We have chosen carbon
monoxide as the adsorbate because it offers the largest
available data set, including structure determination.
Photoelectron spectroscopy (PES) is a very sensitive
tool with which to monitor the change in the electronic
structure, which is why it is the method of choice to
shed light on this question [170]. Figure 19 shows a set
of photoelectron spectra of CO adsorbates on four
different hexagonally close packed metal surfaces [1711
as well as on two transition metal-oxide surfaces [172-
1731. For comparison we show the spectrum of gaseous
[174] and condensed CO [175]. The binding energy
(& = &in - hv) refers to the vacuum level, which al-
lows us to put adsorbates on metals, on insulators, and
molecular solids on the same energy scale. (Often the
binding energy is referenced to the Fermi level (EF)of
the system. The binding energy with respect to the
vacuum level and the binding energy with respect to
the Fermi level are connected via the work function CD
of the system.) The region where we expect photo-
electron emission from the three outer valence levels of
CO, i.e. 50, ln, and 40 levels, is shown, and most of the
following discussion will concentrate on these levels.
From the bottom to the top the heat of adsorption in-
creases from 19 kJmol-' to 142 kJmol-' for the metal
surfaces. This is accompanied by clearly recognizable
changes in the photoelectron spectra. There are several
interesting differences in binding energies, line inten-
sities and line shapes between gas phase [174], con-
densed phase [175] and adsorbate phases [172, 173,
176-1791, which we shall comment on in the following.
We shall start with the adsorbates on the metal surfaces
References see page 938
928 5 Elementary Steps and Mechanisms
25 20 1'5 10
binding energy
Figure 19. Photoelectron spectra of CO adsorbed on metal and
metal oxide surfaces in comparison with gaseous and condensed
CO. The spectra are taken in normal electron emission.
[176-1791, and later turn to the oxide surfaces [172,
1731 because bonding considerations are rather differ-
ent for these systems.
In order to systematically approach an under-
standing of molecule-metal bonding and to relate the
conceptual considerations to experiment we briefly re-
fer to Fig. 20 [171]. In this figure the molecule-metal as
well as the molecule-molecule interaction effects are
illustrated on the basis of a one-electron level diagram
for the valence electrons It shows on the right-hand
side a diagram for an isolated CO molecule correlated
with a one-electron level diagram for a CO molecule
interacting with a single metal atom. On the left-hand-
side the band structure of an isolated CO overlayer is
Figure 20. Schematic diagram for the bonding of an isolated CO
molecule to a metal atom (right hand side) and a free two-
dimensional array of CO molecules (left panel) to a metal surface
(middle).
displayed, and compared in the middle with the full
band structure of the CO adsorbate interacting with the
compact metal substrate with (1 11) orientation. Both
aspects, the molecule substrate as well as the inter-
molecular interactions, have consequences for the ob-
served spectra, but the main effect we shall dwell on
first is the molecule-metal interaction. What happens
electronically can easily be explained in the so-called
Blyholder model [IXO]. The carbon lone pair of CO is
donated into empty d or s levels of the metal atom, es-
tablishing a a-metal-molecule interaction; synergeti-
cally, metal d electrons are donated into empty mo-
lecular orbitals (2.n) of CO forming a n-metal-molecule
interaction. From the view point of the molecule we
can look at this charge exchange process as a .n-dona-
tion-n-backdonation process. This means that the dis-
tribution of electrons among the subsystems, i.e. CO
molecule and metal atom, in the CO-metal cluster
is considerably different to the noninteracting sub-
systems. For example, the electron configuration of the
metal atom in the cluster may be different from the
isolated metal atom, or the electron distribution within
the CO molecule bonded towards the metal atom may
look like the electron distribution of an "excited" CO
molecule rather than the ground state CO molecule
[181]. This scheme has been used to explain the well
known changes in the vibrational properties of ad-
sorbed CO as compared with the gas phase. In addition
to the loss of the rotational fine structure upon ad-

sorption, the CO stretching frequency often shifts by
more than l00cm-' to lower values [182-1851. It is the
filling of the CO antibonding 271 orbital via the back-
donation contribution which weakens the CO bond in
the adsorbate and concommitantly shifts the stretching
frequency to lower values [186]. Also, as a consequence
of this interaction certain electronic levels of the sub-
systems are strongly influenced. Naturally, the dis-
tortions of the molecular as well as the metal levels are
reflected by changes in the ionization energies, their
ionization probabilities, and the line shapes of the ion-
ization bands. In CO/Ag(lll) [176] at T = 20 K CO is
physisorbed as documented by the small adsorption
energy of 19 kJ mol-' . This explains why a spectrum so
similar to condensed CO is observed for this adsorbate.
The splittings in the 40 and 50 ionizations are con-
nected with the formation of a two-dimensional layer
and will not be discussed at t h s point [187]. If com-
pared with the gas phase, however, rather dramatic
changes are found. The bands are shifted by about 1 eV
to lower binding energy and the line widths increase,
which destroys to a large extent the vibrational fine
structure observed in the gas phase, too. Theories have
been developed that allow one to understand these
processes on the basis of hole hopping and relaxation,
i.e. effects in the ionized state, within the quasi-two-
dimensional solid but for the present review we refer
to the literature for details [188-1931. If the heat of
adsorption increases to about 47 kJ mol-' [194], as for
example in the case of CO on Cu(l11) [177], the
features in the spectrum shift and the intensities are
altered. Three lines are still found but their assignment
is very different as compared with the physisorbate
[177].
We only briefly state here that many-particle effects
in the ionized state of the adsorbate due to the presence
of the highly polarizable metal electrons dominate the
spectrum, and this alters the assignment considerably
[177]. If we later turn to the oxide surfaces where such
effects do not occur as strongly but the bond strength is
comparable, we shall see that the interaction may be
directly deduced from the spectrum. We note in passing
that the assignment of the bands to states of different
symmetry has been made on the basis of experimental
investigations using angle resolved photoelectron spec-
troscopy (ARUPS) [170]. Reviews on this subject exist
in the literature [170, 195, 1961. The next step is the
study of the strongly chemisorbed systems with ad-
sorption energies larger than lOOkJmol-'. Out of a
wealth of experimental data [197-2141 we have shown
here only two systems, i.e. CO/Ni(l11) and CO/
Pd(l11). In these cases the spectra show two bands,
whose binding energies are almost independent of the
particular system under consideration as long as inter-
molecular interaction does not play an important role.
The bands are shifted by more than 2eV with respect
5.1 Chemisorption 929
to the gas phase. Via angle resolved photoelectron
spectroscopy [170] it has been shown that the two
bands really contain three components as indicated in
the figure as well as expected from the simple bonding
considerations made above [197]. The carbon lone pair
is shifted close to the 171 ionization due to the strong
charge exchange and is actually located at higher
binding energy. The overall shift of the bands to lower
binding energy is a consequence again of the relaxation
in the ionized state of the adsorbate due to the presence
of the highly mobile metal electrons. Therefore, the
experimental observation are in line with our simple
charge-exchange model for CO-metal bonding but one
has to be careful in the interpretation not to forget the
effects of the probe, in the case of PES the creation of a
hole in the system [170].
We now come to the comparison of the electronic
structure of the adsorbates on the metal surfaces with
those on the oxide surfaces [215, 2161. Very detailed
electronic structure calculations [217-2241 have re-
cently shown that the interaction of molecules with
oxide surfaces differs considerably from the interaction
with metal surfaces in the sense that in the latter case
interaction, at least on the regular surfaces, is much
weaker. However, it is not necessarily a physisorptive
interaction. Briefly, on the (100) surface of the strongly
ionic NiO the interaction of a CO molecule is not gov-
erned by short-range charge-exchanges processes as in
the case of the metal surface but rather by electrostatic
interaction between the multipolar moment of the mo-
lecular electron density and the multipolar moment of
the ionic surface. The reason for this behavior is that
due to the presence of the closed shell oxygen ions in
the (100) surface the molecule cannot approach the Ni
site close enough to exchange charge. Pauli repulsion
sets in at rather large distances from the surface and
repels the molecule. The balance between the electro-
static attractive forces and the Pauli repulsion results in
a rather weak chemisorptive bond of CO on a typical
oxide surface. In addition, due to the rather weak in-
teraction there is no longer a strong preference for one
given orientation of the molecule with respect to the
surface. For example, the molecular axis may be either
perpendicular or tilted, or there may be interaction
either with the carbon end or the oxygen end of the
molecule with the surface. In other words, from an
experimental point of view, we have to check in each
case individually which situation is adopted by the
system [172].
To a certain extent, the vibrational spectra [225, 2261
again provide a clue towards a verification of the gen-
eral statement made above. On oxide surfaces, in gen-
eral, the observed shifts of the stretching frequencies
References see page 938
930 5 Elementary Steps and Mechanisms
Figure 21. Schematic proposed arrangement of CO on a CrzOl
surface [173].
are considerably smaller as compared to adsorbates on
metals [183-1851. The vibrations may be either red or
blue shifted depending on the interaction. The small
red shift observed in some cases may be interpreted by a
limited charge transfer from the oxide to the adsorbed
molecule in the same sense as for adsorbates on metal
surfaces. The often observed blue shift, however, has a
different origin. It can be explained by the so called
“wall effect” [219, 2201 in which the weakly held C O
molecule vibrates against the hard wall of the substrate
which shifts the stretching frequency to higher values,
thus leading to a blue shift. The statement made above
concerning the interaction of CO with the oxide surface
can now also be verified via the photoelectron spectra
in Fig. 19 [172]. We find the binding energy of the
oxygen lone pair located very close to the energy in the
condensed C O film indicating that there is no strong
intermingling between the oxgen long-pair density and
the surface electrons. The same is true for the CO 71-
bond electrons. However, we see a pronounced shift
of the carbon lone pair electrons originating from
the strong Pauli repulsion with the surface electronic
charge. The relaxation shift found for the metal oxide
systems is rather small also because the response of
the oxide surface towards the creation of holes on the
molecule in the ionization process is less pronounced
than with the metal surface. Comparing the spectrum
for the CO/Ni0(100) system with the last example, i.e.
CO/Cr203(111) [ 1731 indicates a similar situation as far
as the overall position of the adsorbate induced fea-
tures are concerned. However, a detailed analysis of
this fi x fi)-ordered adsorbate system shows that the
individual ionizations are considerably shifted with re-
spect to the CO/Ni0(100) system. The reason is simple,
and it can be proved by angle resolved photoelectron
measurements or X-ray absorption measurements, that
the orientation of the molecule with respect to the sur-
face has changed. CO is no longer vertically oriented
on the surface but rather strongly inclined. A schematic
model of the local bonding situation is shown in Fig.
21. The analysis of the chromium oxide system under-
lines the necessity of determining individually the ori-
entation of the molecular axis before we discuss the
2060
term;”‘-p f
205C
r
- 1900
Y
>
5 1880
5
a,
’
0
1860
18LO
1820
0.2 OIL
I
0.6
CO coverage Gco
Figure 22. Stretching frequency of CO adsorbed on Ni(l11) as a
function of CO coverage. The surface was dosed at 9 0 K and
subsequently annealed to 240 K [ 1841.
details of the electronic structure of a system, in par-
ticular on an oxide surface.
The next step in the discussion of structure sensitivity
of chemisorption is to consider the site of adsorption
on a given surface and to answer the question as to
whether and how the site changes as the coverage of
the adsorbate is increased. Figure 22 shows the famous
dependence of the C O stretching frequency on cover-
age for the system CO/Ni( 111) [184]. This dependence
has been interpreted as being due to two effects, namely
a change of adsorbate site upon increase of coverage
and additionally a shift caused by the coupling of the
dynamic dipoles which depends on intermolecular dis-
tance [183-1851. Figure 22 indicates the adsorbate
geometry deduced for the various coverage ranges
based on the stretching frequency data. In recent years
it has become more and more clear, however, that a
structural assignment based on vibrational data has to
be taken with caution. At low coverage a CO stretching
vibration at 1816cm-’ shows up. This is replaced by a
band at 1831 cm-’ if the coverage increases and even-
tually shifts to 1905cm-’ at @ = 0.5 corresponding to
a c(4 x 2) structure. On the basis of the suggestions by
Eischens and Pliskin [182] the band at 1816cm-’ has
been interpreted to be due to adsorption in a threefold
hollow site at low coverage and the band shifting in the

u w w the intermolecular distance would be 2.5 A if the
Figure 23. Schematic drawing of the geometric arrangement CO
on Ni(ll1) in the c(4 x 2) superstructure. Values are distances as
determined by XPD [228].
range 1831-1905cm-* to a CO bridge site. In a very
convincing study based on the analysis of X-ray pho-
toelectron diffraction data Bradshaw and coworkers
[227-2291 have shown that the adsorbate site over the
whole coverage regime remains the same and is a
threefold hollow site as indicated in Fig. 23. The ob-
served shift in the stretching frequency is then purely
due to intermolecular dynamic dipole coupling. Note
that both the inequivalent threefold hollow sites (fcc
and hcp) are occupied in this structure [228]. Another
important factor in chemisorption becomes obvious by
looking at the structures in Fig. 23, namely the coop-
erativity of the process. There is a 3% expansion
($0.07 A) of the outermost Ni-Ni lattice spacing. This
is meant here to stress the finding that although the
surface provides a particular site for adsorption, the
final geometry is determined via the interaction with
the adsorbate and therefore depends on its chemical
identity. This phenomenon is important in connection
with the well known adsorbate induced reconstructions
of surfaces [135]. If the reactivity of the surface towards
another adsorbate changes through the reconstruction
then cooperative phenomena are essential for the over-
all chemical reactivity in the system.
Whereas in the above example the local structure
remains the same for increasing coverage, there are
other cases where intermolecular interaction changes
the geometry of the adsorbate. In the case of CO on
Ni(ll0) at low coverage CO molecules adsorb in two
different adsorption sites, namely on atop and on
bridge sites with vertically oriented axis, as shown in
Fig. 24 [230-2331. The molecule-substrate bond in this
case is so strong that the system can tolerate even large
lateral intermolecular stress. At a coverage of 0 = 1
5.1 Chemisorption 93 1
Figure 24. Geometric arrangement of CO molecules on a
Ni(l10) surface at low coverage (right) and high coverage (left).
molecular axis remained perpendicular. Therefore the
molecular axis tilts from the normal orientation in
order to enlarge the average distance between mole-
cules [234-2361. The equilibrium structure assumed is
shown in Fig. 24, on the left. This system has been
studied in some detail in order to understand the elec-
tronic structure of the system [237-2401. In line with
Fig. 20 where the schematic band structure of an ad-
sorbate system is shown, the present system has been
studied with angle resolved photoelectron spectroscopy
and the band structure has been experimentally deter-
mined [187]. Figure 25 shows the complete experi-
mental band structure in the occupied region, i.e. of the
50, In, and 40 levels [237, 239, 2401. Included is the
band structure in the unoccupied region as determined
by inverse photoemission. The 50, ln, and 40 levels
lead to twice the number of bands due to the non-
symmorphic space group symmetry of the system with
two molecules per unit cell [234]. Following the bands
through the Brillouin zone shows that the energetically
close 50 and In bands hybridize. Also, one can clearly
identify the C0(2n)-Ni(3d) backbonding states below
the Fermi edge. The unoccupied 271 derived levels are
located above the Fermi edge. It is interesting to note
the different magnitudes of the band dispersions for the
different levels. This is clearly due to the variations in
interaction strength for the different molecular orbitals
depending on directionality and spatial extent. The
largest dispersions are exhibited by the n orbitals. In
fact, the 271 orbital shows the largest effects because
they are most diffuse and show large electron density
off the molecular axis. To summarize, the strong inter-
molecular interaction is reflected in the adsorbate band
structure and mainly due to 71-n interaction.
We now turn to the question of how the adsorption
properties of a given molecule changes when we change
the geometric structure of the surface keeping its
References see page 938
932 5 Elementarv Stem and Mechanisms

CO(2x11 p 2 mgI NI(1101

0
0
. .
0 1
LEE0
-6 X I

top l a y e r
-L- second Lay

-3. (100)
- 2-

(110)

Figure 26. Structure of the close packed surfaces of iron (bcc).

surfaces. The adsorption of nitrogen on iron is chosen

124
A r
I
I:
+X
Y
[lo01 r 1701
Kl,
Figure 25. Measured band structure in the range of occupied
and unoccupied levels for CO(2 x l)p2mg/Ni(llO). The wave
vector K is determined along the two orthogonal directions in
the surface Brillouin zone as shown at the top and its energy de-
pendence according to Kll= ( 2 meh-2Ek,n)''2sin 8.
chemical constitution constant. There are many exam-
ples in the literature. Again CO adsorption could be
chosen [241]. Also, hydrogen chemisorption [73] or
oxygen chemisorption [242, 2431, which has been stud-
ied and reviewed in detail by Christmann [73], Wandelt
[242], Brundle [243] and others are prominent examples
for the structure sensitivity of chemisorption on metal
because of its importance in connection with ammonia
synthesis [244]. In particular, Ertl and co-workers [125,
1261 have investigated the structure sensitivity of dis-
sociative nitrogen adsorption on the low-index surfaces
of iron, i.e. the (loo), (110) and (111) surface ori-
entations. Figure 26 shows the arrangement of these
surface structures on top of the body-centered cubic
iron crystal. The (110) surface has a very low sticking
coefficient for dissociative adsorption while the most
open (1 11) surface has a much higher sticking coeffi-
cient [125, 1261. With a combination of photoelectron
spectroscopy [245] and vibrational spectroscopy [246-
2481 the important factors influencing this face specif-
icity have been uncovered. Briefly, on Fe(ll1) high
resolution electron energy loss spectra (HREELS)
[248] are observed as a function of temperature (Fig.
27). At about liquid nitrogen temperature a dominant
feature with a stretching frequency at 2100cm-' is
found. With angle resolved photoelectron spectroscopy
[245] it has been shown that this species is oriented
perpendicular to the surface. It is most likely to be
bound to an atop site. The same species is found on all
low-index iron surfaces [244]. It is weakly held by the
surface. Upon heating the system slightly above 100 K
a second molecular nitrogen species shows up in the
vibrational spectra of Nz/Fe( 11 1) at a lower stretching
frequency (1415 cm-'). Again, photoelectron spectros-
copy has been used to show that this species is bound
in a strongly tilted geometry, in line with the low
stretching frequency typical for side-on bonded dini-

15
N, IFe(ll1)
I
I 2100
z T=7L K
.-v,
c
C
01
c
U
N
0
E0
I\ I I
0 800 1600 2LOO
electron energy Loss [cm-']
Figure 27. Electron energy loss spectra of 15N2on Fe(ll1) as a
function of surface temperature [248].
top layer
second layer
third layer
W the activation barrier is based on experimental date
Figure 28. Proposed arrangement of N2 on Fe(ll1) [245].
trogen complexes. This species only exists on the sur-
face within limited temperature range. Above 160 K
the stretching frequency typical for molecular nitrogen
species disappears and only atomic nitrogen (460 cm-')
is present on the surface. This scenario is typical for the
(1 11) surface, while the existence of the intermediate
species cannot be detected on the other low-index
planes, i.e. (110) and (100) [244]. It is now generally
accepted that the intermediate with the low stretching
frequency is a precursor to nitrogen dissociation, and it
is thought that the (1 11) surface provides the sites,
necessary to assume the strongly tilted geometry [244].
Figure 28 shows the bonding geometry for the inter-
mediate species [245]. The nitrogen molecule can do-
nate both its lone pair as well as the 17c electrons into
empty metal orbitals, and at the same time establish a
5.1 Chemisorption 933
I
0
position on Fe(ll1) surface
Figure 29. Two-dimensional potential energy diagram for the
convesion of y-N2 (vertically adsorbed) to r-N2 (side-on bonded)
[246-2481 on the Fe(l11) surface [249].
back-donative bond via the unoccupied 7c orbital. The
back donation will weaken the nitrogen-nitrogen bond
which finally leads to dissociation. Since both nitrogen
atoms are already in close contact with the metal sur-
face, this picture appears to provide a natural pathway
to dissociation. It is believed to explain the observed
strong face specificity of dissociative nitrogen chem-
isorption on Fe surfaces. Figure 29 shows a semi-
empirical potential energy diagram for N2/Fe( 11 1)
where the pathway from the molecular precursor to the
dissociative adsorption is shown [249]. The value for
[250].
Finally, we would like to have a look at the structure
sensitivity of transition metal oxide surfaces [215]. For
such systems [251, 2521 it is necessary to resort to some
basic considerations about the electrostatics of ionic or
partly ionic systems with respect to surface stabilities.
Figure 30 schematically shows the arrangements of
planes in a crystal of rock salt (AB) structure for the
termination of (100) type on the left and of (1 11) type
on the right [254]. The (100) surface of an AB-type
solid is the typical case for a nonpolar surface with
vanishing dipole moments between the planes and full
charge compensation within the planes. This arrange-
ment leads to a converged, finite electrostatic surface
energy. Upon going to the (1 1 1) surface of an AB-type
lattice we create a polar surface. In this case there is no
charge compensation within each layer and there is
also a dipole moment within the repeat unit perpen-
References see page 938
934 5 Elementary Steps and Mechanisms
stoblelnon-polar stablelpolar
AB -type AB2-type
repeat
vacuum VOCUUm units
Figure 30. Stable and unstable surfaces of AB-type and ABz-type ionic crystals [253]
dicular to the surface. Consequently, the surface energy
does not converge but increases unbound as the num-
ber of repeat units increases. In general, polar surfaces
are not unstable, as illustrated for the ABz-type solids.
Even though there is no charge compensation in the
plane, the dipole moment in the repeat unit perpendic-
ular to the surface vanishes, thus leading to a stable
situation. Returning to polar surfaces of the AB-type to
consider the surface potential V in more detail [253],
2.n
V = -[Nb(20 - 1) + (1 - ~ ) b ]
S Ni( 100)
where S is the area of the unit cell. Equation 31 gives
the surface potential as a function of the number of
layers N, their separation b, and the parameter CJ which
describes the difference in charge of the surface layer
with respect to the bulk layer. It is quite obvious that
the reduction of the surface charge such that CJ = 1/2
leads to the disappearence of the first term in eq 31,
and thus to a converging surface potential independent
of the number of layers. While this is only a qualitative
argument, it shows possible routes for the system to
respond in order to stabilize polar AB-type surfaces.
Surface-charge reduction may be accomplished by re-
constructing [254], i.e. removing half of the ions, or by
the creation of steps. The latter leads to the coexistence
of A-terminated and B-terminated patches on the same
surface and thus to a microscopic charge compensa-
tion. Also, relaxations in the layer distance are ex-
pected to occur in the near surface region which could
help to reduce the surface potential. In certain cases
other causes of stabilization may be considered. Upon
adsorption of H+, provided by exposure to water, for
example, OH- may form on an oxygen terminated
surface thus effectively reducing the surface charge
[252]. Thus one would predict a strong structure sen-
sitivity of water adsorption on oxide surfaces which,
indeed, has been observed [254, 2551 and is exemplified
in Fig. 31. The O(1s) XP spectra are shown for three
different samples [255]. The lower trace shows the
spectrum of a cleaved NiO(100) surface with very low
defect concentration (sharp LEED pattern). The fea-
ture is symmetric after cleavage and it remains sym-
metric even after exposure to water at room temper-
unstableipoiar
AB-type
repeat
vacuum unlts
LEED XPS 01s
grazing excidence
NiO( 1 1I)/
Ni( 1 1 1)
NiO( loo)/
NiO( 100)
cleaved
515 550
Binding energy [ eV ]
Figure 31. O(1s) XP spectra of a cleaved NiO(100) crystal: (a) a
grown NiO(100) film; (b) a grown NiO(l1l) film. The corre-
sponding LEED patterns are shown [254, 2551.
ature [256]. At low temperature an ice layer forms
which can be removed without residue by heating to
room temperature. This indicates that a NiO(100) sur-
face does not dissociatively chemisorb water. The sit-
uation is different for a NiO(100) surface containing
defects, as indicated by the much broader LEED spots
as compared with the cleaved surface [252, 2551. Here a
small feature is found at 2.2 eV higher binding energy.
It becomes particularly pronounced in the spectra if
they are recorded at grazing electron excidence in order
to amplify the surface sensitivity of the method. EELS
investigations have shown that the feature is due to
hydroxyl groups on the surface. These hydroxyl groups
may be removed from a NiO(100) surface by thermal
treatment. Exposure of the cleaned surface to water
leads to the reappearence of hydroxyl, indicating that
water dissociatively chemisorbs on defect sites of a
NiO( 100) surface. Dissociative chemisorption becomes
even more pronounced on the NiO(ll1) surface. The

Figure 32. Schematic drawing of OH-terminated (left), bulk-terminated (middle) and octopolar reconstructed (right) NiO( 1 1 1) surfaces.
upper trace in Fig. 31 indicates a rather high concen-
tration of hydroxyl groups at the surface. When (1 11)
polar surfaces are prepared they often become OH
stabilized, due to the electrostatic instability discussed
above. In favorable cases such as NiO( 11l), the hy-
droxyl groups can be removed from the NiO(ll1) sur-
face as water by thermal treatment. As a consequence,
the OH-free unstable surface reconstructs. The most
stable reconstruction of a polar surface of an ionic
crystal is, according to Lacman [257] and to Wolf
[258], the so-called octopolar arrangement, shown in
Fig. 32 in comparison to the ideal (1 x 1) surface. The
octopolar reconstruction leads to p(2 x 2) unit cell on
the surface and is characterized by the removal of three
out of four oxygen ions in the first layer (in the case of
an oxygen terminated surface) and one out of four
nickel ions within the second layer [254, 2591. The third
layer contains then again a complete hexagonally close
packed oxygen layer. A p(2 x 2) reconstruction has
been observed for iron oxide and nickel oxide but only
in the latter case are there clear indications that an
octopolar reconstruction has actually taken place [255].
Readsorption of water leads to a lifting of the re-
construction and the reoccurrence of the (1 x 1) struc-
ture [254]. Note for completeness that the reconstructed
surface exhibits a considerably higher chemical activ-
ity, for example in the DeNO, reaction, than the
hydroxyl-covered surface which is basically inactive
towards further chemisorption [256]. In other words,
water desorption and readsorption leads to a strong
change in the chemical activity of certain crystallo-
graphic planes of oxide surfaces which may be relevant
with respect to the catalytic activity of powders of real
samples.
Previous sections have discussed the interaction be-
tween adsorbed species in connection with the coverage
5.1 Chemisorption 935
dependence of adsorbate properties and especially en-
ergetics. Coadsorption of different chemical species is
the general case in connection with the discussion of
intermolecular interaction. Intermolecular interaction,
however, is the basis for the understanding of chemical
reactions between adsorbed species. There is such a
vast literature on the subject [260] that a compre-
hensive and exhaustive review of the field cannot be
provided here. Nevertheless, we would like to briefly
address two coadsorbate systems where a broad
knowledge has been accumulated over the years. To
represent the limiting cases we resort again to carbon
monoxide as one component and study its coadsorp-
tion with an electropositive additive, and also with
electronegative additives. Needless to say, all aspects
discussed above for chemisorbate systems in general
are important, even at a more complex level, for
coadsorbate systems. In the latter case it is necessary
to consider the different chemical identities of the ad-
sorbed species, and more importantly their influence on
the electronic structure of the substrate, and on each
other. In other words, the aspect of cooperativity that
adsorbates and substrate interfere and determine each
others properties becomes particularly noteworthy.
The most prominent and most frequently studied
electropositive additives are alkali metals. Several
comprehensive reviews have been published on the
subject which provide more detailed information [260-
2621. Characteristically, adsorption of alkali leads to
dramatic changes of the work function of the system
[260-2621. An example, K on Pt(l1 l), is shown in Fig.
33 [263]. In general, small alkali coverages already
lower the workfunction considerably before monolayer
coverage is reached (in the present case more than
References see page 938
936 5 Elementary Steps and Mechanisms

1.5-
KIPt(ll1)

-2
T=300 K
aI

-
c 00-

Ib
0 I
I I
Q

0 1.5
-

0.0-
0.1 0.2 0.3 I I I
600
I I I
potassium coverage 8 300 LOO 500
Figure 33. Work function of Pt(ll1) as a functionof potassium
temperature [K]
coverage at 300 K [263]. Figure 34. Thermal desorption spectra of (a) clean and (b) po-
tassium covered (0 = 0.015 ) Pt(1 l l). Various CO coverages are
plotted indicating the population of sites close to the alkali at low
4eV). Before completion of the first monolayer cov- CO coverage. On the unmodified Pt(ll1) surface at higher cover-
erage the work function reaches a minimum, turns ages similar sites are observed [265].
around and then approaches, for increasing coverages,
the value of the work function of the bulk alkali [260-
2621. We are concerned only with the regime of alkali 1565
coverages below or close to monolayer coverage. It is
generally accepted that in the low coverage regime the
alkali atoms transfer charge towards the substrate, set-
ting up a strong adsorbate-surface dipole which lowers
the work function, hinders the alkali atoms to cluster
on the surface, and allows them to adsorb as isolated
atoms well separated from each other [264]. The energy
needed to remove the alkali from the surface has been
determined from TDS and calorimetric investigations
to vary between 130 and 250 kJ mol-'. Coadsorption
of CO onto such an alkali-precovered surface leads to
considerable effects on the energetics of the CO-sub-
strate interaction as compared with the pure CO ad-
I
sorbate. TD spectra of the pure and the coadsorbate I I I I 1 1 !
0 1000 2000 3000
system are shown in Fig. 34 [265]. The molecule still energy LOSS [cm-'1
adsorbs associatively on the surface but note that the
dissociative sticking coefficient increases considerably Figure 35. Electron energy loss spectra of CO on clean (lower
in the coadsorbate as has been observed for several trace) and K-modified (upper trace, OK = 0.02) Pt(l1l) [267].
CO-alkali coadsorbates [260]. The adsorption en-
thalpy increases for a typical CO-metal system from
130 kJ mol-' to 197 kJ mol-' for the alkali-adsorbed adsorbate. The explanation is straightforward: elec-
system [266]. There are coverage dependences as well, trons from the electropositive additive are transferred
but we shall concentrate here on a single coverage. To either directly or via the substrate surface into the un-
learn more about how the observed energetic changes occupied CO antibonding orbitals thus weakening the
come about, consider the vibrational spectra of the CO bond [261]. Simultaneously, this stabilizes the CO-
system shown in Fig. 35 [267]. As compared with the alkali interaction on the substrate surface and enhances
pure CO adsorbate the CO stretching frequency in the CO substrate interaction. It turns out, however, to
the coadsorbate is lowered by several hundred wave- be rather difficult to exactly partition the interaction
numbers, indicating a weaker C-0 bond in the co- strength between CO-alkali and CO-substrate. It was
 5.1 Chemisorption 937

Table 5. Desorption energies for some transition metal surfaces modified by electronegative additives.

Adsorbate Surface Modifier E: (kJmol-') References

co Ni( 100) - 140 270

P(2 x 2)s 90
110
C(2 x 2)s x 30
P(2 x 2)O 120
P(2 x 2)N x 90 27 1
(2 x 2)P4gc 93 272
Ni( 111) - 140 273
P(2 x 2)s 91 273
P(2 x 2)O 105 274
Pd( 100) - 160 275
P(2 x 2)s 86
60
Pt(l11) - 154 276
P(2 x 2)s 106 276
P(2 x 2)Se 110 271
Ru(0001) - 170 278

H Ni( 100)
P(2 x 2)s
-
P(2 x 2)s
105
102 5*
84 10
278
279

C(2 x 2)s 48 & 16

Fe( 100) - 87 50 280
P(l x 1)O 60 & 10 280
Pd( 100) - 85 28 1
0.15s 49 28 1

believed for some time that in the coadsorbate the CO-
substrate interaction changes dramatically, leading to a
change in the CO bonding geometry on the surface, i.e.
from a vertically bound CO in the pure adsorbate to a
side-on-bonded CO in the coadsorbate [268]. Near
edge X-ray absorption fine structure (NEXAFS) has
again been instrumental in showing that this is not the
case [269]. In fact, CO remains vertically bonded on
the surface and possibly interacts side-on with the
coadsorbed alkali atom. The side-on geometry was
particularly attractive because, similar to the case of
nitrogen adsorption, this geometry could easily explain
the increased dissociative sticking coefficient [260].
However, as it stands today, either the molecules tran-
siently pass through such a side-on geometry before
dissociation, and the concentration is so low that it
cannot be identified, or dissociation can also start from
vertically oriented, but electronically strongly modified
CO. The described interaction between alkali and CO
in a coadsorbate may be considered as special case of
alkali promotor action, which is well established in
catalysis [260].
A completely different situation is encountered when
we coadsorb carbon monoxide with an electronegative
species. Table 5 [260] collects desorption energies for
carbon monoxide absorbed on transition metals modi-
fied by electronegative additives. In general, and op-
posite to the effect observed for the electropositive
modifier, coadsorption with electronegative modifiers
leads to a decrease in the desorption energy. This may
have different reasons. It could be due to repulsive
modifier-CO interaction, or it could be due to the fact
that the modifier blocks those sites of the surface lead-
ing to the strong CO-substrate interaction for the clean
surface [260]. As judged from the vibrational data the
influence of an electronegative additive onto the CO
stretching frequency is much less pronounced if com-
pared to the electropositive additives [282]. Often, in-
stead of a strong red shift as observed for electro-
positive coadsorbates, a weak blue shift is observed
which in certain cases may even lead to stretching fre-
quencies higher than in the gas phase [282]. In this case
it is even more difficult to disentangle the various con-
tributions, i.e. direct and substrate-mediated inter-
actions. The wealth of data presently available suggest
than an electronegative additive mainly influences the
substrate locally, i.e. in its direct vicinity, in the sense
that (a) the adsorption sites which involve substrate
atoms directly coordinated to the modifier are blocked,
and (b) the adsorption sites sharing some substrate
atoms with the modifier are substantially perturbed.
This means, as schematically shown in Fig. 36 [260],
that for a fcc(lO0) plane and a modifier residing in a
fourfold site four atop sites and four bridge sites are
blocked, and eight bridge sites, four close and four re-
mote fourfold sites are perturbed. With increasing
References see page 938
938 5 Elementary Steps and Mechanisms
Figure 36. Schematic representation of the influence of an ad-
sorbed electronegative modifier in a fourold hollow site (black
circle) on the neighboring sites. Neighboring fourfold hollow sites
which are strongly (+) or less strongly (-) influenced are in-
dicated [260].
modifier coverage, the number of affected sites per
modifier atom decreases, because of overlap of effects.
Following this simple model, obviously the most
tightly bound states of the coadsorbate associated with
the unaffected surface should be rapidly removed with
increasing modifier coverage. In the case of Ni(100)
where CO favors the atop site on the clean surface, one
S atom, for example, is found to remove four atop CO
molecules [270]. In addition, as evidenced by EELS
data and TD spectra [270], CO is pushed to occupy the
close bridge sites and the close fourfold hollow sites, in
which the coupling with the substrate is reduced. Gen-
eralizing this observation if can be stated that co-
adsorption of electronegative additives often leads to a
situation where a coadsorbate accesses the less strongly
bound surface sites, which in turn leads to a strong
modification of the reactivity in the presence of the
modifier [260]. In catalysis such coadsorbate-induced
effects are summarized as poisoning, particularly in
connection with sulfur. However, as we have tried to
indicate, due to the complexity arising from the specif-
icity of the modified systems and the variety of possible
interactions in the coadsorbed layer, there is still no
uniform model for the mechanism of the poisoning ac-
tion, even for the idealized model systems considered in
this section [260]. The situation becomes even more
complex if the additive induces surface reconstruction
or even surface compound formation. Also, if there is a
lack of a tendency to ordering and occupation of a
definite adsorption site, coadsorption phenomena be-
come increasingly complicated. In the case of modify-
ing the surface with C , N, or 0 as opposed to S, P etc.,
transition metal surfaces may reconstruct or carbides,
nitrides, and oxides are formed [260]. Then, of course,
the activity of the surface is determined by the proper-
ties of the new types of compounds formed. Often
island formation is encountered in these systems which
leads to a considerable reduction in the relative number
of modified surface sites because the effect is restricted
to the neighbors of the modifier island boundary.
Summarizing, in such systems the problem of coopera-
tivity, i.e. the phenomenon that the adsorbate and
coadsorbate create their own active sites which are not
present on the clean surface becomes particularly im-
portant. The future study of these effects in chem-
isorption is essential, even under ambient conditions, in
order to identify which are the key effects that operate
during catalysis at a microscopic level.
References
1. G. Wedler, Adsorption, Chemische Taschenbucher, Vol. 9,
Verlag Chemie, Weinheim, 1970,
2. I. Langmuir, J. Amer. Chem. SOC.1916,38, 2221.
3. I. Langmuir, J. Amer. Chem. SOC.1918, 40, 1361.
4. W. Friedrich, P. Knipping, M. v. Laue, Sitzungsber. Buyer.
Akad. Wiss. 1912, 3031367.
5. T. N. Rhodin, G. Ertl (Eds), The Nature of the Surface
Chemical Bond, North-Holland, Amsterdam, 1979.
6. A. Zangwill, Physics at Surfaces, Cambridge University
Press, Cambridge, 1988.
7 . K. Christmann, Introduction to Surface Physical Chemistry,
Topics in Physical Chemistry, Vol. 1, Steinkopff Verlag,
Darmstadt, 1991.
8. G. Ertl, J. Kuppers, Low Energy Electrons and Surface
Chemistry, 2nd ed. Verlag Chemie, Weinheim, 1985.
9. A Clark, The Theory of Adsorption and Catalysis, Academic
Press, New York, 1970.
10. J. G. Dash, Films on Solid Surfaces, Academic Press, New
York, 1975.
11. S. Cerny in The Chemical Physics of Solid Surfaces and
Heterogeneous Catalysis (Eds: D. A. King, D. P. Wood-
ruff), Elsevier, Amsterdam, 1983, Vol. 2, pp 1-57.
12. W. A. Steele, The Interaction of Gases with Solid Surfaces,
Pergamon Press, Oxford, 1974.
13. K. Christmann, J. E. Demuth, Surf: Sci. 1982, 120, 291.
14. 0. Beeck, W. A. Cole, A. Wheeler, Discuss. Faraday SOC.
1950,8, 314.
15. J. Bagg, F. C. Tompkins, Trans. Faraday. SOC.1955, 51,
1071.
16. D. Brennan, D. 0. Hayward, B. M. W. Trapnell, Proc. Roy.
SOC.A (London) 1960,256, 81.
11. G. Wedler, Z. Phys. Chem. (FrankfurtlM.) 1960,24, 73.
18. F. J. Brocker, G. Wedler, Discuss. Faraday SOC.1966, 41,
81.
19. G. Wedler, H. Strothenk, Ber. Bunsenges. Phys. Chem.
1966, 70, 214.
20. G. Wedler, I. Ganzmann, D. Borgmann, Ber. Bunsenges.
Phys. Chem. 1993,97,293.
21. C. Pluntke, G. Wedler, G. Rau, Surf: Sci. 1983, 134, 145.
22. K. Christmann, G. Ertl, 0. Schober, J. Chem. Phys. 1974,
60, 4719-4127.
23. D. A. King, Phys. Scr. 1993, T49, 560-562.
24. J. C. Tracy, J. Chem. Phys. 1972, 56, 2736-2747.
25. J. B. Benziger, R. J. Madix, Surf: Sci. 1979, 79, 394.
26. M. Kiskinova, D. W. Goodman, Surf: Sci. 1981, 108, 64.

27. B. E. Koel, D. E. Peebles, J. M. White, Surf: Sci. 1983, 125,
709.
28. K. Klier, A. C. Zettelmoyer, H. Leidheiser Jr., J. Chem.
Phys. 1970, 52, 589.
29. R. S . Bordoli, J. C. Vickerman, J. Wolstenholme, Surf: Sci.
1979, 85, 244.
30. F. Labohm, C. W. R. Engelen, 0. L. J. Gijzeman, J. W.
Geus, G. A. Bootsma, J. Chem. SOC.,Faraday Trans. 1982, I
78, 2435.
31. J. C. Tracy, P. W. Palmberg, J. Chem. Phys. 1969, 51,
4852-4862.
32. R. J. Behm, K. Christmann, G Ertl, M. A. van Howe, J.
Chem. Phys. 1980, 73,2984-2995.
33. G. Ertl, J. Koch, Z . Naturforsch. 1970,25a, 1906-1911.
34. H. Pfniir, P. Feulner, H. A. Engelhardt, D. Menzel, Chem.
Phys. Lett. 1978, 59, 481-486.
35. G. Lauth, T. Solomun, W. Hirschwald, K. Christmann,
Surf: Sci. 1989,210, 201-224.
36. J. C. Tracy, J. Chem. Phys. 1972,56, 2748-2754.
37. G. Wedler, I. Ganzmann, D. Borgmann, Appl, Surf: Sci,
1993,68, 335-340.
38. N. Al-Sarraf, J. T. Stuckless, C. E. Wartnaby, D. A. King,
Surf: Sci. 1993,283, 427-437.
39. K. Christmann, 0. Schober, G. Ertl, M. Neumann, J.
Chem. Phys. 1974,60,4528-4540.
40. H. Rinne, PhD thesis, Universitat Hannover, 1974.
41. H. Conrad, G. Ertl, E. E. Latta, Surf: Sci. 1974,41,435-446.
42. R. J. Behm, K. Christmann, G. Ertl, Surf: Sci. 1980, 99,
320-340.
43. M. Ehsasi, K. Christmann, Surf: Sci. 1988, 194, 172-198.
44. G. Lauth, E. Schwarz, K. Christmann, J. Chem. Phys. 1989,
91, 3729-3743.
45. K. H. Ernst, PhD thesis, Freie Universitat Berlin, 1990.
46. C. E. Borroni-Bird, N. Al-Sarraf, S . Anderson, D. A. King,
Chem. Phys. Lett. 1991,183, 516.
47. C. E. Borroni-Bird, D. A. King, Rev. Sci. Instr. 1991, 62, 1.
48. H. J. Dixon-Warren, M. Kovar, C. E. Wartnaby, D. A.
King, Surf: Sci. 1994, 16, 307-309.
49. R. J. Behm, K. Christmann, G. Ertl, Surf: Sci. 1980, 99,
320-340.
50. R. Miranda, S . Daiser, K. Wandelt, G. Ertl, Surf: Sci. 1983,
131, 61.
51. M. Grunze, P. A. Dowben, R. G. Jones, Surf Sci. 1984,
141, 455.
52. M. Golze, M. Grunze, W. Unertl, Progr. Surf Sci. 1986,22,
101.
53. G. Ehrlich, Ado. Catal. 1963, 14, 255.
54. P. A. Redhead, Vacuum 1962, 12, 203.
55. D. A. King, Surf Sci. 1975, 47, 384.
56. D. Menzel in Interactions on Metal Surfaces, Topics in
Applied Physics Vol. 4, Springer, Heidelberg, 1975.
57. L. A. Petermann, Progr. Surf: Sci. 1972, I, 2.
58. J . W. Niemantsverdriet, Spectroscopy in Catalysis; An
Introduction, Verlag Chemie, Weinheim, 1993.
59. M. Polanyi, Trans. Faraday SOC.1932,28, 314.
60. C. M. Chan, R. Aris, W. H. Weinberg, Appl. Surf: Sci.
1978, I, 360.
61. E. Habenschaden, J. Kiippers, Surf: Sci. 1984, 138, L147.
62. A. M. de Jong, J. W. Niemantsverdriet, Surf Sci. 1990, 233,
355.
63. J. L. Falconer, R. J. Madix, Surf Sci. 1975, 48, 393.
64. J. E. Lennard-Jones, Trans. Faraday SOC.1932,28, 333.
65. V. Bortolani, N. H . March, M. P. Tosi (Eds.), Interaction of
Atoms and Molecules with Solid Surfaces, Plenum Press,
New York, 1990.
66. M.-C. Derjonqueres, D. Spanjaard Concepts in Surface
Physics, Springer Series in Surface Science Vol. 30, Springer,
Heidelberg, 1993.
5.1 Chemisorption 939
67. M. Scheffler, K. Horn (Eds) Handbook of Surface Science,
in preparation.
68. P. S. Bagus, G. Pacchioni J. Electron Spectrosc. Relat. Phe-
nom. 1994,69, 1-109.
69. H. A. Michelsen, C. T. Rettner, D. J. Auerbach in Surface
Reactins (Ed.: R. J. Madix), Springer Series in Surface Sci-
ence Vol. 34, Springer, Heidelberg, 1994, p. 185.
70. G. Ertl, Ber. Bunsenges. Phys. Chem. 1982, 86, 425.
71. J. C. Polanyi, Science 1987,236, 680.
72. G. Comsa, R. David, Surf: Sci. Rep. 1985, 5, 145.
73. K. Christmann, Surf: Sci. Rep. 1988, 9, 1.
74. M. Balooch, M. J. Cardillo, D. R. Miller, R . E. Stickney,
Surf: Sci 1974, 46, 358-392.
75. G. D. Kubiak, G. 0. Sitz, R. N. Zare, J. Chem. Phys. 1985,
83, 2538-2551.
76. P. Anger, A. Winkler, K. D. Redulic, Surf: Sci. 1989, 220,
1-17.
77. H. F. Berger, M. Leisch, A. Winkler, K. D. Rendulic,
Chem. Phys. Lett. 1990, 175, 425-428.
78. J. K. Nerrskov, P. Stolze, Surf: Sci. 1987, 189/190, 91-105.
79. J. Harris, S. Anderson, Phys. Rev. Lett. 1985, 55, 1583-86.
80. J. Harris, Appl. Phys. 1988, A47, 63-71.
81. J. Koutecky, Trans. Faraday SOC.1958, 54, 1038.
82. (a) T. B. Grimley, Adv. Catal. 1960,12, I ; (b) T. B. Grimley,
Ber. Bunsenges. Phys. Chem. 1971, 75, 1503.
83. T. L. Einstein, J. R. Schrieffer, Phys. Rev. 1973, B7, 3629.
84. G. Ertl in Ref. 5.
85. R . L. Park, T. L. Einstein, A. R. Kortan, L. D . Roelofs
in Ordering in Two Dimensions (Ed.: S . K. Sinha), North
Holland, Amsterdam, 1980, p. 17-24.
86. B. N. J. Persson, Surf Sci. 1991,258, 451.
87. T. L. Hill, Introduction to Statistical Thermodynamics,
Addison Wesley, Reading, 1962.
88. W. Gebhardt, U. Krey, Phasenubergange und kritische Pha-
nomene, Vieweg, Braunschweig, 1980.
89. G. Kortiim, Einfuhrung in die chemische Thermodynamik,
6th ed., Vandenhoeck und Ruprecht, Gottingen, 1972.
90. E. Bauer in Phase Transitions on Single Crystal Surfaces
and in Chemisorbed Layers in Structure and Dynamics of
Surfaces ZZ, (Eds: W. Schommers, P. von Blankenhagen)
Springer, Berlin, 1987, p. 115-179.
91. J. P. Biberian, M. A. van Hove, Surf: Sci. 1982, 118, 443.
92. J. Pritchard, Surf Sci. 1979, 79, 231.
93. D. A. King, M. G. Wells, Surf: Sci. 1972,29, 454.
94. D. A. King, M. G. Wells, Proc. Roy. SOC.London Ser. A.
1974,339, 245.
95. K. D. Rendulic, G. Anger, A. Winkler, Surf: Sci. 1989, 208,
404.
96. C. T. Rettner, L. A. DeLonise, D. J. Auerbach, J. Phys.
1986, 85, 1131.
97. C. T. Rettner, H . E. Pfniir, D. J. Auerbach, Phys. Reu. Lett.
1985, 54, 2716.
98. M. B. Lee, Q.Y. Yang, S . T. Ceyer, J. Chem. Phys. 1987,
87, 2724.
99. A. K. Hamza, H.-P. Steinriick, R. J. Madix, J. Chem. Phys.
1987,86, 6506.
100. A. C. Luntz, D. S. Bethume, J. Chem. Phys. 1989,90, 1274.
101. M. J. Temkin, V. Pyshev, Acta physicochim. USSR 1940,
12, 327.
102. S. Brunauer, P. H. Emmett, E. Teller, J. Amer. Chem. SOC.
1938, 60, 309.
103. H. Freundlich, Colloid and Capillary Chemistry, Methew,
London, 1926.
104. J. Zeldowitsch, Acta physicochim. USSR 1934, 1, 961.
105. D. M. Young, A. D . Crowell, Physical Adsorption of Gases,
Butterworth, London, 1962.
106. J. Lapujoulade, K. S . Neil, Surf: Sci. 1973, 35, 288-301.
107. K. Christmann, Z . Naturforsch. 1979, A34, 22-29.
940 5 Elementary Steps and Mechanisms
108. J. Lapujoulade, K. S . Neil, Chem. Phys. 1972,57,3535-3545.
109. V. Penka, K. Christmann, G. Ertl, Surf: Sci. 1984, 136,
307-318.
110. H. J. Robota, W. Vielhaber, M. C. Lin, J. Segner, G. Ertl,
Surf: Sci. 1985,155, 101-120.
111. K. Christmann, G. Ertl, T. Pignet, Surf: Sci. 1976,54, 365-
392.
112. B. Poelsema, L. K. Verheij, G. Comsa, Surf. Sci. 1985, 152/
153,496-504.
113. B. Polesema, K. Lenz, L. S . Brown, L. K. Verheij, G.
Comsa, Verhandl. DPG 1986, 5:1378.
114. T. E. Madey, Surf: Sci. 1973,36, 281.
115. P. Hofmann, R. Unvin, W. Wyrobisch, A. M. Bradshaw,
Surf: Sci. 1978, 72,635-644.
116. P. J. Holloway, J. B. Hudson, Surf: Sci. 1974, 43, 123-140.
117. F. P. Netzer, R. A. Wille, Surf: Sci. 1978, 74, 547-567.
118. J. Benzinger, R. J. Madix, Surf: Sci. 1980, 94, 119.
119. R. J. Behm, G. Ertl, V. Penka, Surf: Sci. 1985,160, 387.
120. T. Engel, J. Chem. Phys. 1978,69, 373-385.
121. G. Ertl, M. Neumann, K. M. Streit, Surf: Sci. 1977, 64,
393-410.
122. T. E. Madey, J. T. Yates, Nuovo Cimento 1969,5,483.
123. B. J. Hopkins, S . Usamiin in The Stucture and Chemistry of
Solid Surfaces, (Ed.: G. A. Somorjai), Wiley, New York,
1968.
124. M. Bowker, D. A. King, J. Chem SOC.Faraday Trans. 1,
1979, 75, 2100.
125. F. Boszo, G. Ertl, M. Grunze, M. Weiss, J. Catal. 1977, 49,
18.
126. G. Ertl, J. Vac. Sci. Technol. 1983, A l , 1247.
127. K. D. Rendulic, Appl. Phys. 1988, A47, 55.
128. J. N. Russel Jr., I. Chorkendorff, A. M. Lanzilotto, M. D.
Alvey, J. T. Yates Jr., J. Chem. Phys. 1986, 85, 6186-6191.
129. M. A. Morris, M. Bowker, D. A. King in Simple Processes
at the Gas Solid Interface (Eds: C. H. Bamford, C. F. H.
Tipper, R. G. Compton), Elsevier, Amsterdam, 1984, p. 1-
179.
130. W. H. Weinberg in Kinetics of Interface Reactions, Springer
Series in Surface Sciences Vol. 8, Springer, Heidelberg,
1987, p. 94.
131. P. J. Kisliuk, J. Phys. Chem. Solidr 1957, 3, 95.
132. P. J. Kisliuk, J. Phys. Chem, Solids, 1958, 5, 78.
133. G. Ertl, J. Koch, Z. Naturforsch. 1970,25a, 1906.
134. C. T. Campbell, G. Ertl. H. Kuipers, J. Segner, Surf: Sci.
1981,107,207.
135. K. Heinz, Progr. SurJ Sci. 1988,27, 239; Vacuum 1990,41,
328.
136. G. Ertl. Science 1991,254, 1750.
137. M. G. Tenner, E. W. Kuipers, A. W. Kleyn, J. Chem. Phys.
1979,89, 6552.
138. E. W. Kuipers, M. G. Tenner, A. W. Kleyn, S . Stolte,
Nature 1988, 334, 420.
139. G. H. Fecher, N. Bonsering, M. Volkmer, B. Pawlitzky,
K. Heinzmann, Surf: Sci. Lett. 1990,230, L169.
140. G. H. Fecher, M. Volkmer, B. Pawlitzky, M. Bowering,
K. Heinzmann, Vacuum 1993,41,265.
141. H. Miiller, B. Dierks, F. Hamza, G. Zagatta, G. H. Fecher,
N. Bowemig, K. Heinzmann, Surf: Sci. 1992,2691270,207.
142. H. Miiller, G. Zagatta, N. Bowernig, K. Heinzmann, Chem.
Phys. Lett. 1994,223, 197.
143. H. P. Bonzel, Surf: Sci. 1971,27, 387.
144. H. P. Bonzel, R. Ku, J. Chem. Phys. 1973,59, 1641.
145. H. P. Bonzel, E. E. Latta, Surf: Sci. 1978, 76, 275.
146. R. Butz, H. Wagner, Surf: Sci. 1979, 87, 6984.
147. R. Butz, H. Wagner, Surf: Sci. 1979,87, 85.
148. G. Ehrlich, in T. S . Jayadenaah, R. Vanselow (Eds), Surface
Science: Recent Progress and Perspectives, CRC Press,
Cleveland, Ohio, 1974.
149. J. Hokl, G. Porsch, Thin Solid Films 1975,28, 93.
150. P. Schrammen, J. Holzl, Surf: Sci. 1983,130,203.
151. H. M. Love, H. D. Wiedrerick, Can. J. Phys. 1%9,47,657-
663.
152. L. Schmidt, R. Gomer, J. Chem. Phys. 1965, 42, 3573-
3598.
153. A. Polak, G. Ehrlich, J. Vac. Sci. Technol. 1977, 14,407.
154. M. Bowker, D. A. King, Surf: Sci. 1980,94,564-580.
155. D. R. Mullins, B. Roop, S . A. Costello, J. M. White, Surf:
Sci. 1987, 186, 67-74.
156. R. Lewis, R. Gomer, Surf: Sci. 1%9,14, 333-345.
157. E. G. Seebauer, A. C. F. Kong, L. D. Schmidt, J. Chem.
Phys. 1988,88,6597-6604.
158. H. Rothermund, S. Nettesheim, A. von Oertzen, G. Ertl,
Surf: Sci. Lett. 1922, 275, L645.
159. J. R. Chen, R. Gomer, Surf: Sci. 1979, 79, 413.
160. S . C. Wang, R. Gomer, J. Chem. Phys. 1985,83, 4193.
161. A. A. Deckert, J. L. Brand, M. V. Arena, S . M. George,
Surf: Sci. 1989,208, 441.
162. Z. Rosenzweig, I. Farbman, M. Asscher, J. Chem. Phys.
1993, 98, 8277.
163. B. Boddenberg, V. Grundke, G. Auer, Ber. Bunsenges.
Phys. Chem. 1990,94, 348.
164. L. Kevan, M. K. Bowman, Modern Pulsed and Continuous-
wave Electron Spin resonance, Wiley, New York, 1990.
165. H. Schlienz, M. Beckendorf, U. J. Katter, Th. Risse, H.-J.
Freund, Phys. Rev. Lett. 1995, 74, 761.
166. L. Boltrmann, Wien. Ann. 1904,53, 53.
167. W. M. H. Sachtler, Le Vide 1973, 164, 67.
168. V. D. Planck, W. M. H. Sachtler, J. Catal. 1967, 7, 300; J.
Catal. 1968, 12, 35.
169. A. A. Balandin, Ado. Catal. 1958, 10, 86.
170. S . D. Kevan (Ed.), Angle Resolved Photoemission: Theory
and Current Applications, Studies in Surface Science and
Catalysis Vol. 74, Elsevier, Amsterdam, 1982.
171. H.-J. Freund, M. Neumann, Appl. Phys. 1988, A47,3.
172. D. Cappus, J. Klinkmann, H. Kuhlenbeck, H.-J. Freund,
Surf: Sci. Lett. in press.
173. C. Xu, B. Dillmann, H. Kuhlenbeck, H.-J. Freund, Phys.
Rev. Lett. 1991, 67, 3551.
174. D. W. Turner, C. Baker, A. D. Baker, C. R. Brundle, Mo-
lecular Photoelectronic Spectroscopy, Wiley, New York,
1970.
175. W. Eberhardt, H.-J. Freund, J. Chem. Phys. 1983, 78, 700.
176. D. Schmeisser, F. Greuter, E. W. Plummer, H.-J. Freund,
Phys. Rev. Lett. 1985, 54, 2095.
177. H.-J. Freund, W. Eberhardt, D. Heskett, E. W. Plummer,
Phys. Rev. Lett. 1983, 50, 786.
178. C. L. Allyn, Thesis, University of Pennsylvania, 1978.
179. G. Odorfer, Diplomarbeit, UniversitAt Erlangen-Nurnberg,
1987.
180. G. Blyholder, J. Phys. Chem. 1968, 68, 2772; J. Vac. Sci.
Technol. 1974, 11, 865.
181. C. M. Kao, R. P. Messmer, Phys. Rev. 1985, B31, 4835.
182. R. P. Eischens, A. W. Pliskin, Ado. Catal. 1958, 10, 1.
183. A. M. Bradshaw, F. Hoffmann, Sur$ Sci. 1978, 72, 513.
184. C. Surner, F. Xu, J. T. Yates, Surf: Sci. 1988,201, 1.
185. J. G. Chen, W. Erley, H. Ibach, Surf: Sci. 1989,223, L891.
186. G. McElhiney, H. Papp, J. Pritchard, Surf: Sci. 1976, 54,
617.
187. H.-J. Freund, H. Kuhlenbeck in Applications of Synchrotron
Radiation (Ed.: W. Eberhardt), Springer Series in Surface
Sciences Vol. 35, Springer Verlag, Heidelberg, 1995, p. 9.
188. T. C. Chiang, G. Kaindl, A. E. Eastman, Solid State Com-
mun. 1980,36, 25.
189. C. B. Duke, Surf: Sci. 1978, 70, 675.
190. C. B. Duke, W. R. Salaneck, F.-J. Fabisch, J. J. Ritsko,
H. R. Thomas, A. Paton, Phys. Rev. 1978, B18, 5717.
~~
191. J. W. Gadzuk, S. Holloway, K. Horn, C. Mariani, Phys.
Rev. Lett. 1982, 48. 1288.
192. U. Hofer, M. J: Breitschafter, E. Umbach, Phys. Rev. Lett.
1990. 64, 3050.
193. M. Bertolo, W. Hansen, K. Jacobi, Phys. Rev. Lett. 1991,
67, 1898.
194. J. Kessler, F. Thieme, Surf: Sci. 1977, 67, 405.
195. N. V. Richardson, A. M. Bradshaw in Electron Spectros-
copy (Eds: C. R. Brundle, A. D. Baker), Academic Press,
New York, 1982, Vol. 4.
196. E. W. Plummer, W. Eberhardt, Adv. Chem. Phys. 1982, 49,
533.
197. C. L. Allyn, T. Gustafsson, E. W. Plummer, Chem. Phys.
Lett. 1977, 47, 127.
198. K. Horn, A. M. Bradshaw, K. Jacobi, Surf: Sci. 1978, 72,
719.
199. H. Kuhlenbeck, M. Neumann, H.-J. Freund, Surf: Sci.
1986, 173, 194.
200. F. Greuter, D. Heskett, E. W. Plummer, H.-J. Freund,
Phys. Rev. 1983, B27, 71 17.
201. E. S. Jensen, T. N. Thodin, Phys. Rev. 1983, B27, 3338.
202. N. D. Shinn, J. Vac. Sci. Technol. 1986, A4, 1351.
203. R. Miranda, K. Wandelt, D. Rieger, R. D. Schnell, Surf:
Sci. 1984, 139, 430.
204. K. Horn, A. M. Bradshaw, K. Hermann, I. P. Batra, Solid
State Commun. 1979, 31, 257.
205. G. Borstel, M. Neumann, G. Seitz, W. Braun, Proc. 4th In-
tern. Conf: Solid Surfaces, Cannes, 1980, Vol. 1, p. 357.
206. P. Hoffmann, J. Gossler, A. Zartner, M. Glanz, D. Menzel,
Surf: Sci. 1985, 161, 303.
207. D. Heskett, E. W. Plummer, R. A. DePaola, W. Eberhardt,
F. M. Hoffman, Surf: Sci. 1985,164, 490.
208. M. Steinkilberg, Thesis, Technische Universitat Miinchen,
1977.
209. W. Braun, G. Meyer-Ehmsen, M. Neumann, E. Schwarz,
Surf: Sci. 1979, 89, 354.
210. H. Kuhlenbeck, Diplomarbeit, Universitat Osnabriick,
1984.
211. CW. Seabury, T. N. Rhodin, M. Traum, R. Benbow, Z.
Hurych, Surf:Sci. 1979, 97, 363.
212. D. Rieger, R . D. Schnell, W. Steinmann, Surf: Sci. 1984,
143, 157, P. Hoffmann, S. R. Bare, D. A. King, Surf: Sci.
1982,117, 245.
213. G. Apai, R. S. Wehner, R. W. Williams, J. Stohr, D. A.
Shirley, Phys. Rev. Lett. 1976, 37, 1497.
214. R. J. Smith, J. Anderson, G. J. Lapeyre, Phys. Rev. Lett.
1976, 37, 1081. G. J. Lapeyre, J. Anderson, R. J. Smith,
Surf: Sci. 1979, 89, 304.
215. H.-J. Freund, E. Umbach (Eds.), Adsorption on Ordered
Surfaces of Ionic Solids and Thin Films, Springer Series in
Surface Sciences Vol. 33, Springer, Heidelberg, 1993.
216. V. E. Henrich, P. A. Cox, The Surface Science of Metal
Oxides, Cambridge University Press, Cambridge, 1994.
217. M. Pohlchen, V. Staemmler, J. Chem. Phys. 1992, 97, 2583.
218. V. Staemmler in Adsorption on Ordered Surfaces of Ionic
Solids and Thin Films (Eds: H.-J. Freund, E. Umbach),
Springer Series in Surface Sciences Vol. 33, Springer, Hei-
delberg, 1993, p. 169.
219. G. Pacchioni, G. Cogliandro, P. S. Bagus, Surf: Sci. 1991,
225, 344. G. Pacchioni, T. Minerva, P. S. Bagus, Surf: Sci.
1992,275,450.
220. G. Pacchioni, Surf:Sci. 1993, 281, 251; G. Pacchioni, P. S.
Bagus, in Adsorption on Ordered Surfaces of Ionic Solids and
Thin Films (Eds: H,-J. Freund, E. Umbach), Springer Series
in Surface Sciences Vol. 33, Springer, Heidelberg, 1993,
p. 180.
221. G. Pacchioni, K. M. Neymann, N. Rosch, J. Electron
Spectrosc. Relat. Phenom. 1994,69, 13.
5.1 Chemisorption 941
222. K. M. Neymann, N. Rosch, J. Chem. Phys. 1992, 168,267;
Surf: Sci. 1993,297, 223; Chem. Phys. 1993, 177, 561.
223. C. Pisani, R. Dovesi, R. Nada. S. Tamiro, Surf: Sci. 1989,
216, 267.
224. M. A. Nygren, L. G. M. Petterson, J. Electron Spectrosc.
Relat. Phenom. 1994, 69, 43.
225. S. M. Vesecky, X. Xu, D. W. Goodman, J. Vac. Sci. Tech-
nol. 1994, A12, 2114; M. C. Wu, C. M. Truong, D. W.
Goodman, J. Phys. Chem. 1993, 97, 4182; C. M. Truong,
M. C. Wu, D. W. Goodman, J. Amer. Chem. SOC.1993,
115, 3647.
226. J. Heidberg, L. Cabigon, E. Kampshoff, M. Kandel, R.
Kiihnemuth, D. Meine, B. Redlich, 0. Schonekas, M. Suh-
ren, H. Weiss, D. Wetter in Adsorption on Ordered Surfaces
of Ionic Solids and Thin Films (Eds: H.-J. Freund, E. Um-
bach), Springer Series in Surface Sciences Vol. 33, Springer,
Heidelberg, 1993, p. 46.
227. D. P. Woodruff, A. M. Bradshaw, Rep. Progr. Phys. 1994,
57, 1029.
228. K. M. Schindler, Ph. Hofmann, V. Frische, S. Bao, S. Kul-
karni, A. M. Bradshaw, D. P. Woodruff, Phys Rev. Lett.
1993, 71, 2045.
229. A. M. Bradshaw, D . P. Woodruff in Applications of Syn-
chrotron Radiation (Ed.: W. Eberhardt), Springer Series in
Surface Sciences Vol. 35, Springer, Heidelberg, 1995, p. 127.
230. B. A. Gurney, W. Ho, J. Vac. Sci. Technol. 1985, A3, 1541.
231. J. C. Bertolini, B. Terdy, Surf: Sci. 1981, 102, 131.
232. M. Nishijima, S. Masuda, Y. Sakisaka, M. Onchi, Surf: Sci.
1981, 107, 31.
233. P. R. Mahaffy, M. J. Dignam, Surf: Sci. 1980,97, 377.
234. R. M. Lambert, Surf: Sci. 1975,49, 325.
235. R . J. Behm, G. Ertl, V. Penka, Surf: Sci. 1985,160, 367.
236. D. A. Wesner, F. P. Creuen, H. P. Bonzel, Phys. Rev. Lett.
1988,60, 1045; Phys. Rev. 1989, B39, 10770.
237. H. Kuhlenbeck, M. Neumann, H.-J. Freund, Surf: Sci.
1986, 173, 194.
238. H. Kuhlenbeck, H. B. Schalfeld, U. Buskotte, M. Neumann,
H.-J. Freund, E. W. Plummer, Phys. Rev. 1989, B39, 3475.
239. N. Memmel, G. Rangelov, E. Bertel, V. Dose, K. Kometer,
N. Rosch, Phys. Rev. Lett. 1989,63, 1884.
240. G. Rangelov, N. Memmel, E. Bertel, V. Dose, Surf: Sci.
1991,2511252, 965.
241. D. A. King, D. P. Woodruff (Eds), The Chemical Physics of
Solid Surfaces and Heterogeneous Catalysis, Elsevier, Am-
sterdam, 1990, Vol. 3.
242. K. Wandelt, Surf: Sci. Reports 1982, 2, 1.
243. C. R. Brundle, J. Q. Broughton in Ref. 241, p. 131.
244. R. Reval, M. A. Harrison, D. A. King in Ref. 241, p. 39.
245. H.-J. Freund, B. Bartos, R. P. Messmer, M. Grunze, H .
Kuhlenbeck, M. Neumann, Surf: Sci. 1987,185, 187.
246. M. Grunze, M. Gloze, W. Hirschwald, H.-J. Freund, H.
Pulm, U. Seip, N. C. Tsai, G. Ertl, J. Kiippers, Phys. Rev.
Lett. 1984, 53, 850.
247. M. C. Tsai, U. Seip, F. C. Bassignana, J. Kiippers, G. Ertl,
Surf: Sci. 1985, 155, 387.
248. L. J. Whitman, C. E. Beartosch, W. Ho, G. Strasser, M.
Grunze, Phys. Rev. Lett. 1986, 56, 1984.
249. D . Tomanek, K. H. Bennemann, Phys. Rev. B 1985, 31,
2488.
250. M. Grunze, M. Golze, J. Fuhler, M. Neumann, E. Schwarz,
Proceedings of the 8th Intern. Congr. on Catalysis, Berlin,
1984, Vol. IV, p. 233.
251. F. Winkelmann, S. Wohlrab, J. Libuda, M. Baumer, D.
Cappus, M. Menges, K. Al-Shamery, H. Kuhlenbeck, H.-J.
Freund, Surf: Sci. 1994, 307-309, 1148.
252. D. Cappus, C. Xu, D. Ehrlich, B. Dillmann, C. A. Ventrice,
K. Al-Shamery, H. Kuhlenbeck, H.-J. Freund., Chem. Phys.
1993, 177, 533.
942 5 Elementary Steps and Mechanisms
253. P. W. Tasker, J. Phys. C 1978,12,4977;Phil. Mag. A 1979,
39, 119.
254. F. Rohr, K. Wirth, J. Libuda, D. Cappus, M. Baumer, H.-J.
Freund, Surf: Sci. 1994, 315, L299.
255. D. Cappus. M. HaDel, E. Neuhaus, M. Heber, F. Rohr,
H.-J. Freund, Surf Sci., 1995, 337, 268.
256. G. Illing, Thesis, Ruhr-Universitat Bochum, 1990.
257. R. Lacman, Colloq. Znt. CNRS 1965, 152, 195.
258. D. Wolf, Phys. Rev. Lett. 1992, 68, 3315.
259. C. A. Ventrice Jr., H. Hannemann, Th. Bertrams, H. Ned-
dermeyer, Phys. Rev. B 1994, 49, 5773.
260. M. P. Kiskinova, Poisoning and Promotion in Catalysis
Based on Surface Science Concepts and Experiments, Studies
in Surface Science and Catalysis Vol. 70, Elsevier, Am-
sterdam, 1992.
261. H. P. Bonzel, Surf Sci. Rep. 1987, 8, 43.
262. H. P. Bonzel, A. M. Bradshaw, G. Ertl (Eds), Physics and
Chemistry of Alkali Metal Adsorption, Material Science
Monographs Vol. 51,Elsevier, Amsterdam, 1989.
263. M. P. Kiskinova, G. Pirug, H. P. Bonzel, SurJ Sci. 1983,
133, 321.
264. M. Scheffler, Ch. Droste, A. Fleszar, F. Maca, G. Wa-
chutka, G. Barzel, Physica 1991, 1728, 143.
265. L. J. Whitman, W. Ho, J. Chem. Phys. 1989, 90, 6018.
266. H. Pfniir, P. Feulner, D. Menzel, J. Chem. Phys. 1983, 79,
4613.
267. J. E . Crowell, E. L. Garfunkel, G. A. Somojai, Surf: Sci.
1982,121, 303.
268. T. E. Madey, C. Benndorf, Surf Sci. 1985,164, 602.
269. F. Sette, J. Stohr, E. B. Kollin, D. J. Dwyer, J. L. Gland, J.
L. Robbins, A. L. Johnson, Phys. Rev. Lett. 1985, 54, 935.
210. J. L. Gland, R. J. Madix, R. W. McCabe, C. DiMaggio,
Surf: Sci. 1984, 143,46.
271. W. M. Daniel, J. M. White, Surf Sci. 1986, 171, 289.
212. J. C. Bertolini, B. Tardi, Surf: Sci. 1981,102, 131.
213. M. Trenary, K. J. Uram, J. T. Yates, Jr., Surf: Sci. 1985,
157, 512.
274. Xu Zi, L. Surnev, J. T. Yates, Jr., unpublished results.
215. T. Yamada, Z. Runsheng, Ya. Iwisawa, K. Tamaku, Surf
Sci. 1988, 205, 82.
276. M. Kiskinova, A. Szabo, J. T. Yates, Jr., J. Chem. Phys.
1988,89,1599.
211. M. Kiskinova, A. Szabo, J. T. Yates, Jr., Surf: Sci. 1990,
226, 231.
278. J. L. Brand, M. V. Arena, A. A. Deckart, S . M. George, J.
Chem. Phys. 1990,92, 4483.
279. S . Johnson, R. D. Madix, Surf Sci. 1981, 108, 77.
280. J. Benziger, R. J. Madix, Surf: Sci. 1980, 94, 119.
281. M. L. Burke, R. J. Madix, Surf: Sci. 1990, 237, 1.
282. J. G. Chen, W. Erley, H. Ibach, Surf: Sci. 1989,224, 215.
5.1.2 Chemisorption Theory
R. A. VAN SANTEN
AND M. NEUROCK
5.1.2.1 Introduction
Computational quantum chemistry and solid state
physics have reached the stage where quantitatively re-
liable predictions on the interaction of small and mod-
erate sized molecules with transition metal clusters or
surfaces are now possible. Computation therefore be-
gins to enrich experimental information by providing
results of model systems that are not easily accessible
to experiment. In addition they allow us to probe the
nature of transient surface species such as short-lived
intermediates and activated complexes. Nonetheless,
the qualitative theoretical framework that has been es-
tablished over the past 50 years remains highly rele-
vant. It provides a sound framework for conceptual
analysis and interpretation. Computation and experi-
ment then can be used to test ideas on the electronic
and structural parameters which control the geometry,
energetics, and dynamics of the chemisorbed molecule.
Two different schools of thought in chemisorption
theory can be distinguished. The first is born out of the
solid-state physics community, while the second origi-
nates from the theoretical chemistry community. For-
mal chemisorption theory dates back to the 1950s and
1960s where the relevant electronic factor for chem-
isorption was considered to be the local electron den-
sity of states at the Fermi level. This is especially true
in catalysis. Magnetic and conductivity measurements
were usually interpreted in these terms. With progress
in solid state physics, theory became more refined and
surface physicists developed a more physically realistic
view of the surface chemical bond. Koutecky [l],
Newns [2], Grimley [3], and Schieffer [4]are considered
to be the founders of formal chemisorption theory.
Many of their concepts remain today and are the basis
of our current views on chemisorption. Formal chem-
isorption theory is therefore the subject of the fist sec-
tion. One of the most important results derived from
formal chemisorption theory was the rationalization of
strength of adsorbate-surface interaction in terms of
the ratio of adsorbate-surface-atom strength versus the
interaction energy between the surface atoms. It raised
the issue of the existence of the concept of a surface
adsorption complex, hence identifying chemisorption
physics with surface complex chemistry.
The theoretical chemical application of surface
chemical bonding theory, highlighted next, is related to
formal chemisorption theory as developed in surface
physics, but concentrates on quantum chemical con-
cepts as the electron distribution over bonding and
antibonding orbital fragments [5, 61. It will be seen that
both approaches complement each other. The notion
of a surface molecule relates to the surface physicists’
concept of surface state.
The final section provides an overview of the current
understanding of the factors that govern the physical
chemistry of chemisorption. Our understanding of the
factors that determine the site preference of surface
dependence of chemisorption is summarized. We dem-
onstrate many of those concepts through a series of
first-principle quantum chemical results on different
example systems. The results allow us to specifically
quantify different aspects of the interaction, such as
 5.1 Chemisorption 943

lateral interactions and adsorbate and surface structure

relaxation. The first-principle nature of the calculations
leads to very good overall comparisons with available
experimental data. This chapter focuses specifically on
chemisorption to transition metal surfaces.

5.1.2.2 Formal Chemisorption Theory

A Introduction
Quantum theoretical models of chemisorption were in-
itially developed within the framework of two idealized Figure 1. Parameters in the tight-binding model of chem-
theories of the electronic structure of the solid state: isorption.

(i) the tight-binding approximation;

(ii) the jellium model. In general @+ds and +j are nonorthogonal:

In the tight-binding approximation discussed first, the (4)

one-electron wave function Yk are considered to be a
linear combination of atomic orbitals yi:
I
This approximation applies best to the d electrons of
the transition metals and has been used effectively by
the theoretical chemistry community, to study chem-
isorption. In the free electron or jellium model, dis-
cussed in the second section, the one-electron wave
functions are, instead, considered as linear combina-
tions of plane waves:
This nonorthogonality has to be explicitly included
into chemisorption theory to account for the Pauli re-
pulsive part of the interaction energy. The quantum-
mechanical solution of the chemisorption problem
has been found, when the coefficients af and eigenvalue
energies Ei are known. A very useful expression is the
local density of states p , ( E ) , that gives the probability
of finding an electron with energy E in adsorbate orbi-
tal i. In the case where the adsorbate is described by
a single atomic orbital yo with energy ZO, the corre-
sponding expression is

The jellium model best applies to the s and p valence

electrons as in the alkali metals. The consequences of
electron-electron interactions are explored effectively.
The two approaches can be coupled to form a hybrid
model as in effective medium theory [7], where the
tight-binding description is used for the d electrons and
the jellium approach is used to describe the s and p
valence electrons. We will limit the discussion in this
chapter to a review of the main results and concepts. A
more extensive treatment of the theoretical basis of
these methods and derivations is available [6].
B The Tight-Binding Quantum Theoretical Model of
Chemisorption
Within the tight-binding approximation (Fig. l), the
general form of the one-electron wave function that
describes the electronic structure of the chemisorption
system is
(3)
1 i
The wave functions $j correspond to the undisturbed
substrate, and qldSare the one electron wave functions
T ( E ) and A ( E ) are the line width and energy shift
functions and are given by
i
(7)
Figure 2 depicts the energy dependence of po(E)
sketched as a function of the interaction energy be-
tween adsorbate and surface. In the weak adsorption
limit, T ( E ) and A ( E ) can be considered energy inde-
pendent. The maximum electron density of states on
the adsorbate appears to be shifted by A and is broad-
ened by a Lorentzian distribution over width 2r.
The dependence on the interaction energy can be
deduced by substituting the expression for tight binding
wave function in the bulk
~

of the adsorbing molecule. References see page 957

944 5 Elementary Steps and Mechanisms

upper band edge

E 1 upper hand edge

I, lower band edge

----f ---c ----c
lower band edge p (El p (E) p (El
no interaction weak interaction intermediate
interaction

increasing adsorbate -
metal surface interaction
’ a b C
Figure 3. Adsorption-induced changes in the surface local den-
surface sities of states (schematic).
weak intermediate
free atom
adsorption adsorption mo’ecuie
limit

Figure 2. Adsorbate local density of states as a function of

adsorbate-metal surface interaction (schematic).

into eq 5-7 for T ( E ) and A ( E ) .In the situation where

I

‘s
A s ( E )= -( E - E,) (15)
the adsorbate orbital yo interacts only with one surface zb - 1
atomic orbital eqs 6 and 7 reduce to where z , is the number of nearest atom neighbors of
the surface atom, Zb is the number of nearest atom
neighbors in the bulk, and p the overlap energy integral
A ( E )=
1 dE’- T ( E ’ )
E-E’
between metal atomic orbitals (see Fig. 1). The result-
ing energy dependence of ps(E)is sketched in Fig. 3.
The bulk electrons are contained in an electron band
of width

The symbol $’ in eq 10 means that only the principal
part of the integration has to be computed. P’ is given by
P’ = (ddSI/HlylL)
This is the overlap energy integral between the adsorb-
ate orbital and surface atomic orbital qsps(E) is the
surface local density of states. Its general form, shown
here, is similar to eq 5:
4Jzb-Ii~i
This width describes the degree of delocalization of the
bulk electrons. It is proportional to the overlap energy
integral of the metal atomic orbitals and increases with
(12) the number of metal atom neighbours.
The delocalization of the electrons in the surface is
less and, hence, the surface bandwidth is smaller. Sub-
stituting expression (1 3) into (9) leads to the following
line width function:
or2

In the weak interaction limit,

T s ( E )and As(,?) contain the information on the elec-
tronic structure of the solid itself, and x s is the energy
of the surface atomic orbital. Expression 13 can only
be evaluated analytically by making further simplify-
ing assumptions about the bulk electronic structure.
Useful analytical approximate expressions for Ts( E )
and A s ( E ) ,correct up to the second moment of the
electron energy distribution (eqs 6 and 7) are:
R12
provided that as z 10.The more general case is dis-
cussed in subsection D. It is useful to study the inter-
action system in this limit.
The local electron density of states on the adsorbate
in the weak adsorption limit is sketched in Fig. 2. The

adsorbate orbital electron density is broadened into
a Lorentzian distribution around a slightly displaced
adsorbate orbital energy ao. Electrons on the adatom
with energy less than a0 gain in energy compared to
the nonchemisorbed situation. Electrons with energy
higher than NO are destabilized with respect to NO. The
chemisorption energy will therefore depend on the
electron distribution of the electrons on the substrate,
which effectively determines their distribution over
bonding and antibonding fragment orbitals.
In the situation where the single electron of the ada-
tom and the Fermi level (the highest occupied surface
orbital) of the metal have the same energy a0 (prior to
chemisorption), the interaction energy equals
This is the interaction energy corresponding to the
weak adsorption limit.
The interaction energy decreases with increasing
delocalization of the surface electrons. The larger the
width of the surface local density of states, the lower
the interaction energy. This result also implies that the
energy of interaction with a coordinatively unsaturated
surface (small z,) is larger than to a higher coordi-
natively saturated surface. When the parameter
p=-
zsP2
which measures the relative value of adatom surface
atom interaction versus that between the metal atoms,
increases, the Lorenzian electron energy distribution on
the adatom broadens and changes shape (see Fig. 2). A
double peaked structure evolves, which can become
wider than the valence electron energy band of the
surface. The doubly peaked energy distribution in-
dicates the formation of a surface molecule. The low-
energy peak corresponds to the binding part of the
electron density, the high-energy portion of the distri-
bution corresponds to the antibonding part of the
electron density. Similar features arise in the local den-
sity of states of the surface atom due to its interaction
with the adatom (Fig. 3). The bonding and antibond-
ing electron density in the surface molecule is also
controlled, to a significant extent, by the contribution
of the surface electron density. In the surface molecule
limit,
>>I
ZSP2
with a0 be equal to EF the interaction energy becomes:
5.1 Chemisorption 945
This is the same result that would be derived from the
recombination of a free surface atom and the adsorbate
orbital. Now, as also sketched in Fig. 2, two sharply
defined energy states appear outside the surface elec-
tron energy band. The energies correspond to the
bonding and anti bonding molecular orbitals of the
surface complex. The first order correction to eq 19 is
the localization energy of an electron on a surface
atom. This relates to the sublimation energy of a sur-
face atom into the gas phase and can be easily deduced.
In the near surface molecule limit the interaction en-
ergy becomes
E,"n;s'(a0 = E F )= 20' - (20)
FZ 28' -a,& x 0 (21)
where a is a numerical constant that depends on sur-
face electron occupation.
The localization energy is proportional to the square
root of the number of surface atom neighbors and the
metal-metal overlap energy integral. Again, we find
that the chemisorption energy decreases with increas-
ing delocalization of the surface electrons. In the sur-
face molecule limit, chemisorption becomes corrosive.
The metal-metal bonds next to the surface complex
weaken.
The results discussed so far are generally valid even
when more sophisticated computational approaches
are used. Critical to the results discussed so far is the
assumption that only bonding orbital fragments be-
tween adsorbate and surface are occupied. We will now
show that the results can qualitatively change when
antibonding orbital fragments become occupied. This
becomes especially relevant when one is interested in
differences in adsorption energy as a function of sur-
face atom coordination number. This is the subject of
the next topic which analyzes chemisorption to the
(111) surface of a face centred cubic metal. The ad-
sorbing molecule can choose between four different
adsorption sites: atop, twofold, and two threefold sites.
For higher coordination sites, eqs 9 and 10 are no
longer valid for evaluating I-(E). One now has to find
the linear combination of surface atomic orbitals, that
correspond to the local symmetry of the adsorption
site. These are the group orbitals q~i[8, 91. The surface
atom local density of state in eqs 9 and 10 are now re-
placed by group orbital local density of states p t ( E ) .
For instance, in twofold coordination, two s atomic
orbitals from each surface atom are combined to form
the surface group orbital
References see page 957
946 5 Elementary Steps and Mechanisms

-3.6

-3.2

-2.8

energy
-2.4
Figure 4. Group orbitals and their energies.

4
-2.0

and the group orbital density of states

(4;lw -w4;) -1.6

-
The overlap energy integral becomes

(VOIw;) = m’ 0 4 0.8 1.2 1.b

%I
2.0

The general expression for T ( E )is now
T ( E )= nnp’2p:(E)
where n is the coordination number of the adsorbate.
For this more general case, the interaction parameter p
becomes
Figure 5. Bethe lattice results. Interaction energies of onefold,
twofold, and threefold adsorbed hydrogen, as a function of Bethe
lattice atomic orbital electron occupation. For the points (a,b ) ; a
is the hydrogen atom coordination number; and b the number of
Bethe lattice neighbor atoms nearest neighbor of atom involved
in the adsorption bond.

The binding energy is proportional to the square root

It might be expected that the linear dependence of p on
n would lead to a linear dependence of the interaction
energy between adsorbate and surface coordination
number n. This is, however, only the case in the limit
of weak adsorption ( p << 1) and when bonding adsor-
bate-surface orbital fragments are occupied.
It is found in the surface molecule limit that the
bonding orbital is the low-energy solution of the
of the adatom coordination number. The apparent
preference for higher coordination sites strongly de-
pends on the distribution of electrons over bonding
and antibonding absorbate-surface atom fragment
orbitals. When the electron occupation of antibonding
orbitals dominates, low coordination sites may become
preferred.
This is illustrated by a set of model calculations
which use the Bethe lattice approximation to describe

I
equation the embedding of a surface cluster in a metal lattice.

I
n = 1; x(n) = 0 Each metal atom is represented by one s atomic orbital
npI2 and the interaction energy with a hydrogen atom is
=O n=2; x(n)=l
(No -E)-
as + x(n)p - E n = 3; x(n) = 2
computed for varying metal electron occupation. This
simulates the analogous change in d bond filling of
(24) transition metals. This implies that the metal Fermi
Equation (24) corresponds to the clusters depicted in level now varies as a function of surface orbital elec-
Fig. 4. The group orbital and its group orbital energy tron occupation. The electron-electron interaction be-
are also listed for the situation where an asymmetric p tween the electrons on the adsorbate therefore had to
orbital interacts with surface s atomic orbitals in a be explicitly included in the calculation. The calcu-
twofold coordination site. lation results are shown in Fig. 5.
In the surface molecule limit the bonding orbital According to the Newns-Anderson model [2] in the
fragment energy becomes weak adsorption limit ( p << l), the paramagnetic state
on hydrogen becomes most stable. In Fig. 5 the regime
E,” = a, + fib’ (a, = a, = a ) (25) where this occurs (high metal valence band electron

0.6 1 3
o'2k
0 ''/ -14 -12 -10 -8 -6
I
-4
I I
-2
Figure 6. Group orbital local density of states on Bethe lattice
representing (1 11) surface of face centred cubic lattice [ 181:
I . Surface group orbital local density of states of single atomic
orbital.
2 . Surface group orbital local density of states of group orbital
1
-(d + ulil
Jz
3. Surface group orbital local density of states of group orbital
1 traction potentials due to positive nuclei are replaced
-(ul3 + $4+ $4)
Jz
4. Surface group orbital local density of states of group orbital
1 equal but opposite in sign to the electron-electron re-
occupation) is indicated with broken lines. Results pre-
sented are for hydrogen adsorbed atop, twofold, and
in threefold coordination. The metal surface atom co-
ordination number is also varied, in order to compare
adsorption between open and more dense surfaces.
Within the surface electron density regime presented,
it is found that the interaction energy for the atop de-
scribed configuration moves through a maximum value
(actually a double maximum) and that the interaction
energies for the twofold and threefold coordination
sites decrease with metal valence electron occupation.
At low valence-electron occupation where only bond-
ing adsorbate-metal fragment orbitals are occupied,
the expected preference for bonding is for high coordi-
nation sites.
This leads to higher interaction energies for the open
surface. However, as the valence electron occupation
increases, both these trends invert. This inversion arises
from the now very different distribution of electrons
over bonding and antibonding surface fragment orbi-
tals for the same metal valence electron occupation.
This is due to the very different group orbital local
density of states of the surface electrons that interact
with the adatom when surface coordination is varied.
The corresponding group orbital LDOS at different
coordination sites are shown in Fig. 6.
5.1 Chemisorption 947
Embedding of the surface complex clusters into the
A Bethe lattice leads to a broadening of the group orbi-
/l
tals electron densities. This is due to electron delocali-
zation. The maximum surface group electron density
- 1
EF
I
1
/I
/ . I
moves to lower energy with increasing adsorbate coor-
dination number, similar to the corresponding group
orbital energy in the isolated metal clusters (Fig. 4). As
a consequence, when comparing bonding at the same
surface, i.e. constant value of E F , the antibonding ad-
0
I I
2
I
4
4
I
6
sorbate-metal surface fragment orbitals for the atop
geometry are less occupied than those that correspond
to the high coordination geometries. Therefore when
the electron occupation increases, the bond energies for
the high-coordination sites will decrease more readily
E
than those for the low-coordination sites.
The emphasis of the tight-binding approximation is
on covalent binding aspects. Effects due to electrostatic
screening are more appropriately handled by in the
jellium model descriptions, to be discussed next.
C The Free-Electron Approach to Chemisorption:
Jellium Model
Within the free-electron approximation the discrete at-
by a continuous electrostatic background. The charge
density is chosen such that attraction by the electrons is
pulsion energy. This is the so-called jellium model. It is
possible to combine a free-electron calculation for the s
and p valence electrons with a tight-binding description
for the interaction with the d valence electron bond.
This is the essence of effective medium theory [9]. The
advantage of free-electron theory is that explicit ex-
pressions for kinetic, exchange, and correlation energy
are known as a function of electron density. In contrast
to the tight-binding approaches, electrostatic interac-
tions can be described rigorously within free-electron
models. Density functional theory (DFT) approaches,
which include accurate descriptions of the discrete
atomic potentials have been implemented in sophisti-
cated computational schemes and predict adsorption
properties in close agreement with experiment. They
are regarded as computational extension of the free-
electron/tight-binding methods [ lo]. The exchange and
correlation energy functionals used are based on jel-
lium-type expressions. Using the variational principle,
differential equations for the wave functions can be
formulated, and solved for limited and infinite systems.
The first success of the jellium model was the pre-
diction of the work function changes as a function of
adsorbate concentration. The presence of a surface can
be simulated within this model by choosing a coor-
dinate, such that the positive background potential
Vt,(r) is positive when x < 0, but equal to zero when
~ ~~
References see page 957
948 5 Elementary Steps and Mechanisms
rn x=o
"t
x-
ioc
Figure 7. Potential energy at a surface with low electron density
atom adsorbed according to the free-electron model (schematic).
I
x=o
Figure 8. Electron density profile near a surface, x is the surface
plane.
x > 0. If the electron density remained unchanged up
to the surface defined by x = 0, vb(r) would be equal
to -n+. This is sketched in Fig. 7 .
However there is a spillover of electrons due to their
finite kinetic energy and the real electron density gen-
erates a surface dipole. This is depicted schematically
in Fig. 8 .
Adsorption of an atom, as for instance an alkali
atom, can be simulated by adding additional potential
density (see Fig. 7), corresponding to that of the ada-
tom. A lower potential will result in electron backflow
from the adsorbate into the metal, resulting in a re-
duced work function. A higher potential will attract
electron density with an opposite effect. The charge
that develops on the adsorbate is screened by electron-
electron attractions. For instance a negative overall
charge in the regime 0 < x < d , will push away surface
electrons so that a positive charge 6 develops. The re-
sulting image potential attraction is
6=
Ei,
4r + 2
=-
w
c 2
-2
-4
0 0.01 0.02 0.03
no(a.4
Figure 9. Embedding energies of He, H, Li, B, N, C, 0, and Ne
in a homogeneous electron gas as a function of electron gas den-
sity [I I].
where r is the adsorbate surface distance and 2 the free-
electron screening length. A similar result is found for a
positive charge, which will attract electrons. This image
charge screening effect will therefore alter the effective
energy levels of adsorbate molecules bound to con-
_c
ductive surface. Charges become stabilized due to the
induced image potentials. The basic idea of effective
medium theory [9] is to replace the density of an ad-
sorbing atom by that of the jellium substrate. The em-
bedding energy,
AE = AEernb[no(?)]
is equal to the difference in energy between the com-
bined atom and host system minus that of the sepa-
rated atom and the host, where no(?) is the density of
the host. Curves for A(E)emb are shown in Fig. 9. As
expected the interaction with Ne is repulsive, due to
Pauli repulsion even at very low densities. For oxygen,
the attractive interaction reaches a maximum for a
particular density because it will readily accept elec-
trons. Above the substrate density that corresponds to
the interaction energy maximum antibonding adsor-
bate metal surface orbitals become occupied.
In effective medium theory, the interaction energy
includes an induced image potential interaction term
computed from the free-electron part. The contribu-
tions from d-electron interactions computed within the
tight-binding model discussed in subsection B are ex-
plicitly included. The effective medium theory, while
still approximate, is quite powerful because of the in-
clusion of electrostatic terms. It has especially been
very useful in the analysis of alkali coadsorption effects
WI.
 5.1 Chemisorption 949

D The Quantum Chemistry of Chemisorption a

- Energy diagram Bond Order Overlap Molecular Orbital
I l
Hoffmann has been instrumental in developing and
demonstrating the extension of the tight-binding ap-
proximation to relate the surface electronic structure to
the nature of the chemisorbed state [ 5 ] . This is im-
plemented in his extended Huckel (EHT) molecular
orbital approach, which has been extensively used to
analyze and understand chemisorption in many dif- a :
ferent catalytic systems. The simplicity of the EHT -\

approach makes it quite easy to explore how changes

in electronic structural parameters affect changes in
chemisorption. This is certainly an advantage for
this approach. The main limitation of the approach
is that it can only provide qualitative results and
trends. Quantitative results require more extensive, b
-
The bond energy equals: 2E + 2e - 4a = 4 ( P - as )S
first-principle theoretical treatments. Nonetheless many I.S*
important insights and concepts about chemisorption Definition of pk= ,kcks
Bond Order Overlap
can be explored in the framework of the EHT approach. - IJ I 1 lJ

Many of these concepts can then be further verified Figure 10. Extended Hiickel molecular orbitals and energies for
by more detailed quantitative ab initio or density func- hydrogen.
tional methods as will be discussed. We review here the
chemisorption of CO and CH3 to illustrate the quan-
tum chemical description of chemisorption. The EHT &-2 ~

method uses the same functional form of the one-elec- &.I - E

T-..
tron wave functions described for the tight-binding ap- donation- .
~.
proach in eqs 1 and 3. The extended Huckel method

+-
EF
explicitly includes nonorthogonality of the atomic backdonatiopc ~.
-.
orbitals (eq 4). The important interaction term that &+1
depends on orbital overlap, S,, is Pauli repulsion. This 11 Pauli repulsion Eb
E+2
is responsible for the repulsive part of the potential
energy curve for the interaction energy of a chemical adsorbate surface metal surface
orbitals electron distribution
bond. Pauli repulsion arises for the exclusion principle
that states that electrons of the same spin cannot Figure 11. Frontier molecular orbital scheme for adsorbate-
occupy the same orbital. surface interaction.
As illustrated in Fig. 10, inclusion of orbital overlap
leads to larger destabilization of antibonding orbitals
than bonding orbitals. Hence when bonding as well as given in eq 28. It is a relation that depends on the
antibonding orbitals are occupied, the overall inter- group orbital density of states around the Fermi level
action becomes repulsive. In the interaction between s multiplied by terms that depend on the electron occu-
atomic orbitals. Pauli repulsion is linear in the coordi- pation of the valence electron bond:
nation number of the adsorption site:
E,,,(Pauli) = 4n($ - xS)S
(27)
z 4ns2
Hence Pauli repulsion favors bonding to low coordi-
nation sites. where w* =weak interaction limit. Expression 28 pro-
The attractive part of the chemisorption energy can vides an approximate relationship between the electron
be found from the interaction between occupied frag- density near the Fermi level and the chemisorption en-
ment orbitals and unoccupied fragment orbitals (fron- ergy. ~ , ( E Fis) the group orbital local density of states
tier molecular orbital theory). This is illustrated in Fig. at the Fermi level averaged over an energy width of
11.
In the weak interaction limit, the attractive part of npI2
% 2-
the potential energy curve can be computed from sec- fiiai
ond-order perturbation theory. For the simple atomic
metal orbital system, the corresponding expression is References see page 957
950 5 Elementary Steps and Mechanisms
around the Fermi level. Relationship (28) agrees with
the computed electron occupation dependence of the
chemisorption energies for the different adsorption
modes in Fig. 5. It parallels approximately the group
orbital local density of states dependence shown in Fig.
6 . Additional parameters that control the energy dif-
ference between adsorbate and metal surface orbitals,
are the surface work function, adsorbate image poten-
tial, electron affinity, as well as the unoccupied and
occupied surface electron density of states. Note that
we explicitly introduced the image potential term,
which arises from the screening of charge on the ad-
sorbate that is generated by electron donation and
backdonation. The first term of eq 28 is the contribu-
tion to the bond energy due electron donation of elec-
trons from bonding adsorbate orbitals into empty sur-
face orbitals. The second term arises from electron
backdonation of surface metal electrons into the un-
occupied adsorbate orbitals.
We will analyze the quantum chemistry of the sur-
face-chemical bond further by the use of extended
Huckel calculations of the electron energy distribution
of CO chemisorbed to a large Rh cluster [13], simulat-
ing a Rh (1 11) metal surface. In the next section we will
discuss in detail the consequences of the cluster size
choices on the variation in the chemisorption bond
strength. Clusters of 50 metal atoms or more reproduce
the electronic structural features found for calculations
on extended surface. The electronic properties of inter-
est are the local density of states of the adsorbate orbi-
tals or metal atomic orbitals, given by
PdE) = (V,lfi(E - H ) l V i ) (29)
and the bond-order overlap population (eq 31). The
bond-order overlap energy density of states ng(E) rep-
resent the atomic orbital interference terms as a func-
tion of energy and provides information on the bond-
ing and antibonding character of the adsorbate-surface
fragment orbitals.
In figure 12 the computed local density of states of
the highest occupied molecular orbital (HOMO), 5 0
and lowest unoccupied molecular orbital (LUMO)
2n' molecular orbitals of CO adsorbed atop, twofold,
or threefold coordinated to the (111) surface of Rh
are presented. Due to the interaction with surface
orbitals the local density of states (LDOS) distributions
broaden with respect to their sharp gas-phase values.
As expected, the increase in coordination extends the
width of the LDOS of the 2n* orbital. The change in
the 50 orbital is much less coordination dependant be-
cause of the directed nature of this orbital. At the atop
site it will overlap significantly with surface dZ2 and p
orbitals, as can be seen from the corresponding bond
order overlap population densities (Fig. 12). In twofold
and threefold coordination, however, the 5 0 is directed
to the surface normal instead of the surface orbital and
hence has a small overlap with the surface atomic
orbitals.
q ( E ) orbital occupation provides a bonding contri-
bution to the surface chemical bond. When nii(E) is
negative, orbital occupation tends to antibonding and
repulsive contributions. For the corresponding discrete
bond order overlaps P:, this is illustrated for H2 in
Fig 10. The functions nij presented in Fig. 12 enable
analysis into the extent to which s, p, and d electrons
contribute to the bond energy. Note that the 5 0 - d ~ ~
interaction has both bonding as well as antibonding
character. A further increase in the valence electron
occupation (higher E F )will increase the occupation of
antibonding orbitals and hence weaken the interaction
energy. This is a very general result for the interaction
of adsorbate orbitals that are doubly occupied and
have a nearly filled d valence electron bond. Similar to
the results depicted in Fig. 5 , the interaction is a max-
imum when only bonding orbital fragments occupied.
The interaction decreases when the d valence elec-
tron occupation is such that bonding orbital fragments
become depleted. Analysis of the Tc5,,diz -orbital (Fig. 12
(E)) shows that the downwards-shifted 5 0 orbital be-
comes part of the bonding surface orbitals. However,
the broadened and upwards shifted pdz2 surface atomic
local density of states becomes part of the antibonding
surface orbitals.
Opposite behavior is found for the 2n* electron den-
sity. The 2n* local density of states has bonding surface
electron density in the energy regime of the surface d
electron density. The antibonding nature of the surface
electron density near the gas phase 2n* electron energy
level is also reflected in the upwards shift of this den-
sity. The 2n*-surface interaction is mainly of bonding
character. For most metals only bonding surface orbi-
tal fragments are occupied. Whereas the 2n* orbitals
for the gas-phase CO molecule are not occupied, the
interaction of CO with the surface broadens the CO
2n* local density of states and leads to a partial occu-
pation of these orbitals. The CO 2n* orbitals are anti-
bonding and hence their occupation will weaken the
C-0 interaction energy, which is experimentally re-
flected in a lowering of the C-0 stretching frequency.
The 2n* LDOS broadens more when CO is in higher
coordination sites. This leads to a larger contribution
of low-energy bonding orbitals and a resulting in-
creased 2n" electron occupation.
 5.1 ChemisorDtion 951

t f
A tEF EF EF

ki A
twofold twofold

threefold threefold

;?:'
-30
I ,
-20
I /Ll
-10
;-I I I I
0
~
edge

-30
I I I I
-20
I I I
-1 0 0
CO-onefold

-30 -20 -10 0

Energy (eV) Energy (eV) Energy (eV)

A.
- 0
.A Energy (eV) Energy (eV)

Figure 12. The electronic interactions of CO with the Rh(l11) surface [13].
A. The CO, 5 0 local density of states p 5 , ( E ) on different adsorption sites.
B. The CO, 2n* local density of states p2n.( E ) on different adsorption sites.
C. The local density of states of the surface atomic orbital dZ2 for CO chemisorbed atop before and after chemisorption (b).
D. Bond order overlap density of states of the CO 5u orbital for CO chemisorbed atop: (a) ~5~ dz2; (b) nsa s; (c) ~ 5 , , ~. ,
E. Bonder order overlap density of states of the CO 2n' orbital for CO chemisorbed atop: (a) ~ 2 ~ d , (b)
, ; n2n,p,.

The 5a orbital has a strong bonding interaction with
the surface s and p valence electrons which is depen-
dent on the relative values of Fermi level and 5 0 posi-
tion. The antibonding orbital fragments can also be-
come occupied when interacting with the surface s
electrons. For CO adsorbed atop, the 2n* CO orbitals
have only a weak interaction with the surface s atomic
orbitals, because of zero overlap between 2n* and its
next-neighbor s atomic orbital. For this reason p-type
orbitals drive the surface adsorbates to twofold and
threefold coordination sites where group orbitals of the
required symmetry can favorably interact.
The resulting bonding scheme is summarized in Fig.
13. The consequence of the high occupation of anti-
bonding 5a-surface orbital fragments is that CO will
prefer the low coordination atop site to relieve Pauli
repulsion. However, the 271*-p orbital interaction will
favor the higher coordination sites.
A consequence of these opposing forces is that the
differences in bond energies of Co adsorption at differ-
ent coordination sites is small.
In practice, bonding with atoms such as C, 0, or S ,
is dominated by the interaction with the adatom p
atomic orbitals and strongly favors high coordination.
Such atoms bind an order of magnitude stronger to the
References see page 9.57
952 5 Elementary Steps and Mechanisms
‘I--
__c
I
I) 50
~
metal ‘:obi co
d valence metal valence
electron d valence atomic electrons
distribution orbitals
Figure 13. Orbital scheme (schematic) of the interaction of the
CO 5 g and 27r* orbitals with surface dX2 and dZ2orbitals.
metal surface because of the relatively low energy of
the unoccupied atomic orbitals.
A general bonding trend for the transition metals in
a given row of the periodic system is that there is an
increase in the interaction energy in moving from right
to the left (e.g. from Cu, Fe, or Ag to Ru). Three fac-
tors contribute:
(i) the d valence electron occupation decreases;
(ii) the ionization potential (work function) of the
metal decreases (only true for Group VIII and
early transition metals);
(iii) the spatial extent of the d valence electron in-
creases.
The lower adsorption energies for Group IB elements
compared to Group VIII elements, is due to the com-
plete filling of the d valence electron band for the IB
metals. This leads to large repulsive interactions with
occupied adsorbate orbitals (e.g. the CO 5a orbital)
and occupation of antibonding orbital fragments.
5.1.2.3 Concepts in Chemisorption
A Introduction
So far we have discussed the electronic nature of the
surface-chemical bond based on parametrized model
considerations. The current status of computational
chemistry and physics has reached a state where quan-
titatively reliable studies on small clusters and even
transition metal slabs are feasible.
Herein, we simply outline the available first-princi-
ple-based methods for analyzing chemisorption sys-
tems. A more detailed review is available elsewhere
[14]. The first is based on the ab initio Hartree-Fock
approach and its successive improvements by the ap-
plication of configuration interaction techniques [ 151.
Outstanding representatives of this approach in catal-
ysis are Siegbahn [16], Bagus [17], Goddard [18] and
their co-workers. The other first principle approach
is based upon the application of density functional
theory. With the incorporation of nonlocal gradient
corrections to the exchange-correlation functional and
the application of energy minimization techniques, this
technique has the potential to become widely adapted
in chemisorption theory, due to the reliability of the
results and its significant savings in CPU. In addition,
the development and application of DFT-slab pro-
grams allow one to treat very realistic catalytic sys-
tems. Baerends [lOa], Salahub [lob], and Ziegler [lOc]
have made significant contributions to the development
of these approaches.
Within the ab initio computational scheme frame-
work, Whitten [19] has developed an interesting and
useful approach, whereby small clusters (e.g. the sur-
face complex) are embedded into a larger cluster model
of the bulk, which is described by lower quality wave
functions. An alternative embedding scheme has been
developed by Pisani [21]. Next to the embedding
schemes density functional theory applications have
appeared to use extended sates to simulate the tran-
sition surfaces. For example, te Velde and Baerends
were recently able to deduce highly accurate adsorption
values for the adsorption of CO to Cu [21] and resolve
the long-standing theoretical question of the coordi-
nation of CO to Cu, with their DFT-slab approach.
Feibelman [22] used a related approach to analyze the
vibrational motion of hydrogen.
Finally we have to mention the important empirical
approach developed by Shustorovich [23], the bond
order conservation-Morse potential (Boc-MP) method.
This is a parameterized scheme that enables the esti-
mate of adsorption energies of complex molecules
based on parameters derived for those of the free mol-
ecules or adsorbed atoms.
BOC-MP theory is approximately valid for many
systems. Its expressions work most satisfactory for sit-
uations where bonding is nondirectional. Each atom is
assigned a particular valency . When an atom becomes
coordinated to neighbors this valence is distributed
over the different chemical bonds. Because the valency
is constant, the larger the number of the neighbors, the
smaller the energy per bond. This phenomenon is often
observed in chemisorption. When a molecule adsorbs,
intramolecular bonds as well as metal atom bonds of
the surface atoms involved in the adsorption complex
weaken. Ultimately this may lead to surface recon-
struction or surface corrosion.
In this section we will revisit many of the concepts
presented earlier and review their validity within the con-
text of results based on first-principle density functional
theory calculations. Elsewhere [6] we provided an exten-
sive review of first-principle-based calculations adsorp-
tion and reactivity on transition clusters and surfaces.
 +
5.1 Chemisorption 953

B Chemisorption to Clusters
It is now widely recognized that the interaction be-
tween an adsorbing molecule and metal particle may
be strongly cluster dependent. Interestingly, the cluster-
size dependence is most significant for adsorption on
I’c,co= 0 081
small unoptimized metal clusters. Such clusters have Pc,co= 0.276 Pc,co= 0.209
been studied extensively as models of surfaces. We E = 127 kJlmol E,,,, = 21 8 kJ/mol E ini = 2 10 kJ/i1101
illustrate this cluster-size effect here for the interaction
of CO with Co and Rh clusters, based on density I.

functional theory results [24]. The question we address

is what is the preferred coordination site of CO. On the
dense surfaces such as Rh and Co, CO prefers atop
coordination. However, as illustrated in Fig. 14(a), CO
prefers the higher coordination site on small clusters
that model the dense surface.
The result has been shown to be due to the response Pc,c0=0.282 0 240 P&,= 0 100

of the cluster electrons to the disturbance with CO. E i n r = 200 kJiiiioi E = 263 kJ/mol E lnt = 241 kJ/mol
A bond order overlap analysis of the CO nearest-
neighbor atom interaction indicates that the preference
for CO bonding is for onefold coordination. Coordi-
nation differences for the Co atoms in the respective
clusters result in different Co-Co interactions. This
ultimately controls the overall adsorption energetics.
The low Co atom coordination numbers in the clusters
biases CO bonding, to high coordination sites. This is
attributed to the Co electron higher electron local- E Int = I G O kJ/inol E ,nt = 140 l\J/iiioi E = I23 k.lhiol
ization energy. The choice of the spherical clusters Col3
(Fig. 14(b)), enables the study of the three CO coordi- Figure 14. Interaction energies and bond orders of CO adsorbed
to different Co clusters [24]. The bandorder overlap occupations
nation modes on the same clusters confirm this analy- are computed from
sis. Now, the atop coordination of CO is found to be occ

preferred [24]. P,=2CP,:

Bond order overlap population analysis confirms
the explanation proposed earlier on the basis of the
extended Huckel method. The repulsive interaction
between the doubly occupied CO 50 orbital and the Co
3d valence orbitals with approximately eight electrons
per atom leads to occupied antibonding admolecule-
surface atom orbitals. This preferred atop interaction is
only partially counteracted by the backdonating inter-
action with the CO 2z* orbitals that prefer the high
coordination site.
The spherical cluster is but one approach to reduce
cluster-size anomalies, a second method would be to use
extended surface clusters, as discussed later, whereby
a large enough top layer surface exists to remove
edge localization at the central sites. The importance of
cluster atom coordination number is again found in
density functional theory calculations of CO interact-
ing with tetrahedral R h and Rhlo clusters [25]. On the
R h cluster, CO coordination is preferred in twofold
coordination, instead of the onefold surface coordina-
tion site, again the result is due to a low coordination
of the atoms. On the large Rhlo cluster, CO is found to
preferentially adsorb atop. The two inequivalent Rh
atoms of Rhlo cluster enable us to test the relation be-
tween admolecule bond energy and surface atom local
density of states bandwidth. The LDOS distribution of
k
the valence electrons of the Rh atom, with three Rh
neighbors and the Rh atom in the middle of the tetra-
hedral edge with six Rh neighbors are also shown in
Fig. 15. Also given are the interaction energies of CO
on the two Rh atoms. CO is found to interact most
strongly with the coordinatively most unsaturated Rh
atom. Its local density of states distribution has the
smallest degree of delocalization.
It follows from the analyses in the previous section,
that the cluster-size dependence will be stronger for the
more weakly bound adsorbates (as molecules) than the
strongly interacting atoms, when the surface complex
energy is large compared to the electron localization
energy. The computed dependence of adsorbate inter-
action energy on cluster size is now well documented
experimentally. For small particles fluctuations in
cluster-size interacting energy are observed that appear
to correlate with cluster ionization potential [26]. In-
teraction energies appear to converge to surface values
for clusters of the order of 20 atoms or more.
References see page 957
954 5 Elementary Steps and Mechanisms

Rh atom bonded to Rh atom bonded to the proper spin state is chosen [lob, 271. For example,
CO has three nearest C 0 has six nearest
neighbour atoms neighbour atoms
Neurock et al. [27] found that chemisorption energies
for atomic hydrogen and oxygen on the octahedral Pd
0 0 cluster depicted in Fig. 16 are within 15 kJ mol-' of the
C c experimental values for hydrogen and oxygen on Pd
(11 1). Figure 16 depicts the optimized structures for
Pds bare, Pds with hydrogen, and Pds with oxygen.
The bare Pd6 cluster was found to be most stable as a
triplet. This follows known evidence that the Pd dimer
is also triplet and that palladium bulk is paramagnetic.
Geometry optimization of the bare Pd6 cluster trans-
E int 86 kJimol E int = 65 kJ/mol
=
formed the starting octahedral cluster into one which
a b essentially has C4" symmetry. There are four short
planar Pd-Pd bonds of 2.63A and six long Pd-Pd
bonds of 2.76 A.
Adsorption of H or 0 forms strong bonds with the
surface palladium cluster which subsequently alters
the cluster bond lengths. The strong Pd-H and Pd-0
bonds lead to weaker Pd-Pd cluster bonds between the
Pd atoms involved in the adsorption complex.
The effect of the extended surface was examined by
optimizing a bare Pdlg cluster model and a hydrogen
Pdlg model [27]. Results depicted in Fig. 17 demon-
strate that the Pdlg structures bond by pushing the
central three palladium atoms in the surface upwards.
Qualitatively, the results are very similar to those for
Pd6. Due to constrained nature of the surface, however,
-10 -8 -6 -4 -2 EF 2 4 the changes are less dramatic for the Pdlg cluster. The
Energy (eV) elevation of the surface atoms involved in the adsorp-
tion complex is well established in surface science liter-
ature. Van Hove and Somorjai [28], for example, have

-.s clearly shown that ethylidine adsorbed on Pt leads to

longer Pt-Pt bonds and a lifting of these atoms out of
3
the surface plain by about 0.12 A.
P More generally these changes in the local geometry
o g at (or near) the chemisorption site will lead to local
J stresses which could ultimately lead to surface recon-
. . . ., l....,
n
struction effects [29].
0
C Lateral Interactions
0 a w.0. The interaction between atoms or molecules co-
-10 -8 -6 -4 -2 EF 2 4 adsorbed to a transition metal surface may be repulsive
Energy (eV) as well as attractive. Figure 18 illustrates three different
coadsorption situations for adatoms ( N and 0),adinter-
Figure 15. CO chemisorbed atop to a Rhlo cluster [25].The lo- mediates (OH and NH), and molecules (NH3 and 0 2 )
cal densities of states p of Rh 4d, CO 5a and CO 2 ~ *and , the
bond order overlap population densities RV between the rhodium on a (1 11) surface. The interaction energies were com-
and CO orbitals for onefold CO adsorption on Rhlo. Adsorption puted using density functional theory with nonlocal
on Rh (3nn) is shown on the left (a) and adsorption on Rh (6nn) is gradient corrections for both exchange and correlation.
shown on the right (b). Each graph shows: (- ) LDOS of Rh; They illustrate that the interaction energy between
(- - - - -) LDOS of CO 50; (. ' . ' .) LDOS of CO 2 ~ *(-; )
BOOPD between Rh-CO 5u: (- - - - -) BOOPD between coadsorbed atoms or intermediates is repulsive, when
Rh-CO 2R'. they share a covalent bond with a surface atom. The
repulsive interaction is significantly larger when they
share two rather than one surface atom. The nature of
Interestingly, recent results suggest that interaction en- the interaction energy is a strong function of the rela-
ergies converge at smaller cluster sizes for nonlocal tive position of the adsorbates.
density functional calculations provided that both the In the case of coadsorption of 0 and NH3 (case I),
adsorbate as well as the cluster are optimized and that the interaction energy of NH3 is increased when it is
 4:;
5.1 Chemisorption 955

/-

Pd6 dl

Pd4 Pd4

triplet doublet triplet

Pds-Triplet Pd6H-Doublet PdhO-Triplet

1-3 2.756 1-3 2.918 1-3 3.131

1-6 2.631 1-6 2.898 1-6 3.074
3-6 2.756 3-6 2.918 3-6 2.918
2-4 2.756 2-4 2.667 2-4 2.651
2-6 2.632 2-6 2.756 2-6 2.757
4-5 2.757 4-5 2.661 4-5 2.628

Pd-X 1-7 1.768 1-7 2.068

3-7 1.768 3-7 2.068
6-7 1.767 6-7 2.068
Figure 16. Structures and bondlengths (A) for geometrically optimized Pds clusters. Free hydrogen atom adsorbed and 0 adsorbed clusters
are compared [27]

Case I Case I1 Case I11

3 fold - 3 fold 3 fold - 3 fold 1 fold - 3 fold

Species Case I Case I1 Case 111

NH3 on Cu(8,3)-0 -32

NH3 on Cu(8,3)-0 -17
NH2 on Cu(8,3)-OH - -5
NH on Cu(8,3)-0 21 1 25
N on Cu(8,3)-0 30
N on Cu(8,3)-OH 170 29

Oxygen and hydroxyle are always threefold coordinated

Pd - t1 1.78 1.75
Figure 18. Interaction energies between adsorbed atoms on a Cu
surface [30].
Pd - Pd
Local- local 2.75 2.82
Local-local 2.75 2.72

Figure 17. Pdl8-H cluster: Bond distances (A) are compared for
locally geometry-optimized and not geometry-optimized cluster
atom positions. References see page 957
956 5 Elementary Steps and Mechanisms
CO-S'Ni
-10 -8 -6 -4 -2 0 2
E(eV)-
Figure 19. Linear augmented plane wave results of the local
density of states of CO adsorbed on Ni(ll1) [32]:(a) unsupported
CO film; (b) CO adsorbed on nickel; (c) CO adsorbed with po-
tassium on nickel; (d) CO adsorbed with sulfur on nickel.
bound to a surface atom next to the surface atom
Cu-0 adsorption complex. The adsorption energy is
increased by 30kJmolF' [31]. The weakening of the
Cu-Cu bonds next to the cu atom involved in the sur-
face bond with the oxygen bond results in a decreased
delocalization energy of the Cu electrons associated
with the Cu atom to which NH3 is bonded. Lateral
interactions can be also electrostatic in nature. In the
case of ammonia interacting with the coadsorbed
oxygen atom, 10% of the interaction energy originates
from the direct electrostatic attraction between the
negatively charged oxygen atom and positively charged
ammonia. The covalent interaction energies rapidly
decline with distance. Within the jellium model the
interaction energy has been found to decrease as rP3.
Lateral interactions can also be dominated by electro-
static interactions. This is, for instance, the case when
alkali atoms are coadsorbed to a metal surface. Using
the augmented plane wave approximation for CO and
K coadsorbed to Ni, Wimmer et al. [32] convincingly
demonstrated the existence of a strong electrostatic in-
teraction. This can be deduced from the relative shifts
of the Co electronic levels when CO is adsorbed to Ni
in the presence and absence of potassium (Fig. 19).
The positive charge of potassium lowers the work
function of the metal surface and hence lowers the CO
electronic levels compared to those of the Ni-electrons.
The result is an enhancement of the backdonation of
electrons into the CO 2n"-orbitals. The donating inter-
action for the CO %-orbitals is reduced in the presence
of coadsorbed potassium. The enhancement of the
backdonation into the CO 2n*-orbitals changes the
relative energy enough to alter the preferred CO-coor-
dination sites. For instance on the Pt (111)-surface it
has been shown that CO-shifts from the atop position
to a high coordination site [33, 341.
H H
I I
H H
Mg2' Mg2'
A E int= -13.13 kJ/mol A E i n t = -49.61 klimol
H-H H-H
4
M2' Mg2'
AE int= -21.73 kJimol AE int = -90.2 kJ/mol
Figure 20. Changes in the interaction energies of H2 with Ir4
clusters due to the presence of a Mg2+ ion [35].
When a small metal particle is adsorbed next to a
cation, as in a zeolite or on an oxidic support, the
electrostatic field of such a cation will polarize the
metal particle. This is illustrated in Fig. 20 for an Ir4
particle adsorbed next to a Mg2+ ion. The changes in
the interaction energy of H2 with the Ir4 particle are
shown. The interaction energy of H2 with Ir4 (mea-
sured by DFT calculations) is weak. There is a strong
Pauli repulsion between the double occupied H20g
orbital and the occupied 11-4 orbitals. Bonding becomes
more favorable for H2 adsorbed to a single Ir atom. In
the presence of the Mg2+ ion, there is a polarization of
Ir4 which removes electron density from between the
adsorbed H2 molecule and interacting Ir atoms. As a
consequence Pauli repulsion is reduced and now the
hydrogen atoms can approach the Ir atoms so that
strong orbital overlap and develops backdonation of Ir
electrons into the H2ai becomes possible. The inter-
action energy of Hz with Ir is now enhanced and
adsorption parallel to the Ir particle in threefold co-
ordination becomes favored.
5.1.2.4 The Surface Chemical Bond: A Summary
Chemisorption causes significant changes in local elec-
tronic structure, that in the case of strong interactions
often leads to surface rearrangements. The structural
changes can often be understood as a consequence of
bond order conservation. The more bonds there are to
an atom the weaker the other bonds associated with
this atom become.
The adsorbate bond strength follows from the dis-
tribution of electrons over bonding and antibonding

fragment orbitals. The interaction with the s and p
metal surface valence electrons usually favors the high
coordination site. The d valence electron interaction,
however, can be repulsive, which would favor low co-
ordination. The adsorption energy tends to increase
with d valence electron depletion.
The adsorption energy of a molecule can be consid-
ered as controlled by a balance of donating and back-
donating interaction terms. For backdonation, the in-
teraction between adsorbate unoccupied orbitals and
the metal surface favor high coordination. For dona-
tion, the interaction favors low coordination. The d
valence electron interaction controls the balance of
these two effects. Adsorbed atoms with accessible p
atomic orbitals, favor high coordination. The prefer-
ence for site coordination is much less for adsorbed
molecules.
Weak, as well as strong chemisorption locally dis-
turb the electronic structure and geometry of the
surface atoms close to the adsorbate molecule. The
chemical bonds between surface atoms involved in
the adsorption complex and neighboring surface atoms
weaken, dependent on the strength of the adsorbate-
surface interactions.
The substrate disturbance by an adsorbate is strongly
surface or cluster-size dependent. This strongly affects
the overall interaction energy and leads to surface and
particle size-dependent interaction energies. Surface
relaxation effects enable adsorption complexes to take
on geometries very similar to those in the analogous
organometallic coordination complexes.
As the adsorbate concentration increases the intra-
atomic surface bonds become weaker. This can sub-
sequently lead to surface reconstruction effects. The
lateral interaction between the adsorbates give rise to
nonideal mixing and order-disorder transitions of the
surface layer [36]. Surface atoms with lower surface
atom coordination numbers tend to be more reactive
than surface atoms with a higher coordination numbers
due to the larger electron delocalization associated with
the latter. Atoms bind more strongly and tend to prefer
adsorption in high coordination sites, whereas mole-
cules adsorb more weakly with smaller differences in
the interaction energies of different coordination sites.
When donative interactions dominate and the contri-
butions from the d valence electron band is strong,
there will be a tendency towards atop coordination,
The details strongly depend on the distribution of
the valence electrons over surface metal d, s, and p
electrons.
When bonding adsorbate-surface orbital fragments
are occupied, higher coordination tends to be favored.
Occupation of antibonding adsorbate-surface orbital
fragments favors bonding to lower coordination sites.
It is important to consider changes in Pauli repulsion
(due to the antibonding orbital interactions) as well as
the backdonative and donative interactions when ana-
5.1 Chemisomtion 957
lyzing changes in the interaction energy over different
transition metal surfaces.
A decrease in work function will favor backdonative
interactions, and hence increase the bond energy.
However, donative interactions will decrease and the
overall interaction result will depend on the balance of
the two changes. For adatoms the backdonative inter-
action dominates, so that the bond energy increases.
For molecular adsorbates such as CO, the balance is
more subtle. For instance, CO binds more strongly to
Pt than to Ni, because on Pt the donative interaction
dominates. This agrees with the CO preference for atop
coordination. On Ni, however, the backdonative inter-
action dominates, resulting in a preference of CO for
higher coordination sites.
A decrease of the transition metal d valence electron
occupation will decrease the occupation of antibonding
adsorbate-surface fragment orbitals and hence assist
the increase in bond energy when comparing bonding
to metals across a row (from right to left) in the Peri-
odic Table. This increase in interaction energy is also
partially due to the increase in spatial extension of the
d orbital band, that favors overlap with adsorbate
orbital.
The low reactivity of the Group IB metals (Cu, Ag,
Au) compared to that of the Group VIII Ni, Pd, and Pt
metals stems from the repulsive interaction between
adsorbate orbitals with the filled d valence electron
band of the IB metals. The resulting decrease in the
interaction energy is not compensated for by an in-
crease in the backdonating interactions due to the
lowering of the work function of the latter.
Alloying of a Group VIII transition metal with a
Group IB metal leads to changes in bond energies of
adsorbates, because now mixed surface ensembles of
Group IB and Group VIII surface atoms are formed.
New mixed-metal surface sites also exist. This may re-
duce the interaction energy of adsorbates in high co-
ordination sites, because multiple coordination to the
more strongly interacting Group VIII surface atoms
can now become suppressed (secondary ensemble effect)
~361.
References
1. J. Koutecky, Phys. Rev. 1957, 108, 13.
2. D. M. Newns, Phys. Rev. 1969, 178, 1123.
3. T. B. Grimley in Critical Review in Solid State Science,
CRC, 1976, 239.
4. J. R. Schrieffer in Dynamic aspects of Surface Physics, Proc.
Int. School of Physics ‘Enrico Fermi’, course XVII, Bologna,
1974.
5 . R. Hoffmann, Solids and surfaces, a chemist’s view of bond-
ing in extended surfaces, VCH New York, 1988.
6. R. A. van Santen, Theoretical Heterogeneous Catalysis,
World Scientific, Singapore, 1991.
7. J. R. Norskov, N. D. Lang, Phys. Rev. 1980, BZI, 2136. J.
K. Norskov, Phys. Rev. 1982, B26, 2875.
958 5 Elementarv Stem and Mechanisms
8. R. A. van Santen, J. Chem., SOC.Faraday Trans. I , 1987,
83, 1915.
9. M. C. Zonnevylle, R. Hoffmann, P. J. van den Hoek, R. A.
van Santen, SurJ Sci. 1989,233, 233.
1Oa. E. J. Baerends, A. Rozendal, in Quantum Chemistry: the
challenge f transition metals and coordination chemsitry
(Ed.: A. qillard), Reidel, Dordrecht, 1986, p. 159; P. M.
Boerrigter, G. te Velde, E. J. Baerends, Znt. J. Quantum
Chem. 1988,33, 87.
lob. A. St. Amant, D. R. Salahub, Chem. Phys. Lett. 1990, 169,
387; D. R. Salahub, R. Fournier, P. Meynaski, I. Papai, A.
St. Amant, J. Ushio in Theory and Applications of Density
Functional Approaches to Chemistry, (Eds: J. Lababowski,
J. Andrelon), Springer, Berlin, 1991.
11. M. J. Puska, R. M. Nieminen, M. Manninen, Phys. Rev.
1980, B24, 3037.
12. S. Holloway, J. K. Norskov, N. D. Lang, J. Chem. Soc.,
Faraday Trans. 1 1987, 83, 1893.
13. R. V. van Santen, A. de Koster Stud. Sut$ Sci. Catal. 1991,
64, 1.
14. R. A. van Santen, M. Neurock, Catal. Rev. 1995, 37, 557.
F. Ruette (Ed.), Quantum chemistry approaches to chem-
isorption and heterogeneous catalysis, Kluwer, dordrecht,
1992; E. M. Shustorovich (Ed.), Metal-surface Reaction
Energetics: Theory and Applications to Heterogeneous Cat-
alysis, Chemisorption and Surface Diffusion, VCH, New
York, 1991; R. V. van Santen, E. J. Baerends in Theoretical
treatment of large molecules and their interactions (Ed.: Z. B.
Maksic) Springer, 1991, Part 4, pp. 323-390.
15. A. Szabo, N. S. Ostland, Modern Quantum Chemistry,
MacMillan, New York, 1982.
16. P. E. M. Siegbahn, U. Wahlgren in Metal-surface Reaction
Energetics: Theory and Applications in Heterogeneous Cat-
alysis, Chemisorption and surface difusion, VCH, New
York, 1991.
17. P. S. Bagus, K. Hermann, Phys. Rev. 1986, B33, 2987; P. S.
Bagus, F. Illas, Phys. Rev. 1990, B42, 10852.
18. T. H. Upton, W. A. Goddard, CRC Critical Review in Solid
State and Materials Sciences 1981,261; E. A. Carter, W. A.
Goddard, J. Catal. 1980, 112, 80; E. A. Carter, W. A.
Goddard, J. Phys. Chem. 1988,92, 2109.
19. J. Whitten in Cluster Models for Surface and Bulk phenom-
ena, (Eds: G. Pacchioni, P. S. Bagus, F. Parnigiani,) NATO
AS1 series, 375 1991, A. Chattopadhyay, H. Tang, J. L.
Whitten, J. Phys. Chem. 1990, 94 6397; H. Yang, J. L.
Whitten, J. Chem. Phys 1989, 91, 126.
20. T. B. Grimley, C. Pisani, J. Phys. C. 1974, 7 , 2831; C.
Pisani, Phys. Rev. 1978, B17, 3143; R. A. van Santen,
L. H. Toneman, Iitt. J. Quantum Chem. 1977, Suppl 2,
83.
21. G. te Velde, E. J. Baerends, J. Comp. Phys. 1992, 99, 84.
22. P. J. Feibelman, Surf: Sci. 1994, 2991300, 426.
23. E. H. Shustorovich, SurJ Sci. Rep. 1986, 6, 1.
24. M. C. Zonneviylle, J. J. C. Geerlings, R. A. van Santen, J.
Catal. 1994, 148, 417.
25. W. Biemolt, Ph.D Thesis, Eindhoven, 1995.
26. A. Kaldor, C. M. Cox, M. Zakin, Ado. Chem. Phys 1988,
70, 211.
27. M. Neurock, G. Coulston, D. Dixon, unpublished results.
28. U. Starke, A. Barbieri, N. Materer, M. A. van Hove, G. A.
Somorjai, SurJ Sci. 1993, 286.
29. J. K. Norskov, Rep. Progr. Phys. 1998, 53, 1253; J. K.
Norskov in The Chemical Physics of Solid Surfaces and
Heterogeneous Catalysis, (Eds. D. A. King, A. P. Woodruff,
Elsevier, 1993, vol 6.
30. M. Neurock, R. A. van Santen, W. Biemolt, A. P. J. Jansen,
J. Am. Chem. SOC.1994,116, 6860.
31 W. Biemolt, A. P. J. Jansen, M. Neurock, G. J. C. S. van de
Kerkhof, R. A. van Santen, Surf: Sci. 1993,287/288, 183.
32. E. Wimmer, C. L. Fu, A. J. Freeman, Phys. Rev. Lett. 1985,
55, 2618.
33. G. E. Crowell, E. L. Garfunkel, G. A. Somorjai, Surf: Sci.
1983, 121, 203.
34. R. A. van Santen, Proceedings of the 8th International
Congress on Catalysis, Springer, Berlin, 1984.
35. E. Sanchez Marcos, A. P. J. Jansen, R. A. van Santen,
Chem. Phys. Lett. 1990,167,399;A. P. J. Jansen, R. A. van
Santent, J. Phys. Chem. 1990,94, 6764.
36. V. P. Zhdanov, Elementary Physics Chemical Processes on
Surfaces, Plenum, New York, 1991.
37. W. M. H. Sachtler, R. A. van Santen, Adv. Catal. 1977,26,
69.
5.2 Microkinetics
5.2.1 Rates of Catalytic Reactions
M. BOUDART
5.2.1.1 Introduction
Catalysis is a kinetic phenomenon. The study of
kinetics in catalysis started early in the 20th century in
a phenomenological mode. The first study of this kind
dealt with the rate of decomposition of stibine SbH3 on
the surface of antimony, a product of the reaction act-
ing as a catalyst. The kinetic data of Stock and Bod-
enstein [l] were fitted to what became known as an
empirical power rate law:
u = k[SbH31n (1)
where the rate u is proportional to the concentration of
the gas phase reactant to a fractional power n with
empirical values 0.55, 0.56, 0.63, and 0.65 at 273, 298,
323, and 348K, respectively. The kinetics of this his-
torical reaction have been discussed repeatedly [2].
The carly phenomenological period of catalytic
kinetics was followed by a second mechanistic period
based on the findings of Langmuir. A highlight of this
period was the publication in 1931 of a very influential
book by Schwab that was translated into English by
Taylor and Spence [3]. During the 1920s and 1930s,
the kinetics used in heterogeneous catalysis were based
largely on the Langmuir lattice model of a surface
consisting of identical noninteracting adsorption sites.
These so-called Langmuir kinetics were developed first
by Hinshelwood [4] and then by Hougen and Watson
to become a systematized tool for process research and
development in the 1940s and 1950s [5]. The idea was
to assume a reaction mechanism leading to the best
data fitting rate equation for use in catalytic reactor
design and operation. In time, the data fitting proce-

dure became very sophisticated [6]. By reaction mech-
anism, we mean a set of elementary steps describing the
catalytic cycle. Usually, of course, there is more than
one set that comes to mind. The set that fits the data
best is not necessarily the correct mechanism, as
repeated many times in the literature. As a result,
the Langmuir-Hinshelwood-Hougen-Watson equa-
tion have often been regarded with a great deal of
skepticism, in spite of their usefulness in engineering
applications. Part of this skepticism is due to the
Langmuir model. Other models based on the the
theory of nonuniform surface were developed in the
1940s and 1950s, starting with a celebrated paper by
Temkin and Pyzhev [7], who first obtained a working
rate equation for the catalytic synthesis of ammonia.
Rate equations based on models of non-Langmuirian
surfaces were then used extensively by the school of
Temkin [8]. A summary of this approach can be found
elsewhere [9]. Because of the facile objections against
both models of Langmuir and Temkin, a safe return to
empirical power rate equations similar to eq 1 has been
advocated [lo], but also opposed because of the physi-
cal content, imperfect as it may be, of the Langmuir
[ 111 or Temkin kinetics. The matter remains unresolved
[13] and will not be discussed in this chapter. A collec-
tion of useful rate equations and kinetic data for
almost 100 industrially important catalytic reactions is
available in an informative compendium [14].
Rather than the rate equation, i.e. a function that
tells us how reaction rate varies with temperature,
pressure, and composition of the reacting system, this
chapter discusses the catalytic reaction rate itself with
an emphasis on the catalytic cycle and the rate at which
it turns over. General relationships between the ther-
modynamics and kinetics of catalytic reactions will
be presented. Examples will be borrowed mostly from
heterogeneous catalysis, although the concepts apply
equally well to homogeneous, enzymatic and, suitably
modified to chain reactions. On the basis of these
bridges between thermodynamics and kinetics, impor-
tant features influencing the turnover rate will be in-
troduced and discussed. The most important concept is
that of the rate determining step. Other concepts in-
troduce the kinetically significant steps and the most
abundant reactive intermediates in catalytic cycles. The
question of equilibrium or nonequilibrium between the
reactive intermediates and reactants or products will
then be discussed with the kinetic coupling between
elementary steps in the cycle.
5.2.1.2 Turnover Rate or Turnover Frequency:
Generalities
Fifty years ago, the rate of heterogeneous catalytic re-
actions was frequently expressed in so-called arbitrary
units. Activity was commonly expressed by conversion
5.2 Microkinetics 959
plotted vesus time at a given temperature. In either
case, the available information made it impossible to
reproduce the work, even if enough details were pro-
vided on the preparation of the catalyst. This tradition
was so ingrained that arbitrary units were used by
Beeck in a famous paper [15] reporting his extensive
work on evaporated metal films of transition metals
used as catalysts for the hydrogenation of ethene. Yet,
Beeck had measured the area of his films and he could
have reported areal rates, i.e. rates per unit surface area
of the films.
The measurement of areal rates on supported metal
catalysts became possible after the first attempt by
Boreskov and Karnaukhov to obtain the area of sup-
ported metal particles by chemisorption of dihydrogen
[16]. In this way, areal rates for silica gel-supported
platinum catalysts were obtained for the oxidation of
sulfur dioxide [17] and of dihydrogen [18]. This tech-
nique was then applied for the first time to y-alumina-
supported platinum reforming catalysts containing a
much smaller weight fraction of metal. That study, in
the Esso (now Exxon) laboratories, revealed that the
metal clusters were about 1 nm in size [19]. Sinfelt and
co-workers at Esso then used the technique to report
for the first time areal rates on reforming catalysts [20-
221. In fact, these were called specific rates, but this
expression is now reserved to rates per unit mass of
catalyst. The systematic use of areal rates in comparing
catalytic activity of metals used in the form of evapo-
rated films, large single crystals, and supported par-
ticles or clusters, ushered in the era of quantitative
measurements of catalytic activity. Yet, ultimately the
rate should, if possible, be referred to the number of
active sites, because such a rate expresses the rate at
which the catalytic cycle turns over: it is a turnover rate
or turnover frequency.
The definition goes back to the early days of enzyme
catalysis when the rate of reaction was referred to the
amount of enzyme and called turnover number. Thls
appellation was unfortunate, as “turnover number” is
not a number but has the dimension of one over time.
In addition, turnover number in enzymatic catalysis
usually denotes the maximum value of the rate per
catalytic site, at saturation of the enzyme by the react-
ing substrate, as defined by Michaelis-Menten kinetics
[23]. This unfortunate limitation is totally unnecessary
and provides another cogent reason to avoid com-
pletely the use of turnover number in catalysis. A
qualitative comparison of the activity of a solid catalyst
with that of an enzyme in terms of a turnover number
first appeared in the literature in 1963 [24]. It appears
that a quantitative turnover number for heterogeneous
catalysis was first used in 1968 to denote the rate of
reaction referred to the number of surface platinum
atoms titrated with dihydrogen on a supported plati-
References see page 971
960 5 Elementary Steps and Mechanisms
num catalyst [25]. Subsequently, the rate referred to the
number of catalytic sites became known as turnover
rate, ut, or turnover frequency (TOF). It is simply de-
fined as the number of revolutions of the catalytic cycle
per unit time, generally the second [26]. It is a chemical
reaction rate, a differential quantity depending on
temperature, pressure and concentrations. Like all cat-
alytic rates, it is hard to measure. Frequently, turnover
rate is replaced by a related, but generally not identical
quantity, the site time yield (STY), defined as the
number of molecules of a specified product made per
catalytic site and per unit time [27].
The difficulty in measuring TOF is not only in de-
termining the rate but in counting active sites. Besides,
sites may not be all identical. In spite of these experi-
mental and conceptual difficulties, there are many
advantages in attempting to report a TOF. This will be
illustrated by the following examples dealing with solid
acids or metallic catalysts.
5.2.1.3 Examples of Turnover Rate Measurements
The arduous problems faced in the correct measure-
ment of the rate of heterogeneous catalytic reactions
that are not under the influence of heat and mass
transfer, poisoning, activation and deactivation, are
not considered in this chapter. The focus is on the de-
termination of the amount of catalytic sites. Clearly the
first task is that they be identified. Next come two
questions as to whether the sites are identical and
whether they interact. Generally, there exist no un-
ambiguous answers, especially because identification
and counting must be ideally carried out in situ, i.e.
during the catalytic reaction.
But in spite of the difficulties, there are cases for
which true turnover rates have been determined con-
vincingly. The first example deals with zeolites and re-
actions catalyzed by their protonic Br~stedacidic sites.
Catalysis by zeolites is a vast subject recently sum-
marized in its science and technology by Haag [28].
Zeolites are silicoaluminates that are now available by
synthesis in the form of pure single crystals of micro-
metric size. Their crystallinity can be excellent. If the
+
atomic ratio of A1 to (Si Al) remains small, i.e. below
~ 0 . 1 interactions
, between A1 ions and also inter-
actions between their associated protons remain negli-
gible. This situation has been examined and reviewed
by Barthomeuf and found to depend on the topology
of the zeolitic framework [29]. With sodium-free ZSM-
5 (MFI) zeolites, the lack of interaction between pro-
tonic sites has been checked by measurements of the
specific rate for n-hexane cracking on samples with
Si :A1 atomic ratios between 15 and almost 100 000, a
range of almost 4 orders of magnitude [28]. In that
range, rate, at constant temperature and pressure, was
found to be strictly proportional to the A1 content.
Clearly, all catalytic sites are identical and non-
interacting, in the range of composition covered in this
study. For seven other acid-catalyzed reactions, a lin-
ear correlation between activity and concentration of
Br~nstedsites was also found, albeit in a more limited
range of Si:Al ratios. Haag concludes: “The possi-
bility to synthesize zeolite catalysts with a well-defined
pre-determined number of active sites of uniform ac-
tivity is certainly without parallel in heterogeneous
catalysis.” Haag also adds: “Turnover frequencies
(TOFs) for a variety of acid catalyzed hydrocarbon
reactions could be determined for the first time.” It
must be noted that this achievement in catalytic science
was driven by its many industrial applications, yet was
made possible by the synthesis of pure single crystals of
the catalytic material.
Now, let us turn our attention to catalysis by metals.
The situation depends on whether a given reaction is
structure insensitive or structure sensitive. An opera-
tional definition of structure sensitivity is that the areal
rate of the reaction or its TOF depends on surface
crystalline anisotropy revealed by working on different
faces of a single crystal or on clusters of varying size
between 1 and 10nm. Historically, the lack of effect
of particle size was first noted for the hydrogenation
of cyclopropane on supported Pt [26]. The effect of
particle size was first observed for the synthesis of
ammonia on supported iron [30]. But the concept
of structure insensitivity and sensitivity received un-
equivocal confirmation from studies of the two above
reactions on single crystals of platinum [31] and iron
[32], respectively.
Consider the case of structure-insensitive reactions.
For a number of well investigated reactions catalyzed
by metal, the areal rate or the TOF under fixed con-
ditions does not depend, or depends only slightly, on
surface crystalline anisotropy as expressed on clusters
of varying size or on single crystals exposing different
faces. Moreover, in many cases, identical or almost
equal values of TOF were obtained with metal clusters
supported on one or several carriers and on single
crystals of the same metal. These striking results have
been discussed in detail elsewhere [26].
How is it possible to avoid significant effects of sur-
face crystalline anisotropy on areal rates or values of
TOF? First, the catalytic site involved in the rate de-
termining step, if there is one, should consist of only
one surface metallic atom or of two adjacent ones at
the most [33]. Otherwise, structure sensitivity should be
recognizable. However, even with a catalytic site con-
sisting of single metal atom, structure sensitivity might
still be observable.
Second, it is possible that the surface coverage during
reaction should be close to saturation and surface re-
construction at the few remaining isolated sites might

~~
have erased surface anisotropy altogether. Such an ex-
planation has been proposed to account for the structure
insensitivity of palladium in the oxidation of carbon
monoxide on single crystals and supported clusters at
pressures between lo-' and 102mbar [34]. Indeed, the
surface of a Pd tip used in field ion microscopy re-
constructs as a result of CO adsorption at 1 mbar [35].
Reconstruction of catalytic surfaces in adsorption or
catalysis so as to minimize surface free energy was
proposed and advocated by Boreskov as a general
principle in heterogeneous catalysis [36]. More re-
cently, surface reconstruction has been shown to be one
of the possible mechanisms accounting for chemical
oscillations in heterogeneous catalysis [37]. In every
case, the question is whether surface reconstruction, if
thermodynamically favored, can be reached kinetically
in a catalytic run. This question is reexamined below,
in connection with structure-sensitive reactions.
Another explanation of structure insensitivity is
based on the formation of a reactive hydrocarbon
overlayer on a metal surface during a catalytic reaction
involving hydrocarbons [38]. If, for instance, the rate
determining step in the reaction is the dissociation of
hydrogen on this overlayer, insensitivity to the sub-
jacent metal structure becomes understandable [26].
Irrespective of the true explanation of structure in-
sensitivity when it is observed on a supported metal,
a single crystal, or both, the areal rate does not seem
to be the best way to report the rate data. Indeed, on
certain faces of a crystal, there may be sites that are
inaccessible to reactants. Why not report a turnover
frequency referred to the number of one or several
types of surface atoms? For supported metals, it is
not the surface area of the metal that is measured,
but rather the number of surface atoms counted by a
fully described titration by chemisorption following a
method that has been calibrated by means of indepen-
dent physical techniques. Such a TOF value is mean-
ingful since, for a structure-insensitive reaction, all
accessible surface atoms can be considered as equally
active sites. Such a statement is strengthened if the
same value of turnover frequency has been measured
for the same reaction under identical conditions on
several faces of a single crystal of the same metal. The
availability of TOF values on single crystals at pres-
sures equal to those used with supported metals, as
pioneered by Kahn, Petersen, and Somorjai [31], must
therefore be regarded as a critical step forward in the
evolution of catalysis by metals toward a quantitative
science. Indeed, even with the best reproducible work
on supported metals, it is not possible to be sure that
all of the possible support effects have been eliminated,
both in the measurement of rate and in the counting
of sites. Thus, work with large single crystals becomes
the standard by which the quality of the work on sup-
ported metals can be judged.
5.2 Microkinetics 961
Consider next the case of structure-sensitive reac-
tions. The best example is ammonia synthesis on iron,
a reaction that continues to be of great industrial im-
portance. It is also another example of the decisive
results that can be obtained by studying a catalytic
reaction on large single crystals at high pressures in-
vestigated by Somorjai and co-workers [32]. Thus, it
was reported that by far the most active face of iron
single crystals in ammonia synthesis at 20 bar was the
(1 11) plane, by more than two orders of magnitude for
the areal rate. This was attributed to special sites with a
coordination number of seven, as suggested carlier in
work on supported iron clusters [39]. With these re-
sults, a value for the TOF on Fe(ll1) can be obtained
and extrapolated to TOF data on ammonia synthesis
at 1 bar on a multiply promoted industrial catalyst.
The two TOF values agree within a factor of two [40].
Because of the uncertainties in the extrapolation and in
the counting of iron sites on the industrial catalyst, this
comparison suggests that the industrial catalyst exposes
predominantly the optimum seven-coordinate sites that
are by far by far the most active ones in the reaction.
If this tentative conclusion if firmed up by further
observations on a working industrial catalyst, it will be
the first time that a true TOF has been reported for a
structure-sensitive reaction on a complex commercial
metallic catalyst [41]. This will be also the first docu-
mented example of a structure-sensitive reaction on a
metal surface that is reconstructed so as to expose the
most active sites. This reconstruction is brought about
or stabilized by catalyst promoters and/or by ammonia
used in the reduction of the catalyst. The surface re-
construction of supported iron clusters with appear-
ance of (111) facets after exposure to ammonia was
reported earlier following studies with Mossbauer effect
spectroscopy [39]. Surface reconstruction, if complete,
leads again to the possibility of reporting what is be-
lieved to be a true TOF.
5.2.1.4 Comparison of Rate Data
Today, rate data in arbitrary units or plots of con-
version versus time are slowly becoming the exception
rather than the rule in the scientific literature of catal-
ysis. With only one unit, namely the second, a value
of turnover frequency offers a straightforward way to
compare data obtained in different laboratories. Com-
parisons between supported and Wilkinson homo-
geneous rhodium catalysts for the hydrogenation of
cyclohexene show very close values of TOF under
similar conditions [42].
For the same reaction, TOF values in the gas phase
and liquid phase are collected in Table 1 for supported
References see page 971
962 5 Elementarv Stem and Mechanisms
Table 1. Turnover frequency for hydrogenation of cyclohexene
at 298K and H2 at atmospheric pressure for gas and liquid
phases. In the latter case, cyclohexane was the solvent, and the
rate was zero order with respect to cyclohexene.
Supported Metal exposed TOF (s-I)
metal (o/O)a
Gas phase Liquid phase
Ni
Rh
36-100
5-100
*
2.0 0.5
6.1 -t 1
0.45 k 0.15
1.3 -t 0.1
Pd 11-76 3.2 0.8 1.5 0.2
Pt 14-100 2.8 k 0.1 0.6 k 0.1
"Determined by H2 chemisorption.
metals: Pt [43, 441, Pd [45], Ni [46, 471, and Rh [42].
Two facts emerge from this comparison. First, the
TOF values for Ni, Rh, Pd, and Pt differ by less than
an order of magnitude. Second, for a given metal, the
TOF in the liquid phase is lower than that in the gas
phase by less than an order of magnitude.
For the oxidation of carbon monoxide on palladium,
TOF values at low and high pressures and low and
high temperature have been compared in the case of
large single crystals, and of clusters supported both on
~-Al2O3single crystals and on high specific surface
area ?-A1203 [34]. These comparisons have led to the
discovery of a new support effect consisting of the sur-
face diffusion of CO adsorbed on the support to the
interface with supported palladium and subsequent
reaction with oxygen [48]. It is hard to imagine how
this effect would have been identified without the use
of TOF values. Indeed, it is because of anomalous
TOF values that the supply of molecules of CO by
surface diffusion could be assessed and explained
quantitatively.
Another application of TOF values deals with the
frequent occurrence of poisoning. Thus, for the hydro-
genation of cyclohexene, some supported Ptly-Al203
catalysts were found to be poisoned with sulfur origi-
nating from sulfates on the support. For the hydro-
genation of cyclohexene, a typical structure-insensitive
reaction as can be seen in Table 1, values of TOF
calculated by counting surface platinum atoms not
covered with sulfur by means of hydrogen adsorption
were found to remain constant [49]. Again, it must
be pointed out that the extensive TOF data for the
hydrogenation of cyclohexene on many platinum
catalysts supported on different supports and consisting
of metal clusters of size between 1 and lOnm, agreed
under identical conditions with TOF values obtained
on a platinum single crystal [26].
A major obstacle in the correct measurement of in-
trinsic kinetic data in heterogeneous catalysis is the
ubiquitous parasitic effect of heat and mass transfer,
especially inside the pores of high specific area mate-
rials. Here again, values of TOF come to the rescue in
the form of a useful experimental criterion first pro-
posed by Koros and Novak [50]. Thus, in the case of
supported metals, if the same value of TOF is obtained
for a given reaction at fixed conditions on two catalytic
samples containing different amounts of metal on the
porous support, the kinetic data are not disguised by
heat or mass transfer [50]. To be on the safe side, the
TOF should be measured at two different temperatures
on both samples.
Another application of TOF deals with new catalytic
materials. It is frequently said that a new material is
attractive because of its high activity in a certain cata-
lytic reaction. What kind of activity is it? And how
does that activity compare with that exhibited by prior
catalysts? While a comparison of specific rates may be
adequate, the use of TOF values gives a more direct
comparison, for instance in the case of molybdenum
carbide as compared to ruthenium for the hydroge-
nolysis of alkanes [51]. In each case, of course, the
method used in counting sites must be fully described,
especially with new materials for which there is little
information on the nature of the sites.
Finally, TOF values are most helpful in assessing the
role of promoters in catalysis. The classical example is
ammonia synthesis on metallic iron. Values of TOF
reveal that promotion by alumina is only textural, i.e.
it maintains high specific area without changing the
rate per exposed iron atom. By contrast, potassium
oxide promoters do change the TOF for the reaction,
per iron atom exposed, at least at high pressure and
high values of conversion [41].
In conclusion, TOF values are very useful. Why are
they not used more often? First, a TOF is a reaction
rate, i.e. a differential quality, that is always hard to
measure. Second, in heterogeneous catalysis, it is even
harder to measure a rate because of problems of heat
and mass transfer, and catalyst deactivation. Finally, a
turnover rate is even harder to obtain because of the
necessity to determine or estimate the number of sites.
If they are not all equally active, the turnover rate will
have an average value. Hence, many prudent inves-
tigators omit to report a TOF even when they could
estimate one because the rate may have been disguised
by mass and heat transfer, or because the rate was not
measured: all that was measured was a yield. In that
case, a site time yield, STY, can still be reported. It is
the number of molecules of a specified product made
per catalytic site per unit time, real time in a batch re-
actor or space (residence) time in a flow reactor. For
those who hesitate to estimate the number of sites,
there is a last alternative, namely the space time yield,
the number of molecules of a specified product made
per unit volume of reactor per unit time. In the units
popularized by Weisz [52], the space time yield for
large catalytic processes is, in order of magnitude.

1 pmol cmF3s-l, which translates roughly into a STY
equal to 1 s-l, also in order of magnitude [53]. These
order of magnitude values are dictated by two limiting
considerations when porous catalysts are used. First, if
rates were much smaller, the reactor would have to be
too expensive for a given productivity. Second, if rates
were much larger, the catalyst and thus also the reactor
volume would be poorly utilized because the heteroge-
neous reaction would be confined to the mouth of the
pores of the catalyst, as a result of diffusional limi-
tations into and out of the catalyst grains. Both con-
straints give the technologist a “window on reality”
with the above value at its center [52].
5.2.1.5 Relationships between Thermodynamics and
Kinetics
These relationships are first considered for elementary
steps. Then, they are generalized to the case of catalytic
cycles consisting of these elementary steps. For in-
stance, a widely accepted catalytic cycle for ammonia
synthesis is as follows:
Stoichiometric
number a
N2 + 2* + 2N* 1 rates u + ~and u - ~ to
H2 + 2* + 2H* 3 A i and the rate of step i at equilibrium ~ i , The
N* + H * + NH* + * 2 A is related to the Gibbs energy G by the definition
NH* + H* + NH2 * + * 2 Ai = -(aGi/dti),;,
NH2 * +H* + NH3 * + * 2 where ti is the extent of reaction. Note that for any
*
NH3* NH3 + * 2 reaction to proceed ( u > 0), Aoneed not be positive but
NZ+ 3H2 = 2NH3
Each elementary step is written as it is believed to take
place at the molecular level. Thus the dissociation of
dinitrogen on two adjacent catalytic sites denoted by
the asterisk * is written as
N2 + 2* + 2N*
but not as
;N2+*+N*
as this is not a molecular event. The dissociation could
take place in two successive elementary steps, via the
precursor N*N:
N2+*+N*N
N *N + * + 2N*
Other possibilities come to mind. Kinetically, what
matters is that the sum of the elementary steps as writ-
ten add up to the stoichiometric equation for the over-
all reaction or catalytic cycle. In this summation, each
5.2 Microkinetics 963
Table 2. Thermodynamics and kinetics for an elementary step i.
Ii k+ijk-i = exp(AP/RT)
IIi v + , / v - i = exp(Ai/RT)
IIIi ui = u+i [I - exp(-Aj/RZ‘)]
IVj if A,/RT c< 1: then vi = U ~ , ~ ( A , / R T )
step must be taken a times, where a is the stoichio-
metric number of that step. In the example at hand, a is
1, 2, or 3. Each step (subscript i ) is in principle a two-
way step: its net rate is
ui = u+i - u-i
But, kinetically it could be one-way (or irreversible), if
u+i >> v - i . Alternatively, it could be quasiequilibrated if
u+i and u-i are both much larger than ui. At the kinetic
steady state accociated with the name of Bodenstein,
aiu = v + ~- u-i (2)
where u = v+ - v- is the net rate of the overall reaction
at the Bodenstein steady state. If the rate of every ele-
mentary step can be described by transition state
theory, the four relationships of Table 2 between ki-
netics and thermodynamics are easy to obtain [9]. For
the ith step, they relate rate constants k+i and k-i, or
, the standard affinity A:, the affinity
~ . affinity
(3)
A must be. Also at equilibrium Ai = 0, hence Z)+i = U - i
as follows from equation I1 in Table 2, a relationship
attributed to De Donder [54, 551. The third relation-
ship says that the net rate vi can be expressed in terms
of the forward rate u+i and the thermodynamic poten-
tial A i . The last relationship, near equilibrium, is linear
between flux and force.
These four equations are straightforward, but how
do they apply to a catalytic cycle? The answer is sim-
ple: they do if the catalytic cycle contains one step,
called the rate determining step (rds), subscript d. If
there exists such a step, all the other steps in the cycle
are, by definition, in equilibrium, or at least in quasie-
quilibrium, as defined above ( U + i and U - i >> V i ) .
Thus, if there exists a rds, Ai = 0 for all values of i
except for the rds. All (or almost all) of the affinity for
the cycle A is dissipated in the rds, so that
Ad = A / o ~ (4)
So, the four equations of Table 2 become those of
Table 3. Equation I is obtained readily form equation
References see page 971
964 5 Elementary Steps and Mechanisms
Table 3. Thermodynamics and kinetics for a catalytic cycle with
od for rds.
I k + / k - = exp(AO/CdR T )
IIi v+/v- = exp(Ao/ad R T )
I11 U+ - V- = U- [eXp(Ao/CdR T ) - 11
IV if A / a d RT << 1: then v = V,(A/Ud R T )
I1 if the rate equation can be expressed, within a certain
approximation, in the form of a power rate law (eq 1).
In that case, from equation 11, at the standard value of
the concentrations v+ = k, and v- = k-. Also, A be-
comes Ao, so that equation I follows directly. Finally,
equation I11 and IV follow from I1 in Table 3, as they
do in Table 2.
Applications of the bridges between thermodynamics
and kinetics, as collected in Tables 2 and 3, vary from
a check of good behavior (for I), a measure of the
approach to equilibrium (for II), the possibility of
obtaining the net rate, knowing only the forward rate
or the reverse rate (for 111), and the simplicity of a lin-
ear rate law near equilibrium (for IV).
The equality I between the ratio of rate constants
and equilibrium constants is an expression, at equilib-
rium, of the principle of microscopic reversibility; at a
distance away from equilibrium it extends microscopy
reversibility, relying on the validity of transition state
theory. Insofar as thermodynamic data are more ac-
cessible than kinetic data, the possibility of obtaining
one rate constant from its opposite and the corre-
sponding equilibrium constant can always be helpful.
The De Donder equation I1 will be used later in ex-
plaining catalytic coupling between elementary steps in
a catalytic cycle.
Finally, the linear relationship IV first obtained by
Horiuti [56] was used by him to pinpoint the rate de-
termining step of the ammonia synthesis on iron. If its
catalytic cycle is the one shown above and if the stoi-
chiometric number (Td of the rds is found equal to
unity, the rds must be the chemisorption of N2. The
value of CTd was obtained by measuring v near equilib-
rium and obtaining v, from the rate of the isotopic
equilibration
l4NI4N+ lSNH3e lsNI4N+ 14NH3
taking place in a mixture of N2, H2, and NH3 at equi-
librium except for the isotopic composition, since ki-
netic isotope effects are negligible in this situation. An-
other kinetic use of equation IV in Table 3 is for the
catalytic removal of pollutants down to the ppm level,
since the rate described by the linear relationship IV
must be first order with respect to any variable meas-
uring the distance away from the end of reaction [57].
Historically, the linearity of IV was first verified [58] for
the dehydrogenation of cyclohexane to benzene and its
reverse on a nickel catalyst, prior to Horiuti’s work.
However, a later statistical analysis of the linear kinetic
data on both sides of equilibrium revealed that de-
parture from linearity occurred considerably beyond
the validity of the expansion of the exponential of
equation I11 to its first-order term [59]. The data anal-
ysis suggests that (Td is equal to 3. An appropriate rds
for the overall reaction
c6H12 = C6H6 -k 3H2
might then be the desorption of H2 taking place three
times in each turnover of the catalytic cycle [60].
5.2.1.6 Most Abundant Reactive Intermediates and
Kinetically Significant Steps
According to Campbell,
“Proof of a mechanism of a steady-state, solid catalyzed
reaction lies in directly observing all of the elementary
reactions and their kinetics, and then folding these
microkinetics into a global kinetic model for the net
reaction (. . .). The beauty of knowing a mechanism is
that it gives an intelligent way to extrapolate kinetics
to unknown condition. Also, it gives a detailed, fun-
damental understanding of the reaction that might be
valuable in devising schemes for improving activity
or selectivity, be they by changing reaction conditions
or by modifying the catalyst” [61].
This vibrant declaration expresses what we want to
know in heterogeneous catalysis. Fortunately, what we
need to know in practice to develop, control, and im-
prove solid catalysts, is much less than what we want to
know. The reason is that few elementary steps affect
the turnover rate of a catalytic cycle through their rate
constants, forward, reverse, or both. Such steps that do
are said to be kinetically significant steps. Other steps
that are in quasiequilibrium may affect the rate of the
cycle, but not through their rate constants. Thus what
may be a deplorable situation from the standpoint of
knowledge, is a great simplification in practice, with
the warning that such simplifications may not be ap-
plicable outside the scope of process variables within
which they were invoked.
An example of the advantage of simplifying a cata-
lytic cycle is the catalytic ammonia synthesis. Suppose
that the catalytic cycle given in Section 5.2.1.5 with six
elementary steps proceeding in two directions: this is a
cycle with 12 rate constants and five reactive inter-
mediates. But, assume now that nitrogen adsorption-
desorption is the rate-determing step in Horiuti’s sense,
so that all five subsequent steps are in quasiequili-
brium. Suppose further that the most abundant re-
active intermediate is N*, so that
“*I >> [NH*],“H2.1, [NH3*],[H*]

Then, all five quasiequilibria following the rate deter-
mining step can be added side by side to yield an overall
quasiequilibrium designated by the symbol #:
2N* + 3H2 # 2NH3 + 2* (5)
with an overall equilibrium constant K . With the two
assumed simplifications, the turnover rate is deter-
mined by the two rate constants of the rds, and the
equilibrium constant K . The rds with dd = 1 proceeds
at a rate Ud = Ud+ - Vd-. The rds is the only step for
which we can write
u+ = Vd+; 0- = v&
with v = v+ - v- as the turnover rate of the cycle. The
rds is the only kinetically significant step (kss).
However, there exist many situations besides that
first recognized by Horiuti, for which there also exists
only one kss in the catalytic cycle, but not all other
steps are in quasiequilibrium. Consider the rate of oxi-
dation of carbon monoxide on palladium or platinum.
Langmuir reported this kinetic study at a 1921 discus-
sion of the Faraday Society [62]. Ertl and co-workers
extended all earlier studies by systematic kinetic in-
vestigations on large single crystals [63]. Next to the
two-way ammonia synthesis, the one-way oxidation of
carbon monoxide is today the best understood catalytic
reaction.
With almost equimolar mixtures of CO and 0 2 , be-
tween 400 and 600 K, from lop6 to lo2mbar, the im-
portant steps in the catalytic cycle seem to be
COa + 0, -
kLn
CO2.g
where, using the common notation of surface science,
the subscripts a and g denote surface or gaseous species
respectively. Thus, CO chemisorbed without dissocia-
tion, is in quasiequilibrium (equilibrium constant K,);
0 2 is chemisorbed dissociatively in a one-way step (rate
constant ka); and COa reacts in one-way with 0, to
form COZ in a verijied Langmuir-Hinshelwood (LH)
step (rate constant ~LH). Because of the existence of
two nonequilibrated steps, there is no rds in the classi-
cal sense of Horiuti. But what about the kinetics?
Two temperature regimes are recognized: between
% 400 and z 500 K with a normal exponential increase
of the rate with temperature, and above % 500 K (T,,,
where the rate decreases exponentially with T . Between
400 and 500K, and over eight orders of magnitude of
total pressure [64], the chemisorption of 0 2 occurs on a
surface practically saturated with CO,. The only rate
constant affecting the turnover rate is k,, correspond-
ing to the chemisorption of 0 2 . The latter is then the
only kinetically significant step.
5.2 Microkinetics 965
By contrast, above T,, at low pressure, CO is ad-
sorbed on the free palladium surface that is sparsely
covered with CO, still in quasiequilibrium and CO,
diffuses to the periphery of islands of 0, where the one-
way LH step occurs. Now the LH step is the only ki-
netically significant step, If we accept the proposition
[65] that the only kinetically significant step in a cata-
lytic cycle is still called the rate-determining step, even
if not all others are in quasiequilibrium, and if the cus-
tomary symbol A adopted for the rate determining step
is kept as in the case of Horiuti, below Tmaxwe have
COeCOa
02 ++ 20,
COa + 0, +CO2
and above T,,,
CO +COa
0 2 ++ 20,
co, + 0, +co2
Thus the rate-determining step shifts from the first one-
way step in the cycle to the second one as the tem-
perature rises. Another possible way to handle the
situation would be to state that both one-way steps
determine the rate of the catalytic cycle. We believe
that this simple prescription carries less useful infor-
mation than the one we favor. Consider next the case
of catalytic cycles with two, one-way steps.
Consider a one-way reaction A * B with two, one-
way steps: the entry step (in) and the exit step (out).
There may exist many intermediate steps, reversible,
irreversible, or quasiequilibrated, but there is no need
to consider them explicitly if adsorbed B is the most
abundant reactive intermediate, so that [*] + [B*] =
[L],where [L] is the total density of sites. To obtain the
rate u for the overall one-way reaction, the only ki-
netically significant steps are the first and the last in the
cycle:
kin
*+A+ ...
B*k"",. * + B
At the quasisteady state,
v = Ui" = V,"t
so that, in general, the turnover rate v will depend on
both kin and kOut.
This situation is observed in the low pressure one-
way decomposition of ammonia at temperatures that
References see page 971
966 5 Elementary Steps and Mechanisms
are neither low nor high, at low pressure on several
metals [66]. Thus, according to our definition, there is
no rate determining step in this general case. But at low
temperatures, the rate is described simply as u = uOut
[67], with desorption of nitrogen B being the rate de-
termining step, while at the other limiting case of high
temperature, u = Uin [68], with adsorption of ammonia
A becoming the rate determining step.
There are many examples of first-order rate equa-
tions corresponding to adsorption on a sparsely cov-
ered surface being the rate determining step, and of
zero-order rate equations corresponding to desorption
from a saturated surface being the rate determining
step. An example of the first kind is the oxidation of
propylene to acrolein [69]. An example of the second
kind is the decomposition of germane on a germanium
film saturated with adsorbed H [70]. The latter example
is discussed in Section 5.2.1.8.
In both cases, the original definition of Horiuti can-
not be used. We could say that there exists an unknown
number of nonequilibrated steps, but that does not
carry useful information. Indeed, we have no kinetic
information on intermediate steps between the entry
step and the exit step: some may be in quasiequili-
brium, some may not be. Even in the general case
treated above, we may say that both the entry step and
the exit step are kinetically significant, because they are
nonequilibrated, but what about the intermediate ones?
How many are also nonequilibrated? What is the
meaning of these nonequilibrated steps?
By contrast, our proposed definition of the rate de-
termining step is consistent with common usage, except
that common usage usually lacks precision. Thus it can
be said that the rate determining step in the oxidation
of propylene to acrolein is the abstraction of a hydro-
gen atom from propylene to form an allylic surface in-
termediate. This conforms to our definition because the
measured overall rate is that of the entry step, as con-
firmed by the observed kinetic isotope effect in the ratio
of kin,H to k,,,D. The latter is equal to what is expected
if the transition state of the rds corresponds to the for-
mation of the light or heavy allylic intermediate [69].
It must be noted that with the extended definition of
the rate determining step being adopted as being the
only kinetically signficant step, the stoichiometric num-
ber of Horiuti’s rds, C Q , must be replaced in Table 3,
by Ternkin’s [70] average stoichiometric number 0 of
the catalytic cycle, as defined by
the summation applying to all the steps that are not in
quasiequilibrium. Clearly 0 = Dd for the Horiuti sit-
uation. Unfortunately, there is no way to determine 0.
Besides, its value changes with extent of reaction.
However, the relationships of Table 3 with 0 sub-
stituted in them are correct and retain their usefulness,
even if 0 remains unknown.
Ultimately, as the microkinetics methodology [71-
741 develops, the concept of the rate determining step
may be replaced by the tool of parametric sensitivity,
as proposed, without using this name, by Campbell
[61]. In microkinetics, a catalytic (or chain) reaction
or a network of such reactions is described by a set of
known or plausible elementary steps; the correspond-
ing rate constants are collected from experimental data,
theoretical calculations, or semiempirical estimates.
The ordinary differential equations are then solved by
computer. The activity and selectivity are obtained.
Clearly, it is important to know which steps are kineti-
cally significant, in order to obtain a physical insight in
the fog of a very complex situation. As put by Camp-
bell, a degree of rate control Xi could be obtained from
each step:
(7)
where the partial derivative of the overall rate with re-
spect to the rate constant ki of the step is taken holding
constant the equilibrium constant for step i and the
rate constants k, for all other steps. Campbell notes
that the rate determining step as defined by Horiuti and
with the extended definition being the only kinetically
significant step, has a degree of rate control of unity,
whereas the latter is equal to zero for all other steps.
Clearly, all intermediate values of Xi are possible and
probable in the majority of cases.
As more microkinetic calculations are performed, it
will be interesting to review the situation. In the mean-
time, it is convenient to continue talking about a rate
determining step in the sense of Horiuti, or perhaps
about a single kinetically significant step with all other
steps being equilibrated or not. These simplifications
are helpful in theory and in practice.
5.2.1.7 Kinetic Coupling in Catalytic Cycles: Effect
on Rate
In catalytic or chain cycles, the reactive intermediates,
surface adducts or free radicals, can be in equilibrium
with stable reactants or products. Alternatively, their
concentrations at the kinetic steady state can be lower
or higher than equilibrium values. The latter situation
is the rule rather than the exception in chain reactions.
Unfortunately, in the customary approach to catalytic
kinetics, common assumptions are that the reactants
and products are in equilibrium with catalytic inter-
mediates, and the rate determining step is a so-called
Langmuir-Hinshelwood step between two such inter-
mediates. These assumptions may well be inadequate
or incorrect in many situations.

To understand how catalytic cycles turn over, it
is important to distinguish between equilibrium and
steady-state values of concentrations. The best example
is borrowed from the classic work of Bodenstein and
Lind [75] on the one-way thermal reaction between H2
and Br2 that remains one of the best understood non-
elementary reactions. Under the conditions used by
Bodenstein and Lind, there is, at the steady state, an
equilibrium between bromine atoms and molecules
[Br],, = [Br],. The propagation cycle is
k+1
Br+H2 + HBr+H
k-i
HfBrzkz - + H B r + B r
It consists of a first two-way step followed by a second
one-way step. The reason for the difference is that the
first one is strongly endothermic while the second one
is strongly exothermic. In both cases, the change in
entropy is very small. It follows that the extent of the
endothermic step will be limited by equilibrium with
the product of the reaction HBr. In fact, its standard
affinity AO is negative.
However, the affinity A of the first step is positive
until the end of the reaction. Thls is due to the kinetic
coupling [54] between the propagation steps, itself a
consequence of the De Donder relation I1 in Table 2.
Indeed, it can be shown [55] that, at the steady state,
v1 = 02 and, at half reaction, [Hz] = [Brz] = [HBr]:
However, according to the data of Bodenstein and
Lind, as interpreted by others, k2/k-1 = 10. Thus the
concentration of H in equilibrium with H2, HBr, and
Br in step 1, namely [H]l,ehas been reduced to a value
at the steady state [HI,, by a factor of 11; this also ac-
counts for a positive value A1 instead of a negative
value of A!. This reduction is the consequence of
kinetic coupling between steps 1 and 2 in the cycle: be-
cause k2/k-1 = 10, the atoms of hydrogen produced in
step 1 are pumped away be step 2 ten times faster than
they return to H2 by the reverse of step 1. In other
words, kinetic coupling is expressed by
02 = 10 x u-1
In the absence of kinetic coupling, step 1 would be in
quasiequilibrium:
02 = V+I = v-I, and [HI,, = [HI,,,
Kinetic coupling helps a catalytic cycle to turn over, in
spite of unfavorable thermodynamics affecting one or
several elementary steps through the principle of Le
Chatelier: by accumulating reactants or evacuating
products. Hence, if calculations reveal that some steps
in a catalytic cycle are thermodynamically unfavored
5.2 Microkinetics 967
(A! < 0), it does not mean that the cycle will not turn
over, as it may well do so with the help of catalytic
coupling. Finally, let us note that in the example of the
HBr reaction, while [Br],, = [Brie, the concentration
of H atoms at the steady state, while inferior to that
corresponding to its equilibrium concentration corre-
sponding to the first step (eq 6) is many orders of
magnitude larger than that corresponding to the equi-
librium between H2 and H in the system.
Let us now consider examples of catalytic reactions
where the steady state concentration of a surface spe-
cies is very different from the value expected if equi-
librium prevailed between this surface species and a
fluid-phase species. They will be examples of kinetic
coupling and also of a useful concept, that of fugacity
of a surface species in equilibrium with a fluid phase
species not at its steady state fugacity but at a different
virtual fugactity.
The first example is a reaction similar to the first
catalytic reaction studied kinetically by Stock and
Bodenstein [ 11, namely, the decomposition of gaseous
germane GeH4 on a mirror of germanium acting as the
catalyst for the decomposition to germanium and di-
hydrogen [65]. The following facts must be explained:
(i) the rate does not depend on germane or hydrogen
pressure; (ii) decomposition of GeH4 in the presence of
D2 does not yield HD; (iii) decomposition of mixtures
of GeH4 and GeD4 yields H2, D2, and H D but no
GeH,D4-, species; (iv) the equilibration of H2 and D2
takes place rapidly on a germanium mirror at the tem-
perature where GeH4 decomposes on its own sepa-
rately. These observations can be explained by two,
one-way steps:
*+GeH4-+ ... (in)
H * +H * -+ H2 + 2 * (out)
If these steps are one way and the surface is saturated
with hydrogen (or deuterium) atoms: (i) the measured
rate is zero order; (ii) D2 has no access to the surface
during decomposition of GeH4; (iii) GeH4-GeD4 mix-
tures yield H D but no isotopically mixed reactants; (iv)
H2-D2 mixtures equilibrate on Ge but not under de-
composition of GeH4. During decomposition of GeH4,
the fugacity of H* is very high, much higher than it
would be in the presence of H2 at a steady state pres-
sure of H2 generated in the decomposition. In sum-
mary, the kinetic coupling between the in step and the
out step results in a surface coverage by H atoms close
to unity; only a much higher virtual pressure of H2
could lead to such saturation with H2 alone in the gas
phase. The only kinetically significant step is the de-
sorptions of H2, and H2 does not inhibit the decom-
position since it cannot compete with GeH4 for access
to the surface.
References see page 971
968 5 Elementary Steps and Mechanisms
The second example is the platinum-catalyzed de-
hydrogenation of methylcyclohexane (M) to toluene
(T) and H2. It was studied far from equilibrium and,
indeed, the rate was found not to be inhibited by tol-
uene. The data led to a catalytic cycle with, as in the
previous example, only two, one-way steps, both of
them kinetically significant:
*+M-+ ...
T* +T+*
where T* is the most abundant reactive intermediate.
As chemical intuition teaches that both toluene and
benzene are more strongly held on platinum than
methylcyclohexane, substantial amounts of benzene
were added to the feed but were found to have no ap-
preciable effect on the rate of reaction of toluene. Thus
it appears that the steady state surface coverage by
toluene is very much higher than that corresponding
to equilibrium with gas phase toluene. Again, this is
a result of kinetic coupling: it explains the absence of
inhibition by toluene at its steady state partial pressure
as well as the absence of inhibition by added benzene.
As above, everything happens as if the steady state
fugacity of adsorbed toluene were in equilibrium with a
virtual fugacity of gas-phase toluene much higher than
its steady state fugacity or that of added benzene.
In a third example, the enhanced coverage at the
steady state as compared to an equilibrium coverage
could be measured quantitatively. The reaction was the
decomposition of ammonia at low pressure and high
temperature on a molybdenum foil heated to high
temperature in an ultrahigh vacuum chamber acting
as a continuous stirred tank reactor [76]. At the low
pressures used, it was possible to measure adsorption
isotherms of N2 on Mo by determining equilibrium
surface coverage [N*] by Auger electron spectroscopy.
The latter was also used to determine the steady state
concentration "*Is during the decomposition of am-
monia. The results show that "*Is >"*] at equal
pressures of N2. Again it appears that, because of
kinetic coupling, the fugacity df N* at the steady state
corresponds to a virtual fugacity of gaseous N2 con-
siderably higher than that prevailing at the steady state.
Unfortunately, under the conditions of this work, there
are two kinetically significant steps, in and out, as dis-
cussed in Section 5.2.1.6, so that the affinity is dis-
sipated over two one-way steps and there is no way to
calculate the virtual fugacity of N* to compare it with
experimental values.
However, this comparison can be done when am-
monia is decomposed over an iron catalyst at medium
temperatures and pressures. These were the conditions
for the work analyzed by Temkin and Pyzhev in their
famous paper already mentioned in the Introduction
[7]. In this paper, Temkin and Pyzhev introduced the
idea of surface fugacity of adsorbed nitrogen together
with the assumption that nitrogen adsorption and de-
sorption were rate determining in the sense made
clearer by Horiuti a few years later. Thus, in ammonia
synthesis, gaseous N2 is not in equilibrium with ad-
sorbed nitrogen. In ammonia decomposition gaseous
N2 is not in equilibrium with adsorbed nitrogen or ab-
sorbed nitrogen that forms iron nitride Fe4N. Consider
the case of decomposition, with desorption of N* as
rds, N* as most abundant reactive intermediate and, as
a consequence, the reverse of eq 5 followed by the rds:
2NH3 + 2 * -2N* + 3H2
2N * h N 2 + 2*
Since the desorption (subscript d)-adsorption (sub-
script a) step is rate determining, with Od = 1, the
affinity of that step is equal to the affinity A of the
overall reaction
2NH3 = N2 + 3H2
Hence, from the De Donder equation, applied to the
rds,
vd
- = exp(A/RT) (9)
0,
Now, u, must be proportional to the fugacity of gas-
eous N2 at the steady s t a t e f ~ ~the
, ~ ;constant of pro-
portionality is a certain function d, the value of which
depends on the fugacity of surface nitrogen at the
steady state f ~ * , ~ :
Ua = d( fNt,s) x fN2,s (10)
Consider now that desorption at the steady state is in
equilibrium with a virtual fugacity of N2, f ~ ~i.e., the
~ ,
fugacity of gaseous N2 that would be required at equi-
librium to maintain the fugacity of surface nitrogen at
its steady state valuef&. Then, for this virtual equi-
librium,
vd = va)v
where va,v is a rate of adsorption which is the product
OffNz,v and the function 4 with its value at the fugacity
of adsorbed nitrogen at the steady state. Hence:
Ud = Ua,v = '$( fN*.s) x fN2,v (11)
Substituting eqs 10 and 11 into 9 yields
fN2,v = fN2.s x exp(A/RT) (12)
from which it is easy to calculate fN2,v. Indeed, eq 12
can be rewritten as
where K is the equilibrium constant for the overall re-
action

2NH3 = N2 + 3H2
From tables, at 673 K, the value of K is 6 x 1013Pa2.
The value of the virtual pressure,
obtained from eq 13 with the use of partial pressures,
can be checked from experimental thermodynamic
data in the literature. This was done by lchikawa [77],
who used equilibrium data for two equilibria at 673 K
andpH2 = 1.01 x 1o5Pa (1 atm):
iN2 + 4 F e e F e 4 N
(15)
NH3 + 4Fe Fe4N + HZ (16)
The data give p~~= 3.10 x lo8 Pa and P N H ~ = 0.736 x
lo5Pa. Note that it takes a pressure of N2, more than
1000 times larger than that of NH3 to maintain Fe4N
at equilibrium. Moreover, the calculated value of P N ~
is identical to that calculated for p ~from ~ eq. 14~using
the above values for p ~ ~ N~ H ,and ~ , K. The result is
p ~= 3.15 ~ x, lo8~Pa, identical to the thermodynamic
experimental value.
The reason for summarizing here the calculations
of lchikawa is that the concept of virtual pressure, or
fugacity, first introduced by Temkin and Pyzhev and
generalized later by Kemball [78], is often regarded
with suspicion for the simple reason that it is called
virtual. The calculation of lchikawa vindicates quanti-
tatively the use of virtual pressure or fugacity obtained
from the De Donder equation. The concept of virtual
fugacity is potentially useful whenever an adsorption-
desorption step in a catalytic cycle at the steady state is
not in quasiequilibrium.
Another application of virtual fugacity is found in a
study of aromatization of light alkanes on a zeolite
catalyst: as surface hydrogen and gas phase hydrogen
are not in equilibrium at the steady state, a high
fugacity of adsorbed hydrogen leads to cracking of ad-
sorbed hydrocarbon fragments and a decrease in se-
lectivity for aromatization [79, 801. To abate the high
fugacity of surface hydrogen, gallium is added to the
zeolite to provide an escape route for surface hydrogen.
In conclusion, kinetic coupling between steps in a
catalytic cycle at the steady state can result in a striking
departure of surface concentrations from their equilib-
rium mode.
5.2.1.8 Kinetic Coupling between Catalytic Cycles:
Effect on Selectivity
Kinetic coupling between catalytic cycles is responsible
for a generally observed phenomenon, namely that,
if two catalytic reactions take place at the same time,
their individual rates will not be the same as those
5.2 Microkinetics 969
prevailing if they are run separately with the same cata-
lyst under the same conditions. This is due to the face that
reactants or products in each reaction compete for the
same sites or reactive intermediates. Hence selectivity,
or ration of rates, can be studied only in a situtation
when the two reactions proceed at the same time.
A classic example is the hydrogenation in parallel of
two different alkyl aromatic compounds A1 and A2 on
a Raney nickel catalyst [81]. Separately, each com-
pound saturates the surface and each rate is equal to a
different rate constant kl or k2. However, in simulta-
neous hydrogenation, each compound shares the satu-
rated surface as determined by two adsorption equilib-
rium constants K1 and K2. The ratio of rates is now the
product of the ratio of concentrations [Al]/[A2] multi-
plied by a selectivity factor
Si ;2 = ki Ki lk2K2
where the kinetic coupling is determined both by the
ratio of rate constants and of the ratio of equilibrium
constants. The sum of the rates when the reactions are
run separately is simply kl < k2, whereas in parallel
hydrogenation the situation can be much more com-
plex. Indeed, suppose that k2 < k2 but that K1 > K2.
Then the slow component A1 will monopolize most of
the surface, but as it reacts away, the sum of the rates
will increase as the faster component A2 gets access to
the surface. Many examples of this kind were obtained
by Jungers et al. in the 1950s [82].
As a second example, consider the consecutive hy-
drogenation of an alkene A1 to an alkene A2 to an
alkane A3 on a heterogeneous catalyst. If K1, the
adsorption equilibrium constant of A1 is much larger
than K2, the corresponding equilibrium constant for
A2, selective yields of A2 are possible even when k2, the
rate constant for the surface hydrogenation of A2, is
much larger than that for A1. Indeed, the two reactions
are coupled kinetically as both A1 and A2 compete for
the same sites. A striking example of this situation is
shown in Table 4. As can be seen, the hydrogenation is
not only regioselective but also stereoselective, and the
yields. i.e. the products of conversion and selectivity,
are impressive, thanks to kinetic coupling.
Table 4. Products of reaction between 2-butyne and deuterium
at 287 K on palladium, from the data of Meyer and Burwell [83].
Product Yield (mol YO)
2-Butyne 22.0
cis-2-Butene-2,3-d2 77.2
trans-2-Butene-2,3-d2 0.7
1-Butene 0.0
Butane 0.1
References see page 971
970 5 Elementary Steps and Mechanisms
As a third example, bifunctional catalysis uses kinetic
coupling between catalytic cycles, but in this case the
catalytic cycles are different catalysts. An example with
kinetic details is discussed elsewhere [26], In the iso-
merization of n-pentane, n-C5, iso-pentane, i-C5, an
intermediate n-pentene, n-CF, is first found by de-
hydrogenation of n-Cs on a platinum catalyst, in close
proximity to an acidic catalyst, an alumina that ac-
tually supports the platinum clusters. Alumina then
isomerizes n-Cy to i-Cy and the latter diffuses back to
the metal where it is hydrogenated into the final prod-
uct, i-CS. The process relies on the proximity of the two
catalytic functions.
A last example deals with the enantiomeric selectivity
for reaction of an alkene A with hydrogen B on a soluble
organometallic chiral catalyst *. This reaction network
was studied exhaustively by Landis and Halpern [84].
Because all six kinetically significant rate constants
were measured, this is perhaps the best catalytic net-
work with which to illustrate the effects of kinetic cou-
pling in and between catalytic cycles [85]. There are
two branches in the reaction network, left and right:
On each branch, the addition with * leads to the two
most abundant reactive intermediates A5 and AT, via a
first step that may be one-way, two-way, or quasiequi-
librated. These intermediates then react via a second
one-way step with B to give ultimately the diaster-
eoisomers S and R via steps that are not kinetically
significant. The two branches are kinetically coupled
through the concentration of free sites *, and the two
steps on each branch are kinetically coupled through
the steady-state relationship
v+1 - u-1 = v2
In particular, if 212 << v-1, the first step will be in qua-
siequilibrium, whereas in the opposite situation,
u2 >> u-1, the first step will be one way. These two ex-
treme situations correspond to a vanishingly small
pressure of H2, p ~ or~an , infinite one, respectively.
This is because, on each branch, 212 = k2[A*] x P H ~ .
Thus, the enantiomeric selectivity, expressed as [(R)]/
[(S)]will depend on p ~ decreasing
~ , with increasing
pressure. Why should that be? At the high pressure
limit, the first steps are one way and their rate con-
stants determine the selectivity
At the low pressure limit, the first steps are in quasie-
quilibrium and the selectivity is determined by the
equilibrium constants of the first steps and the rate
constants of the second steps:
Table 5. Enantiomeric selectivity ( R ) / ( S ) from the data of
Landis and Halpern at 298 K [84].
PHI (atm) (R)/(S)
0 54
0.35 49
1.1 24
9.07 6.6
a, 2.0
Now, the chemistry is such that, from experimental
data,
Given this chemistry, at all values of p~~ the favored
product is the ( R ) enantiomer in spite of the fact
that the more abundant intermediate is A:. As p~~
increases however, the kinetic pull to the right
(k2( R )/k2(4) is weakened by the relatively growing im-
portance of the thermodynamic pull to the left
) . net result is shown it Table 5 and de-
( K i S ) / K i R )The
tails are given in the original paper [84] and a review of
the work [85]. At the intermediate values of the pres-
sure, the first step is neither one way nor quasiequili-
brated. The corresponding commonly-made approx-
imations in catalytic kinetics must always be regarded
with caution, convenient as they may be.
5.2.1.9 Conclusions
It will have been noted that this chapter does not con-
tain any rate equation except the empirical one in the
introduction, as a historical tribute to Bodenstein. Of
course, this is no accident. Rate equations in heteroge-
neous catalysis are useful in the design and operation
of catalytic reactors; how to obtain rate equations and
fit kinetic data properly is common knowledge [6].
The emphasis in this chapter has been on the rate it-
self rather than on the rate equation. The turnover rate
is the kinetic signature of a catalyst. To understand
how the catalytic cycle turns over, a number of sim-
plifications can be envisaged. Beyond the steadystate
approximation, these simplifications include the rate
determining step, or a single kinetically significant step,
and the most abundant reactive intermediate. Steps in
the sequence can be one-way, two-way, or quasiequili-
brated. Sometimes, the rate can be obtained from a
minimum of experimental knowledge. For instance, in
the case of ammonia synthesis on iron catalysts, the
knowledge that the rate determining step is the ad-
sorption-desorption of nitrogen permits calculation of
the rate from the rate of desorption only by means of
equation I11 in Table 2, without any knowledge of
the rate equation. Thus it should be possible to use
temperature programmed desorption data of nitrogen

on single crystals of iron and on commercial catalysts,
to obtain rates of ammonia synthesis. Attempts to
do that have been discussed [86-881 and appear to be
remarkably successful.
In the last analysis, the extraordinary success of in-
dustrial heterogeneous catalysis is a fascinating reality.
How has it been possible to invent so many catalysts
from on a priori quantitative, or even qualitative,
knowledge of the elementary steps in their catalytic
cycle, or of the nature of the active sites? It is suggested
that this apparent good luck is accounted for, in part,
by the phenomenon of catalytic coupling between steps
in the cycle. In this way, some individual steps may
well be thermodynamically unfavored, but the cycle
will turn over nonetheless. Catalytic cycles turn over
through the interplay between thermodynamics and
kinetics, as stressed many times in this chapter. The
example of alkane isomerization on a bifunctional alu-
mina supported platinum catalyst, introduced in Sec-
tion 5.2.1.8, reinforces this point. Indeed, the thermo-
dynamic concentration of the alkene intermediate
under the reaction conditions is so small that, were it
not for the kinetic coupling through the alkene between
the two catalytic functions, the process would not have
worked. As usual, the process was developed first be-
fore the bifunctional coupling was recognized chemi-
cally and kinetically. In the future, foresight, based in
part on kinetic principles, may lead to new practical
processes. The methodology has been called kinetics
assisted design of catalytic cycles [89].
References
1. A. Stock, M. Bodenstein, Ber. deut. Chem. Ges. 1907, 40,
570.
2. M. Boudart in Structure and Properties of Solid Surfaces,
(Eds: R. Gomer, C. S. Smith), University of Chicago Press,
Chicago, 1953, p. 400.
3. G.-M. Schwab, Catalysis from the Standpoint of Chemical
Kinetics, translated by H. S. Taylor and R. Spence, Van
Nostrand, New York, 1937.
4. C. Hinshelwood, Kinetics of Chemical Change, Oxford Uni-
versity Press, Oxford, 1940.
5. 0. A. Hougen, K. M. Watson, Chemical Process Principles.
Part Three. Kinetics and Catalysis, John Wiley & Sons, New
York, 1947.
6. G. F. Froment, L. Hosten in Catalysis: Science and Technol-
ogy, (Eds: J. R. Anderson, M. Boudart), Springer, Heidel-
berg, Vol. 2, Chapter. 3, p. 97, 1980.
7. M. I. Temkin, V. Pyzhev, Acta Physicochim. URSS, 1940, 12,
217.
8. S. K. Kiperman, Introduction to the Kinetics of Heterogeneous
Reactions, Nauka, Moscow, 1964.
9. M. Boudart, G. Djega-Mariadassou, Kinetics of Heteroge-
neous Catalytic Reactions, Princeton University Press,
Princeton, NJ, 1984, Chapter 4.
10. S. Weller, AIChE J. 1956, 2 59.
11. M. Boudart, AlChE J. 1956,2, 62.
12. A. Coma, F. Llopis, J. B. Monton, S. W. Weller, Chem.
Eng. Sci. 1988, 43, 785.
5.2 Microkinetics 971
13. M. Boudart, Ind. Eng. Chem. Res. 1989,28, 379.
14. R. Mezaki and Hakunai Inoue, Rate Equations of Solid Cat-
alyzed Reactions, University of Tokyo Press, 1991.
15. 0. Beeck, Rev. Mod. Phys. 1945, 17, 61.
16. G. K. Boreskov, A. P. Karnaukhov, Zhur. Fiz. Khim. 1952,
26, 1814.
17. G. K. Boreskov, V. S. Chesalova, Zhur, Fiz. Khim. 1956, 30,
2560.
18. G. K. Boreskov, M. G. Slin’ko, V. S. Chesalova, Zhur. Fiz.
Khim. 1956,30, 2781.
19. L. Spenadel, M. Boudart, J. Phys. Chem. 1960,46,204.
20. J. H. Sinfelt, H. Hurwitz, R. A. Shulman, J. Phys. Chem.
1960,64, 1559.
21. D, J. Yates, W. F. Taylor, J. H. Sinfelt, J. Phys. Chem. 1964,
68, 2962.
22. W. F. Taylor, D. J. Yates, J. H. Sinfelt, J. Phys. Chem. 1965,
69, 95.
23. B. C. Gates, Catalytic Chemistry, Academic Press, New
York, 1992.
24. R. L. Burwell, Jr., J. Am. Chem. SOC.1963,85,2877.
25. M. Boudart, A. Aldag, J. E. Benson, N. A. Dougharty, C. G.
Harkins, J. Catal. 1966, 6, 92.
26. M. Boudart, G. Djega-Mariadassou, Kinetics of Hetero-
geneous Catalytic Reactions; Princeton University Press.
Princeton, NJ, 1984, Chapters 3 and 5.
27. M. Boudart, Cat. Rev.-Sci. Eng. 1981,23, 1.
28. W. 0. Haag in Zeolites and Related Microporous Materials:
State of the Art 1994 (Eds: J. Weitkamp, H. G. Karge, H.
Pfeifer, W. Holderich), Elsevier, Amsterdam, 1994, p. 1375.
29. D. Barthomeuf, J. Phys. Chem. 1993, 97, 19902.
30. M. Boudart, A. Delbouille, J. Dumesic, S. Khammouma, H.
Topsere, J. Catal. 1975, 16, 90.
31. D. R. Kahn. E. E. Petersen. G. A. Somorjai, J. Catal. 1974,
34, 294.
32. N. D. Spencer, R. C. Schoonmaker, G. A. Somorjai, J.
Catal. 1982, 74, 129.
33. M. Boudart, Proceedings of the 6th International Congress on
Catalysis (Eds: G. C. Bond, P. B. Wells, F. C. Tompkins),
The Chemical Society, London, 1977, Vol. 1, 1.
34. L. Kieken, M. Boudart, Catal. Lett. 1993, 17, 1.
35. A. Gaussmann, N. Kruse, Catal. Lett. 1991, 1, 305.
36. G. K. Boreskov, Heterogeneous Catalysis, Nauka, Moscow,
1990.
37. G. Ertl, Top. Catal. 1994, 1, 305.
38. S. J. Thomson, G. Webb, J. Chem. SOC.,Chem. Comm. 1976,
256.
39. J. A. Dumesic, H. Topsere, M. Boudart, J. Catal. 1975,37,513.
40. M. Boudart, D. G. Loffler, J. Phys. Chem. 1984,88,5763.
41. M. Boudart, Top. Catal. 1994, 1, 405.
42. M. Boudart, D. J. Sajkowski, Faraday Disc. 1991, 92, 57.
43. E. Segal, R. J. Madon, M. Boudart, J. Catal. 1978, 52, 45.
44. R. J. Madon, J. P. O’Connell, M. Boudart, AlChE J. 1978,
24, 904.
45. E. E. Gonzo, M. Boudart, J. Catal. 1978, 52, 462.
46. M. Boudart, W.-C. Cheng, J. Catal. 1987, 106, 134.
47. M. Boudart, C. M. McConica, J. Catal. 1989, 117, 33.
48. L. Kieken, M. Boudart, Proceedings of the 10th International
Congress on Catalysis, 1993, Elsevier, Amsterdam, Vol. B,
1313.
49. G. Leclercq, M. Boudart, J. Catal. 1981, 71, 127.
50. R. J. Madon, M. Boudart, I&EC Fundam. 1982,21, 438.
51. J. S. Lee, S. Locatelli, S. T. Oyama, M. Boudart, J. CataL
1990, 125, 157.
52. P. B. Weisz, CHEMTECH, 1992, July 424.
53. M. Boudart, J. Mol. Catal. 1985, 30, 27.
54. M. Boudart, J. Phys. Chem. 1983,87,2786.
55. M. Boudart, I&EC Fundam. 1986,25, 70.
56. J. Horiuti, J. Res. Inst. Catal. Hokkaido Univ. 1953, 2, 87.
57. M. Boudart, I&EC Res. 1994, 33, 8, 2900.
972 5 Elementary Steps and Mechanisms
58. I. Prigogine, P. Outer, C1. Herbo, J. Phys. Colloid Chem.
1948,52, 321.
59. M. Boudart, D. J. Loffler, J. C. Gottifredi, Int. J. Chem. Kin.
1985,17, 1119.
60. E. Iglesia, personal communication.
61. C. T. Campbell, Top. Catal. 1994, 1, 354.
62. I. Langmuir, Trans. Faraday SOC.1922, 17, 621.
63. T. Engel, G. Ertl, Adv. Catal Rel. Subj. 1978, 28, 1.
64. S . M. Landry, R. A. Dalla Betta, J. P. Lu, M. Boudart, J.
Phys. Chem. 1990, 94, 1203.
65. M. Boudart, K. Tamaru, Adv. Catal. 1957, 9, 699.
66. K. Tamaru, Acc. Chem. Res. 1988,21, 88.
67. J. C. Jungers, H. S . Taylor, J. Am. Chem. SOC.1935,57, 679.
68. D. G. Loffler, L. D. Schmidt, J. Catal. 1976,44, 244.
69. C . R. Adams, T. J. Jennings, J. Catal. 1964,3, 549.
70. M. I. Temkin, Int Chem. Eng. 1971, 11, 79.
71. J. A. Dumesic, B. A. Milligan, L. A. Greppi, V. R. Balse,
K. T. Sarnowski, C . E. Beal, T. Kataoka, D. F. Rudd, A. A.
Treviiio, Ind. Eng. Chem. Res. 1987, 26, 1399.
72. J. A. Dumesic, D. F. Rudd. L. M. Aparicio, J. E. Rekoske,
A. A. Treviiio, The Microkinetics of Heterogeneous Catalysis,
American Chemical Society, Washington, DC. 1993, p. 155.
73. W. Feng. E. Yvynckier, G. F. Froment, Ind. Eng. Chem. Res.
1993,32, 2997.
74. S . C. Korra, M. T. Klein, R. J. Quann, Ind. Eng. Chem. Rex
1994,34, 101.
75. M. Bodenstein, S . C. Lind, 2. Phys. Chem. 1907, 57, 168.
76. M. Boudart, S . Egawa, S . T. Oyama, K. Tamaru, J. Phys.
Chem. 1982, 78, 987.
77. S . Ichikawa, Chem. Eng. Sci. 1989, 44, 2754.
78. C. Kemball, Disc. Faraday SOC.1966,41, 190.
79. E. Iglesia, J. E. Baumgartner, G. D. Meitzner, Proceedings of
the 10th International Congress on Catalysis 1992, (Eds.: L.
Guczi, F. Solymosi, P. T6tenyi) Elsevier, Amsterdam, 1993,
p. 2353.
80. E. Iglesia, J. E. Baumgartner, Catal. Lett. 1993,21, 55.
81. J. P. Wauquier, J. C. Jungers, Bull. SOC.Chim. 1957, 1280.
82. J. C. Jungers, J. C. Balaceanu, F. Coussemant, F. Eschard,
A. Giraud, M. Hellin, P. Leprince, G. E. Limido, CinPtique
Chimique AppliquPe, Technip., Paris, 1958.
83. E. F. Meyer, R. L. Bunvell, Jr., J. Am. Chem. SOC. 1963, 85,
2877.
84. C. R. Landis, J. Halpern, J. Am. Chem. SOC.1987,109, 1746;
6217.
85. M. Boudart, G. Djega-Mariadassou, Catal. Lett. 1994,23, 7.
86. L. M. Aparicio, J. A. Dumesic, Top. Catal. 1994, 1, 233.
87. P. Stoltze, J. K. Nmskov, Top. Catal. 1994, 1, 253.
88. B. Fastrup, Top. Catal. 1994, I , 213.
89. M. Boudart, in Catalyst Design: Progress and Perspectives
(Ed.: L. L. Hegedus), Wiley, New York, 1987, Chapter 6,
p. 141.
5.2.2 Dynamics of Surface Reactions
C.B. MULLINSAND W. H. WEINBERG
5.2.2.1 Introduction
This chapter attempts to systematize much of the ex-
isting experimental data regarding the chemistry of
surfaces within the context of a few general “rules” or
concepts. In other words we shall attempt to answer
the question, “How should one think about surface re-
activity?” This question is addressed both from the
point of view of macroscopic kinetics (thermally aver-
aged rate coefficients) and microscopic phenomena (the
dynamics of energy exchange and accommodation, and
elementary surface reaction rate coefficients). The or-
ganization is as follows. Section 5.2.2.2 groups surface
reactions, such as chemisorption or bimolecular cata-
lytic reactions, into two classes, which are defined as
“direct” and “trapping-mediated”, and it is argued in-
tuitively that trapping-mediated reactions should, in
general, be more commonly observed experimentally.
Section 5.2.2.3 presents some elementary results of
transition state theory that support the point of view
that trapping-mediated reactions should be preferred
over direct reactions. It is also explained when direct
reactions would be expected to become important.
Section 5.2.2.4 discusses the concept of trapping, an
understanding of which is necessary in order to appre-
ciate fully the meaning and mechanism of a trapping-
mediated surface reaction, and a detailed discussion of
the initial rate of dissociative chemisorption of nitrogen
on the W( 100) surface is presented. Quantitative rate
parameters are deduced for the trapping-mediated
reaction. It is demonstrated that industrial catalytic
surface chemistry involving this system would be do-
minated by the trapping-mediated reaction channel,
but data is also presented showing when the direct re-
action channel would become important. The latter
involves translationally energetic molecular beams of
nitrogen. Section 5.2.2.5 provides a synopsis of the
earlier sections.
5.2.2.2 Direct Versus Trapping-Mediated Surface
Reactions
A direct reaction in t h s context is defined as one which
occurs at the surface in a single collision from the
gas phase with a timescale typically less than 10-l2s.
Two examples of such a reaction are direct dissocia-
tive chemisorption, e.g. A2(g) + 2A(a), and the so-
called Eley-Rideal surface reaction mechanism, e.g.
A(a) + AB(g) + A2B(g). (Here, and hereafter, (a) de-
notes an adsorbed species, and (g) denotes a gas-phase
species. Later in the chapter (a) is replaced by (p) or (c),
and this implies physical adsorption or chemisorption,
respectively.) In this type of reaction, the most relevant
temperature is that of the gaseous reactant. Since the
reactant is not trapped at the surface, it cannot ac-
commodate to the surface temperature.
The other, quite different, type of reaction is a trap-
ping-mediated surface reaction. In this case there is a
“real” (bound) intermediate in the reaction, which is
usually termed a precursor. The lifetime of the pre-
cursor is usually sufficiently long compared to a vibra-

(a)
X,Y
Figure 1. (a) One-dimensional potential energy diagram (along
the reaction coordinate p ) for the hypothetical surface reaction
Az(g) + Az(p) + Az(c) + 2A(c). (b) Potential energy surface in
the form of a contour diagram for the reaction in Fig. l(a). The
reaction coordinate p is shown as a dashed line. Note that the
carves labeled I and 1 1 represent motions of the dissociated sur-
face species, i.e. A(c), perpendicular and parallel (x, y directions)
to the surface, respectively, and that the transition states are de-
noted by the symbols t and $.
tional period of the trapped molecule perpendicular to
the surface that it accommodates thermally with the
surface. Although the elementary surface reaction rate
coefficients will be functions of surface temperature
only, the measured reaction rate could be a function of
both the gas and the surface temperature. This is be-
cause the trapping probability into the precursor state,
a dynamic effect, is primarily a function of the gas
temperature.
The most general case of trapping-mediated dis-
sociative chemisorption may be written as
A2(g) * A2(P) * A2(C) 2A(c)
+ (1)
Here, the physically adsorbed A2 molecule is a pre-
cursor to the molecularly chemisorbed state, which, in
turn, is a precursor to the dissociatively chemisorbed
products of the reaction. Projections of a possible
potential energy surface describing this reaction are
shown in Fig. 1. In Fig. ](a) the potential energy along
5.2 Microkinetics 973
the reaction coordinate p is shown; the potential min-
ima corresponding to the physically adsorbed Az, the
molecularly chemisorbed A*, and the dissociatively
chemisorbed A2 are noted explicitly. The two barriers,
both of which lie below the “vacuum zero” of energy
(the gas-phase A2 molecule infinitely far from the sur-
face and at rest), have associated with them two tran-
sition states which are denoted by t and $. T o the left
of the minimum corresponding to dissociative chem-
isorption in Fig. l(a), there are two branches of the
potential, one of which is denoted by i and the other
of which is denoted by 11. The Ibranch corresponds to
motion perpendicular to and into the surface, which
is repulsive due to the Pauli Principle. The 11 branch
corresponds to motion parallel to the surface, which is
periodic on a single-crystalline surface and describes
adatom hopping on the surface. A contour plot which
captures the important features of the potential energy
surface is shown in Fig. l(b). The intramolecular
(A-A) spacing is plotted on the ordinate, and the
spacing between the surface and the center of mass of
the A2 molecule is plotted on the abscissa. Note that
the three potential minima and the two barriers (tran-
sition states) are labeled in the same way as they are in
the one-dimensional representation of Fig. 1(a). The
final dissociatively chemisorbed state corresponds to
two A adatoms typically separated by one lattice con-
stant on the surface. Since this is a large separation
compared to most molecular bond lengths, one would
in general expect “late” barriers to dissociative chem-
isorption, i.e. barriers that are will into the exit channel
of the reaction, as is indicated by the 3 in the contour
plot of Fig. l(b).
It can be deduced from Fig. 1 that trapping-medi-
ated dissociative chemisorption is far more probable
than direct dissociative chemisorption except in un-
usual cases. In order for the A2 molecule to chemisorb
dissociatively in a direct fashion, it must have sufficient
kinetic energy to scatter inelastically from the repulsive
part of the potential surface and redirect sufficient en-
ergy along the reaction coordinate (the A-A intra-
molecular stretching mode in this case) in order for the
reaction to occur. Note that the other inelastic channels
in this scattering event are dissipative and do not lead
to reaction. These dissipative channels include phonon
and electron-hole pair excitation in the solid, and other
intramolecular energy exchange within the A2 mole-
cule, e.g. rotational degrees of freedom and trans-
lational degrees of freedom parallel to the surface.
Hence, a considerable apparent activation barrier to
direct dissociative chemisorption would be observed
experimentally for this system, although there is no
barrier at all with respect to the vacuum zero of energy
for trapping-mediated dissociative chemisorption.
References see puge 983
974 5 Elementary Steps and Mechanisms

Reference Energy Zero

where CiBA is the two-dimensional concentration of

w potential energy surface, dp is a differential length
.-
1 S-A+S-BA
”
2
2s +
CL Reaction Coordinate, p
Figure 2. One-dimensional potential energy diagram (along the
reaction coordinate p ) for the hypothetical Langmuir-Hinshel-
wood reaction, S-A + S-BA + S-A-B-AI -+ S + A2B(g),
where S denotes a surface binding site.
The other previously mentioned example of a direct
surface reaction is the Eley-Rideal mechanism. This
is to be contrasted with the Langmuir-Hinshelwood
mechanism, which implies a reaction between two ad-
sorbed species, and which clearly fits into our definition
of a trapping-mediated reaction. It is probably fair to
state that the Langmuir-Hinshelwood reaction mech-
anism is the more correct way to describe most surface
reactions, which is another way of saying that trap-
ping-mediated reactions are usually more important
than direct reactions. In the next section some simple
theoretical arguments are presented that put this con-
clusion on a firmer foundation.
5.2.2.3 Transition State Theory of Surface Reaction
Rates
In this section the rates of Langmuir-Hinshelwood and
Eley-Rideal surface reactions are formulated within
the context of an elementary, equilibrium statistical
mechanical treatment of transition state theory. This
allows a quantitative assessment of the relative impor-
tance of these two mechanisms: the trapping-mediated
Langmuir-Hinshelwood reaction versus the direct
Eley-Rideal reaction.
“activated complexes” at the transition state on the
along the reaction coordinate at the transition state, I,
is the center of mass velocity (dpldt) of reacting species
along the reaction coordinate, andf(p) is a normalized
velocity distribution function.
The concentration of activated complexes CiBAmay
be evaluated from the equilibrium reaction that is as-
sumed in eq 2 via the equilibrium coefficient of this re-
action K ( T ) ,i.e.
where CA and CBAare the two-dimensional concen-
trations of the reactants, p is the area of an adsorption
site on the surface, q is a single-particle canonical en-
semble partition function, $, is the single translational
degree of freedom of the ABA complex along the re-
action coordinate, and the prime on (qaBA)’reminds us
that the one degree of freedom along the reaction co-
ordinate (4;) has been removed from this partition
function, i.e. there are 3n - 1 (not 3n) degrees of free-
dom in (qaBA)’,where n is the number of atoms per
activated complex. For example, if A and B are taken
to be atoms in this example, then there are eight de-
grees of freedom in (qaBA)’.The degree of freedom
contained in ( q i ) can very loosely be thought of as
what would be the frustrated translational degree of
freedom perpendicular to the surface of the ABAl
complex were it bonded to the surface in our usual
picture of adsorption.
Writing eq 4 as

A Langmuir-Hinshelwood Reaction Mechanism and substituting eq 5 into eq 3 allows us to write, for

Consider the following surface reaction which obeys the rate.
the Langmuir-Hinshelwood mechanism:
S-A + S-BA + ( S . . . A . . . B . . .A)’ + A2B(g) (2)
where S denotes a surface binding site. The one-
dimensional potential energy diagram along the re- In order to evaluate ( b ) , we perform the averaging
action coordinate p is shown in Fig. 2 for this activated with the properly normalized Maxwell-Boltzmann
and (assumed) exothermic reaction. The rate of the re- distribution function and the rate expression of eq 6
action is just the rate (usually written as a flux) at becomes
which the reactants cross (irreversibly) a hypothetical
surface at the saddle point of the potential energy sur-
face (the barrier in Fig. 2) which separates the reactants
from the products. Hence, it follows that

In order to evaluate qb, we make use of the classical
phase integral for the one-dimensional translation
along p which gives
where A is an effective de Broglie wavelength. Sub-
stituting eq 8 into eq 7 allows us to write the rate of
reaction as
If, as is essentially always the case, the reactants oc-
cupy specific sites, then this localization implies that
each of the partition functions that appears in eq 9 is a
vibrational partition function. This gives
and
where we have assumed that the vibrational degrees
of freedom are separable, and the electronic ground
states are nondegenerate. The “Boltzmann factors”
that result from the electronic partition functions con-
tain the potentials Ut and U which are defined in Fig.
2. Within the harmonic approximation, the individual
vibrational partition functions are given by
e - h ~ i / 2 kTs
~ alization of the reactants and the transition state.
qv,i = 1 - e-hvi/kBT, (12)
We define a new vibrational partition function qv,i as
follows:
Q , = (1 - e-hvi/kBTs
V,I - 1-l (13)
and we include the zero-point-energy contribution ex-
plicitly in the Boltzmann factor, i.e.
and
Combining eqs 10-15 into eq 9 gives
8 S-A
r=l
X
5.2 Microkinetics 975
where AU; = U i - 170 > 0. We might, for complete-
ness, multiply the right-hand side of eq. 16 by ( K ) ,
the so-called dynamical correction to transition-state
theory [l], which taken into account the possibility of
barrier recrossing by the product molecule.
We can write the reaction rate (in units of flux,
recall) in terms of a Langmuir-Hinshelwood reaction
rate coefficient as
RL-H -
kLPHCAC J ~ A (17)
where
kL-H =
- k(0)L-He-AUi/kBT, (18)
In order to obtain a “physical feeling’’ for the mag-
nitude of the preexponential factor of the Langmuir-
Hinshelwood reaction rate coefficient, let us assume
that T, = 400 K, p = 6.7 x cm2 (a site concen-
tration of 1.5 x 1015cm-2), and the ratio of the vibra-
tional partition functions (with the factor involving the
zero-point energy having been extracted) is approx-
imately unity. In this case we find that
k(o)L-H
z 5.6 x 1 0 - 3 ( ~ ) ~ m
s-l2 (19)
and (K) < 1. Hence, a “normal” value of the pre-
exponential factor would be expected to be approx-
imately lop3cm2 s - l . When “unusual” values of the
preexponential factor are observed experimentally,
they are frequently rationalized in terms of various ad
hoc assumptions concerning the localization or deloc-
In fact, most experimentally observed “unusual” pre-
exponential factors are probably not unusual at all.
Rather, they are generally a consequence of an artifact
in the analysis of the data. The root of the problem lies
in something known as the compensation effect, which
has been explained recently in terms of a temperature-
dependent distribution of reacting surface configura-
tions [2, 31.
B Eley-Rideal Reaction Mechanism
We shall now derive an expression for the rate of an
Eley-Rideal reaction in order to compare it with the
results of eqs 17 and 18 for the rate of a Langmuir-
Hinshelwood reaction. In this way, we shall be able to
assess the relative contribution of each under different
experimental conditions. The Eley-Rideal analog of
the reaction given in eq 2 may be written as
+ BA(g) * ( S . .. A . . . B . . . A)i + S + A2B(g)
(20)
References see page 983
976 5 Elementary Steps and Mechanisms
Recall that this is a direct reaction since BA is not
adsorbed on the surface. The rate of this Eley-Rideal
reaction is defined to be (in flux units, as before)
where e A / & the ratio of the fractional surface coverage
of A to the site area, is the two-dimensional concen-
tration of A, CBA,g is the three-dimensional (gas-phase)
density of BA, and kE-R is the Eley-Rideal reaction
rate coefficient (in units of volume divided by time),
which may be written as
kE-R - k(0)E-Re-AUit/kBT8
(22)
where T, is the gas-phase temperature. Since it is not
quite clear what the appropriate temperature ( T s , T,,
or something intermediate between the two in non-
isothermal systems) is to use in connection with the
partition function of the activated complex, in what
follows we shall assume T, = Tg = T when evaluating
the preexponential factor of the Eley-Rideal reaction
rate coefficient. We shall, however, use T, in the
Boltzmann factor for this direct chemical reaction, al-
though this also is clearly an approximation.
We next introduce the probability of reaction
Pr(BA),which is a function of the fractional surface
coverage of A, as follows:
where F ‘ B ~is the impingement flux of BA onto the
surface. Assuming that BA is an ideal gas and equating
the two expressions for the reaction rate in eqs 21 and
23 allow us to write the rate coefficient of the Eley-
Rideal reaction as follows:
where j,as before, is the area of an adsorption site of
reactant A, and q r , B A is the canonical ensemble parti-
tion function of a single translational degree of free-
dom of a BA molecule, which in this case is given by
The rate of this Eley-Rideal reaction may also be
written, in units of flux, as
in exactly the same way the Langmuir-Hinshelwood
rate was written in eq 3. Following the same derivation
presented earlier in eqs 4-8, we find that the Eley-
Rideal reaction rate may be written as
~~
Reference Energy Zero
‘
S+A p W
Reaction Coordinate, p
Figure 3. One-dimensional potential energy diagram (along the
reaction coordinate p ) for the hypothetical Eley-Rideal reaction,
+
S-A + BA(g) --t (S-A-B-At + S A>B(g),where S denotes a
surface binding site.
where the notation is the same as that employed earlier,
and (ti), the dynamical correction to transition-state
theory is included explicitly here. The one-dimensional
potential energy diagram along the reaction coordinate
for this reaction is shown in Fig. 3. Note, by comparing
eqs 21 and 27, that the Eley-Rideal reaction rate co-
efficient is given by
where q B A in this case refers to the single-molecule
canonical ensemble partition function of a gaseous BA
molecule. If we assume that BA is an ideal gas, we can
replace C B A by
, ~ P B A / k B T ; we also extract one of the
translational degrees of freedom from qBA in eq 27, i.e.
one factor of qt,BA given by eq 25. This allows us to
rewrite eq 27 as
where the primes on the partition functions remind us
that one degree of freedom has been extracted.
As in the case of the Langmuir-Hinshelwood
mechanism discussed earlier, we assume that both the
reactant A and, of course, the transition state are
localized. This allows us to write
and
where the potentials associated with the electronic part
of the partition function (the Boltzmann factors) are
shown in Fig. 3, and qf,BAis the partition function of

the two rotational degrees of freedom. Extracting the
zero-point energy from the various vibrational partition
functions, as was done in connection with eqs 12-15,
allows us to rewrite the Eley-Rideal reaction rate as
r 1
where AUA is defined below eq 16, and the subscript
zero implies that the zero-point energy is included in
this potential.
In order to obtain a physical feeling for the pre-
exponential factor of the rate coefficient implied by eq
32, which in this case is dimensionless, let us assume
that the ratio of the vibrational partition functions
is unity, T = 400K, p = 6.7 x cm2, and BA is
taken to be carbon monoxide whch implies that
I/qtBA z 2.7 x 10-'*cm2 and q:,BA x 150. With these
assumptions, eq. 32 becomes
REPR% 2.7 X 10-5(K)e-AUalikBTg6AFBA (33)
It is important to note that eq 33 is of nearly the same
form that would be derived for the initial rate of
chemisorption of BA with a localized transition state,
but with an additional factor of 6~ on the right-hand
side. This, of course, is as it should be since the chem-
isorption reaction can occur at any unoccupied site on
the surface, whereas the Eley-Rideal reaction can only
occur at a site that is occupied by an A adatom.
The Eley-Rideal reaction rate coefficient, given by
eq 28, may be written as
[ 1
-%
':=l qv,ABA.i e-A u,'/kB Tg
(34)
(7fi3_1qv,A,j)(q:,BAq?,BAqV,BA)
If we make the same assumptions as in the previous
paragraph concerning the temperature (400 K), the
value of l / q t . B A ( 1.6 x cm), the value of q;,BA
(1 50), and the cancellation of the vibrational partition
functions, eq 34 becomes
for the physically relevant case of a localized surface
reactant and a localized (at the saddle point of the po-
tential energy surface) transition state. The reaction
probability contained in eq 24 may be written as
5.2 Microkinetics 977
and we note that, whereas P, is a function of e A , the
ratio P r ( e A ) / d A is not a function of I ~ A .Using our
same assumptions concerning qt,BA, p, and T that were
introduced earlier, and making use of the order-of-
magnitude estimate for kE-R given by eq 35, we find
that
which, of course, could have been written down di-
rectly from eq 33. If we ignore (i.e. take to be unity) the
e A in the denominator on the left-hand side of eq 37,
for the reasons discussed below eq 33, eq 37 gives the
initial probability of chemisorption of the BA mole-
cule. Furthermore, this order-of-magnitude estimate,
which applies to a localized transition state, is not just
valid for direct chemisorption. It also applies to trap-
ping-mediated chemisorption (with the gas temperature
replaced by the surface temperature in the Boltzmann
factor), and to both molecular and dissociative chem-
isorption. Indeed, depending on where the potential
curves cross with respect to the reference energy (the
vacuum zero), i.e. the curves for physical adsorption
and molecular chemisorption for the case of molecular
chemisorption or for molecular adsorption and dis-
sociative chemisorption for the case of dissociative
chemisorption, A @ could be either greater than or less
than zero; e-Aui/ksTcould be either less than or greater
than unity. When the Boltzmann factor e-Aui/kBT is
greater than unity, which corresponds to unactivated
chemisorption with respect to the gas-phase energy
zero, then it would appear that the probability of
chemisorption (see eq 41) could exceed unity. This is,
of course, not the case. When the probability, calcu-
lated from an expression such as eq 37, appears to
exceed unity, then its value is in fact just unity, and the
adsorption reaction is flux limited. In this case an
equilibrium approximation, implicit in the derivation
of the probability of chemisorption that is analogous to
eq 37 for the Eley-Rideal reaction probability, simply
breaks down.
C Langmuir-Hinshelwood versus Eley-Rideal
Reaction Rates
It is possible to compare quantitatively the rates of the
Eley-Rideal and Langmuir-Hinshelwood mechanisms
by taking the ratio of eqs 32 and 16. Noting that
Ci= ei/p, and if we assume that T, = Tg= T , equiv-
alent dynamical corrections to transition state theory
for both cases ( K ) and that the activation energies
of the reaction AU: are the same for the Eley-Rideal
References see page 983
978 5 Elementary Steps and Mechanisms
and the Langmuir-Hinshelweed mechanisms, we find
that
1 of one of the reactants in forming the transition state
Next, we make the same assumptions that we made
earlier, namely that T = 400 K and that BA is a carbon
monoxide molecule, which implies that q$A =
2.7 x 1 0 - l ~cm2 and x 150. If we furthermore
assume that the ratio of vibrational partition functions
in eq 38 is approximately unity, then we find that
RE-R
RL-H
with the flux in units of cm-2 s-’. If the flux is in units
of site-’ s-I, and assuming that z 1.5 x 1015cm-2
is the site concentration, then eq 39 may be written as
RE-R
-
RL-H = 3.2 x lo-”( 2) (40)
With the flux on a per site basis,a pressure of 10-6T,
for example, corresponds approximately to FBA =
1 site-‘ s-l.
The implication of eqs 39 and 40 would seem to be
that the rate of the direct Eley-Rideal reaction is
always completely negligible compared to that of the
trapping-mediated Langmuir-Hinshelwood reaction.
In fact, except in rather unusual cases, the Langmuir-
Hinshelwood reaction rate dominates that of the Eley-
Rideal reaction; however, the domination is not quite
so dramatic as that which seems to be implied by eqs
39 and 40. Let us next try to understand intuitively why
our calculated ratio REPR/RLpHis so very small, and
then we shall examine our derivation more carefully
in order to decide when the Eley-Rideal rate could
become important compared to the Langmuir-Hin-
shelwood rate.
When two reactants are adsorbed adjacent to one
another, as in the Langmuir-Hinshelwood mechanism,
they can attempt to react on the order of 10’2-10’3
times each second. Think of this as either an asym-
metric atom surface stretching frequency or as a frus-
trated translational frequency of an adsorbed molecule
parallel to the surface. In a direct reaction, such as the
Eley-Rideal mechanism, this “attempt frequency” is
replaced by the impingement flux which, for example,
is approximately 1 site-’ s-l at lop6Torr. Such high
pressures would be required to overcome this limita-
tion, that adsorption (perhaps even condensation) of
the reactant would become extremely important, thus
favoring the Langmuir-Hinshelwood mechanism. We
shall return to this point later. Indeed, the situation is
even worse than implied by this argument based on
attempt frequencies, because there is a “loss of entropy”
in the Eley-Rideal mechanism. This loss of entropy is
reflected, for example, by a prefactor of 2.7 x
(rather than unity) in eqs 33 and 37. This argument can
also be couched in terms of “effective” or “virtual”
pressures of an adsorbed molecule. For example, satu-
ration coverage (0 = 1) on a lattice, the site concen-
tration of which is 1.5 x 1015cm-2, corresponds to an
effective three-dimensional pressure of an ideal gas at
400 K of approximately 3000 atm!
We can write eq 40 in an alternate pedagogic form
if we make the approximation that the adsorption-
desorption equilibrium of the BA molecule is not per-
turbed significantly either by the Langmuir-Hinshel-
wood or Eley-Rideal reactions or by the presence of
the A adatoms. Furthermore, if the surface temper-
ature is sufficiently high that ~ B A= 1, which is in the
spirit of our trying to decide when the Eley-Rideal
mechanism would be most competitive with the Lang-
muir-Hinshelwood mechanism, we may write
F~~ ~ B A
BA(g) BA(a) (41)
~ , B A
where 5BA is the probability of trapping of the BA
molecule into the chemisorbed state, and kd,BA is the
rate coefficient of desorption of the adsorbed BA. In
the pseudoequilibrium, physical situation we are imagi-
ning
~QBA
--
dt
- FBA~B
-A
which implies that
(43)
The rate coefficient for desorption kd,BA, may be writ-
ten in a Wigner-Polanyi form as
(44)
where a normal value of k d(0), B A , the preexponential
factor of the first-order desorption, rate coefficient, is
approximately 1013s-’, Ed,BA is the activation energy
of desorption of the adsorbed BA molecule, and the
temperature that appears in the Boltzmann factor is the
surface temperature. Substituting eqs 43 and 44 into
Eq. 40 gives
(45)
Within the context of eq 45, we can now inquire into
when might the direct Eley-Rideal reaction become
competitive with the trapping-mediated Langmuir-
Hinshelwood reaction. Clearly, within the approxima-

~~~ ~
tions inherent in eq 45, which have been delineated
explicitly, the crossover from the Langmuir-Hinshel-
wood rates being greater to the Eley-Rideal rates being
greater occurs at
SBAeEd,BAlkBTs M 3.2 x (46)
and the Eley-Rideal rate becomes more significant at
lower values of (BAeEd’BA1kBTs.
When might we expect rBAeEdBAlkBTs to be suffi-
ciently small for the Eley-Rideal mechanism to be-
come important? The answer to this question is when
the AB molecule is bound very weakly to the surface
and the surface temperature is very high (which tends
to make eEdBAlkBTssmaller), and when both the surface
and especially the gas temperature are sufficiently high
to make SBA very small indeed. It has been observed
experimentally on numerous occasions that the trap-
ping probability decreases slightly with increasing sur-
face temperature and decreases precipitously with in-
creasing gas temperature [4-161. Hence, we tentatively
conclude that very high gas temperatures would tend to
emphasize the Eley-Rideal mechanism. Indeed, if the
gaseous reactant were supplied with a very high trans-
lational energy and also a very narrow translational
energy distribution, then it is possible that the reaction
might only follow the Eley-Rideal mechanism. For
example, we could picture the surface temperature be-
ing sufficiently low that the rate of the Langmuir-
Hinshelwood reaction is negligible, whereas the gas
“temperature” could actually be greater than the ac-
tivation energy of the reaction. Although this peculiar
set of experimental conditions could lead to a greater
relative rate of the Eley-Rideal mechanism compared
to the Langmuir-Hinshelwood mechanism, we would
nevertheless expect the absolute rate of the direct re-
action to be quite low.
An equivalent way of considering this issue is to
imagine that the barrier for reaction is sufficiently high
with respect to the binding energy of the adsorbed re-
actant that the surface temperature would have to be so
high in order to allow the surface reaction to occur that
the lifetime of the adsorbed reactant would be too short
both to accommodate to the surface temperature and,
of more importance, to execute even one successful
hop on the surface. A physical situation to which this
argument may apply is the direct versus the trapping-
mediated dissociative chemisorption of methane on the
Ni(l11) surface [17-191. A critical assessment of the
assumptions that lead to eqs 39, 40 and 45 is presented
in Ref. 20.
The arguments above tell us that any contribution of
a direct reaction is almost always negligible, even under
the most favorable circumstances. In almost all tech-
nological applications, one can ignore the direct chan-
nel of the reaction compared to the trapping-mediated
channel.
5.2 Microkinetics 979
Thus, trapping-mediated chemisorption, even in the
case of activated dissociative chemisorption, will almost
always be far more important than direct chemisorp-
tion. In both cases, the form of the rate will involve
a preexponential factor multiplied by a Boltzmann
factor. The preexponential factor will be larger for the
trapping-mediated case, however, due to the loss of
entropy in going from the reactant to the transition
state in the direct reaction. Moreover, the apparent
activation energy will be greater for the direct reaction
since the barrier almost always lies in the exit channel
on the potential energy surface, see Fig. 1. In this case
also, direct (dissociative) chemisorption would only be
expected either within the context of a translationally
hot supersonic molecular beam experiment or when the
trapping-mediated barrier is sufficiently large that the
necessary surface temperature to surmount it precludes
a sufficient lifetime of the trapped reactant.
Since essentially all industrial catalysis is trapping
mediated rather than direct, we shall concentrate al-
most entirely on describing and quantifying this kind of
surface reactivity in the remainder of this chapter.
5.2.2.4 Trapping-Mediated Surface Reactions
A Trapping
Since trapping at a surface is the first step in all trap-
ping-mediated reactions, it is important to have a rea-
sonably complete picture concerning this dynamical
phenomenon. The reason we refer to trapping as a dy-
namical phenomenon is because it involves momentum
and energy exchange between the gas-phase atom or
molecule and the repulsive part of the gas-surface po-
tential (see Fig. l(b)). Tentatively, we might expect that
an atom or molecule is trapped at the surface if it ex-
changes sufficient energy in the gas-surface collision to
have negative total energy (kinetic plus potential) after
the collision - negative with respect to our usual zero
of energy which is the gas molecule infinitely far from
the surface and at rest. A schematic picture that helps
clarify this discussion is shown in Fig. 4, which depicts
a gas with a three-dimensional Maxwell-Boltzmann
velocity distribution interacting with a one-dimensional
potential of (say) physical adsorption. Only the dis-
sipative collision of a molecule with the most probable
velocity is shown explicitly, but keep in mind that each
molecule with its own particular velocity (kinetic en-
ergy) collides with the repulsive barrier and undergoes
energy exchange.
There are a number of dissipative (inelastic) channels
available in the gas-surface collision. If the impinging
gas-phase particle is an atom, e.g. a rare gas atom, then
~ ~~
References see page 983
980 5 Elementary Steps and Mechanisms

1.o

.8
-
f
z0
n
2 .6
n
0
C

.4

2
UI

Figure 4. Illustration of the trapping mechanism of gas-phase .2

atoms or molecules (represented by the Maxwell-Boltzmann dis-
tribution of kinetic energies f(T ) a t a surface via inelastic inter-
actions with the repulsive part of the surface potential (curve 0
shown on the left). 0 .2 .4 .6 .8
E cos la6 eI
Figure 5. Trapping probability of xenon on Pt(ll1) as a function
of E, cos’ 0, (in eV) at a surface temperature of 85 K (note that
the “inelasticity” is restricted to the solid. Electronic 1 eV is approximately equal to 96.8 kJ mol-’).
excitations of the rare gas are considered to be in-
accessible. In this case, the principal inelastic channels
are phonon excitations and electron-hole pair creation,
i.e. vibrational and electronic excitations. The detailed
nature of the substrate, e.g. the Debye temperature and
the densities of phonon and electronic states. will dic-
tate the dynamics of the individual collisions. If the
impinging gas-phase particle is a molecule, then there
are additional dissipative channels that are available.
In this case, there can be intramolecular energy transfer
that converts translational energy to either rotational
or vibrational energy. Hence, to first order we can
define the thermally averaged trapping probability to
be the fraction of molecules in the Maxwell-Boltzmann
distribution of Fig. 4 that lose sufficient energy to fall
below the zero of energy.
Rettner et al. [21], employing molecular beam tech- Figure 6. One-dimensional potential energy diagram (along the
niques, measured the trapping probability of xenon on reaction coordinate p ) proposed for the dissociative chemisorp-
Pt( 11 1). Numerous kinetic energies were employed at tion of Nz on W(100).
three different angles of incidence, to determine the
trapping probability at a surface temperature of 85 K. B Example of a Trapping-Mediated Surface Reaction
The experimental results are shown in Fig. 5. Not sur- The potential energy of the reaction
prisingly, the trapping probability decreases strongly
with increasing translational energy (gas “temper-
ature”). Notice also, however, that the trapping prob-
N2(g) 2 N2(P) 4 2N(c) (47)

ability has a strong dependence on the angle of in-
cidence, increasing with increasing angle of incidence.
It has been found that the normal component of mo-
mentum of a gas-phase particle frequently has signif-
icantly more importance than the parallel component
and, thus, increasing the angle of incidence increases
the probability of trapping [5, 21-30]. There are only
minor deviations in the trapping probability as a func-
tion of surface temperature relative to the variations
that are observed with changing gas temperature (in-
cident kinetic energy) [ S , 231. Thus, to first order, we
can think of the trapping probability as only a function
of the gas temperature.
along the reaction coordinate on W(100) is similar to
the solid curve of Fig. 6 . If the potential curves for the
physical adsorption of molecular nitrogen and for the
chemisorption of two nitrogen atoms do indeed cross
below the zero of energy, as indicated in Fig. 6, this has
an important qualitative consequence on the proba-
bility of dissociation of an impinging nitrogen mole-
cule, namely, if the surface temperature is increased,
then the reaction probability should decrease. This
statement can be rendered quantitative by considering
the initial probability of dissociative chemisorption of
nitrogen, i.e. the reaction probability on a clean sur-
face. If the surface temperature is sufficiently high that
 5.2 Microkinetics 981

the concentration of physically adsorbed nitrogen is P y + t in the limit of low T, 1

negligibly small, then a pseudo-steady state approx-
imation implies that

- --
dt
+
dep <F - (kd k,)6, % 0
where 6, is the fractional surface coverage of the
physically adsorbed nitrogen, [ is the trapping proba-
bility of molecular nitrogen into the well for physical
adsorption, F is the impingement flux of nitrogen (on a
per site basis), and kd and k, are the rate coefficients for
molecular desorption and reaction (dissociative chem-
isorption), respectively.
From eq 48 we find that the pseudo-steady state
fractional coverage of physically adsorbed nitrogen on
the surface is given by
(49)
and the reaction rate is given by
-)kr
R, = k,6, = < F (
kr + kd
Note that the form of eq 50 is intuitively predictable,
namely the rate of the reaction is given by the rate of
production of reactants ( ( F ) multiplied by the rate
coefficient for reaction, which is normalized by the sum
of the rate coefficients of reaction and desorption for
this two-channel “reaction”. The probability that an
impinging nitrogen molecule adsorbs dissociatively is
simply the rate of reaction divided by the impingement
flux, as defined above, i.e.
The form of eq 51 proves our earlier assertion that
for the potential energy diagram of Fig. 6, the reaction
probability decreases as the surface temperature in-
creases. Since both rate coefficients are of the Polanyi-
Wigner form, namely
(However, T, must not be so low that E, cannot be
surmounted!)
It is apparent from eq 51 that the reaction proba-
bility is composed of two different ingredients: a “dy-
namics” part describing the gas-surface collision (em-
bodied by 5 ) and a surface “kinetics” part describing
the competing dissociative chemisorption and desorp-
tion reactions (embodied by kd/k,). The dissociative
chemisorption of N2 on the W(100) surface has been
studied with the aim of both verifying the assumptions
inherent in the derivation of eq 51 and determining the
parameters of fundamental ph sical significance con-
tained therein, namely 5 , kd(0) l k70)
, , and Ed - E,. These
results are summarized next.
Rettner et al. [31] employed a method similar to the
reflectivity technique of King and Wells [32] to measure
the initial probability of dissociative chemisorption of
nitrogen on the W( 100) surface as a function of impact
energy, parametric in surface temperature. The results
of these measurements for surface temperatures be-
tween 300 and lOOOK, and impact energies up to
% 116 kJmol-’ (1.2 eV) are shown in Fig. 7. The first
thing to notice in Fig. 7 is that there is a minimum
P,(Ei). Below approximately 50 kJmol-’ (0.5 eV), P,
decreases with increasing Ei, whereas above this impact
energy, P, increases with increasing Ei. At lower values
1.0 - I I I I I
-
N2/W( 100)
ei=oo
-
.
.- 300K
n 800K
? 0.6 A1000K -
t
0.4

then eq 51 can be written as follows: 0.2

0.01 I , I I I 1
0.0 0.4 0.8 1.2 1.6
Kinetic Energy (eV)
Since Ed > E, in this case, the second term in the de-
Figure 7. Initial probability of dissociative chemisorption P, of
nominator becomes progressively smaller as the surface Nz W(100) as a function of impact energy E, at variable sur-
temperature decreases, which gives rise to a larger face temperatures (note that 1 eV is approximately equal to
value of the reaction probability. The corresponding 96.8 kJmol-’)
maximum and minimum values of the reaction proba-
bility for the case when Ed > Er are given by References see page 983
982 5 Elementary Stem and Mechanisms
of the impact energy, the dissociative chemisorption is
dominated by the trapping-mediated mechanism, and
is a strong function of the surface temperature, as may
be seen in Fig. 7. At higher impact energies, the prob-
ability of dissociative chemisorption increases because
the "direct" mechanism of chemisorption dominates in
this regime. At sufficiently high impact energies, where
the trapping probability approaches zero, only the di-
rect channel contributes, and the probability of chem-
isorption is approximately independent of temperature.
This regime is being approached at the highest impact
energies in Fig. 7. Indeed, at surface temperatures far in
excess of 1000 K, there is very little trapping-mediated
chemisorption at any impact energy and, therefore,
data of this type would allow us to assess the relative
importance of the two channels (trapping mediated and
direct) at the other surface temperatures as a function
of the impact energy. We should also note that the
minimum in P,(Ei) moves to lower impact energies as
the surface temperature increases (see Fig. 7) because
the probability of the direct reaction is not a significant
function of surface temperature, whereas the proba-
bility of the trapping-mediated reaction decreases rap-
idly with increasing surface temperature.
Rettner et al. [33], with further experimentation and
modeling, have determined that for impact energies less
than 9.7 kJ mol-' (0.1 eV) there is no detectable con-
tribution from the direct channel to chemisorption.
Thus, if we restrict our analysis to an incident energy
less than 9.7 kJ mol-' (0.1 eV), then a strong test of the
validity of eq 51 can be made, insofar as its proposed
description of trapping-mediated chemisorption is con-
cerned. For this purpose we rewrite eq 51 as follows:
_5 _ 1 = -kd (55)
n
r r 4 Fig. 6 that the potential curves cross, in this case,
and we note that 5 and P, have not been measured
independently. Mullins and Weinberg have performed
similar studies of the trapping-mediated dissociative
chemisorption of C2H6 on the Ir(ll0) surface in which
the trapping probability 5 and the chemisorption prob-
ability were independently measured [34-361. However,
Rettner et al. [23] have shown 5 to be a weak function
of surface temperature. If eq 55 is valid, then a plot
of In(E/P, - 1) as a function of reciprocal surface
temperature should be linear with a slope equal to
-(Ed - E , ) / ~ B
and an intercept equal to ln[kd , 1.
(0)/ k (0)
The data of Fig. 7 are plotted in this Arrhenius form in
Fig. 8, assuming 5: = 0.6, and a straight line is observed
over the temperature range employed, 300-1000 K.
This result validates the simple model that was employed
to describe the trapping-mediated chemisorption see
eqs 48-54.
The slope and intercept of Fig. 8 imply that the fol-
lowing rate parameters describe the trapping-mediated
dissociative chemisorption of nitrogen on the W( 100)
NdW(100)
1 '
-7
0'
I
VP -1
2
v
5 -2
u
-3
I I I
0 1 2 3 4
1000/ TI (K-')
Figure 8. Verification of the trapping-mediated mechanism pro-
posed in eq 51 for the dissociative chemisorption of N2 on W(100).
surface:
Ed - E, = 15.5 kJmol-' (56)
and
k r ) / k i o )= 25 (57)
where there is an uncertainty (one standard deviation)
of k1.9 kJ mol-' in the difference in the activation en-
ergies, and an uncertainty of &8 in the ratio of the
preexponential factors. Since the rate coefficient of de-
sorption of nitrogen in the precursor state has not been
measured, the rate coefficient of the elementary surface
reaction involving rupture of the molecular nitrogen
bond cannot be determined precisely. It is clear from
15.5 kJ mol-' below the gas-phase energy zero, and
that the apparent activation energy with respect to this
latter energy zero is -15.5 kJmol-I.
The fact that kf'jk!') > 1 for this reaction is com-
pletely expected [37]. Using the formalism developed in
Section 5.2.2.3, the ratio of the preexponential factors
of desorption and reaction may be written as
where both the electronic part of the partition functions
(the Maxwell-Boltzmann factor) and the zero-point
energy contribution of the vibrational partition func-
tions have been extracted from the two transition-state
partition functions that appear in eq 58. Whereas the
constrained and localized partition function for re-
action (qi): will contain four degrees of vibrational
freedom we expect that the partition function for de-
sorption (qJ); will contain two rotational degrees of
freedom (with one degree of freedom being vibra-

NdW(1OO) T=, 800 K
-
2 0.60 - x r 00
.-
0 0 m450
a 0 =60°
2
n of methane with a surface at a temperature of l000K
c
._ 0.40 -
.- ::o
p T
5
a
0.20 - 0 methane on the surface is only about seven vibrational
0.00
t’
0 1 2 3 4 5
Kinetic Energy (eV)
Figure 9. Initial probability of dissociative chemisorption via
the direct mechansim of N2 on W(100) as a function of impact
energy E, at a surface temperature of 800K, parametric in in-
cident angle. The trapping-mediated component of chemisorption
has been subtracted from the total chemisorption probability to
construct this plot [32] (note that 1 eV is approximately equal to
96.8 kJmol-’)
tional). On this basis, for a nonspherically symmetric
reactant we would expect k!’/k(O’ to lie between ap-
proximately 10 and 1000 (in the absence of any quan-
titative information concerning the relative magnitudes
of (Kd) and (K~)). The measured value of k?’/k(’) is
completely consistent with this expectation.
Finally, we note that Rettner et al. [33] have mea-
sured the initial probability of direct dissociative
chemisorption of nitrogen on W(100) at a high surface
temperature (Ts= 800 K) (the contribution from the
trapping-mediated channel was subtracted). These re-
sults are shown in Fig. 9, where the initial probability
of chemisorption is plotted as a function of impact
energy. Notice that the threshold value of the normal
energy required for detectable direct reaction is ap-
proximately 9.7 kJ mol-’ (0.1 eV), which is approx-
imately 25 kJ mol-’ greater than the minimum energy
pathway from reactant to product (i.e. the trapping-
mediated mechanism). The reason for this is clear from
Fig. 1, and it was discussed in connection with that fig-
ure. In the direct mechanism excess total energy is
necessary in order to transfer sufficient energy to the
reaction coordinate in order for the reaction to occur,
i.e. for there to be “funneling” of reaction products
into their final ground state, rather than a reflection of
the reactants from the repulsive wall of the potential. It
is for this reason that most of the surface chemistry
that occurs in the “real world” is trapping mediated
rather than direct.
As mentioned in Section 5.2.2.3, one of the few
exceptions to this general rule is likely to be the dis-
sociative chemisorption of methane. The reason for
5.2 Microkinetics 983
this is that methane typically has both quite large en-
ergy barriers for activation (with respect to the gas-
phase energy zero, i.e. Er - Ed > 0 for methane) as
well as quite small well-depths for physical adsorption,
of the order of 17 kJ mol-’ or so, on most transition
metal surfaces. For example, consider the interaction
(gas temperature equal to surface temperature in
a “bulb” rather than a “beam” experiment). At this
temperature and with a well-depth for physical ad-
sorption of 17 kJmol-*, the residence time of the
i periods, i.e. of the order of 1 ps. On a time-scale that is
this short, the concept of a trapping-mediated reaction
6
becomes ill-defined, and we would expect significant
contributions from the direct channel, i.e. the direct
reaction of methane molecules in the high-energy tail
of the Maxwell-Boltzmann distribution. This is an in-
efficient way for the molecule to react, but in the case
of methane this is likely to be the only path. We must
emphasize, however, that this is an exceptional case.
Very few chemical reactants have a binding energy on
a catalytic surface which is as small as that of methane
(or, more precisely, there are very few systems for
which Er - E d is as large as that of methane).
5.2.2.5 Synopsis
The concepts of trapping-mediated and direct surface
reactions were defined and discussed in Section 5.2.2.2
of this chapter, and we argued intuitively that trapping-
mediated reactions should be observed with greater
frequency. These arguments were quantified in Section
5.2.2.3 within the framework of elementary transition
state theory, and the trapping-mediated channel (the
Langmuir-Hinshelwood mechanism) was shown to
dominate the direct channel (the Eley-Rideal mecha-
nism) under industrial catalytic conditions. Finally, in
view of the importance of trapping-mediated reactions,
trapping dynamics and the dissociative chemisorption
dynamics of N2 on W(100) were discussed in the limit
of low surface coverages in Section 5.2.2.4.
References
1. A. F. Voter, J. D. Doll, J. Chem. Phys. 1985, 82, 80.
2. K. A. Fichthorn, W. H. Weinberg, Langmuir 1991, 7, 2539.
3. H. C. Kang, T. A. Jachimowski, W. H. Weinberg, J , Chem.
Phys. 1990, 93, 1418.
4. C. T. Rettner, E. K. Schweizer, H. Stein, J. Chem. Phys.
1990, 93, 1442.
5. C. B. Mullins, C. T. Rettner, D. J. Auerbach, W. H. Wein-
berg, Chem. Phys. Lett. 1989, 163, 11 1.
6. A. V. Hamza, H.-P. Steinruck, R. J. Madix, J. Chem. Phys.
1987, 86, 6506.
7 . C. B. Mullins, Y. Wang, W. H Weinberg, J. Vuc. Sci. Tech-
nol. 1989, A7,2125.
984 5 Elementary Steps and Mechanisms
8 A. C. Luntz, M. D. Williams, D. S . Bethune, J. Chem. Phys.
1988, 89,4381.
9. H. P. Steinruck, R. J. Madix, Surf: Sci. 1987, 185, 36.
10. S . L. Tang, J. D. Beckerle, M. B. Lee, S . T. Ceyer, J. Chem.
Phvs. 1980. 84. 6488.
11. M.’ P. D’Evelyn, H. P. Steinruck, R. J. Madix, Surf: Sci.
1987, 180. 41.
12. J. Harris,’A. C. Luntz, J. Chem. Phys. 1989,91, 6421.
13. P. Alnot, A. Cassuto, D. A. King, SurJ Sci. 1989, 215, 29.
14. K. D. Rendulic, A. Winkler, H. Karner, J. Vac. Sci. Technol.
1987, A5, 488.
15. C. R. Arumainayagam, M. C. McMaster, G. R. Schoofs, R.
J. Madix, Surf: Sci. 1989,222, 213.
16. B. E. Hayden, D. C. Godrey, Surf: Sci. 1990,232, 24.
17. M. B. Lee, Q.Y. Yang, S. L. Tang, S. T. Ceyer, J. Chem.
Phys. 1986,85, 1693.
18. M. B. Lee, Q.Y. Yang, S. T. Ceyer, J. Chem. Phys. 1987,87,
2724.
19. T. P. Beebe, Jr., D. W. Goodman, B. D. Kay, J. T. Yates, Jr.,
J. Chem. Phys. 1987,87, 2305.
20. W. H. Weinberg in Dynamics of Gas-Surface Interactions
(Eds: C. T. Rettner, N. W. R. Ashfold), Royal Society of
Chemistry, London, 1991.
21. C. T. Rettner, D. S. Bethune, D. J. Auerbach, J. Chem. Phys.
1989, 91, 1942.
22. G. Comsa, R. David, Surf: Sci. Reports 1985, 5 , 145.
23. C. T. Rettner, E. K. Schweizer, H. Stein, D. J. Auerbach,
Phys. Rev. Lett. 1988, 61, 986.
24. K. C. Janda, J. E. Hurst, Jr., C. A. Becker, J. P. Cowin, L.
Wharton, D. J. Auerbach, Surf: Sci. 1980, 93, 270.
25. C. T. Campbell, G. Ertl, H. Kuipers, J. Segner, Surf: Sci.
1981,107, 220.
26. C. T. Rettner, E. K. Schweizer, C. B. Mullins, J. Chem. Phys.
1989,90, 3800.
27. J. E. Hurst, Jr., C. A. Becker, J. P. Cowin, K. C. Janda, L.
Wharton, Phys. Rev. Lett. 1979,43, 1175.
28 K. C. Janda, J. E. Hurst, Jr., C. A. Becker, J. P. Cowin, D. J.
Auerbach, L. Wharton, J. Chem. Phys. 1980, 72, 2403.
29. L. K. Verheij, J. Lux, A. B. Anton, B. Poelsema. G. Comsa,
Surf: Sci. 1987, 182, 390.
30. C. T. Rettner, C. B. Mullins, D. S . Bethune, D. J. Auerbach,
E. K. Schweizer, W. H. Weinberg, J. Vac. Sci. Technol. 1990,
A8, 2699.
31. C. T. Rettner, H. Stein, E. K. Schweizer, J. Chem. Phys.
1988,89, 3337.
32. D. A. King, M. G. Wells, Surf: Sci. 1972, 29, 454.
33. C. T. Rettner, E. K. Schweizer, H. Stein, J. Chem. Phys.
1990, 93, 1442.
34. C. B. Mullins, W. H. Weinberg, J. Vac. Sci. Technol. 1990,
A8, 2458.
35. C. B. Mullins, W. H. Weinberg, J. Chem. Phys. 1990,92,3986.
36. C. B. Mullins, W. H. Weinberg, J. Chem. Phys. 1990,92,4508.
37. C. T. Campbell, Y.-K. Sun, W. H. Weinberg, Chem. Phys.
Lett. 1991, 179, 53.
5.2.3 Theoretical Modeling of Catalytic
Reactions
J. K. N 0 R S K O V AND P. STOLTZE
5.2.3.1 Introduction
A catalyst is a substance which modifies the rate or the
selectivity of a chemical reaction. A particular catalyst
and reaction system is therefore characterized by the
rates for different elementary steps in the reaction, and
the theoretical modeling of catalytic reactions must
therefore, as the end product, have the calculation of
the kinetics.
The kinetics of a catalytic reaction is usually mea-
sured in a reactor under conditions relevant to the
industrial process. The measured overall rates can then
be fitted to a mathematical model, the macroscopic
kinetics. This is extremely convenient for process design
purposes [l].
If the aim is to explore the mechanism of the re-
action and to understand which are the important
parameters of the catalyst determining the activity,
then a microkinetic model is needed. A microkinetic
model is based on a detailed mechanism and indepen-
dent information about the rates of the elementary
steps involved and the stability of the intermediates. It
can be said that the microkinetic model is the synthesis
of all the basic knowledge about a reaction over a
given catalyst.
The input into a microkinetic model is usually mea-
sured rates of elementary steps and measured heats of
adsorption together with thermodynamic data for
the gas (or liquid) phase above the catalyst. The input
parameters may be measured over the catalyst [2]. This
gives information of direct interest for the catalyst sys-
tem considered, but often the interpretation of the
experiments is difficult because the state of the active
surface is not known and may vary with the conditions
of the experiment. Alternatively, the input can be taken
from measurements on model systems. If the structure
of the catalyst is known and there is a suspicion which
is the active phase, then it is possible to isolate this
phase, usually in the form of a single-crystal surface
and do experiments on this model using the chemical
and structural characterization tools available to sur-
face science [3]. A third possibility for obtaing input
has become available during the early 1990s. Electronic
structure theory has developed to a point where realis-
tic bond energies and activation barriers can be calcu-
lated [4]. Typically, the model catalysts used in such
calculations are even more idealized than in the surface
science experiments (perfect surfaces, ordered over-
layers etc.), but the insight into the details of the po-
tential energy surface of the reaction is much greater.
The modeling of a surface chemical reaction can
take place at many different levels of complexity.
Section 5.2.3.2 gives an overview of the various ap-
proaches ranging from a complete description of the
dynamics of the reaction over simulations which in-
clude the adsorbate-adsorbate interactions to the sim-
plest mean field approaches. The latter is by far the
most widespread in use and in Section 5.2.3.3 a de-
tailed description is given of this approach using the
synthesis of ammonia over an iron catalyst to exem-
plify the concepts introduced. Finally, Section 5.2.3.4

I
I
Full dynamics
I
I
1 neglect details
ofdynamics
I Monte Carlo description I
mean field
approximation
Continuum description
neglect adsorbate-
adsorbate interactions
Langmuir-Hinshelwood
description
Figure 1. Hierarchy of methods for treating surface reactions.
briefly points to a few examples of reactions that have
been treated in this way.
5.2.3.2 Different Approaches to Simulations of
Surface Reaction Kinetics
As mentioned above, it is possible to attack the prob-
lem of simulating the rate of a surface chemical re-
action at several different levels of complexity [ 5 ] . At
the most fundamental level the time development of
the reaction system is followed in detail, and at the
most approximate level only gross averages are con-
sidered. The hierarchy of methods for dealing with
surface reactions is illustrated in Fig. 1, and the fol-
lowing discusses the different levels individually.
A The Full Reaction Dynamics
At the most fundamental level the time development
of the system is followed in detail. The reactants are
started in a specific initial (quantum) state and the
equation of motion are propagated to give the final
state. The equation of motion of the system is the time-
dependent Schrodinger equation, or, if the atoms in-
volved are heavy enough (not H or Li), Newton’s
equation. The starting point is the adiabatic potential
energy surface on which the process takes place. For
some reactions, electronic excitations during the reac-
tion are important and must be included in addition to
the electronically adiabatic dynamics [6].
The adiabatic potential energy surface is the ground-
state electronic energy of the system as a function of all
the degrees of freedom of the system. Figure 2 shows as
an example a two-dimensional cut, through the poten-
tial energy surface for Hz dissociation on a Cu( 1 1 1) sur-
face 171. The two dimensions shown are the distance of
5.2 Microkinetics 985
2.25
1.I5
’ 0.6 0.8 1 1.2 1.4
Figure 2. A calculated ab initio potential energy surface for H2
on Cu(l11). The contours of constant potential energy are shown
in a plane containing the distance of the molecule (lying parallel
to the surface perpendicular to the bridge site) above the first Cu
plane, and the H-H bond length.
the molecule over the surface and the H-H distance.
There are another four Hz degrees of freedom plus all
the degrees of freedom of the surface atoms for a
complete description, but the two degrees of freedom
included in the cut contain the most important in-
formation [8, 91. The barrier for dissociation is about
0.7 eV (or 70 kJ mole-’) and the vibrational excitation
of the incoming molecule will help surmounting the
barrier (this follows from the fact that the barrier is to a
substantial extent along the H-H coordinate).
Both conclusions are in agreement with detailed
molecular beam studies [lo]. From the potential the
dissociation probability Q,d,v , j : T,) can in prin-
ciple be calculated as a function of the translational
energy &n, angle of impact (0, d), vibrational state v,
and rotational statej of the incoming molecule, and the
temperature T, of the surface [l 11. At this level of detail
it is possible to see whether the reactivity may be
enhanced by selectively exciting the vibrational or
rotational degrees of freedom of the reactants.
If we are only interested in the rate for a gas in
thermal equilibrium at a temperature T over a surface
of the same temperature, the adsorption rate can be
calculated as
XS(Ekin,e,d,v,j, T)sinedQdddEkin (1)
Such a detailed description could in principle be
made of every elementary step in a given reaction,
while a dynamical simulation of a whole chemical
process including several elementary steps is usually
impossible. Typically, intermediates in a reaction can
References see page 990
986 5 Elementary Steps and Mechanisms
have lifetimes which are many orders of magnitude
larger than typical times for a dynamical simulation (a
few picoseconds). Another point is that if the aim is to
obtain an elementary reaction rate for a system at a
given temperature, the full dynamical approach may be
too detailed. Clearly, the Boltzmann factor in eq 1 will
single out only the lowest energy processes, that is, the
processes involving molecules with energies just around
the minimum barrier E,. A determination of the bar-
rier will therefore usually give the rate as
y ( T ) = AePEa/lkT (2)
Estimating the prefactor is in principle not an easy
task, but often simple absolute rate theory suffices [12].
Given a rate expression as in eq 2 for the elementary
steps, it is possible to simplify the description of the
rate problem substantially.
B The Kinetic Monte-Carlo Approach
The main difficulty in setting up a set of rate equations
for a surface chemical reaction based on rate expres-
sions such as eq 2 for the elementary steps is that the
activation energies and prefactors may depend strongly
on the surroundings of the reacting molecule. This is a
well-known problem in gas- and liquid-phase reactions,
but it is even more severe for surface reactions. The
reason is that when atoms and molecules are con-
densed onto a solid surface they are forced into a
framework with a typical spacing between adsorbates
of only a surface unit cell dimension. The interactions
between adsorbates can therefore be quite substantial,
leading to ordered adsorbate structures at low tem-
peratures. The best procedure is therefore to include
the interactions between the adsorbates as directly as
possible.
If the surface can be viewed as a lattice on which the
reaction proceeds, then the problem can be formulated
in the following way. Instead of considering the posi-
tion of the adsorbates on the surface as a continuous
variable we can associate each adsorbate to a lattice
site. This is often possible even when adsorption leads
to a reconstruction of the surface. There are basically
five kinds of processes that can take place on this lat-
tice: adsorption, dissociation, diffusion, recombination,
and desorption. Figure 3 shows schematically the pos-
sible processes. At each lattice site the configuration of
surrounding adsorbates will determine the activation
energies for the different processes, if the form and
strength of the interactions are known.
The kinetics can then be calculated using the kinetic
Monte-Carlo method [5, 13-15]. In this method the
different processes discussed above are now simulated
on the lattice by randomly letting adsorbates arrive
on the surface, dissociate, diffuse, recombine or desorb.
For each attempted process the probability of a suc-
cessful event is taken to be proportional to the rate (eq
,I<
"1 adsorption
cfldesorption
Figure 3. A schematic overview of the five elementary steps in a
surface reactions: adsorption, dissociation, diffusion, recombina-
tion, and desorption.
2). Since the detailed configuration during each event is
known, the actual activation energy including all local
interactions can be calculated and used. Adsorption
processes will have a rate where the prefactor is related
to the gas-phase pressure.
If the interaction energies are known from indepen-
dent experiments or calculations, this will give a very
detailed description of the process. The computational
effort needed is, however, very substantial compared to
the simpler mean-field description described in the next
section. At low temperature, where diffusion is slow
and ordered structures appear, this is the only possible
accurate description of the system. At temperatures
high enough that diffusion is fast and the adsorbates
are evenly and randomly dispersed over the surface,
such an approach is usually not necessary [ 13, 151.
C The Continuum Description
Instead of considering the detailed configuration of the
system as outlined above, a considerably simpler mean-
field description is possible. Here, only average prop-
erties are considered. The rate of a given elementary
step involving adsorbates A and B are assumed, given
by
R = reAeB (3)
where OA and d B are the coverages of A and B, respec-
tively, and the rate constant r is given by an expression
such as eq 2. The activation energy and prefactor may
still depend on the surroundings, but only in an aver-
age way through the coverages.
Even at this level the kinetics may become quite
complex. There are several examples [16, 171 where the
adsorption probability of one of the reactants depends
strongly on the coverage of another. This can lead to a
strongly nonlinear set of equations for the total rate of
a reaction and to very interesting phenomena such as
kinetic oscillations and chaos [13, 15-17].
D Langmuir-Hinshelwood Description
At the simplest level even the coverage dependence of
the parameters entering the rate constants is neglected.
This is obviously an oversimplification, but is still an
extremely useful way of analyzing a reaction, and it has
been shown in a number of cases to work remarkably
 5.2 Microkinetics 987

well. The next section discusses this approach in more At equilibrium the kinetic equation must go
detail. smoothly into the appropriate equilibrium equation.
The rate constants k+ and k-, and the equilibrium
constant K, are thus related by
5.2.3.3 Simplest Mean-Field Approach
(9)
This section first discusses in some detail how to con-
struct a microkinetic model at the simplest level, where and the activation energies E i and E!, and the energy
interaction between adsorbates are only included of reaction AE, are related by
through the saturation coverage. The use of such a
model is then considered. E$ - EL = AE (10)
The kinetic model be formulated using kinetic equa-
A Constructing the Model tions for all steps or using equilibrium equations for all
To make the following discussion more explanatory, but the slowest steps. The latter approach reduces the
ammonia synthesis is used as an example as this is one computational effort and leads to a kinetic expression,
of the simplest catalytic reactions from a modeling which is far easier to analyze. However, if a step, which
viewpoint: is slow in reality, is modeled by an equilibrium equa-
+
N2(g) * + N2* tion, the microkinetic model becomes unrealistic and it
may in some cases be the simplest to treat a problem-
+
N2* * + 2N* atic step by a kinetic equation.
N* + H a + NH* * + For the NH3 synthesis the kinetic scheme has been
NH* + H* + NH2* + * (4) treated by both approaches, [21-261, and identical
NH2* + H* + NH3* * + results are obtained [26], when the same set of input
parameters is used.
NH3* + NH3(g) +* When the mechanism has only one slow step, the
H2(g) + * + 2H* system of equilibrium equations and the rate equation
may be solved with respect to 8* and the coverages by
In this mechanism, the second step is known experi- intermediates may be expressed by 8, and the partial
mentally to be slow [18-201. pressures of the reactants and products. If the mecha-
The competition for adsorption sites is very im- nism has two or more slow steps [27-291, the solution
portant for the kinetics of a heterogeneous catalytic for the coverage by free sites is more difficult. For three
reaction. For this reason sites, *, are included as a re- or more slow steps, numerical solution is likely to be
actant in the kinetic model. As a site must be either free the only option.
or occupied by one of the surface intermediates, there is The equilibrium constants are expressed in terms of
a conservation law for the coverages, the molecular partition function, e.g.
C8x=1 (5)
where 8x is the coverage by the intermediate X.
In writing this equation it is implicitly defined that The molecular partition function Z is a product of
8x = 1 represents saturation. With this convention, contributions from each mode for the molecule. The
coverages may be interpreted as probabilities. number and nature of the modes is known from the
Each of the equilibrium steps gives an equilibrium symmetry of the intermediate.
equation, e.g. The enthalpy of formation for reactants and inter-
PNz = 8 ~ ~ *
K1-8* mediates may be calculated from the expression for the
(6) partition function
PO
A slow step gives a kinetic equation, e.g. dln Z
H = kBT2-
2 dT
r2 = k + 2 8 ~ ~ ,-
&k-28N* (7)
The enthalpy obviously depends on temperature.
which expresses the net rate r as the difference between However, even at T = 0 the enthalpy of an inter-
the forward rate r+, and the backward rate r - . mediate may differ substantially from the electronic
The condition for equilibrium is that both the for- binding energy due to contributions of iAo from the
ward and the backward rate are much larger than the zero-point motion in vibrational degrees of freedom.
net rate
Ir, - r-1 << min(r+,r - ) (8) References see page 990
988 5 Elementary Steps and Mechanisms
Simulation of adsorption is an important tool for the
determination of rate constants. Complicated adsorp-
tion experiments may be simulated through a numer-
ical integration of the relevant part of the reaction
mechanism once the parameters in the model are
known.
However, for determination of parameters it is pos-
sible to obtain an expression for the sticking coefficient
0, by equating the rate of adsorption in the kinetic
model to the conventional expression used to define 0.
For instance, the sticking coefficient for N2 at zero
coverage, 00, is
where d is the density of adsorption sites and m is the
mass of a Nz molecule, and the activation energy for 00
is
The first term comes from the thermodynamics of the
molecular precursor, the second term is the activation
energy for the dissociation, while the third term comes
from the temperature dependence of the flux. The acti-
vation energy for the rate determining step is obviously
significantly different from the activation energy for the
sticking coefficient.
While sticking coefficients and more generally the
kinetics of adsorption is an important source for rate
constants, simulation of desorption experiments is the
primary source of binding energies, e.g.
for N2 desorption in the limit of infinite pumping
speed. The ground state energy for N* is calculated
from K2 after K2 has been determined from experiment.
Simulation of more complicated desorption mecha-
nisms such as disproportionation is possible through
numerical solution of the relevant equations.
In the case where there is one slow step in the re-
action mechanism, the solution for the rate of the cat-
alytic reaction is straightforward [24, 251:
The parenthesis has a simple interpretation in terms of
the thermodynamics of the gas phase while the inter-
esting part of the kinetics comes from the factor 6:.
When the reaction mechanism has more than one
slow step [27-291, the solution for the rate is more
complicated, but the final expression is remarkably
similar.
10-3 1
Experimental output
Figure 4. Comparison of calculated and measured ammonia
production over a commercial iron-based catalyst for a broad
range of temperatures, pressures, N : H ratios and gas flows; from
Ref. 23.
While the number of parameters may be large, only
a few of the parameters are usually significant. The
significant parameters are easily determined by calcu-
lation of the sensitivity of the calculated rate at typical
conditions to a small variation in the value of each
parameter. For NH3 synthesis, for instance, the rate of
dissociation of Np and the binding energy for N* are
the most significant parameters and the kinetics of de-
sorption of N2 in a temperature programmed de-
sorption (TPD) experiment is rather closely related to
the kinetics of NH3 synthesis [30, 311.
Fragments of the kinetic model may be tested by
detailed simulation of the same experiments as have
been used to determine the parameters in the model.
Failure to reproduce these experiments is entirely pos-
sible if a wrong reaction mechanism has been assumed.
A sensitive test can be made by simulation of a re-
actor through numerical integration of the rate ex-
pression. This will test the model at higher pressures,
frequently at higher coverages, in the presence of all
reactants, intermediates and products, Fig. 4.
The model can be further tested for internal con-
sistency. Steps treated through kinetic equations should
be slow at least under some conditions or the model
may obviously be simplified. The significant parame-
ters should obviously be determined rather directly
from experiments. The coverages of the intermediates
calculated under reaction conditions should not be
greatly different from the coverages in the experiments
used to determine the parameters.
Once the kinetic model passes the tests, the model
may be used to understand details of the catalytic re-
action, such as the origin of the macroscopic activation
energy, the reaction orders, or the structure of empiri-
cal kinetics.
The activation enthalpy for the catalytic reaction may
be calculated from the microkinetic model as [24, 251

For a mechanism with a single slow step, the activation
enthalpy for the catalytic reaction equals the activation
energy for the slow step plus the averaged binding en-
ergy for the intermediate. The average binding energy
is formed by weighing the desorption energy for each
intermediate through equilibrium steps with the cover-
age for the intermediate. Intuitively, these two con-
tributions represent the increase in the reaction rate at
increased temperature through increase in the rate lim-
iting step and increase in the number of free sites, re-
spectively. For models with more than one slow step
[27-291 the contribution to the average desorption en-
ergy is further weighted by the relative rates of the slow
steps.
As the desorption enthalpy for the intermediates de-
pends on temperature (eq 12), and as the coverage by
intermediates may vary significantly with the operating
conditions for the catalyst, the microkinetic model will
often predict regions of operating conditions where the
activation enthalpy is significantly larger than usual.
Except at conditions where the coverage of all inter-
mediates is negligible, the activation enthalpy for the
catalytic reaction will be larger than the activation en-
thalpy for the slow step.
B Using the Model
In the preceding section we developed the simplest
methods for building a microkinetic model using the
ammonia synthesis as an example. For this reaction it
turns out that if all input data are taken from surface-
science experiments on single crystal Fe(ll1) surfaces
[20] and the catalyst is characterized only by the active
Fe area, then the model gives a reasonable account of
the kinetics of the real catalyst under industrial con-
ditions [23-251. Due to its simplicity, the model is far
from perfect; several details of the kinetics are not accu-
rately described. Also, the quantitative aspects of the
model calculation are no more accurate than the input
parameters. For the ammonia synthesis reaction, for in-
stance, there is still some uncertainty about the sticking
probability and about the absolute value of the nitro-
gen binding enthalpy [20, 30, 32, 331. Also, the mea-
sured values for the adsorption rate are known only for
low coverages. While the former problem does not ap-
pear to affect the results significantly [30], the effects of
the latter are still not understood in detail [31].
In spite of the shortcomings of the modeling, the real
strength is that it can be used to understand variations
in the catalytic activity from one system to another.
The stability of the intermediates and the activation
barriers are among the input parameters for the mi-
crokinetic model, and it is straightforward to calculate
the effects of changes in stability for some or all the
intermediates. Adsorbed promotors (alkali atoms, for
5.2 Microkinetics 989
-01
Elel
- 02 i-1
1
iI
I ~
0 5 10
Number of d electrons
Figure 5. Trends in dissociative energies and activation energies
for dissociation as a function of the number of d electrons. The
results are calculated in the Newns-Anderson model including
the coupling between an adsorbate level E, and the metal d band.
All energies are measured relatively to the Fermi energy in units
of the d-band width and E, = 1.5 corresponds to a chemisorbed
atom with a level well below the d band, whereas E, = 0.0 corre-
sponds to a dissociating molecule with its antibonding level par-
tially filled around the Fermi level.
instance) are known to change the stability of the in-
termediates and, by including such changes in the
model, one can immediately see if these changes also
account for the observed effects of promoters in the
real catalyst system. This turns out to be the case for
the potassium promotion of the iron-based catalysts
used for the ammonia synthesis, where it is the K-
induced lowering of the barrier for N2 dissociation and
lowering of the NH, stability on the surface which
dominates [30].
The model can also be used to begin understanding
which metals are active and which are not. It is known
roughly how the stability of the intermediates and the
barrier for Nz dissociation varies through the periodic
system [34]. Figure 5 shows the result of a simple model
calculation of these variations as a function of the
number of d electrons in the metal. If these variations
are included in the kinetic model, variations in the
ammonia activity shown in Fig. 6 are obtained. The
modeling clearly shows the volcano shape, which has
been observed experimentally for many reactions [35].
For elements to the left of Fe, the kinetics is similar
to the kinetics of Fe, but the rate is low due to the
strong bonding of N*. For elements to the right of Fe,
the rate is low due to the low sticking coefficient for N2,
References see page 990
990 5 Elementarv Stem and Mechanisms
7 observed for the catalyst and the microkinetic models
0.30 X
, 00-lNi
5 6 7 8 9 1
Number of d electrons
Figure 6. The calculated ammonia production for a fixed set of
reaction conditions as a function of the number of d electrons;
from Ref. 35.
the coverage by N* is low, and the predicted kinetics
are somewhat different to the kinetics of Fe. Iron (and
the other elements with approximately seven d elec-
trons) is the optimum choice not because the sticking
probability is high or because there is much free sur-
face, but because it is the best compromise between the
two effects.
5.2.3.4 Selected Examples
There are now a number of microkinetic studies of in-
dustrially important reactions. This section provides
pointers to some examples.
Microkinetic modeling has been applied in mecha-
nistic studies of the synthesis of HCN [36], the hydro-
genation of ethylene over Pt [37] the reactions CO + 0 2
and CO + NO over Rh [38], the reaction between NO,
and NH3 [39], and the oxidation of CH4 over oxides
[2]. Microkinetic modeling of nonisothermal reactions
has also been used in studies to understand the ignition
of H2 and 0 2 over Pt [15, 40-431 and the highly exo-
thermic partial oxidation of CH4 [44, 451.
Microkinetic modeling of the oxidation of CO over
Pt [16, 17, 46, 471 has demonstrated quantitatively that
reconstruction of the Pt surface under reaction con-
ditions leads to a coupling between reactivity and
coverages by intermediates, which causes the reaction
condition to display deterministic chaos.
The catalytically active structure for the Cu-based
catalysts used in the water-gas shift reaction and, in
particular, in methanol synthesis, has been contro-
versial. Recent microkinetic models of these reactions
[27-29, 481 starting from experimental data for Cu
single crystals, show a good agreement between the re-
sults calculated by the microkinetic model. The kinetics
thus suggest that the active structure in the catalyst is
the basal planes of metallic Cu.
References
1. J. Rostrup-Nielsen, in Elementary Reaction Steps in Hetero-
geneous Catalysis, Kluwer Academic Publishers, 1993, p 441.
2. J. A. Dumesic, D. F. Rudd, L. M. Aparicio, J. E. Rekoske,
A. A. Trevino, The Microkinetics of Heterogenous Catalysis,
The American Chemical Society, 1993.
Cu
3. G. A. Somorjai, Introduction to Surface Chemistry and Cat-
alysis, John Wiley and Sons, 1994.
0 4. R. A. van Santen, Theoretical Heterogeneous Catalysis,
World Scientijic Lecture and Course Notes in Chemistry,
Word Scientific, 1991, vol. 5.
5. H. Chuan Kang, W. H. Weinberg. Surf: Sci. 1994, 2991300,
755.
6. S. Holloway, J. K. Nsrskov, Bonding at Surfaces, Liverpool
University Press, 1991.
7. B. Hammer, M. Scheffer, K. W. Jacobsen, J. K. Nsrskov.
Phys. Rev. Lett. 1994, 73, 1400.
8. D. Halstead, S. Holloway, J. Chem. Phys. 1990, 93; 2859.
9. S. Kuchenhoff, W. Brenig, Y. Chiba. SurJ Sci. 1993, 245,
389.
10. C. T. Rettner, D. J. Auerbach, H. A. Michelsen. Phys. Rev.
Lett. 1992, 68, 1164.
1 1 . H. A. Michelsen, C. T. Rettner, D. J. Auerbach, R. N. Zare.
J. Chem. Phys. 1993, 98, 8294.
12 G. Wahnstrom, B. Carmeli, H. Metiu. J. Chem. Phys. 1988,
88, 2478.
13 J. W. Evans, J. Chem. Phys. 1992,97, 572.
14 M. W. Deem, W. H. Weinberg, H. C. Kang. Surf; Sci 1992,
276, 99.
15 V. P. Zhdanov, B. Kasemo. Surf: Sci. Rep. 1994,20, 111.
16 G. Ertl, Ado. Catal. 1990, 37, 231.
17 G. Ertl, Surf: Sci. 1993, 2871288, 1 .
18 P. H. Emmett, S. Brunauer. J. Am. Chem. Soc. 1934, 56, 35.
19 M. Grunze in The Chemistry and Physics of Solid Surfaces
and Heterogeneous Catalysis, (Eds: D. P. Woodruff, D.
King.) 1982, Vol 4, p. 413.
20 G. Ertl in Catalysis Science and Technology (Eds: J. R.
Anderson, M. Boudart) 1983, Vol 4, p. 209.
21 M. Bowker, I. Parker, K. C. Waugh. Appl. Catal, 1985, 14.
101.
22 M. Bowker, I. Parker, K. C. Waugh. SurJ Sci. 1988, 197,
L223.
23 P. Stoltze J. K. Narskov. Phys. Rev. Lett. 1985, 55, 2502.
24. P. Stoltze. Phys. Scr. 1987, 36, 824.
25. P. Stoltze, J. K. Nsrskov. J. Catal. 1988, 110, 1.
26. J. A. Dumesic, A. A. Trevino. J. Catal. 1989, 116, 119.
27. C. V. Ovesen, P. Stoltze, J. K. Nsrskov, C. T. Campbell. J.
Catal. 1992, 134, 445.
28. P. B. Rasmussen, P. M. Holmblad, T. S. Askgaard, C. V.
Ovesen, P. Stoltze, J. K. Nsrskov, I. Chorkendorff. Catal.
Lett. 1994, 26, 313.
29. T. Askgaard, J. K. Nsrskov, C. V. Ovesen, P. Stoltze. J.
Catal. 1995, 156, 229.
30. P. Stoltze J. K. Nsrskov. Topics in Catal. 1994, I , 253.
31. L. M. Aparicio, J. A. Dumesic, Topics in Catal. 1994, 1 ,
233.
32. C. T. Rettner, H. Stein. Phys. Rec. Lett. 1987, 59, 2768.
33. C. T. Rettner, H. Stein. J. Chem. Phys. 1987, 87, 770.
34 S. Holloway, B. I. Lundquist, J. K. Narskov in Proc. In?
Congr. Catalysis, Berlin, 1984, Vol. 4, p. 85.

35. J. K. Nerrskov, P. Stoltze. Surf: Sci. 1987, 1891190, 91.
36. N. Waletzko, L. D. Schmidt. AZChEJ. 1988,34, 1146.
37. J. E. Rekoske, R. D. Cartright, S . A. Goddard, S . B. Sharma,
J. A. Dumesic. J. Phys. Chem. 1992, 96, 1880.
38. S. H. Oh, G. B. Fisher, J. E. Carpenter, D. W. Goodman, J.
Catal. 1986, 100, 360.
39. J. A. Dumesic, N. Y. Topsere, T. Slabiak, P. Morsing, B. S .
Clausen, E. Tornquist, H. Topsere in Proc. loth Znt. Congr.
Catal., Budapest, 1992.
40. B. Helsing, B. Kasemo, V. P. Zhdanov. J. Catal. 1991, 132,
210.
41. M. Fassihi, V. P. Zhdanov, M. Rinnemo, K-E. Keck, B.
Kasemo. J. Catal. 1993, 141, 438.
42. W. R. Williams, C. M. Marks, L. D. Schmidt. J. Phys. Chem.
1992,96, 5922.
43. M. P. Zum Mallen, W. R. Williams, L. D. Schmidt. J. Phys.
Chem. 1993, 97, 625.
44. D. A. Hickman, L. D. Schmidt, AZChEJ. 1993,39, 1 164.
45. D. A. Hickman, L. D. Schmidt, Science, 1993,259, 343.
46. K. Krischer, M. Eiswirth, G. Ertl. J. Chem. Phys, 1992, 96,
9161.
47. K. Krischer, M. Eiswirth, G. Ertl, J. Chem. Phys, 1992, 97,
302.
48. C. T. Campbell, Topics in Catal. 1994, I , 353.
5.2.4 Theory of Surface-Chemical
Reactivity
R. A. VAN SANTEN
AND M. NEUROCK
5.2.4.1 Introduction
The relationship between the overall rate of reaction
and the reactivity of a catalytically active surface is
complex due to competing phenomena at the elec-
tronic, atomic, and molecular levels. The catalytic re-
action is a cycle composed of a series of fundamental
elementary steps in which the state of the active surface
is maintained by regenerating active surface sites upon
each cycle turnover. The chemical state of individual
reaction sites and the overall active surface are readily
altered as a function of reaction conditions. Surface
coverage, for example, can change with conversion.
This affects the distribution of vacant sites available
for reaction, and can subsequently lead to surface re-
construction. Temporal changes in either the chemical
state or the surface coverage ultimately alters surface
reactivity.
The overall reaction rate of a catalytic cycle is con-
trolled by a delicate balance of the strength of the
adsorbate-surface interaction. A strong interaction
between the reactant and surface tends to favour the
initial activation of the adsorbate and its subsequent
dissociation. This is desirable for systems which are
activation-limited and have no secondary decomposi-
tion routes. By increasing the reactant adsorption,
however, we are also likely to increase the product
5.2 Microkinetics 991
adsorption. This can change the nature of the limiting
step and in some cases lead to desorption-limited
kinetics. The optimum is, therefore, a balance of the
adsorbate-surface chemical bond strength; the reac-
tants readily adsorb and the products desorb. This is
the well-known Sabatier’s principle where the hghest
reaction rates are achieved at intermediate adsorbate-
surface chemical bond strengths. It is the basis for
Volcano-structure-reactivity plots which depict a
maximum in reactivity at a given metal-adsorbate
bond strength.
The direct extrapolation of detailed information on
surface chemical bonds to predications about the over-
all catalytic reactivity, while useful, must be done with
extreme caution so as to properly account for all rele-
vant competing phenomena. The surface science com-
munity has been very careful in their attempts to bridge
well defined single crystal results, performed at ultra-
high vacuum (UHV) conditions, to results for real
catalytic surfaces at operating conditions. Schlogl [ 11
has cited differences in the surface structure (materials
gap), reactor operating pressures (pressure gap), re-
actor transport phenomena (transport gap), and the
single-crystal model of the polycrystalline surface
(model gap) as important issues in mapping funda-
mental information on the nature of the adsorbate-
surface interaction to quantitative kinetic descriptions
of real catalytic phenomena.
While there are still a number of issues which need to
be resolved, we are moving closer towards narrowing
these gaps. There is a growing number of examples
whereby the gaps have actually been closed. Ertl [2],
Goodman and co-workers [3], and Nsrskov [4],for
example, have been instrumental in demonstrating how
well-defined surface science data and theoretical de-
scriptions of the binding and reactivity of adsorbates
on surfaces can be used to predict real catalytic systems
under operating conditions for various example sys-
tems. Ertl demonstrated the well-established ammonia
synthesis mechanism from a series of probing UHV
experiments. Goodman and co-workers [3] presented
a number of relevant catalytic examples, for both
structure-insensitive and structure-sensitive reactions,
whereby a pressure-modified UHV chamber was used
to resolve pressure-gap issues. Some examples include
CO methanation, CO oxidation, alkane adsorption,
and hydrogenolysis on different metal surfaces. N0r-
skov [4]applied statistical mechanics to predict ammo-
nia synthesis production on an iron-based potassium-
promoted catalyst over various temperatures and up to
300 atm, and water-gas shift kinetics on copper.
While high-pressure designed UHV chambers, such
as that described by Goodman, allow us to probe
binding and surface reactivity of molecular adsorbates
References see page 1004
992 5 Elementary Steps and Mechanisms
on transition metal surfaces from experiments, careful
first-principle quantum chemical calculations coupled
with atomic reactivity simulations provide a comple-
mentary computational approach to modeling pressure-
gap concerns. A fundamental understanding of (i) the
nature of the adsorbate-surface interaction and (ii)
how reaction probabilities change as a function of sur-
face composition changes is therefore essential. Both
of which can be gained from advanced computations.
Surface coverage changes lead to changes in the nature
of the active surface which can ultimately alter the
nature of the rate limiting step of the catalytic reaction
cycle.
The selectivity of a reaction is controlled by both the
energetics of the available reaction channels, as well as
the temporal surface composition. By way of illus-
tration, consider the balance between elementary dis-
sociation and desorption events. At low surface cover-
ages, the step with the lowest energy barrier proceeds.
At higher coverages, however, the number of available
vacant sites become a critical concern. For dissociation
to proceed, there must be at least on vacant site ad-
jacent to the adsorbed species. Dissociation is, there-
fore, highly dependent upon surface concentration. The
reaction rate of an isolated molecule is then an explicit
function of surface coverage.
Surface coverage or surface composition can also act
indirectly and change surface reactivity by electroni-
cally altering the adsorbate-chemical bond strength
through lateral surface interactions. Lateral surface in-
teractions lead to nonideal mixture properties. Effective
attractive interactions can lead to ordered-disordered
phenomena, whereas effective repulsive interactions
can lead to phase separation with surface island for-
mation. In the latter case, the overall reaction rate is
significantly reduced because reactions between adsor-
bates can only occur at the boundary of the islands
formed.
Island formation is usually accompanied by surface
reconstruction [ 5 ] . Interestingly, Ertl and co-workers
[6]have demonstrated that, at least at UHV conditions,
“hot” ad-surface atoms generated by dissociation of
an adsorbed molecule can migrate over an 80 8,distance
before finally adsorbing. The energy released by dis-
sociation dissipates slowly on the time-scale of transla-
tional motion of the “hot” atoms, and thus allows
these species to move considerable distances before
finally chemisorbing.
In this chapter we examine a set of important con-
cepts of the adsorbate-surface chemical bond deduced
from quantum chemical calculations and how these
fundamental ideas can be coupled with statistical me-
chanics to provide energetic and kinetic analyses. Cur-
rent first-principle quantum chemical analyses com-
bined with transition-state theory can help to elucidate
overall reaction pathways, competing elementary steps
of the catalytic cycle, and actual specifics about the
controlling reaction mechanism. As was suggested,
coupling of first-principle electronic and kinetic de-
scriptions enable us to begin to bridge some to the
pressure gap and material gap issues discussed. By way
of illustration, we examine overall reaction cycles
as well as the controlling reaction coordinates for vari-
ous reaction mechanisms for various types of surface
chemistries.
Historically, most of the theoretical efforts aimed at
heterogeneous catalysis have focused on small mole-
cule reactivity over different transition metal surfaces.
There has been a wealth of valuable information from
formal chemisorption theory, semiempirical and ab
initio studies on the principles that govern chem-
isorption. Collectively, these systems provide a very
valuable resource and qualitative description of the
relevant principles which control the nature of the sur-
face-adsorbate bond and small molecule reactivity on
clean transition metal surfaces.
In Section A.5.1.2 we focused on the information
and insights derived from formal chemisorption theory,
tight-binding approaches, and current first-principle
results for these systems. The explosive growth of both
methods development as well as the computational
hardware capabilities today enable us to examine sig-
nificantly larger transition metal systems, as well as
more complex adsorbates, and provide quantitatively
reliable values for adsorption energies and activation
barriers. In addition to CPU advances (which enable
larger cluster calculations) recent two- and three-
dimensional density functional theory (DFT) codes are
enabling the first-principles treatment of real extended
transition metal surfaces and slabs and their inter-
actions with adsorbates [7].
Our knowledge about transition metal reactivity and
the current success of fmt-principle calculations now
put us in a favorable position to begin to answer ques-
tions posed at the chemistry of other complex mate-
rials, such as transition metal sulfides and transition
metal oxides. In many respects, the chemistry of these
systems is considerably more challenging due to the
unknown surface structures, the ease of surface re-
construction, and the ill-defined nature of the active
centers. Current first-principle attempts at modeling
these systems are discussed.
We, therefore, structure this chapter around the im-
portant elements that help to define surface reactivity
and subsequently present recent examples from (i)
classical small molecule-transition metal surface re-
activity, (ii) transition metal sulfide reactivity, (iii)
transition metal oxide reactivity, and (iv) well-defined
metal oxide (zeolite) reactivity. This diverse scope in
chemistry provides a more comprehensive treatment of
the theory of surface reactivity, whereby we extend our
knowledge base and cover important elementary cata-

lytic steps such as C-0 dissociation, C-H activation,
C-S bond scission, O-mediated bond activation, metal
atom insertion, and proton transfer pathways.
5.2.4.2 Outline
The first part of this Section describes the electronic
and structural features that determine the rates of dis-
sociation and association reactions on transition metal
surfaces. We elucidate the lowest energy reaction paths
and discuss the utility of two important theoretical
concepts for analyzing transition metal surface chem-
istry. The first is the principle of least metal atom
sharing which helps to describe adsorbate bonding on
transition metal surfaces and provides a qualitative de-
scription of minimum energy reaction paths over sur-
faces. The second is the concept of negative ion for-
mation. In various systems, barriers for dissociation can
be substantially reduced if the dissociated adsorbate-
surface species is allowed to take on a negative charge.
Both surface structure and surface electron delocali-
zation play important roles in controlling surface re-
action rates. We discuss how changes in delocalization
of the metal s, p, and d valence electrons affect re-
activity. A more in-depth analysis is provided in Sec-
tion A.5.1.2, on chemisorption, and elsewhere [8, 91. In
addition, we analyze the adsorbate-surface structure at
various active sites and how it changes along operative
reaction coordinates. The Brernsted-Polanyi relation-
ship is discussed in terms of relating changes in the
activation energy with changes in reaction energy to
predict trends in the reaction rate constants. CO and
NO are used as prototype adsorbates for bond activa-
tion and dissociation. Their dissociation paths are an-
alyzed in terms of transition state reaction rate theory
to determine surface reaction rate constants.
The fundamentals of X-H bond activation is con-
sidered next and analyzed in terms of methane and
ammonia activation. A dissociating molecule that is
coordinatively saturated initially has a repulsive inter-
action with the surface which is only overcome when
the molecule deforms itself. Hydrogen, for example,
requires a significant activation of the H-H bond be-
fore it will coordinatively bind to a transition metal
surface. On surfaces of low reactivity, reaction with
coadsorbed oxygen, either atomic or molecular, can
provide low energy paths to drive the dissociation
which would otherwise not proceed. This is discussed
in terms of ammonia activation by oxygen over copper.
We examine both precursor atomic and molecular
oxygen surface species. Coadsorption of ammonia with
oxygen (either atomic or molecular) demonstrate sig-
nificant lateral attractive surface interactions. In addi-
tion, we comment on the repulsive lateral interactions
for adsorbates bound too close to one another.
5.2 Microkinetics 993
In the follow-up section, we illustrate the utility of
reaction energy diagrams in identifying slow reaction
steps and estimating the surface composition of the re-
acting surface overlayer. In the ammonia oxidation
example, the reaction energy diagram is used to exam-
ine the N2 versus NO production.
Some of the simple ideas discussed for adsorbates on
metals are then applied to sulfide catalysis. The analy-
sis of very small adsorbates on transition metal surfaces
to the treatment of larger adsorbates, such as thio-
phene, dihydrothiophene, butadiene, on model tran-
sition metal sulfides relevant in hydroprocessing sys-
tems is expanded. The use of the reaction energy
diagram for the analysis of thiophene activation by the
Ni3SZ is extended. This example nicely illustrates the
importance of including complete structure relaxation
for adsorbates, as well as clusters in establishing the
controlling intermediates and reaction pathways.
The final section covers recent studies on different
amorphous reducible transition metal oxides and well-
ordered crystalline metal oxides. The current status
of theoretical calculations on transition metal oxide
selective oxidation catalysis is assessed and proton
transfer chemistry in zeolites is examined. The latter
illustrates the concept that the intermediate geometries
in surface reactions often bear little resemblance to
analogous gas-phase intermediates. Reducing the acti-
vation barrier on a surface typically requires a sig-
nificant interaction between the reactant and active
surface site and thus substantial adsorbate-surface re-
construction. In the proton transfer reaction, a charge
separation energy has to be minimized.
Most of the theoretical results referred to in this chap-
ter are based on the use of first-principleelectronic struc-
ture calculations. The details for these calculations are
described in more detail in their original references. We
focus instead upon the chemical concepts of reactivity.
The current status of first-principle computational
quantum chemistry is such that interaction energies can
be predicted within reasonable accuracy, i.e. about
20kJmol-', for systems of catalytic interest [8]. Such
theoretical results are typically accurate for the cluster
chosen as a chemical model. Discrepancies between the
calculations and experiment then represent differences
in the chosen model and the actual experiment.
This includes deviations due to limited-size models
(clusters) of bulk phenomenon, as well as deviations
due to competing phenomena inherent in the actual
experiment. A critical discussion on the use of different
quantum chemical methods and cluster-size effects is
given in Ref. 8. We have chosen to discuss selected ex-
amples of the surface-chemical reactions representative
of main classes of surface chemistry and reactivity in
heterogeneous catalysis.
References see page 1004
994 5 Elementary Steps and Mechanisms
P i / O -
i
C
I 0
bending stretching
,o
c” I
C- I Rh
0 Rh
no yes
FCC(I I I )
E,,,
FCC(100)
E,,,
Figure 1. Dissociation paths of CO (schematic): (a) bonding and
stretching of the CO bond; (b) CO dissociates not along a metal-
metal bond the C and 0 atom generated by dissociation require
~
a surface ensemble; (c) dissociation paths and activation energies
on Rh(ll0) and Rh(ll1) surfaces.
5.2.4.3 Transition Metal Surface Chemistry
A Dissociation of CO and NO
Semiempirical ASED extended Huckel calculations
were used to provide the low-energy CO dissociation
paths over Rh(l1 l), Rh(100), and Rh(ll0) surfaces
as modeled by large 50 atom clusters [lo]. Some of
the more favorable dissociation paths determined are
illustrated in Fig. 1. CO binds end-on with the carbon
atom attached to one or several surface metal atoms.
In order to dissociate, CO bends in such a way that
both the carbon as well as the oxygen develop stabiliz-
ing bonds with the metal surface atoms. On the (1 1 1)
and (100) surfaces, the lowest energy path involves
stretching the C - 0 bond over a metal atom center with
a combined reduction of the 0-C-surface angle. The
result is an intermediate with a C-0 bond nearly par-
allel to the surface. The carbon and oxygen atoms
produced move towards the higher coordination ad-
sorption sites that share only a single surface atom
center. An alternative path was found on the (1 10)
surface. Here the ASED calculations indicate a re-
action path in which the CO molecule crosses over a
metal-metal surface bond. For NO dissociation over
Cu, the results of a series of first-principle density
functional calculations on a Cu(9,4) model of the Cu
(100) surface indicate the same low energy metal-metal
bond crossing reaction path [Ill. A minimum energy
barrier for dissociation requires CO or NO bond
weakening to be made possible by electron transfer
from the surface into the antibonding 271* CO or NO
orbitals. The position of the molecular CO axis over a
surface atom allows for the overlap of one of the two
271* orbitals with surface atomic orbitals of the same
symmetry, i.e. the d,, or dyzmetal orbitals.
If, however, the CO axis is placed perpendicular,
over a metal-metal bond, both 2n’ bonds can become
activated via the interaction with the asymmetric sur-
face d atomic orbitals. The overlap of the 271* orbital
normal to the surface, as well as the activation of the
C - 0 bond, however, are now somewhat less.
The energy barrier for C - 0 activation can be low-
= I80 kJImo1 ered by lowering the work function of the metal.
Clearly, the lower the work function, the easier it is to
form a negative ion surface species. This acts to en-
hance surface reactivity. This observation agrees well
= 140 kJimol with experiment. More open surfaces, which corre-
spond to lower work function surfaces, tend to be more
reactive. Coadsorption of promoters, such as alkali
species or oxides which act to lower the work function,
also help to enhance CO dissociation.
In Section A.5.1.2, the relevance of the d valence
band width and the degree of electron delocalization
to adsorption energetics and surface reactivity is dis-
cussed. We note here that the adsorbate-surface
chemical bond strength tends to increase with an in-
crease in the degree of coordinative unsaturation of the
surface atoms. The greater degree of coordinative sur-
face unsaturation reduces the delocalization of surface
electrons and narrows the d valence band width. Ad-
sorption energies thus increase accordingly.
The surface reaction energy is a function of the ad-
sorbed initial state, the adsorbed dissociated products,
and the transition-state surface complex. This situation
can be simplified by applying the Brmsted-Polanyi
relationship which indicates a linear correspondence
between the activation energy and the overall reaction
energy:
6Eact t AEreact (1)
The reaction energy depends only on the overall dif-
ference in energy between the adsorbed initial state and
the final reaction state, the adsorbed dissociated prod-
uct fragments.
Due to the compensating changes in donating and
backdonating interaction terms, the changes in the
surface atom reactivity affect the adsorption energy for
molecules much less than the changes in the binding
energies for the corresponding adatom products. Hence
on a more reactive surface, the reaction energy for dis-

sociation tends to be more favorable. A comparison of
the reactivity of the nickel surface with the corre-
sponding platinum surface, for instance, indicates that
CO binds more strongly to Pt than Ni, but that the rate
of CO dissociation is faster on Ni. The lower work
function of Ni results in significantly larger C and 0
atom binding energies to Ni than to Pt. This favors a
lower dissociation barrier on Ni than on Pt.
Whereas molecules such as CO and NO usually ad-
sorb perpendicular to the surface, chemisorbed oxygen
or nitrogen molecules chemisorb parallel to the surface.
The increased backdonation of electrons results in a
significant negative charge and population of bond-
weakening antibonding orbitals. The parallel binding
geometry further enhances the increase in backdona-
tion. The molecular bond can now stretch considerably
to accommodate the increase in backdonation and in-
teract more strongly with different metal surface atoms.
The activation energy for the dissociation of different
diatomics from this state is low. Examples include the
dissociation of 0 2 on the (110) surface of Ag [12], and
dissociation of N2 on the (111) surface of Fe [13].
Based on calculations with the N2 dimer, Blomberg
and Siegbahn [14] proposed that N2 would dissociate
by crossing perpendicularly over the metal-metal
bond.
Transition-state reaction rate theory can be used to
predict reaction rate constants from computed vibra-
tional frequencies of ground state and transition state
structures. Computed preexponents for dissociation
from the adsorbed state are typically [l 11
AdkS
pre z 10" s-l
Surface recombination of adatoms to form an adsorbed
molecular intermediate results in a preexponential
value of the order of
1013 s-l
Pre
A typical value for the preexponent of the rate-con-
stant of desorption from the same immobilized state as
used for dissociation is
AdeS
pre 1015s-l
The preexponent value of 1013 for the recombination
reaction indicates that the entropy of the transition
state complex for the recombination path is close to the
entropy for the adsorbed atoms. The situation is there-
fore described as a tight transition state. The consid-
erably elongated molecular intermediate has very lim-
ited mobility due to its strong interaction with the
surface. This is in contrast with the transition state for
desorption where the molecule is freely rotating and
has two translational degrees of freedom. The large
difference in the preexponents of the rate constants
for desorption and dissociation help to explain why
5.2 Microkinetics 995
the rate of dissociation is much lower than that of de-
sorption, even though their activation energies are
close to one another. This is the case, for example, for
the Rh(l11) surface [lo]. It also helps to explain the
low sticking coefficients that are found for dissociative
adsorption. At low surface coverage the apparent rate
constant for dissociation is
kdiss(app;6 << 1) = kzif X &ds (2)
The apparent activation barrier is the intrinsic barrier
minus the energy for adsorption. The apparent barrier
is, therefore, significantly lower due to adsorption. In
some instances the apparent barrier can even become
negative. Despite this lowering of the barrier, the rate
of dissociation is still often a limiting step in the over-
all catalytic reaction cycle. An example of this is the
dissociation of N2 on Fe for the ammonia synthesis
reaction. The low rate in this system is attributed to the
low preexponent factor for the rate constant for dis-
sociation [13].
Reaction paths for dissociation require an ensemble
of surface atoms to carry out the dissociation, so that
the adatoms generated as reaction products have sites
whereby they can strongly adsorb. Dissociation reac-
tions are suppressed when the availability of suitable
surface ensembles is reduced due to site blocking from
coadsorbate species or additives. Alloying the metal
surface with an inert has the same effect and also acts
to reduce available reaction sites and measured dis-
sociation rate constants.
Because of their short lifetimes (femto- to pico-
seconds), transition state complexes are only indirectly
accessible to experiment. Detailed molecular beam and
high resolution spectroscopic techniques can be used
to probe the vibrational, rotational, and translational
velocity distribution of molecules that recombinatively
desorp from a surface, to elucidate structural evidence
on different surface transition-state complexes. Using
infrared emission spectroscopy, Coulston and Haller
[ 151 were able to deduce information on the elementary
reaction path on the structure of the transition state for
CO oxidation on Pd, Pt, and Rh foils. From differences
in the measured vibrational-rotational emission spec-
tra, they were able to conclude that the transition state
is a bent COi- complex on Pd, whereas on Pt and Rh
the complex is nearly linear. The work functions of
each of these metals are surprisingly similar, and is
therefore not a likely explanation these differences. The
density of states at the Fermi level, however, is nearly
three times higher for Pd than for Pt or Rh and was
suggested as the reason for the change in transition-
state surface structure. The higher density of state for
Pd increases the amount of charge transfer to the
References see page I004
996 5 Elementary Steps and Mechanisms
Figure 2. Transition state for dissociation of CH4 molecule on
Nil3 cluster [16].
weakly bound COz adsorbate. In an effort to accom-
modate the higher negative charge, the bound COz
distorts and, therefore, takes on a more bent structure.
The speculated reaction path on each of these metals is
considered to be the same, and one in which the CO
migrates atop an adsorbed surface oxygen.
B CH4 and NH3 Dissociation
The activation energy of a surface dissociation reaction
is minimized by the interaction of the dissociation
products with the surface. The tetrahedral carbon atom
in methane is physically shielded from the surface
atoms by its outer shell of hydrogen atoms. Only by
deforming the molecule (or the outer hydrogen shell)
by stretching one of the C-H bonds is the carbon
allowed to approach the surface with an attractive
interaction. This is clearly seen in the computed tran-
sition state complex of CH4 on a Nickel cluster, shown
in Fig. 2 [16]. While the C-H bond is significantly
stretched, the rest of the molecule is essentially un-
altered and maintains its rotational freedom. The
methyl radical and hydrogen atom products have sig-
nificantly less interaction with the surface metal atoms
than that of the oxygen and carbon adatom frag-
ments from CO dissociation (some 250 kJ g-atom-'vs.
600 kJ g-atom-'), and hence the dissociating molecule
in the transition state maintains considerable mobility.
The computed preexponent for dissociative adsorp-
tion is only 0.1 times the hard sphere collision rate.
the weak adsorption energy of CH4 implies a high
activation energy for dissociation (= 80 kJ mol-I).
Dissociation then occurs by direct activation of the
colliding molecules or, when adsorbed, by the energy
transfer from the impacting molecules [16]. An ex-
tensive study of CH4 and CH, activation in general by
transition metal atoms and surfaces can be found in the
work of Siegbahn and co-workers [17].
Figure 3. Contour plot of the electron density difference for
threefold adsorption of Mg2+, p(Mg2++Ir4)-p( Ir4- p(Mg2').
Dashed lines show a decrease, solid lines an increase of the elec-
tronic density, except for the solid lines next to dashed lines which
depict nodal surfaces [ 181.
The weak initial interaction of CH4 with the metal
surface is due to the strong repulsive interaction be-
tween the occupied metal and methane orbitals. The
unoccupied methane orbitals are very high in energy
and have little orbital overlap with the occupied metal
orbitals. Attractive interaction terms are therefore very
small, and the interaction is predominantly repulsive.
As discussed in Section A.5.1.2, for the interaction of
H2 with small Ir clusters, the polarization of such a
particle by a charged cation will pull charge towards
the cation and deplete the atomic orbitals located in
atoms opposite the Mg2+ ions, as shown in Fig. 3 [18].
This reduces the electron density between H2 and co-
ordinated Ir atoms. The hydrogen molecule can now
more closely approach the cluster and lead to a sub-
stantial increase in the attractive interaction energy.
Protons or cations adsorbed in zeolites, near to small
metal particles can in the same way promote alkane
activation.
Activation of the N-H bonds in ammonia is more
facile than the activation of C-H bonds in CH4, be-
cause of the basic nitrogen atom with its free lone pair
orbital, and the hydrogen atom of the molecule which
can now interact with the surface atoms. Ammonia
binds more strongly to transition metals to the left in
the periodic table primarily due to their substantially
lower work functions and unfilled d valence electron
bands. Under conditions of higher temperature many
of these surfaces will readily dissociate ammonia. Am-
monia binds much more weakly, however, to transition
metal surfaces to the right of the periodic table with
highly filled d bands, such as Cu [19] and Pt [20]. Dis-
sociation over these surfaces is much more difficult.

Surface Reactants A E = 0 kJimol
2 28
Transition Complex AE = 132 kJ/mol
Surface Products A E = +48 kJimol
9 P p0.98
Figure 4. Transition states for the dissociation of NH3 by co-
adsorbed atomic oxygen.
Coadsorbates, such as adsorbed oxygen can, how-
ever, considerably facilitate the dissociation. On both
Cu and Pt, Roberts and co-workers [19, 201 have
shown that with coadsorbed oxygen the ammonia dis-
sociation occurs quite readily even at room temper-
ature to form NH, fragments. As discussed in Section
B.4.6.1, such a reaction with an adsorbed oxygen is
quite common for acidic adsorbate protons, as for the
hydroxylic protons of CH30H or CH3COOH. It is
probable that CH4 oxidation by Pt is also promoted by
an initial reaction of CH4 with adsorbed oxygen. Both
atomic as well as molecular oxygen precursors have
been suggested in promoting NH3 dissociation. We
considered both and examined the dissociation reaction
path of NH3 over Cu(ll1) using nonlocal density
functional calculations [21]. Figure 4 depicts the com-
puted transition state complex for the reaction with
atomic oxygen, whereas Fig. 5 shows the complex with
molecular oxygen. The activation energy of the latter is
nearly half that of NH3 with atomic oxygen. This is
5.2 Microkinetics 997
A E = 0 kJ/mol
I
2.17
2.05k2.17
A E = 67 kJimol
0.98
1.853,
A E = -25 kJimol
0.99
Figure 5. Transition states for the dissociation of NH3 by
molecular oxygen.
primarily due to the enhanced stabilization of the
stretched N-H bond by the interaction with the closer
oxygen atom from the molecularly adsorbed oxygen.
At low coverage, the effective activation energy is the
activation energy of the surface reaction minus the heat
of adsorption (eq 2). Notwithstanding the higher acti-
vation energy for the activation of ammonia by atomi-
cally adsorbed oxygen, the effective activation energy
for this process may be lowest! Molecular and atomic
oxygen have both been cited as a possible precursor in
ammonia oxidation over Cu(ll1) [19].
The overall reactivity of higher alkanes is larger than
that of methane, mainly because of their larger heats of
adsorption. The higher heats of adsorption act to de-
crease the apparent activation energy for C-H bond
scission with respect to that of the gas phase (eq 2). The
barrier is thus more substantially reduced for the lon-
ger alkanes than than for methane.
References see page 1004
998 5 Elementary Steps and Mechanisms
Table 1. Recombination and dissociative adsorption activation
energies for the recombination of adsorbed hydrogen and methyl
species and CH4 on Cox and Nix [16a].
Metal Recombinative Dissociative
desorption adsorption
(kJ mol-‘) (kJ mol-‘
Cluster size 7
co 81 199
Ni 71 86
Cluster size 13
co 95 92
Ni 86 104
The activation of the C-H bonds in alkenes is easier
than in alkanes for two reasons. First, for molecules
larger than ethylene, ally1 formation is possible. This
weakens the C-H bond. Secondly, alkenes bind more
strongly to surfaces as witnessed by both their greater
adsorption energies and their closer approach to the
metal surface. This promotes a stronger interaction
with the transition metal surfaces which acts to lower
the barrier for dissociation. As was the case for alka-
nes, the coadsorption of oxygen can support C-H
bond activation in alkenes. One example is the activa-
tion of ethylene over Ag by the coadsorption of oxygen
PI.
Adsorbed ethylene and other alkenes can react in
consecutive reactions via various different routes on
transition metal surfaces. Highly reactive metal sur-
faces can further dehydrogenate the absorbed ethylene
(alkene) to form vinyl surface species or ethylidine
which can undergo further dehydrogenation to form
acetylene. In addition, the C-C bond scission path
also becomes accessible. In the presence of hydrogen,
adsorbed ethyl fragments can react to form ethane
provided the transition metal surface can dissociate H2.
The mechanism of C-H scission or C-H bond for-
mation in alkenes is similar to that for CH4 activation.
C-H activation is favored by a reaction path such that
the barrier to bond stretching is weakened by electron
back-donation from the surface. The most favorable
reaction path for the dissociating C-H bond is that
which crosses atop over a surface metal atom so that
the asymmetric d atomic orbitals can contribute to the
activation of the C-H bond for cleavage.
The C-C bond scission of alkanes by transition
metals can only occur after one of the C-H bonds has
been cleaved. The activation energy for these reactions
decreases with the hydrogen content of the hydro-
carbon fragment. This is primarily due to the reduction
of the steric repulsion of the C-H bonds by the metal
surface.
Table 1 compares DFT computed activation barriers
for methanation verses carbon chain growth. Reported
are the barriers for the recombination of hydrogen and
C +3H
21 27
CH3 CH+ZH
32
C +3H
CH3
I CH+ZH IT-
Figure 6. The relative energies of CH, species chemisorbed on
(a) a Nil3 and (b) a Col3 cluster [16b].
methyl fragments to form methane over Ni and Co
clusters, and the barrier for the recombination of an
adsorbed carbon atom with a CH2 surface fragment.
The cluster size dependence is discussed in Section
A.5.1.2. The rate of formation of CH4 on Co is signif-
icantly slower on Co than on Ni due to the higher sta-
bility of the reactant CH3 and H fragments on the Co
surface. The lower rate of methanation of surface car-
bon on Co favors the chain growth reaction.
Interestingly, in a set of experimental studies on the
reactivity of CH3 groups generated by low temperature
bombardment of adsorbed CH4, Ceyer and co-workers
[24] found that the decomposition of CH3 on Ni is
an activated processes, but upon further increasing
the temperature, the CH species become the preferred
intermediate. Six CH surface intermediates can sub-
sequently recombine on the transition metal surface to
form benzene when the temperature remains low. To
help elucidate the reactivity of CH, species, Burgh-
graaf [16c] used nonlocal DFT calculations to examine
the relative stability of CH, species on Ni and Co. The
results are depicted in the reaction energy diagram
shown in Fig. 6. This figure clearly shows that CH3 as
well as CH activation are endothermic, but CH2 de-
composition is exothermic. This agrees with the low
“CH2” concentrations observed on surfaces covered
with CH, species.
C Reaction Energy Diagrams
a Ammonia Synthesis
A reaction energy diagram provides an important
overview of the energy changes that accompany each
elementary step. It provides a concise pictorial sum-
mary of the energetics of the complete overall catalytic
 5.2 Microkinetics 999

I
--
N+3H

I I
J l
314

--
I I29
i -
c
s
-400 -

I
389
Y -600 -
NH2 +H P8
-960

I I
1400 ti
-800 -
,1000 -
543 460
1 I
-1200

Figure 8. The reaction energy diagram corresponding to the re-

action skips shown in Fig. 9 [21]:
SR1 NH3.ads + Oads + NH2ads + OHads
Figure 7. A scheme for ammonia synthesis on Fe(ll1) at low SR2 : NH2.ads + Oads + NHads +O h d s
nitrogen coverages; energies in kJ mo1-I [13].
SR3 : NHads + Oads + Nads f OHads

reaction cycle and leads to insights about the control-

ling mechanism and how it may change under various
2 q g s J & + 2 AE = -96 kJ
operating conditions. For the ammonia synthesis re-
action catalyzed by the (1 11) surface of Fe, a complete

-2e
diagram has been established based on the vast number
of fundamental experimental studies aimed at resolv- 3 - 3
ing the controlling steps [13]. This diagram is shown in AE = -44 kJ
Fig. 7.
Decomposition of N2 is found to be exothermic,
whereas the steps involved in the formation of NH, 2
A E = +96 kJ
species by hydrogen addition are endothermic. The first
hydrogen addition step is the most difficult. The acti-
vation energy of N2 dissociation with respect to the gas
phase is only a few kilojoules per mole. Notwithstand- 2
ing this apparently low barrier, N2 dissociation is con- j
A E = +I2 kJ 2 *
sidered to be rate limiting and competes with NH3 de-
sorption. The low rate of nitrogen dissociation here is
attributed to entropic considerations where there is a 2
+ 2
low preexponent for dissociative adsorption, as was A E = -96 kJ
discussed in Section 5.2.4.2.

b Ammonia Oxidation 2 - 1
A DFT-computed overall reaction energy diagram for
AE=-214 kJ
ammonia oxidation by Cu(ll1) is shown in Fig. 8 [21].

3e
In Fig. 9 the corresponding elementary steps are
-+
shown. While only the overall reaction energies are
depicted here, the diagram still provides a very useful
summary of the chemistry and a tool for examining
6
+ A E = + I 17 kJ
changes in this system. The results indicate that the
coadsorption of oxygen and ammonia increases the
Figure 9. A catalytic cycle of elementary reaction steps of the
binding of ammonia by z 40 kJmol-'. This is attrib- 02,NH3 oxidation reaction catalyzed by Cu.
uted to the weakening of the surface complex Cu-Cu
bonds by adsorbed oxygen, and the subsequent
strengthening of the neighboring Cu-N bond. Ammo-
nia prefers to adsorb atop. In the presence of chem- References see page 1004
1000 5 Elementary Steps and Mechanisms
isorbed oxygen, ammonia situates itself at a neighbor-
ing Cu site to enhance the attractive through-metal
surface interaction. Dissociative adsorption of oxygen
is highly exothermic. At low concentrations, the prod-
uct oxygen atoms prefer to sit in threefold coordination
sites which do not share any surface Cu atoms. The
strong interaction of the oxygen atoms with the Cu
surface may ultimately lead to surface reconstruction
to form ordered adsorbed oxygen arrays. This is seen
experimentally [27]. As discussed in subsection B above
coadsorbed oxygen influences the surface dissociation
paths. While the oxygen site is clearly the favored
site for dissociated hydrogen in terms of the overall
thermodynamics, does it play a role in the actual
mechanism or kinetics? Ammonia dissociates by atom-
ically adsorbed oxygen. The endothermicity for am-
monia dissociation is lowered by 128 kJ mol-' (from
176 kJ mol-' for the reaction NH3+ds + NH2.adst Hads
to 48 kJ mol-') when atomic oxygen is introduced into
the reaction mechanism [21].
The first step of ammonia dissociation is the ab-
straction of hydrogen by oxygen, which is endothermic.
Subsequent NH2 and NH dissociation steps, however,
become progressively easier, whereby NH2 is essen-
tially thermoneutral and NH dissociation is exothermic
in the presence of adsorbed atomic oxygen. The favored
adsorption sites for the NH, intermediate species
change as the value of x changes. Those with a higher
value of x (NH3) prefer lower coordination sites,
whereas those with a small value of x prefer higher
coordination sites.
Due to the relatively weak Cu-N bond, associative
nitrogen recombination and desorption is exothermic.
However, the relatively strong Cu-0 interaction leads
to an endothermic recombinative desorption of NO.
The free surface is regenerated by the recombination
of 2(0H),d, to remove H20. Once again, there is a
competition between dissociative adsorption (of NH3)
and desorption processes (HzO). The large entropic
changes associated with recombinative desorption help
to favor water desorption as the temperature is in-
creased. The change in entropy for the ammonia sur-
face dissociation step is quite small and not likely to
enhance the dissociation energetics. This step is there-
fore likely to be rate-limiting. The selectivity for the
formation of N2 versus NO is determined by their rel-
ative rates for recombination as well as their proba-
bility of finding the corresponding ensemble of ad-
sorbed surface atoms. Higher surface coverages of
oxygen preferentially favor the formation of NO. Se-
lectivity is, therefore, expected to be a strong function
of surface coverage. One predicts an increase in nitro-
gen production with increasing conversion. In the Ost-
wald ammonia oxidation process, a Pt-Rh alloy in the
form of a thin gauze is used to carry out this reaction.
These metals have a stronger interactions with NZ than
was found for the N2 interaction with Cu. This drives
N2 formation to become endothermic on these metals.
5.2.4.4 Transition Metal Sulfide Catalyzed
Desulfurization
The theoretical study of desulfurization reactions cata-
lyzed by sulfidic catalysts has been limited by both the
difficulty in defining the active surface and active sites
for hydrodesulfurization (HDS) as well as the lack of
relevant model systems. Recent fundamental experi-
mental evidence is now providing a more complete
picture of the working catalytic surface [28-311. This
catalytically active site corresponds to a distribution of
small Co&like or Ni3Sz-like particles along the edges
of MoS2 basal planes. Studies on highly dispersed sul-
fide particles formed on carbon [31a-c] and those
deposited in the microporous framework of a zeolite
[31d-f] have demonstrated that small metal sulfide
particles themselves are active HDS components.
Previous theoretical studies have, for the most part,
been aimed at understanding the electronic structure of
bulk metal sulfides and how thiophene, a probe HDS
species, interacts [32]. Nearly all of these studies have
been performed at the semiempirical level and have
been aimed at qualitative trends. The recent experi-
mental evidence, which suggests small transition metal
sulfide particles as active components, as well as the
advances in computational quantum chemistry cited
earlier, have motivated first-principles-based analyses
of this chemistry. We report on a series of density
functional studies aimed at elucidating the reaction en-
ergy diagrams for different postulated reaction schemes
for thiophene desulfurization over a model Ni$2 clus-
ter, and also on a plausible mechanism for C-S bond
scission [32, 331.
The overall catalytic cycle for thiophene HDS in-
volves a number of different sequential and competing
elementary steps. To elucidate controlling paths, we
examined a number of different plausible cycles [33].
The two most relevant systems are depicted in Fig. 10.
The first is considered to dominate at very low surface
coverages whereby thiophene adsorbs in a parallel-like
arrangement on the surface. DFT calculations indi-
cated that the q4 mode for the adsorption of thiophene
is more favorable than the completely parallel qs mode.
The cycle involves the adsorption of thiophene which
is highly exothermic, the coadsorption of hydrogen,
hydrogenation to form the stable 2,5-dihydrothiophene
intermediate, scission of the C-S bonds to form buta-
diene which desorbs from the surface, subsequent
hydrogenation, and the associative removal of H2S to
regenerate the starting Ni& cluster. Inspection of the
reaction energy diagram (Fig. 11) indicates that both
the adsorption of thiophene and the hydrogenation of
 5.2 Microkinetics 1001

thiophene are highly exothermic processes which are

likely to occur quite readily. The C-S bond scission
and associative removal of H2S (sulfur removal), how-
ever, are highly endothermic processes and likely to be
rate-limiting steps in the overall cycle. This is consistent
with experimental evidence which has cited both
of these steps as rate-limiting candidates at different
operating conditions [29, 341.
From the single-crystal studies of Gellman and
Somorjai [35] there appears to be a change in the
favored adsorption site and possible surface recontruc-
tion as the sulfur coverage is increased beyond 0 > 0.67
of a monolayer of coverage.
DFT calculations predict a similar change in surface
structure as the sulfur coverage in increased. Threefold
bound sulfur species move to twofold positions in an

ji
effort to lower repulsive interactions.
Figure 10 depicts thiophene as y' adsorbed. This is
+ Butadiene(g) to be compared with y4 adsorption when no hydrogen
is coadsorbed. Coadsorption of hydrogen atoms also
weakens the interaction energy in the thiophene by
almost 6kJmolp1. Wiegand and Friend [36] find a
7 7
similar change in the binding and energetics when
going from a clean Ni surface to one saturated with
sulfur. The binding mode of 2,5-dihydrothiophene is
t also through the I?' mode at higher surface coverages.
Both the C-S bond scission and the sulfur removal
steps are again highly endothermic. Which of these
steps controls the reaction kinetics is highly dependent
upon the operating conditions.
In both situations there is strong competitive ad-
Figure 10. Catalytic hydrodesulfurization cycle via q' adsorbed sorption between H2, HIS, thiophene, and DHT. This
thiophere and dehydrothiophere intermediates. illustrates the complexity of hydrodesulfurization cat-
alysis. Strong inhibition effects due to vacancy sup-
pression of reactant or product molecules also exist due
the inhibition for the dissociation of DHT. Although
0
the metal-sulfur bond ultimately appears to control
both of these situations, the optimum metal-sulfur
-62
bond strength is actually highly dependent upon the
-40
reaction conditions and the operative surface coverage.
- This is consistent with the ideas brought forth by both
z
h
3
Chiannelli and Harris [29] which indicate the optimal
i
5 : -74 metal-sulfur bond strength is an intermediate M-S
-80
E 3:
-
bond, and Nmskov et al. [34] who describe the optimal
metal-sulfur bond as the weakest M-S bond.
6 ;F'
'U

-I

-120
\ -81 /+70
5.2.4.5 Reactivity of Oxidic Surfaces
-160
A Reactivity of Coordinatively Unsaturated Oxide
Surfaces
Electrostatically neutral oxidic surfaces expose both
-200 cation as well as oxygen atom anion sites. Whereas in
Reaction Path the bulk the formal charge on oxygen can often be
Figure 11 The reaction energy diagram corresponding to Fig.
I0 [33]. References see page 1004
1002 5 Elementary Steps and Mechanisms
considered to be equal to -2, this is not the case on a
metal or metal oxide surface. The decreased Madelung
energy on the surface makes charge transfer between
cation and anion on the surface less favorable. This can
give rise to reactive 0- sites, or in the case of the ad-
sorption of 0 2 , 0; anions which are active in non-
selective combustion paths for reacting hydrocarbons
[37]. Dehydrated surfaces have both Lewis acid cat-
ionic sites as well as Lewis base anionic sites. Water will
physisorb at low temperatures and readily dissociate on
nonpolar surfaces. This occurs on reducible as well as
nonreducible oxides. Protons attach at bridging oxygen
sites and behave as Brsnsted acids, whereas the OH-
fragments adsorb to the cation sites and behave as
Br~nstedbases. This is readily deduced from Pauling's
valency rules [9]. Oxides, such as MgO, when heat
treated can activate H2 via a heterolytic dissociation
path at vacancy positions. More reactive oxides, such
as ZnO, can form hydroxy groups and metal hydrides
(ZnH) under mild conditions. This is the analog of the
heterolytic chemistry reported in the previous section
for Ni3S3, where H2 heterolytically splits to form both
the metal sulfhydryl and the metal hydride fragments.
A review of quantum chemical studies applied to the
analysis of selectivity of reducible metal oxides has re-
cently been reported by Witko [37]. We summarize
here the main conclusions which are primarily the re-
sult of a series of earlier theoretical studies by the
Witko and Haber groups. Active selective oxide cata-
lysts, such as Moo3 and V205, contain reducible cati-
ons which have close to octahedral coordination. Five
oxygen atoms are shared with other cations while the
sixth oxygen is end-on coordinated. This oxygen atom
often acts as a spectator, a nonreactive oxygen atom,
because of its strong triple M=O bond character. It
helps to promote the reactivity of the bridging oxygen
sites. Calculations [38] show that the negative charge of
the end-on 0 atom is significantly larger than that for
the more reactive bridging oxygen atoms. In many se-
lective oxidation routes, the bridging oxygen is thought
to be more readily inserted into the reacting organic
molecule. This is then subsequently followed by a facile
rearrangement of coordination polyhedra resulting in
the formation of shear planes [39]. This is the analog of
the surface reconstruction effects observed for many of
the transition metal surfaces when atoms or fragments
adsorb strongly at higher concentrations. This is also
analogous to the cluster reconstruction effects dis-
played by the metal sulfide clusters when sulfur was
either added or removed, as discussed in Section
5.2.4.4.
Activation of the adsorbate C-H bonds associated
with alkanes is proposed to occur by an oxidative
addition of the hydrocarbon to a reducible cation
site. The hydrogen atom becomes bound as a proton
to a bridging oxygen centre. The carbenium ion inter-
mediate formed from the organic molecule which is
initially complexed with the cation converts to a o-
bonded state with another bridging oxygen. This leads
to a surface alkoxy intermediate, quite analogous to
the o-bonded carbenium ion intermediates found in
zeolites (as is discussed in the next section). In the
reducible oxide case, the bridging oxygen can incor-
porate into the hydrocarbon. In this elementary step, a
second hydrogen atom needs to be activated. It can
then recombine with an already present OH group on
the surface to produce H20. The catalytic cycle is
closed by the dissociative adsorption of oxygen. This
does not necessarily occur at the same sites where C-H
activation or selective oxidation reactions occur.
B Proton Transfer in Zeolites
The zeolitic proton is bound to a lattice oxygen atom
that bridges two tetrahedrally-coordinated lattice cati-
ons each of different valency. The first is the Si4+ ion of
the framework, while the second is a trivalent cation
such as A13+ which has substituted and Si4+framework
site. Acidic protons are generated at the cation bridg-
ing surface oxygen sites by the splitting of H20, as
discussed in the previous section. These zeolitic protons
behave as Brsnsted acids. Quantum chemical calcu-
lations indicate a strong covalency between the hydro-
gen and oxygen, with a small dipole moment [40]. The
proton charge is estimated as xO.1e.u. The highly
acidic nature of the proton becomes apparent when it is
disturbed by an adsorbed basic molecule. Due to the
repulsive interaction between the lone-pair nitrogen
atom and the OH electrons, the OH electrons polarize
away with a significant increase in the proton charge.
The zeolitic proton remains attached to the framework
oxygen, as indicated by the short 0 - H distance, and
forms hydrogen-bonding interactions with the acetoni-
trile. The interaction energies of the base-proton pair
follow the protonation values of their counterpart gas-
phase species.
The low dielectric content of a zeolite implies a high
energy cost for charge separation. Reaction paths for
proton activation with reactant molecules are con-
trolled by the need for minimum charge separation of
the lattice basic oxygen and the protonated inter-
mediates [40]. We illustrate this by using methanol
adsorption and activation as a motivating example.
Infrared spectroscopy and DFT quantum chemical
analyses demonstrate that CH30H is hydrogen bonded
to a zeolitic proton in the coordination geometry de-
picted in Fig. 12. The zeolitic proton is coordinated to
the oxygen atom of methanol while the proton asso-
ciate with the OH bond of methanol is coordinated the
neighboring Lewis base oxygen atom that bridges the
A1 and Si. Hydrogen-deuterium exchange between
 5.2 Microkinetics 1003

---T
A

B
t
A

/
1
h

oe
A1 Si C
*
0

Figure 12. Methanol dehydration: (A) groundstate adsorption mode (end-on); (B) transition state for transition A-C; ( C ) intermediate
adsorbed state (side-on); (D) Transition state for dehydration; (E) Product state of methoxy and physisorbed HzO [41].

protonated zeolite and the deuterated methanol occurs TS"

very rapidly, because of the low energy of the sym-
fi\
iI \
metrical transition state corresponding to protonated
methanol, which is only a few kilojoules per mole

\
higher in energy than the ground state structure.
Methanol can also undergo a dehydration path to
form water and the CHT carbenium ion. This reaction, Cluster+
. -.

however, is highly activated. The DFT computed bar-

~

CH,OH ~
+H,O
Methoxide
rier is ~ 2 1 kJmol-'
2 (Fig. 13). The corresponding .......

transition state, structure D in Fig. 12, is very different

from the structure found for the hydrogen-deuterium
exchange reaction. What is similar, is that the zeolitic
proton is directed towards the methanol oxygen. The complex
structure of the activated CH; species is planar and / complex
"side-on
remarkably similar to that of the free CHT species. The "

positive charge that develops on the bound CH: spe-

cies is stabilized by electrostatic interactions between
the carbon atom and a Lewis basic oxygen. In the final 7
... ............
state (structure E), the CH: ion becomes covalently adsorption
com pIex
bound to the bridging oxygen and forms an adsorbed "end-on''
methoxy species. It is important to distinguish the sta-
ble o-bonded ground states of carbenium ions and the Figure 13. The reaction energy diagram corresponding to Fig.
excited carbenium transition states. The protonation of 12
ethylene [42], for example, proceeds in three distinct
steps. In the first, ethylene interacts with the proton to
form a weakly n-bonded intermediate. The ethylene is
then protonated and proceeds through a transition References see page 1004
1004 5 Elementary Steps and Mechanisms
state where the Lewis basic oxygen stabilizes the pos-
itive charge that develops on the nonprotonated carbon
atom. In the third and final step, the formed ethyl
fragment and the lattice oxygen atom combine to form
a stable ethoxylate intermediate.
The identification of carbenium ions with activated
transition states also helps to elucidate the nature of
free carbonium ion intermediates. Carbonium ions are
protonated saturated hydrocarbon intermediates which
can carry out a host of different chemistries [4]. Such
ions have been identified in the gas phase by mass
spectrometry. In the zeolite, however, they are to be
regarded as unstable transition state species [40]. The
heterolytic bond strength of the OH group is approx-
imately 1250 kJ mol-' , whereas the protonation energy
of a hydrocarbon is only a few hundred kJmol-'.
Carbonium ion formation only becomes possible be-
cause of the stabilizing electrostatic interaction between
protonated alkane and negatively charged zeolite
lattice.
5.2.4.6 Conclusions
Clearly, applied quantum chemistry is reaching a stage
that experimentally verifiable predictions can be made.
The main issue is the choice of model cluster systems.
The structure of adsorbed intermediates can be deduced
from a comparison between computed and experi-
mental spectra. A relationship with kinetics becomes
possible, due to the computational accessibility of
transition states. As a result many new insights on the
parameters that determine the reactivity of catalytic
surfaces become available. An important feature of the
activated state is the very specific arrangement of sur-
face site atoms that is needed to provide the reaction
path of lowest energy. Another important feature is the
strong influence of coadsorbates, not only the adsorp-
tion energies, but also on the relative geometries of the
adsorbed molecules, with important consequences for
the reactivity.
The use of reaction energy diagrams enables the
illustration of the kinetically competitive behavior of
surface reactions versus desorption. Due to the large
change in entropy, desorption reactions become fast at
high temperatures, whereas surface reactions tend to
become rate-limiting. At low temperatures desorption
can become rate-limiting.
Theory can be used to compute reaction energy dia-
grams for proposed mechanisms of catalytic reaction
cycles.
Computational techniques are approaching the sta-
tus of experimentalists tools. Extensions of current
capabilities to systems of increasing complexity is
highly desirable.
Acknowledgments
MN wishes to acknowledge Dr. George W. Coulston
and Dr. David A. Dixon from DuPont Central Re-
search and Development for their helpful discussions.
References
1. R. Schlogl, Angew. Chem., Int. Ed. Engl. 1993, 32, 3, 381.
2. G. Ertl, Angew. Chem., Int. Ed. Engl. 1990,29, 1219.
3. (a) D. W. Goodman, Surf Sci. 1994, 299/300, 237; (b) S. H.
Oh, G. B. Fisher, J. E. Carpenter, D. W. Goodman, J. Catal
1986, 100, 360.
4. J. K. Nsrskov, Prog. Surf: Sci. 1991, 38, 1.
5 . (a) J. K. Nerrskov, Rep. Prog. Phys. 1990, 53, 1253; (b) G. A.
Somorjai, M. van Hove, Prog. Surf: Sci. 1991, 30, 201; (c) D.
A. King, Surf Sci. 1994, 299/300, 698; (d) T. Gritsch, D.
Coulman, R. J. Behm, G. Ertl, Phys. Rev. Lett. 1989, 63,
1086.
6. H. Brune, J. Wintterlin, R. J. Behm, G. Ertl, Phys. Rev. Lett.
1992, 68, 624.
7. (a) G. te Velde, E. J. Baerends, Phys. Rev. B 1991, 44, 7888;
(b) G. te Velde, E. J. Baerends, Chem. Phys. 1993, 177, 399;
(c) P. J. Feibelman, Surf: Sci. 1994, 313, L801; (d) P. J. Fei-
belman, Phys. Rev. Lett. 1991, 67, 461; (e) J. C. Boettger, Int.
J. Quant. Chem.: Quant. Chem. Symp. 1993, 27, 147; (f) B.
Hammer, K. W. Jacobsen, J. K. Nsrskov, Phys. Ret.. Lett.
1992, 69, 1971.
8. R. A. van Santen, M. Neurock, Catal. Rev. 1995, 37, 557.
9. R. A. van Santen, Theoretical Heterogeneous Catalysis,
World Scientific, Singapore, 1991.
10. A. deKoster, A. P. J. Jansen, R. A. van Santen, J. J. C.
Geerlings, Far. Disc. Chem. SOC.1989, 87, 263.
11. M. A. van Daelen, Y. S. Li, J. M. Newsam, R. A. van
Santen, Chem. Phys. Lett. 1994,226, 110.
12. (a) M. A. Barteau and R. J. Madix, Surf: Sci. 1980, 97, 101;
(b) D. A. Outka, J. Stohr, W. Jark, P. Stevens, J. Salomons,
R. J. Madix, Phys. Rev. B 1987, 35, 4119; (c) T. H. Upton,
P. Stevens, R. J. Madix, J. Chem. Phys. 1988, 88, 3988; (d)
I. Panas, P. Siegbahn, Chem. Phys. Lett. 1988,153,458.
13. (a) G. Ertl, Solid State and Material Science, CRC Press,
Boca Raton, 1982, 349; (b) G. Ertl, in Catalytic Ammonia
Synthesis, Fundamentals and Practice, (Ed.: J. R. Jendings),
Plenum Press, 1991.
14. M. R. A. Blomberg, P. E. M. Siegbahn, J. Am. Chem. SOC.
1993, 115, 6908.
15. G. W. Coulston, G. Haller, J. Chem. Phys. 1991, 95, 6932.
16. (a) H. Burghgraef, A. P. J. Jansen, R. A. van Santen,
J. Chem. Phys. 1993, 98, 8810; (b) H. Burghgraef, A. P. J.
Jansen, R. A. van Santen, Chem. Phys. 1993, 177, 407; (c)
H. Burghgraef, Ph. D. thesis, University of Eindhoven, 1995.
17. (a) P. E. M. Siegbahn, U. Wahlgren, E. Shustorovich Metal-
Suvface Reaction Energetics: Theory and Appli-cations to
Heterogeneous Catalysis, Chemisorp t ion, and Surface D f-
fusion, VCH, New York, 1991; (b) P. E. M. Siegbahn, M. R.
A. Blomberg, M. Svensson, J. Am. Chem. Soc. 1993, 115,
4191; (c) M. R. A. Blomberg, P. E. M. Siegbahn, J. E.
Backvall, J. Am. Chem. SOC.1987, 109, 4450; (d) M. R. A.
Blomberg, P. E. M. Siegbahn, M. Svensson, J. Phys. Chem.
1994, 98, 2063; (e) M. R. A. Blomberg, P. E. M. Siegbahn,
M. Svensson, J. Phys. Chem. 1992, 96, 5783; (f) P. E. M.
Siegbahn, Theor. Chim. Acta 1994, 87, 277; (g) M. R. A.
Blomberg, P. E. M. Siegbahn, U. Nagashima, J. Wenner-
berg, J. Am. Chem. SOC.1991, 113, 424; (h) J. E. Backvall,

E. E. Bjorkman, L. Pettersson, P. Siegbahn, A. Strich, J. Am.
Chem. SOC.1985,107, 7408; (i) P. E. M. Siegbahn, I. Panas,
Surf: Sci. 1990,240, 37.
18. E. Sanchez Marcos, A. P. J. Jansen, R. A. van Santen, Chem.
Phys. Lett. 1994, 167, 399.
19. (a) M. W. Roberts, J. Mol. Catal. 1993, 74, 11; (b) B. Afsin,
P. R . Davies, A. Pashusky, M. W. Roberts, D. Vincent, Surf:
Sci. 1993, 284, 109; (c) A. Boronin, A. Pashusky, M. W.
Roberts, Cat. Lett. 1992, 16, 345; (d) B. Afsin, P. R . Davies,
A. Pahuski, M. W. Roberts, Surf: Sci. Lett. 1991, 259, L724;
(e) B. Afsin, P. R . Davies, A. Pashuski, M. W. Roberts,
D. Vincent, Surf: Sci. 1994, 284, 109.
20. T. S. Amorelli, A. F. Carley, M. K. Rajumon, M. W. Rob-
erts, P. B. Wells, Surf: Sci. 1994, 315, L990.
21. M. Neurock, R. A. van Santen, W. Biemolt, A. P. J. Jansen,
J. Am. Chem. SOC.1994, 116, 6860.
22. R. A. van Santen, J. P. E. Kuipers, Ado. Catal. 1987, 35,
265.
23. R. A. van Santen, A. de Koster, T. Koerts, Catal. Lett. 1990,
7, 1.
24. S. T. Ceyer, J. D. Beckerle, M. B. Lee, S. L. Tang, Q. Y.
Yang, M. A. Hines, J. Vac. Sci. Techn. 1987, A5, 501; M. B.
Lee, Q. Y. Yang, S . L. Tang, S. T. Ceyer, J. Chem. Phys.
1986,85, 1693.
25. M. C. Wu, D. W. Goodman, J. Am. Chem. SOC.1994, 116,
1364; M. C. Wu, D. W. Goodman, G. W. Zajac, Catal. Lett.
1994,24, 23; M. C. Wu, P. Lenz-Slomm, P. W. Goodman, J.
Vac. Sci. Techn. 1994, A12, 2205.
26. G. Ertl, S. B. Lee, M. Weiss, Surf: Sci. 1982, 114, 515.
27. (a) D. J. Coulman, J. Wintterlin, R. J. Behm, G. Ertl, Phys.
Rev. Lett. 1990, 64, 5432; (b) F. Jensen, F. Besenbacher, E.
Laegaards, I. Steensgaard, Phys. Rev. B 1990, 41, 10233; (c)
F. Jensen, F. Besenbacher, E. Laegaards, I. Stensgaard, Surf:
Sci. Lett. 1991, 259, L774; (d) F. Jensen, F. Besenbacher, I.
Stensgaard 1992, 2691270, 400.
28. H. Topsoe, B. S . Clausen, Catal. Rev.-Sci. Eng. 1984,26,395-
420; N. Topose, H. Topsoe, J. Catal. 1993, 139, 641-651.
29. (a) R. R. Chianelli, Catal. Rev.-Sci. Eng. 1984, 26, 361-393;
(b) S. Harris, R. R. Chianelli, Chem. Phys. Lett. 1983, 101,
603-605; (c) S. Harris, R. R. Chianelli, J. Catul. 1984, 86,
400-412; (d) S. Harris, R. R. Chianelli, J. Catul. 1986, 98,
17-31.
30. (a) S. P. A. Louwers, R. Prins, J. Catal. 1992, 133, 94.
31. (a) V. H. J. de Beer, J. C. Duchet, R. Prins, J. Catal. 1981,
72, 369; (b) J. C. Duchet, E. M. van Oers, V. H. J. de Beer,
R. Prins, J. Catal. 1983, 80, 386; (c) J. P. R. Vissers, J. P. R.,
C. K. Groot, E. M. van Oers, V. H. J. de Beer, R. Prins, Bull.
SOC.Chim. Belg. 1984, 93, 813; (d) Vissers, J. P. R., V. H. J.
de Beer, and R. Prins, J. Chem. SOC.,Faraday Trans. I 1987,
83, 2145; (e) W. J. J. Welters, G. Vorbeck, Zandbergen, J. W.
de Haan, V. H. J. de Beer, R . A. van Santen, J. Cutal. 1994,
150, 155; (f) T. I. Koranyi, L. J. M. van de Ven, W. J. J.
Welters, J. W. de Haan, V. H. J. de Beer, R. A. van Santen,
Cat. Lett. 1993, 17, 105-116.
32. (a) A. B. Anderson, J. J. Maloney, J. Yu, J. Catal. 1988, 112,
392-400; (b) A. B. Anderson, J. Yu, J. Catal. 1989,119, 135-
145; (c) R. P. Diez, A. H. Jubert, J. Mol. Catal. 1992, 73,65-
76; (d) M. C. Zonnevylle, R. Hoffmann, S. Harris, SurJ
Sci. 1988, 199, 320; (e) F. Rouette, N. Valencia, R. Sanchez-
Delgado, J. Am. Chem. SOC.1989, 111, 40-46; (f) J. Ro-
driguez, Surf: Sci. 1992,278, 326-338.
33. M. Neurock, R. A. van Santen, J. Am. Chem. SOC.1994,116,
4427.
34. J. K. Nsrskov, B. S . Clausen, H. Topsoe, Catal. Lett. 1992,
13, 1-8.
35. A. J. Gellman, M. E. Bussell, G. A. Somorjai, J. Cutul. 1987,
107, 103.
5.2 Microkinetics 1005
36. B. C. Wiegand, C. M. Friend, Chem. Rev. 1992,92,491-504.
37. M. Witko, J. Mol. Catal. 1991, 70, 277.
38. J. N. Allison, W. A. Goddard ACS Symp. Ser. 1985,279,23.
39. A. Bielanski, J. Haber, Catal. Rev.-Sci. Eng. 1992, 19, 1 .
40. R. A. van Santen, G. J. Kramer, Chem. Rev., in press.
41. S. Blazkowski, R. A. van Santen, J. Phys. Chem., in press.
42. V. B. Kazansky, I. W. Senchenya, J. Catal. 1989, 119, 108.
5.2.5 Isotopic Labeling and Kinetic Isotope
Effects
K. TAMARU
AND S . NAITO
5.2.5.1 Introduction
Isotopes are defined as elements having the same
atomic number but different atomic weight, and are
divided into two categories: radioactive and stable iso-
topes. Since their chemical properties are practically
identical, it is possible to label particular molecules by
isotopes and follow them through a sequence of chem-
ical or physical changes. This method is known as the
isotope tracer technique. It is applied in many research
fields, such as chemistry, physics, biology, and medical
science. In early research, a great variety of radioactive
isotopes were applied in tracer experiments because
of the facility of tracing them with a Geiger counter.
Recent development of high-resolution mass spectro-
meters has enabled the application of stable isotopes
in tracer experiments, which are much more easy to
handle than radioactive isotopes.
In the study of chemical reaction kinetics, isotope-
labeled reactants are frequently employed to follow a
reaction pathway and to determine the reaction mech-
anism. By applying this technique, it is possible to
identify the position of bond scission in chemical re-
actions. Polanyi and Szabo [l] have employed ‘*O-
labeled water to determine the position of bond scis-
sion in the hydrolysis of esters as follows:
RCOO-R’ + HO*-H + RCOOH HO*R’+
RCO-OR’ + H-O*H -+ RCOO*H + HOR’
Only ‘*O-labeled carboxylic acids were obtained and
no ‘*Owas detected in alcohols, clearly indicating that
the reaction takes place through the second reaction
path.
It is known that the isotopic mass of an atom in the
reactant molecule has an influence on the reaction rate
as well as on the reaction equilibrium. The change in
References see page 1012
1006 5 Elementary Steps and Mechanisms
reaction rate with the substitution of an isotope is
called the kinetic isotope effect and is distinguished
from changes of equilibrium constants, called equilib-
rium isotope effects. It is possible to design isotope-
labeling experiments to test a postulated mechanism in
which a particular bond is broken in the transition
state. If such bond breaking is involved in the rate de-
termining step, the ratio of the rate constants between
light and heavy isotopes would become large.
In the first part of this article, several examples of
isotope-labeling experiments will be described, which
have provided some important conclusions in under-
standing the mechanisms of heterogeneous catalysis.
In the second part, the theoretical background of the
kinetic isotope effect is discussed and some examples of
relevant research are described, which successfully elu-
cidated the postulated reaction mechanisms. A more
thorough review of this field is also available [2].
5.2.5.2 Isotope Labeling in Heterogeneous Catalytic
Reactions
Modern techniques for surface analysis, such as elec-
tron spectroscopies, have enabled investigation of the
structure and electronic state of surfaces under reaction
conditions. Adsorbed species during the reaction can
also be observed by various surface techniques such
as high resolution electron energy loss spectroscopy
(HREELS) and infrared reflection absorption spectro-
scopy (IR-RAS). However, the adsorbed species which
exist on the surface during the reaction are not always
the reaction intermediates. Sometimes, stable adsorbed
species present on the surface during the reaction are
byproducts, which have no relationship with the re-
action pathway. The isotope tracer technique is a use-
ful method with which to investigate whether the
observed surface species is the reaction intermediate
or not. One of the adsorbed species is labeled by an
isotope atom (D, 13C, "N, l8O, etc.) and its rate of
disappearance is followed by some kind of surface
spectroscopy. At the same time, its rate of appearance
in the product molecule is followed by mass spectro-
scopy. When both rates are identical, it can be con-
cluded that the observed adsorbed species is the reac-
tion intermediate.
A Hydrogenation of Carbon Monoxide
Because of the need to develop effective catalysts for
the conversion of synthesis gas (CO-H2) to liquid fuels
(Section B.3.6), a great many studies have been carried
out on the mechanism of the Fischer-Tropsch reaction.
During the hydrogenation of CO over Group VIII
metal catalysts, infrared spectroscopy demonstrates
that most of the surface is covered by adsorbed CO.
To investigate whether or not the adsorbed CO goes
T i m e (min.)
+
Figure 1. "CO H2 reaction on I3C-preadsrobed Ni film at
523K: 0 I3CH4; 0 I2CH4; 0 I2CO2 (from Ref. 2).
directly to methane, Araki and Ponec [3] employed
carbon-13 isotopes as a tracer. First of all, 13C0 was
admitted at 573K to a clean Ni film and the dis-
proportionation
2 I 3 c o -+ 13c(a) + l3 C O ~
took place to form surface carbon-13. After evacuation
of the system at 573 K to remove adsorbed 13C0 from
the surface, "CO and H2 were introduced onto the
surface at 523K. As shown in Fig. 1, the first prod-
uct appearing from this mixture was 13CH4, and the
formation of "CH4 and l2CO2 was accompanied by
an induction period. These facts demonstrate clearly
that surface carbon is used for methanation rather
than CO from the gas phase, and that surface carbon
does not recombine easily with 0 into CO and COz
molecules.
To clarify the role of adsorbed CO for higher hy-
drocarbon formation, Biloen et al. [4] carried out sim-
ilar experiments over silica-supported Ni, Co, and Ru
catalysts. Surface carbon-13 was deposited on the sur-
face of these catalysts via the disproportionation re-
action of I3CO. Subsequent exposure of these catalysts
to I2CO and H2 led to the abundant formation of
13CH4 and of hydrocarbons containing several 13C
atoms within one molecule. These results indicate that
oxygen-free species CH, (x = 0 to 3) are possible in-
termediates in methanation and that they are capable
of being incorporated into growing hydrocarbon
chains.
The detailed mechanism of this surface carbon in-
corporation process over Ru/SiO2 was investigated by
applying the isotope tracer technique as well as infra-
red spectroscopy during the reaction [5, 61. During the
CO-H2 reaction, the accumulation of long straight-
chain hydrocarbons on the Ru surface was demon-

0
1
60
Time
I20
(min.)
I80 ' 1.
Figure 2. I3C content in each hydrocarbon product having dif-
ferent carbon numbers during l2 C 0 + H2 reaction at 423 K after
the accumulation of 13Csurface hydrocarbon x C I ; 0 C2-C3; 0
c 4 - C ~ ;A C S- C ~ ;0 c4Cc (from Ref. 4).
strated by infrared spectroscopy. In order to investigate
the role of these accumulated hydrocarbons in the
overall reaction, carbon-13 labeling experiments were
carried out. After the accumulation of I3C surface hy-
drocarbon species from I3CO-H2 reaction at 423 K,
I2CO was introduced and the molecularly adsorbed
CO was replaced by l 2 C 0 at room temperature. Then
a H2-I2CO reaction was carried out at 423 K and the
I3C content in the hydrocarbon products was analyzed
with time. As shown in Fig. 2, 13C content in the re-
action products was reduced with time, but could be
extrapolated to 100% at the initial stage of the reaction.
Moreover, the I3C content in each of the hydrocarbon
products having different carbon numbers did not
exhibit any difference, suggesting that all the hydro-
carbon products are formed from common building
blocks and no l2C0 insertion is involved in the chain
growth. The role of the accumulated hydrocarbon spe-
cies in the overall reaction was shown to be a reser-
voir of carbon atoms which may be supplied as CI
species through the scission of their carbon-carbon
bonds.
Another set of key products in CO+H2 reactions
are oxygenated compounds such as methanol and
ethanol. The reaction intermediates of these reactions
were also investigated applying the isotope-labeling
technique. Takeuchi and Katzer [7, 81 have studied the
mechanism of methanol and ethanol formation from
CO + H2 over Rh/Ti02 catalysts. When the hydro-
genation of an equimolar amount of 13C160 and
12C1s0was carried out, the isotopic distribution in the
produced methanol was 13CH3160Hand 12CH3NOH,
indicating that methanol synthesis occurs by a non-
dissociative mechanism. On the other hand, the iso-
topic distribution in the produced ethanol was com-
pletely different from that of methanol and fits the
isotopic composition predicted for a fully dissociative
model. In this model all CO molecules are assumed
5.2 Microkinetics 1007
Figure 3. Carbon isotope distribution in products during
"CO + H2 reaction at 393 K over a Rh/SiOz catalyst after
I3CO+ H2 reaction at 392 K (from Ref. 8).
to completely dissociate to adsorbed C and 0 atoms
which then recombine statistically to produce ethanol.
To explain these results, they proposed the mechanism
of CO insertion into an adsorbed carbene followed by
isotopic scrambling in the adsorbed intermediate.
The mechanism of acetaldehyde and ethanol for-
mation from CO + H2 below atmospheric pressure has
been investigated by combining infrared spectroscopic
measurements and 13C0 and C'*O isotopic tracer
studies with reaction kinetics [9]. The rate of acetalde-
hyde and ethanol formation are markedly dependent
on the nature of the metal precursors employed. The
addition of sodium cations depresses the total catalytic
activity, while the selectivity for acetaldehyde increases
considerably. In order to study the mechanistic rele-
vance between acetaldehyde and hydrocarbon forma-
tion, isotopic tracer experiments were carried out as
follows:
(i) the 13CO-H2 reaction was carried out at 393K
over freshly reduced Ru/SiO2, and I3C surface
species (hydrocarbon and acetate species) were
accumulated on the catalyst;
(ii) the catalyst was cooled to room temperature and
a sufficient amount of I2CO was introduced to re-
place the molecularly adsorbed 13CO;
(iii) a 12CO-H2 mixture was introduced and the cata-
lyst was heated to 393 K as rapidly as possible;
(iv) the 13C distribution in the products was analyzed
with time.
As shown in Fig. 3, the 13C distribution in the methyl
group of the acetaldehyde exhibited almost the same
behavior as for the hydrocarbons and was extrapolated
References see page 1012
1008 5 Elementary Steps and Mechanisms
Time (min.)
Figure 4. Gas phase composition for the reaction of propene
with ' * 0 2 over bismuth molybdate at 698 K (from Ref. 9).
to the isotopic purity (90%) of the 13C0 used at the
initial stage of the reaction. On the other hand, the 13C
distribution in the formyl group of acetaldehyde dif-
fered markedly from those of hydrocarbons, and de-
creased rapidly to the natural abundance of 13C(1.1%).
This indicates that during the steady state CO+H2
reaction, there exist common C1 intermediates through
which hydrocarbons and the methyl group of ace-
taldehyde are formed, and acetaldehyde is formed via
CO insertion into these C1 intermediates.
B Oxidation of Propene
The catalytic oxidation of hydrocarbons has been in-
vestigated extensively because of its practical impor-
tance in the chemical industry. In particular. oxidation
(Section B.4.6.2) and ammoxidation (Section B.4.6.6)
of propene to acrolein and acrylonitrile, respectively, is
the most thoroughly studied catalytic system from the
viewpoint of processes catalytic compositions, kinetics,
and mechanisms over various oxide catalysts such as
Cu20 and bismuth molybdate (Fig. 4).
In the early stages of these studies, radioactive 14C
propene was employed as a tracer. Isaen and co-work-
ers [lo, 111 showed that carbon dioxide was formed
from both propene and acrolein, and that an organic
film was formed on the catalyst during reaction, which
appeared to be constructed from the outermost meth-
ylene carbon in propene. Radioactive tracers are
sometimes powerful for the investigation of surface
layers, which are not so easy to detect by certain sur-
face spectroscopic techniques [ 121. Isotopic-labeling
experiments using I3C, deuterium, and l S 0 , have re-
vealed the reaction pathway, the structure of the re-
action intermediates, the rate determining step, and the
source of oxygen in the acrolein product. Voge et al.
[13] employed propene, 13C-labelled in the methyl
group, in the oxidation over cuprous oxide and exam-
ined the isotopes in the produced acrolein. Half of the
13C in the acrolein product was found in the carbonyl
group, indicating the initial removal of a hydrogen
atom from the methyl group to form an ally1 inter-
mediate which subsequently reacts at either end with
equal probability. Confirmation and extension of this
mechanism was obtained by Adams and Jennings [ 141
employing deuterium-labeled propenes. The same iso-
topic ratios were obtained in the deuterated acrolein
products from 1-propene-ld and l-propene-3d, again
indicating the symmetric allylic intermediates.
Keulks [15] used 1 8 0 2 in an attempt to determine
the mode of oxygen atom incorporation into oxygen-
containing products formed in the catalytic oxidation
of propene on bismuth molybdate catalysts. Before
studying the oxidation of propene with " 0 2 , the iso-
topic exchange process of the oxygen atoms of the bis-
muth molybdate with gaseous oxygen (lSO2) was
checked. No exchange in the gas phase composition
was noted at the temperature range of propene oxi-
dation. On the contrary, when propene oxidation was
carried out with 1 8 0 2 , the main products were
C3H4160 and Cl6O2, and "0-containing products
were only 2-2.5% as shown in the gas phase. These
results strongly suggest that not only the oxide ions
in the top layer but also oxide ions in the subsurface
layers may participate in the reaction. Moreover, the
diffusion of oxide ions between the surface and the bulk
must be rapid, and gaseous oxygen, once adsorbed,
must be able to undergo this diffusion into the bulk.
Wragg et al. [16] have confirmed this conclusion by
using "0-enriched bismuth molybdate catalysts.
C Propene-Deuterium Addition and Exchange
Reaction
The catalytic exchange of hydrocarbons with deute-
rium is one of the oldest and most extensively studied
reactions in the isotopic tracer experiments. The inves-
tigation of these processes provides valuable informa-
tion on the adsorbed state of hydrocarbons and the
mechanism of C-H and C-C bond activation on the
catalytic surfaces, which have been reviewed by Kem-
ball [17].
Since microwave spectroscopy has extremely high
resolution and high sensitivity, it can be applied to
analyze the composition of the mixture of isotopic
isomers, which are formed during isotope-labeling
experiments. For example, in the case of the propene-
deuterium exchange reaction, four isomers of propene-
dl and seven isomers of propene-d2 can be analyzed
quantitatively at the same time by this technique [ 181.
Fig. 5 summarizes the possible reaction intermediates
of the propene-deuterium exchange reaction leading to
the isotopic distribution of propene-dl isomers as in-
dicated. Sometimes it is difficult to distinguish the re-
action intermediates only from the isotopic distribution
patterns, especially when more than two intermediates
are operating at the same time. Accordingly, separate

( i ) A s s o c i a t i v e mechanism
1-
/ reaction rates were more than two orders of magni-
H,C - CH +
I tudes larger than that of Cu and the activation energies
*I I , r
S, S,
s, s,
( i i ) D i s s o c i a t i v e mechanism
H,C= CH
t t
CHD-CH-CH, CH2-CD-CH3 CH,-CH-CH,D
( 1 -4) (Z-d,) (3-4)
( i i i ) I n t r a m o l e c u l a r hydrogen s h i f t
4
CHD-CH-CH,( 1 -d,]
I I I I
1 1
CH,D-CH=CH,(3-d,)
Figure 5. Reaction pathway and reaction intermediates of pro-
pene-deuterium addition and exchange reaction.
kinetic investigations of C3H6-D2 and C3H6-C3D6
reactions may be helpful to distinguish whether the
exchange proceeds through associative or dissociative
mechanisms [191. Although some difficulties exist, this
technique is powerful in studying the structural change
of active sites, such as surface composition of alloy
catalysts or particle size effects of supported oxide
catalysts.
In the case of alloy catalysts consisting of active
Group VIII metals and inactive Group IB metals,
alloying has been considered to reduce the number of
atoms in active metal ensemble that are necessary for
a certain type of adsorption or product formation
(Section A.5.3.3). On the other hand, changes in the
electronic structure (ligand effect) due to alloying are
not very pronounced in these systems. The propene-
deuterium exchange reaction mentioned above was
applied with Ni, Pd, Pt, and their alloys with Cu, to
investigate the role of copper in the modification of
active sites and reaction mechanisms of Group VIII
metals [20-221.
Over Group VIII metals deuterium addition and ex-
change proceeded via common reaction intermediates,
namely 1-propyl and 2-propyl adsorbed species. Their
relative activities as well as activation energies were
considerably different on each metal catalyst, as sum-
5.2 Microkinetics 1009
marized in Table 1. When these metals were alloyed
with copper, the distribution patterns became similar to
each other and resembled that over Cu, whereas the
were still close to those of the individual metals. From
CHD-CH-CH,
(C-&i-l-d,) these results, the following model for active sites was
CH, -CH-CH,
(3-4)
D proposed to elucidate the role of the copper component
s, in alloy catalysts. The surface composition (70-80% Cu
in all the alloys employed) indicates a surface where the
Group VIII metals are isolated atomically from each
other and surrounded by copper atoms. In this sit-
uation, copper metal clusters or mixed clusters of Cu-
Group VIII metals become the active sites, which
implies the important role of electronic interaction
between surface copper and Group VIII metal in alloy
t I catalysts.
CH,-CH-CH,D The activity and selectivity of the catalytic reaction
(3-d,l
over supported metal catalysts greatly depend on the
particle size of the metal as well as on the nature of the
CHz-CD-CH,(2-d,)
t support materials. Morphological changes of the metal
with particle sizes and its electronic interaction with
the support may be the main factors governing these
phenomena,
Such effects for supported metal oxide catalysts have
CH,D-CH-CH, ( 3 -d,l been investigated in the cases of C#L-D2 and C3H6-
C3D6 reactions over the catalysts with various particle
sizes of ZnO, TiO2, and ZrOl dispersed on Si02 [23-
251.
Over unsupported metal oxide catalysts, deuterium
addition and exchange took place through different re-
action pathways. The addition process to form propane
proceed via the associative mechanism of propyl inter-
mediates, whose reverse reaction to form propene-dl
was prohibited. Independently, hydrogen exchange of
propene proceeded through a 71-ally1 intermediate over
ZnO and through carbenium ion intermediates over
Ti02 and ZrO2. By dispersing the oxides on silica, the
rate of C3H6-D2 reaction increased considerably with
the decrease of activation energy. Accompanied with
the deuterium addition process to form propane, the
hydrogen exchange process began to take place
through common 1-propyl and 2-propyl intermediates.
Lower loading catalysts of ZnO and Ti02 exhibited
much higher activities predominantly through 1-propyl
intermediates with smaller activation energies. From
X-ray photoelectron (XPS) and X-ray absorption fine
structure (EXAFS) spectroscopies, it was suggested
that small oxide particles were trapped between silica
support particles and particular active sites were pro-
duced on them. In the case of ZrOz/SiOz, XPS analysis
suggests the formation of a Zr02 monolayer over silica,
which exhibits completely different catalytic behavior
as compared to larger unsupported ZrO2 oxide particles.
References see page 1012
1010 5 Elementary Steps and Mechanisms

Table 1. Isotope distribution in monodeuteriopropene formed during c3H6-D~ reaction over various catalysts (P(D2) = 4 kPa,
P(C3H6) = 2 kPa).

Catalyst Reaction ECX, Ehyd. Mean Da cis-1-dl tr~ns-1

-dl 2-dl 3-dl
temp. (K) (kJmoI-‘) (kJmol-’) content (%) (%I (“4 (“w
(4)
Ni 200 41 39 0.03 2 3 95 0
0.24 3 3 94 0
Pd 210 34 33 0.03 35 33 16 16
0.10 28 28 14 30
0.19 23 22 14 41
Pt 235 45 43 0.01 9 57 28 6
0.05 16 48 28 8
0.10 20 42 28 10
Ni-Cu 180 35 33 0.03 11 12 77 0
0.14 11 11 78 0
Pd-Cu 225 36 35 0.02 20 17 58 5
0.20 20 18 57 5
Pt-Cu 248 44 43 0.01 13 9 68 10
0.09 12 10 67 11
cu 263 - - 0.01 14 14 72 0
0.05 13 14 73 0
6
aQ = 1/600 C i x dl (mean deuterium content in propene molecule).
i= 1

5.2.5.3 Kinetic Isotope Effect

A theoretical evaluation of the kinetic isotope effect

requires precise knowledge of the potential energy sur-
face of the reaction, which is quite difficult to obtain in
ordinary chemical reactions. Accordingly, the ratios of
reaction rates of reactions involving different isotopes
is used for the evaluation of the kinetic isotope effect.
Bigeleisen and Wolfsberg [26] developed the following
theoretical derivations, utilizing the activated complex
formulation.
Consider two sets of reactions with different isotope-
labeled reactants A, B, . . . , through activated com-
plexes X1 and X2 to form products L, M, . . .
A2+B2+ . . . +X: +L2+M2+...
(2)
where M , I,, Iy,I, and vi represent the molecular
weight of reactant A, the moment of inertia of the ro-
tation, and vibrational frequencies. The superscript T
represents the activated complex. If we assume that a
particular bond j , which contains an isotope, is broken
by the reaction, the vibrational partition functions of
the A molecule and activated complex will cancel each
other out except for the vibrational mode vj. Con-
sequently, the activation energy may be expressed from
the zero point energy of this vibrational mode as

The ratio of the rate constants kl/k2 can be expressed

as in eq 1, using partition functions of the reactants In the case of molecules which contain C-H and C-D
QA,, Q B ~.,. . and of the activated complexes Q,’ and bonds, the zero point energy of a C-H vibration is
Q,’ 3 about 5 kJ mol-’ larger than that of a C-D vibration.
Since there is no difference in the electronic state of
- - xi Q: Q A ~Q B ~. . .
ki -
(1) isotopic isomers, the same potential energy curves can
k2 x2.Q: QA, QB, . .. be applied to them. Such potential energy curves in the
where ic1 and x2 represent the transmission coefficients. initial and activated states of the reaction are drawn in
When only the A molecule is substituted with isotope the same figure. Figure 6(a) shows the case when a
atoms, partition functions for the rest of the reactants C-H or C-D bond is broken completely in the acti-
can be canceled out, and eq 2 is derived by separating vated state, where the difference of the zero point en-
the partition function of A into translational, rota- ergies between C-H and C-D is zero. As a result, the
tional, and vibrational components activation energy as well as the reaction rate for the

Eg(bond broken)
EH ' ED EH
Figure 6 . Potential energy diagram of vibrational modes in the reaction of molecules having C-H or C-D bonds: (a) complete breaking
of C-H and C-D bonds in the activated state; (b) no change of the C-H and C-D bonds in the activated state; (c) strengthening of
C-H and C-D bonds in the activated state. & ( E D ) = activation energy of the reaction involving C-H(C-D) bond. Eo(C-H),
Eo(C-D) =zero point energy level of C-H and C-D bonds
molecule with C-H bond becomes larger than those
for the molecule with C-D bond. On the other hand,
Fig. 6(b) shows the case that C-H or C-D bonds do
not change when the activated complex is formed. The
difference of the zero point energies between C-H and
C-D vibrations in the activated state is the same as
that in the initial state (5.0 kJ mol-I). Accordingly,
there is no difference in the activation energies and the
reaction rate between these two reactions. In some cases,
C-H and C-D bonds in activated complexes are stron-
ger than those in the initial states, and the difference of
the zero point energies becomes larger than those of the
initial state. In such cases, the activation energy as well
as the reaction rate for the molecule with a C-D bond
is larger than that with a C-H bond, as shown in Fig.
6(c). This is called an inverse isotope effect.
A Ammonia Synthesis
Ammonia synthesis (Section B.2.4) is one of the oldest
and most important catalytic reactions ever inves-
tigated. A number of substances show considerable ac-
tivity as ammonia synthesis catalysts. Fe, Ru, Os, and
Re and nitrides of Mo, W, and U are the best known.
Among them, iron, doubly promoted by A1203 and
K20, is the most important catalyst in industrial use.
The role of these promoters has been investigated
extensively be Strogin et al. [27] using single-crystal
model catalysts.
It is generally accepted that ammonia synthesis from
HZ and N2 proceeds through dissociative adsorption of
hydrogen and nitrogen followed by the subsequent
stepwise hydrogenation of N(a) to NH(a), NHZ(a), and
5.2 Microkinetics 101 1
ED EH ' ED
NH3(a). The rate-determining step of these processes
depends on the kind of catalyst, partial pressures of
H2 and N2, and the reaction temperatures. Isotope-
labeling experiments with I5N and D were carried out
in various catalytic systems to determine the rate
determining step and reaction intermediates.
Ozaki et al. [28] compared the ammonia formation
rates in N2-H2 and N2-D2 reactions and found an in-
verse isotope effect over a doubly promoted catalyst,
where deuterium reacted three or four times faster than
hydrogen. This large inverse isotope effect has been
explained by the difference in adsorption of ammonia
and deuteroammonia to form the imino radical as an
adsorbed species which retards the reaction rate. Aika
and Ozaki [29] have shown a similar inverse isotope
effect in ammonia synthesis over an unpromoted iron
catalyst. This effect was also explained by the difference
in the equilibrium concentration of adsorbed species
that retards the rate of the nitrogen dissociation, i.e. the
rate determining step. The main mechanistic difference
between promoted and unpromoted catalysts is the
most abundant adsorbed species during the reaction.
Over promoted catalysts, NH(a) is the main adsorbed
species, whereas this is N(a) over unpromoted catalysts.
It has been reported that the rate of ammonia for-
mation over transition metals is markedly enhanced by
addition of alkali metal cations. Since the most difficult
step for ammonia synthesis is accepted to be the dis-
sociation of nitrogen molecules, this enhancement is
also expected in the isotopic exchange of nitrogen which
References see page 1012
1012 5 Elementary Steps and Mechanisms
2.1 2.2 2.3 2.4
I 1T (X lo3 K)
Figure 7. Temperature dependence of the rates of methane and
CZ hydrocarbon formation in the reactions of CO-H2 and CO-Dz
over Ru/SiO2: A methane and A CZ+hydrocarbons in CO-H2; 0
methane and 0 C24 hydrocarbons in CO-Dz (from Ref. 27).
requires the dissociation. Morikawa and Ozaki [30]
found that the isotopic exchange over an unpromoted
iron catalyst is much slower than the ammonia forma-
tion rate, which is not enhanced by the presence of H2.
On the other hand, over doubly promoted catalysts,
the exchange rate became comparable to the ammo-
nia formation rate and was accelerated by the presence
of H2.
B Hydrogenation of Carbon Monoxide
The effects of using D2 rather than H2 during CO hy-
drogenation were investigated using alumina-supported
Ru catalysts [3 11. For the alumina-supported catalysts,
the rate of CD4 formation was 1.4-1.6 times faster
than the formation of CH4. Noticeable isotope effects
have also been observed during the synthesis of C2-Cll
alkenes and alkanes. For c 2 - C ~hydrocarbons, the syn-
thesis occurs more rapidly in the presence of D2 rather
than H2, but the magnitude of the inverse isotope
effect declines towards unity with increasing numbers
of carbon atoms in the product. The observed inverse
kinetic isotope effect can be rationalized by the results
of a combination of the kinetic and equilibrium isotope
effects associated with individual elementary steps.
The Hz and D2 kinetic isotope effect was measured
on silica-supported R u catalysts not only for the hy-
drogenation of CO but also for the hydrogenation of
deposited carbon formed by the disproportionation
of CO [32]. Figure 7 shows the inverse kinetic isotope
effects in both cases (kH/kD = 0.57 for methane and
0.43 for higher hydrocarbons), which led to the con-
clusion that the rate-determining step was the hydro-
genation of carbon formed by dissociative adsorption
of CO. The inverse effect can be interpreted in terms of
a higher surface concentration of CD, than of CH,
because of the thermodynamic stability of CD,.
References
1. M. Polanyi, A. L. Szabo, Trans. Faraday SOC. 1934,30, 508.
2. A. Ozaki, K. Akia, Isotopic Studies of Heterogeneous Cutal-
ysis, Academic Press, Kodansha, 1976.
3. M. Araki, V. Ponec, J. Catal. 1976, 44, 439.
4. P. Biloen, J. N. Helle, W. M. H. Sachtler, J. Catal. 1979, 58,
95.
5. Y. Kobori, H. Yamasaki, S. Naito, T. Onishi, K. Tamaru, J.
Chem. SOC., Faraday Trans. I , 1982, 78, 1473.
6. H. Yamasaki, Y. Kobori, S. Naito, T. Onishi, and K. Tam-
aru, J. Chem. Soc., Faraday Trans. I , 1981, 77, 2913.
7. A. Takeuchi, J. R. Katzer, J . Phys. Chem. 1981,85, 937.
8. A. Takeuchi and J. R. Katzer, J. Phys. Chem. 1982,86,2438.
9. H. Orita, S. Naito, and K. Tamaru, J. Catal. 1984, 90, 183.
10. 0. V. Isaev, L. Ya Margolis, I. S. Sozanova, Dokl. Akud.
Nauk. S S S R 1959,129, 141.
11. 0. A. Golovina, 0. V. Isaev, M. M. Sakharov, Dokl. Akad.
Nauk. S S S R 1962, 142.
12. G. F. Berndt, Catalysis, Royal Society of Chemistry, 1983,
p. 144.
13. H. H. Voge, C. D. Wagner, D. P. Stevenson, J. Catal. 1964,
3, 549.
14. C. R. Adams, T. J. Jennings, J. Catal., 1963, 2, 58.
15. G. W. Keulks, J Catal. 1970, 19, 232.
16. B. D. Wragg, P. G. Ashmore, J. A. Hockey, J. Catul. 1971,
22,49.
17. C. Kemball, Ado. Catal. 1967, 17, 223.
18. T. Kondo, S. Saito, and K. Tamaru, J. Am. Chem. SOC. 1974,
96, 6857.
19. S. Naito, M. Tanimoto, J. Catal. 1986, 102, 377.
20. S. Naito, M. Tanimoto, J. Chem. SOC.,Chem. Cornm. 1987,
363.
21. S. Naito, M. Tanimoto, J. Catal. 1989, 119, 300.
22. S. Naito, M. Tanimoto, Chem. Letts. 1988, 411.
23. S. Naito, M. Tanimoto, Chem. Letts. 1990, 2145.
24. S. Naito, M. Tanimoto, M. Soma, Y. Udagawa, Stud. Surf:
Sic. Cutal. 1993, 75, 2043.
25. S. Naito, M. Tanimoto, M. Soma, J. Chem. SOC.,Chem.
Comm. 1992, 1443.
26. J . Bigeleisen, M. Wolfsberg, Adu. in Chem. Phys. 1958, I , 15.
27. D. R. Strogin, S. R. Bare, G. A. Somorjai, J. Catal. 1987,
103, 289.
28. A. Ozaki, S. H. Taylor, M. Boudart, Proc. Roya/ SOC.
(London) 1960,258, 47.
29. K. Aika, A. Ozaki, J. Cutal. 1969, 13, 232.
30. Y. Morikawa, A. Ozaki, J. Catal. 1968, 12, 145.
31. C. S. Kellner, A. T. Bell, J. Catal. 1981, 67, 175.
32. Y. Kobori, S. Naito, T. Onishi, K. Tamaru, J. Chem. SOC.,
Chem. Comm. 1981, 92.
5.2.6 Transient Catalytic Studies
K. TAMARU
5.2.6.1 Importance of In Situ Transient studies
In chemical reactions, generally speaking, if the rate
(rate equation, rate constant, and activation energy) of
each of the elementary steps which make up the overall
reaction is known, the kinetic behavior of the overall
reaction may be understood. In the reaction between
hydrogen and bromine, for instance, the kinetic be-

havior of the overall reaction is rather complicated as
follows:
However, if all the elementary reactions in the over-
all reaction are known, the kinetic behavior of the
overall reaction may be explained on the basis of the
rate constants, activation energies of the forward and
backward reactions of the following elementary steps:
Br2 = 2Br
Br+H2=HBr+H
H+Br2=HBr+Br
In the case of such gaseous reactions the rate constants
and activation energies of each of the elementary steps
are independent of the concentration of reactants, re-
action products and reaction intermediates. This ap-
proximately also holds for reactions in solution.
In the case of heterogeneous catalytic reactions, the
properties of the medium (catalyst surface) where those
reactions take place depend upon the concentrations of
reactants, reaction intermediates, and reaction prod-
ucts and, in some cases, spectator species that do not
participate in the reaction path but which are chem-
isorbed on the catalyst surface. This is because the
properties of metal surfaces, such as work function and
heat of chemisorption, are markedly susceptible to the
amounts and kinds of coadsorbed species. In many
cases surface restructuring may happen along with the
adsorption, which results in marked changes in the re-
activity of the surface. Consequently, even if the sur-
face is well-defined at the beginning, being examined
by various spectroscopic techniques, it can be markedly
different under the reaction conditions. In the case of
oxide catalysts, for instance, the extent of oxidation of
the catalyst surface during the redox reaction is also
influenced by the relative ratio of the rates of reduction
and oxidation. This results in prominent changes in the
properties of catalyst surfaces depending upon the rel-
ative concentration of the reactants and reaction tem-
perature. The fugacity of oxygen over an oxide is one of
its inherent thermodynamic properties and its change
definitely influences the properties of the oxide itself, or
its catalytic activity. In the case of acid catalysts the
acidity of the working catalyst surface, which should be
associated with the catalytic activity, is that under the
reaction conditions and not that under the conditions
far from those prevailing in the actual reaction.
In the case of heterogeneous catalysis the rate con-
stants, activation energies and, sometimes the kinetic
equation as well, of elementary (or simpler) steps which
consist of the overall reaction will change, depending
upon the concentrations of the chemisorbed species on
the catalyst surfaces. The concentrations of the chem-
isorbed species on the catalyst surfaces, depend not
only upon the concentrations of reactants and products
5.2 Microkinetics 1013
in the ambient phase, but also the heats of chem-
isorption of those species and also reaction inter-
mediates and spectator species. Even more important is
that the catalytic reaction proceeds through a certain
number of elementary steps or reaction intermediates,
and the chemical potentials of these intermediates
depend upon which of the steps is rate determining.
Consequently, by estimating the chemical potentials of
the intermediates during the reaction, the rate deter-
mining step may be identified. One of the best-known
examples of this kind is the case of ammonia decom-
positionon on an iron surface. Since the rate of de-
sorption of chemisorbed nitrogen is rate determining in
this reaction, the chemical potential of chemisorbed
nitrogen, or the virtual pressure of nitrogen which
would be in adsorption equilibrium with the chem-
isorbed nitrogen in the working state, may become very
high during the course of the reaction, and iron nitride
is easily formed in the catalyst surface by decomposing
ammonia on iron surface. This is because the chem-
isorbed nitrogen during the course of reaction is not in
equilibrium with the ambient nitrogen gas, but is equi-
librated with the ambient ammonia and hydrogen:
Consequently, the iron catalyst surface during the re-
action may be surface iron nitride, which is quite dif-
ferent from a clean iron surface.
The properties of catalyst surface, consequently, are
determined not only by the concentration of the re-
actants and products in the ambient phase, but also by
the mechanism of the reaction, or by the location of the
rate determining step of the overall reaction. It is, ac-
cordingly, concluded that the kinetic behavior of each
of the elementary steps in heterogeneous catalytic re-
actions is influenced by the reaction itself through the
properties of the catalyst surfaces, which is again de-
termined by the concentration of the adsorbed species
during the course of the reaction.
Under these circumstances the rate and activation
energy of the elementary reaction in heterogeneous
catalysis should be measured under the reaction con-
ditions, since the surface which works as a catalyst is
that under the reaction conditions. In many cases the
rate of desorption of a chemisorbed gas is estimated by
temperature programmed desorption, using the gas and
the adsorbent under vacuum. The reactivity of chem-
isorbed species is similarly estimated by temperature
programmed reaction. The rate of desorption and the
reaetivity of the chemisorbed species thus estimated are
not necessarily the same as those under the reaction
conditions-under the reaction conditions many other
chemisorbed species are present to various extents on
the catalyst surface which modify the properties of the
References see page 1023
1014 5 Elementarv Stem and Mechanisms
catalyst surfaces. In some cases, if all the simpler pro-
cesses which make up the overall reaction are studied
separately, as functions of surface coverages of all the
possible adsorbed species and partial pressures of re-
actants and reaction products, the kinetic structure of
the overall reaction can be elucidated.
Adsorption during the course of reaction may be es-
timated by direct measurements in a closed circulating
system with a comparatively large amount of catalyst.
Such volumetric measurements only provide the total
amount of each of the compounds, and do not identify
the chemisorbed species [l]. Gas chromatography may
also be employed by using a microcatalytic reactor, or
by putting the catalyst in a separatory column, and by
using the flowing reactants (or an inert gas) as the car-
rier gas [2]. Direct studies of adsorbed species during
the course of reaction by spectroscopic techniques,
such as IR and EXAFS gives more information on the
adsorbed species [3].
When a catalytic reaction is a zero-order reaction
with respect to the reactant, the active sites in the cat-
alyst surface are saturated with such species as re-
actants or reaction intermediates, the adsorption being
independent of the ambient gas pressure. The kinetic
behavior does not tell what species are occupying the
active sites, but adsorption measurements may yield
the amount of pressure-independent adsorption and
its composition. This provides important information
on the number of active sites and the kind of the spe-
cies which saturate them. Spectroscopic measurements
of pressure-independent adsorbed species during the
course of reaction may provide additional information
about the saturating species.
In 1992 the book Catalysis Looks to the Future was
published by the National Academy Press, Wash-
ington, prepared by the panel on New Directions in
Catalytic Science and Technology, Board on Chemical
Sciences and Technology, National Research Council,
from which the following quote is taken:
It is desirable to focus on areas in which the extensive
scientific and technological resources of academe and
industry may lead to the fastest practical results. In
order of priority, these areas are: 1. in situ studies of
catalytic reactions; 2. characterization of catalytic sites
at atomic resolution (metals, oxides). . .
It is certainly important to study catalyst surface
under the reaction conditions, especially by spectro-
scopic methods, but even if some chemisorbed species
are observed during the course of the reaction, it does
not mean that it is a reaction intermediate through
which the overall reaction proceeds. It may be just oc-
cupying the active sites as a spectator, even retarding
the rate of the overall reaction. It is, accordingly, not
sufficient to study the catalyst surface in the working
state to elucidate the mechanism of the overall re-
action. In addition to observations of the chemisorbed
species under the reaction conditions, in situ dynamic
or transient approachs should be employed to identify
the real reaction intermediates, for instance by isotopic
methods [3]. Since the isotope has the same chemical
properties as the nonlabeled species, a change in the
concentration from the nonlabeled to the labeled spe-
cies will not perturb the chemical environment or the
total surface concentration of adsorbed species during
the reaction. Such a replacement will propagate to the
successive reaction intermediates, finally reaching the
reaction products. By following this behavior the re-
action path may be followed and the rate of each of the
elementary steps may be estimated. This information
should be compared to the rate of the overall reaction
measured separately. This approach was first proposed
in 1964 [3], and these basic principles of studying cata-
lytic reactions under reaction conditions are described
in other publications [l, 41.
5.2.6.2 Experimental Method
As an example of the experimental techniques for the
in situ transient approach, the apparatus of Chuang
and coworkers is shown in Fig. 1 [5]. The infrared cell
employed in the apparatus is shown in Fig. 2 [6]. It has
four stainless steel flanges, gas inlet and outlet, and two
step CaF2 windows. It is used at temperatures up to
533K and under a pressure of 60atm. The step win-
dows minimize the reactor volume and reduce the op-
tical path length for the gaseous species in the reactor.
The catalyst is pressed into a self-supporting disk and
then placed in the IR cell, and subsequently treated
prior to the reaction studies. By means of the inlet
system shown in Fig. 1 one of the reactants can be
switched to its labeled species in the steady state of the
reaction, total reactant flow being maintained constant
throughout the experiment. The change in the isotope
content in the chemisorbed species after the switch is
monitored by in situ IR spectrometry and the effluent
composition is examined continuously by a mass spec-
trometer, including its isotope abundance. The com-
position of the reaction products is also determined
with time by a gas chromatograph. In this way the
structure of the adsorbed species, the reaction path,
and also the rate of each of the elementary steps on the
catalyst surface under the reaction conditions may be
estimated.
Additional information which may be obtained from
the in situ transient approach is to estimate the chem-
ical potentials of the reaction intermediates during the
course of the reaction. In the steady state of chemical
reactions the chemical potentials of the reaction inter-
mediates prior to the rate determining step may be
approximately equal to that of the reactants, in many

Valve for
Balancing
Pressure
*COIAr
l2 co
C 2H4
Figure 1. Combined in situ IR and mass spectrometry system.
CaF2
Flange Fluid in O-ring window Nut
\
Fluid out
Side view Front view
Figure 2. High-pressure cell for in situ IR studies.
cases, being in partial equilibrium. For instance, for
ammonia decomposition on iron catalyst, the chem-
isorbed nitrogen is in partial equilibrium with the am-
bient ammonia and hydrogen. Consequently, by meas-
uring adsorption during the course of the catalytic
reaction, the chemical potentials of reaction inter-
mediates may be estimated, which gives important in-
formation with which to identify the rate determining
step of the overall reaction.
In some cases, transient response can be monitored.
Under reaction conditions, the chemical potential (or
the pressure) of one of the reactants or the products
may be suddenly changed (or even reduced to zero)
and the subsequent changes in the chemical potentials
(or concentrations) of the reaction intemediates may be
followed. In this manner the rate of simpler steps in the
overall reaction may be estimated [3, 71.
Gas chromatographic or pulse techniques can be
used to study kinetic behavior of adsorbed species
under reaction conditions, as proposed by Tamaru in
1959 [2]. Using the reacting gas as a carrier gas and in
the steady state of the reaction, a pulse of one of the
5.2 Microkinetics 1015
reactants or products was added and its retention time
measured to estimate the value of A x l A p on the work-
ing catalyst, where Ax is the increase of its adsorption,
when its pressure was increased by Ap. If the sample
gas is fully saturated on the catalyst surface during
the reaction, no additional adsorption takes place
( A x / A p = 0).
Various applications of this technique may be
readily proposed. In the simplest case, if the reacting
gas used as the carrier is switched to an inert gas in the
steady state, the adsorbed species on the working cata-
lyst will be desorbed with time. In this case, instead of
inert gas, one of the reacting gases may be employed to
study the reaction between the adsorbed species and
the reacting gas under the reaction conditions.
Gas chromatography is a convenient way of study-
ing both the amount and the kinetic behavior of each
of the adsorbed species on the working catalyst. Vari-
ous types of reactors may be used, such as fixed-bed
reactors, well-stirred or backmixed reactors, and single
pellet reactors. In this way transient response data may
be obtained, which can contribute significantly to the
chemical engineering aspects of reactor design and
catalyst development [2]. However, such overall ad-
sorption kinetic data will not necessarily elucidate the
mechanism of the catalytic reaction at an atomic or
molecular level.
5.2.6.3 Kinetics of Adsorption and Desorption
In Langmuir adsorption isotherm theory, the adsorp-
tion isotherm may be derived by equating the rates of
adsorption and desorption. The rate of desorption, is
assumed to be proportional to the amount of chem-
isorbed species, and the rate constant is independent
of coverage. The rate constant of desorption may be
References see page 1023
1016 5 Elementary Steps and Mechanisms

adsorption of labeled species and desorption of non-

labeled species, as shown in Fig. 3 [9]. In this manner
the absolute rates of adsorption as well as desorption
can be estimated in the presence of the adsorbing gas.
In the case of desorption of hydrogen chemisorbed
on transition metals, the rate was independent of the
ambient gas pressures, as generally accepted. It is of
interest, however, that the rate of desorption of carbon
monoxide on various transition metals is a function

1
0 I
I not only of its coverage, but also of the pressure of the
I
I adsorbing gas-a feature known as adsorption-assisted
1.0 desorption. The rates of adsorption (Ta) and desorption
Total CO
0
(Yd) of carbon monoxide on transitiion metals can be
P expressed by the following equations and, accordingly,
the adsorption isotherm could be obtained:
0.5
ra = k a P / [ 1 + K d / ( 1 - O)]
Yd = kde( 1 + BP")
0 where K and n are constants which are approximately
0 60 120 180 0.2 and 0.8, respectively, and ka and kd are rate con-
Timelsec
stants [lo]. The rate of adsorption is expressed by the
Figure 3. CO, C 1 * 0and total CO coverage during adsorption well-known Kisliuk mechanism [ 1 11. Lombard0 and
and desorption measurements on Pd at 380 K. Bell reported that the rate of adsorption-assisted de-
sorption may be explained by taking repulsive inter-
action among adsorbing species into consideration [121.
dependent on its coverage in many cases and various Kawai and coworkers recently studied the process of
empirical rate equations have been proposed, such as dynamic equilibrium between the gaseous and adsorbed
Frumkin's equation [8]. However, in all cases the rate CO by means of time-resolved infrared reflection ab-
of desorption was considered to be only dependent on sorption spectroscopy using l2CI6O and 13C'80[13].
the coverage and independent of ambient gas pressure. Their results are given in Table 1. It was accordingly
This is why the rate of desorption is usually measured demonstrated that the potential well of the preadsorbed
by temperature programmed desorption under vacuum. CO sites became shallow due to local repulsive inter-
In this technique a gas is chemisorbed at lower tem- action between the preadsorbed CO and the incident
peratures and the temperature is raised under vacuum CO molecules from the gas phase, the rate of desorp-
to observe the desorption. tion consequently being enhanced, as shown in Fig. 4.
Very few experiments have been carried out to mea- It is generally accepted that the molecular beam of CO
sure the rate of desorption while adsorption is taking scatters isotropically, which suggests comparatively
place, or in the presence of adsorbing gas in the am- long residence times for CO molecules as precursors
bient phase. The rate of desorption can be measured by before they scatter. Those long lived precursor species
the transient method, by replacing the adsorbing gas may assist the desorption of preadsorbed CO molecules
by its labeled species and subsequently measuring the due to dynamic repulsive interaction.

Table 1. The rates of the desorption process under equilibrium with gas phase CO, the cross-section for the flux-induced desorption and
the desorption rates into vacuum. The desorption rates into vacuum are calculated from a preexponential factor 1.6 x lOI4 s-' and a de-
sorption energy of 127 kJmol-I.

Temperature (K) +
kl ($)Fa1 ki 01
Rate of the desorption Rate of the desorption Cross-section of the
under the equilibrium into vacuum (SKI) flux-induced desorption
with gas phase CO (Cm2)
(s-')
368 2.9 10-3 1.5 x 3.5 x 10-15
380 4.7 10-3 5.6 x 10-4 5.1 x 10-15
392 7.1 10-3 1.9 x 10-3 6.6 x
400 9.4 x 10-3 4.1 10-3 6.6 x
 5.2 Microkinetics 1017

--+0
15

Desorption
-i
t
v

C
.-
Surface parallel 3 lo
C !i
P
.
c

$ 5
W
Temporal and local shallowing
a
2
r
L

a
Ic
o Incident CO $!

Adsorbed CO / 0 1 2 3
Time (h)
4 5

Figure 5. Decomposition of surface formate prepared from

I3C0 and D2 reaction over Crz03 at 573 K for 1 h: ( 0 ) amount of
Potential energy surface
D13CO0 under evacuation at 533K; (c) amount of DI3COO
during circulation of "CD3OD at 533 K; ( i ) amount of D ' * C 0 0
during circulation of "CD30D at 533 K.
Figure 4. Interaction of incident CO and preadsorbed CO

is not a reaction intermediate (Fig. 5). However, in the

5.2.6.4 Catalysis
In the case of desorption, the bond between the chem-
isorbed molecules and the surface is broken, whereas
in catalysis the bond within the chemisorbed species
is broken or synthesized. Since adsorption-assisted de-
sorption was observed, the reactivity of chemisorbed
species may be influenced by the presence of ambient
gas.
A Catalytic Decomposition of Alcohol
In the case of methanol decomposition on chromia, as
in the case of zinc oxide [14], the reaction proceeds via
formation of methoxide and then formate ion on the
catalyst surface prior to the reaction products [15]. The
formate ion thus formed is stable at reaction temper-
atures under vacuum, which may suggest that formate
presence of methanol in the ambient gas, the surface
formate becomes very reactive to form readily the re-
action products, as shown isotopically [15]. In the
presence of alcohol, the infrared absorption spectra of
chemisorbed formate shifts to suggest weakening of
C-H bond (Table 2).
The decomposition of ethanol proceeds readily on
an N b monomer catalyst on SiOz with high selectivity
to dehydrogenation [16]. This reaction was studied by
means of EXAFS and IR techniques. When ethanol
was introduced onto the catalyst, dissociative adsorp-
tion takes place to form an ethyl radical and a proton
on the catalyst. In the presence of ethanol in the gas
phase, dehydrogenation takes place. If, during the
course of the reaction, the ethanol is removed from the
gas phase, the reaction stops, although chemisorbed
species remained on the catalyst surface, unchanged

Table 2. IR absorption bands for formate ions on Cr203.

Assignment Wavenumber (cm-')

CD30D decompositions DCOOD adsorption

DCOO DCOO DCOO DCOO

only +CD30D only +CD30D
~~

C-D stretching 2184 2160 2182 21 50

0 - C - 0 antisymmetric stretching 1548 1573 1548 1573
0-C-0 symmetric stretching 1329 1337 1331 1337

References see page 1023

1018 5 Elementary Steps and Mechanisms

400 500 600 700 800

Temperature/K

Figure 7. Temperature programmed desorption spectra from

adsorbed ethanol on Nb/SiOl.
Reaction timelmin

Figure 6. Dehydrogenation of C2H50H on Nb/SiOz: (0)H2; (A) H

00 0,
CH3CHO.
Nb/.H, CHI
/ *, CH,
? ?
(Fig. 6). If the ethanol is then reintroduced the reaction
continues, at the same rate as before. In this manner
the presence of the reactant promotes the selective de-
hydrogenation reaction.
The ethanol thus chemisorbed was not reactive at
the reaction temperatures, but if it is treated at tem-
peratures considerably higher than the reaction tem- zm%n%% 52TamZm

peratures under vacuum, dehydration decomposition CH,CHO + H, H

takes place to produce C2H4 and (CtH5)2O, as shown
in Fig. 7. It is of interest that the dehydrogenation re-
action proceeds selectively in the presence of ethanol,
whereas in the absence of ethanol molecules in the gas
phase, the chemisorbed species decomposed to give
dehydration at considerably higher temperatures. The
scheme of the overall reaction is depicted in Fig. 8.
The decomposition of the chemisorbed species is pro-
moted not only by the reactant, but also by the presence
of various species in the gas bhase. It is of interest that
the decomposition of the chemisorbed species is pro-
moted by the presence of various molecules in the am-
bient gas, in particular those with higher donor num-
bers (Fig. 9). It is therefore important to note that the
reactivity of chemisorbed species on the catalyst sur-
faces is markedly influenced by the presence of not only
the reactants, but also other molecules. Consequently,
even if the reactivity of the chemisorbed species on the
catalyst surface is studied by temperature programmed
O0,O\
0 Nb-H-CH
c,n5 0 ?o C H 3
zTm%mzz
Figure 8. Mechanism of the dehydrogenation of ethanol on
Nb/SiO2.
techniques using one of the reactants, this may not give
its reactivity under the reaction conditions.
The mechanism of water-gas shift reaction on MgO
and ZnO was studied by EXAFS and FTIR and the
coadsorbates on the surface species, such as water
vapor, showed a marked effects on not only the re-
activity, but also the selectivity (to dehydrogenation) of
the decomposition of surface formate [ 171.
B Decomposition of Formic Acid
The decomposition of formic acid on various metal
surfaces has been widely studied as a typical catalytic
 5.2 Microkinetics 1019

4.0

0.8

Lo
0.7 -._, 3.0

-8 E
+ -E
7

0.6 .
7'
z
9
E 2.0
.-
C
._
._
I0
2
E
I I I I I I I I I i
D
0 20 40 60 80 1.o
Donor number

Figure 9. Rate of initial dehydrogenation decomposition of

chemisorbed species and donor numbers of nucleophilic reagents.

350 1 PtOoIr
0 log v = -0.8
0 log v 10
0
0 05
Coverage OHCOO-
1.o

400
Figure 11. The decomposition rate of formic acid as a function
of its coverage and temperature over Ni/Si02: (-) under 11
torr formic acid vapor: (- -) under vacuum; ( 0 . 5 ) 293 K: (A, .>)
383 K; (m, 1) 373 K.
450

500
g ported that the rate of the reaction on a silver catalyst
t was different for four different isotopic compounds, i.e.
550
60 70 80 90 100
AH,(kcal eq-')
Figure 10. Catalytic activity of various metals given by the
temperatures to show the same rate and their heats of formation
of formate salts.
reaction. When the catalytic activity of the reaction on
various metal surfaces was plotted against the heats of
formation of their formates, a volcano-shaped curve
was obtained (Fig. 10) [18]. It is therefore concluded
that the stability of the surface formates, through
which the overall reaction proceeds, determines the
catalytic activity: if it is too stable or too unstable, the
activity will be slow and, consequently, demonstrates a
maximum activity (or volcano curve) at an optimum
stability of the formate.
In connection with this reaction, Block and Kral re-
the rates of decompositions of HCOOH, HCOOD,
DCOOH and DCOOD are 3.02, 2.24, 1.78, and 0.61
( lop4ml min-'), respectively [19]. If the decomposition
of the surface formates is the rate determining step of
the overall reaction, these four different isotopic species
will give only two different rates.
The rate of the decomposition of surface formate in
the absence or presence of formic acid in the gas phase
was measured, which clearly showed a marked influ-
I10 ence of the presence of formic acid in the gas phase
(Fig. 11) [20]. This appears to be an adsorption-assisted
process, where the decomposition of surface formate is
promoted by the presence of formic acid in the gas
phase. It has been demonstrated, as mentioned above,
that the reactivity as well as the selectivity of the de-
composition of chemisorbed formate ion are markedly
influenced by the presence of water vapor, which plays
an important role in not only formic acid decom-
position, but also the water-gas shift reaction, since
both reactions proceed via formate ion.
The catalytic decomposition of formic acid on
TiOz(il0) was also studied in detail and it was con-
cluded that the catalytic dehydrogenation reaction
References see page 1023
1020 5 Elementary Steps and Mechanisms
0 0.4
Figure 12. Fractional respones F(t ) of gaseous effluent species during isotopic switching from Ar/'2CO/H2/C2H4 to '3CO/H2/CzH4:
C? = steady-state concentration of the species before the switch; C" = steady-state concentration of the species after the switch; CO
chemisorption = 28.5 pmol CO per gram of catalyst.
proceeds in a bimolecular process, requiring surface for-
mate and a formic acid molecule from the gas phase [21].
C Hydroformylation and Methanation
One recent example of a transient infrared study is
hydroformylation and methanation over Rh-Ce/SiOz
catalyst, [5, 6, 7, 221. In this study the experimental
apparatus mentioned above (see Figs 1 and 2 ) was
employed.
The Rh surface area was measured to be 2 8 . 5 ~
mol g-' by pulse chemisorption at 305 K. The ratio of
CO : H2 :C2H4 used for hydroformylation was 1 : 5 : 5
(instead of 1 : 1 : 1) to enhance formation of the alde-
hyde. The reaction products were C2H6 as the major
species and C2H5CHO and CH4 as the minor species,
with a turnover frequency (TOF) for propionaldehyde
production of 0.43 min-' at 453 K and 0.1 MPa. The
results, given in Fig. 12, were normalized to F(t), which
is the fraction of the final response. The evolution of
the I3CO curve is faster than that for C2Ht3CH0. The
complete switch from "CO to I3CO was achieved at
approximately 0.85 min, whereas the switch from
Hfa)
CO(g)=CO(a)-
C2H4(g)+ iH2 + CzHs(a) -
\
COW ClHSCO(a)
\ H(a)
C2H6(g)
0.8 1.2 1.6 2.0
Time (min)
C2H5CH0 to C2H:3CH0 took approximately 1.6min.
The rate of appearance of 13C0 in the gas phase is in
good agreement with the appearance of adsorbed I3CO
on the catalyst surface. The residence time for the
intermediates involved in propionaldehyde formation
was estimated to be 0.44min; that of linear CO, which
is the precursor for CO insertion, was 0.05 min.
The reaction mechanism for the methanation and
hydroformylation, proposed by Chuang and co-workers
[22], taking the results obtained by Sachtler co-workers
[23] into consideration, is summarized in Scheme 1
where (8) and (a) represent gaseous and adsorbed
states, respectively. In the methanation reaction the
adsorbed CO can dissociate and then hydrogenate to
produce methane. Propionaldehyde is formed by the
insertion of linearly adsorbed CO into the adsorbed
C2HS(a) species which is formed by hydrogenation of
ethylene. The results of this study indicate that the
propionaldehyde may be formed via (i) the insertion of
CO into adsorbed ethyl species to form the acyl inter-
mediate, (ii) hydrogenation of the acyl intermediate to
produce adsorbed propionaldehyde, and (iii) desorp-
H(a)
CH,(a) -+ CH4(g)
-%C2HsCHO(a) - C2HsCHO(g)
Scheme 1
 5.2 Microkinetics 1021

I I
0

"0 60 120 180

Reaction time (min)

Figure 13. Adsorption measurements during the hydrogenation of butene (1-butene fH2) + D2 over ZnO at 513 K: (): H2 (g); (0)
H(a); (I)D2(g); (m) D(a); (0)
HD(g); ( x ) 1-butene (a); (3) butane (g).

Table 3. Deuterium content in the main products produced by the reaction of C4H6 + H2 + DZ (1 : 1 : 1) over ZnO at 29 "C.

Reaction Conversion 1-Butene Hydrogen H2-D2 exchange"

time (min) ("A)
C4H6 C4Hz C4H7D C4H6D2 H2 HD D2 H2 HD D2

10 6.9 1.oo 0.46 0.08 0.46 0.45 0.07 0.48 0.29 0.40 0.31
20 10.4 1.oo 0.51 0.11 0.38 0.42 0.01 0.48 0.26 0.46 0.27
40 16.9 1.oo 0.48 0.17 0.35 0.31 0.18 0.51 0.25 0.49 0.26
80 27.0 1 .oo 0.46 0.20 0.34 0.33 0.19 0.48
160 45.0 1 .oo 0.44 0.26 0.30 0.34 0.26 0.40

"H2-Dl exchange reaction in the absence of 1,3-butadiene under similar conditions.

tion of adsorbed propionaldehyde, and that step (ii) is
the rate determining step. The steady-state rate mea-
surements show that increasing reaction pressure de-
creased the overall activation energy and increased
both rate and selectivity for propionaldehyde. The
coverage of C2H5CO was estimated from TOF and
residence time to be 0.19 min.
D Hydrogenation of Alkenes on Oxides
Hydrogenation of butadiene and butene was studied
over ZnO catalyst, adsorption during the course of the
reaction being measured as given in Fig. 13 [24]. The
adsorption of butadiene as well as hydrogen stayed
unchanged under various pressures of reactants during
the reaction, but the order of the reaction was first
order with respect to hydrogen and zero order with re-
spect butadiene. The products of the reaction between
H2, Dz, and butadiene are given in Table 3. The H2-
D2 exchange reaction proceeds on the catalyst, but
is markedly retarded by the presence of unsaturated
hydrocarbons. The reaction products clearly indicate
that the hydrogen, which is added, has just the same
isotopic composition as that in the ambient gas, which
looks as if it is added in molecular form. It is well
known that hydrogen is heterolytically dissociated to
chemisorb on ZnO, but the behavior of such chem-
isorbed hydrogen seems to suggest that such dis-
sociatively chemisorbed hydrogen does not participate
in the hydrogenation, the hydrogen to be added com-
ing from the gas phase.
The hydrogenation of ethylene over ZrO2 was stud-
ied, which showed similar hydrogenation behavior [25].
In this reaction system various adsorbed species may
be observed by IR techniques, such as OH, Zr-H, n-
bonded ethylene, end-on ethane, and side-on ethane.
Hydrogen was first chemisorbed on ZrO2 at room
temperature and cooled down to 220K and then eth-
ylene was introduced step by step as shown in Fig.
14. The ZrH absorption peak decreased steadily and n-
bonded ethylene peak increased, whereas no absorp-
tion spectra due to C2H5(a) and CzHs(a) were detected.
although n-bonded ethylene increased successively
This behavior indicates that the dissociatively pre-
adsorbed hydrogen is not added to the ethylene in-
References see page 1023
1022 5 Elementarv Stem and Mechanisms
3.0 1
11* . u s
reaction timelmin 0 30 60 90
ethene pressure / Tom P=O.14
1st reaction
Figure 14. Interaction of ZrH(a) and ethene gas.
CH2 DCH2D (g)
DZ r \
x-bonded ethene side-on ethane end-on ethane
Figure 15. Reaction mechanism of ethene hydrogenation over
ZrOz.
troduced to form ethane, but is desorbed as molecular
hydrogen under such conditions. Further important
evidence for nonreactive Zr-H and OH species was
provided by the results from volumetric measurements
which showed that ethylene hydrogenation proceeds
even at temperatures below 223 K where heterolytic
dissociative adsorption of hydrogen does not take
place. The overall reaction was first order as to hydro-
gen and zero order with respect to ethylene, which
is quite similar to the case on ZnO catalyst. Various
evidence suggests the mechanism of ethylene hydro-
genation reaction on ZrO2 to be as shown in Fig. 15.
E Other Catalytic Reactions
The examples introduced above are some of the typical
catalytic reactions studied by in situ transient tech-
niques. Similar techniques have been also applied to
various catalytic reactions, and many important results
have been obtained, demonstrating the effectiveness of
the approach. Some of the examples are as follows.
The rate coefficients for elementary processes occurring
during Fischer-Tropsch synthesis over Ru/Ti02 were
estimated by Komaya and Bell [26]. The oxidation of
carbon monoxide on molybdenum trioxide was also
studied by Iizuka et al. by in situ transient isotopic
0 30 60 90 0 30 60 90
P=0.17 P=0.14
2nd reaction 3rd reaction
techniques [27]. The reaction of adsorbed NO and CO
has been investigated by Chuang and co-workers,
combining in situ infrared and mass spectroscopy on
Rh/SiOz [28].
5.2.6.5 Summary
The in situ transient approach is one of the most or-
thodox ways to elucidate the mechanism of heteroge-
neous catalysis. Since catalysis takes place on the active
catalyst surface, which can differ considerably from the
clean catalyst surface, the properties of the surface
under working conditions should be examined in order
to understand the real nature of heterogeneous catal-
ysis. In addition, the behavior of each chemisorbed
species on each reaction site on the working catalyst
surface should be studied istopically under working re-
action conditions. In this manner not only the reaction
path, but also the rate of each component step of the
overall reaction may be estimated, and the rate de-
termining step may be elucidated. The spectator species
of the reaction which are chemisorbed on the catalyst
surface, but not participating in the reaction, may also
be identified.
In many cases of heterogeneous catalysis small
amounts of poisons retard the rate of reaction, and
promotors enhance the rate considerably. In some
cases the life of the catalyst plays an important role.
especially in industrial catalysts. The nature of such
poisons, promotors, and catalyst life can be inves-
tigated by the in situ transient approach using isotopes.
Any technique which is available for the study the
catalyst surface under the reaction conditions may be
used in the in situ transient approach. Techniques such
as IR spectroscopy can be used to follow the behavior

of labeled species. In situ polarized total-reflection
fluorescence extended X-ray absorption fine structure
spectroscopy can also give new information about the
working catalyst surface [29]. Gas chromatographic
techniques may be widely used to study the kinetic
behavior of adsorbed species under the reaction con-
ditions. Transient methods also provide important
information on catalyst and reactor designs, to assist
with chemical engineering aspects.
In many cases of heterogeneous catalysis the inter-
mediate species on the catalyst surface are too reactive,
and their concentrations become too small to be de-
tected by spectroscopic techniques. It then becomes
necessary to develop increasingly sensitive techniques
with higher resolution to detect small amounts of
chemisorbed species on different sites of the catalyst
surface. The nature of different surface sites, such as
steps, kinks and defects, should also be studied in con-
nection with surface catalysis. The role of those partic-
ular sites in catalysis require to be studied in more de-
tail. It is also of fundamental importance to identify the
real nature of active sites in heterogeneous catalysis.
In this respect recent advances in studying the micro-
scopic structure of solid surfaces using scanning tun-
neling microscopy will undoubtably yield more infor-
mation on the nature of active catalytic sites.
References
1. K. Tamaru, Bull. Chem. SOC.Jpn. 1958, 31, 666.
2. K. Tamaru, Nature 1959, 183, 319; J. Happel, Catal. Rev.
1972. 6,221; H, Kobayashi, M. Kobayashi, Catal. Rev. 1974,
10, 139; C. 0 . Bennett, Catal. Rev. 1976, 13, 121; P. L. Mills,
J. J. Lerou, Reo. Chem. Eng. 1993, 9, 1.
3. K. Tamaru, Ado. Catal. 1964, 15, 65.
4. K. Tamaru in Catalysis: Science and Technology (Eds: J. R.
Anderson, M. Boudart), Springer Verlag, Berlin, 1991, Vol.
9, 38.
5. G. Srinivas, S. S. C. Chuang, M. W. Balakos, AIChE 1993,
39, 530.
6. S. S. C. Chuang, S. I. Pien, J. Catal. 1992, 135, 618.
7. M. W. Balakos, S. S. C. Chuang, J. Catal. in press.
8. A. Slygin, A. Frumkin, Acta Physicochim., U.R.S.S., 1940,
12, 321.
9. T. Yamada, T. Onishi, K. Tamaru, Surf: Sci. 1983, 133,
533.
10. T. Yamada, K. Tamaru, Z . f : Physik. Chem., N.F. 1985, 144,
195; T. Yamada, Y. Iwasawa, K. Tamaru, Surf: Sci. 1989,
223, 527.
11. P. J. Kisliuk, J. Phys. Chem. Solids 1957, 3, 95; 1958, 5, 5.
12. S. J. Lombardo, A. T. Bell, Surf: Sci. 1991, 245, 213.
13. N. Takagi, J. Yoshinobu, M. Kawai, Phys. Reo. Lett. 1994,
73, 292.
14. A. Ueno, T. Onishi, K. Tamaru, Trans. Faraday SOC.1971,
67, 3585.
15. K. Yamashita, S. Naito, K. Tamaru, J. Catal. 1985, 94,
353.
16. M. Nishimura, K. Asakura, Y. Iwasawa, Proceedings of the
9th International Congress on Catalysis, (Eds: M. J. Phillip,
M. Terna), 1988, 4, 1842.
5.2 Microkinetics 1023
17. T. Shido, K. Asakura, Y. Iwasawa, J. Chem. Soc., Faraday
Trans. I 1989, 85, 441; Y. Iwasawa, Elementary Reaction
Steps in Heterogeneous Catalysis (Eds: R. W. Joyner, R. A.
van Santen) Kluwer, 1993, 287.
18. W. M. H. Sachtler, J. Fahrenfort, Actes du Deusieme Congres
lnternationale de Catalyse, Paris, 1960, 1961, p. 831.
19. J. Block, H. Kral, Z. Elektrochem. 1959, 63, 182.
20. K. Takahashi, E. Miyamoto, K. Shoji, K. Tamaru, Catal.
Lett. 1988, 1, 213.
21. H. Onishi, T. Aruga, Y. Iwasawa, J. Catal. 1994, 146, 557.
22. S. S. C. Chuang, R. Narayanan, Appl. Catal. 1990, 57, 241;
G. Srinivas, S . S. C. Chuang, J. Phys. Chem. 1994, 98, 3024.
23. P. Biloen, W. M. H. Sachtler, Adu. Catal. 1981, 30, 165; W.
M. H. Sachtler, M. Ichikawa, J. Phys. Chem. 1986, 90, 4752.
24. S . Naito, Y. Shimizu, T. Onishi, K. Tamaru, Bull. Chem.
SOC.Jpn. 1970, 43, 2274; S. Naito, Y. Sakurai, H. Shimizu,
T. Onishi, K. Tamaru, Trans. Faraday, SOC.1971,67, 1529.
25. J. Kondo, K. Domen, K. Maruya, T. Onishi, J. Chem. Soc.,
Faraday Trans., 1990, 86, 397; 3021; 3665.
26. T. Komaya, A. T. Bell, J. Catal. 1994, 146, 237; B. J. Savat-
sky, A. T. Bell, ACS Symp. Ser. 1982, 178, 105.
21. Y. Iizuka, Y. Onishi, T. Tamura, T. Hamamura, J. Catal.
1980, 64, 437.
28. G. Srinivas, S. S. C. Chuang, S. Debnath, J. Catal. 1994, 148,
748.
29. M. Shirai, T. Inoue, H. Onishi, K. Asakura, Y. Iwasawa, J.
Catal. 1994, 145, 159.
5.2.7 Positron Emitters in Catalysis
Research
G. JONKERS
5.2.7.1 Introduction
Much quantitative information on reaction mecha-
nisms and kinetics is originating from (ultra) high vac-
uum technique studies on catalytic (model) systems,
which may or may not have been pretreated. With re-
spect to the acquired data, the question often is raised
is whether these are also valid for the actual conditions
(e.g. high pressure, temperature and hostile environ-
ment) under which catalysts may operate in commer-
cial units. Therefore, there has always been a strong
incentive for analytical techniques, which may provide
the quantitative data on reaction mechanisms and
kinetics from the catalysts under true operating
conditions.
To carry out in situ kinetic studies on catalysts sev-
eral criteria must be satisfied. First, detection of re-
action and/or interacting molecular species should not
perturb the process under investigation. Detection
should preferably allow both identification of any in-
termediate reaction species and quantification of the
relevant reaction parameters. Second, the kinetic pro-
References see page 1032
1024 5 Elementary Steps and Mechanisms
cess being studied should characterize the entire inner
and outer surface of a real catalyst system. Finally,
the surface detection technique should be capable of
resolving continuous changes in the surface process.
However, it should be realized that no single technique
will be able to provide the ultimate answer to the
questions raised and that the solution should be sought
in multidisciplinary studies, taking into account the
scope and limitations of detection techniques, and cat-
alyst preparation and characterization techniques.
Several analytical techniques are potentially capable of
generating information about the state of a catalyst
surface under operating conditions:
Fourier transform infrared and Raman spectroscopies
(FTIR, F T R S ) : as the fundamental spectroscopic
parameters of adsorbed species are unknown and
many distinct components may be present, even
qualitative evaluation of the recorded spectra will be
difficult. To date, therefore, the extraction of quanti-
tative information has turned out to be impossible.
Mossbauer (emission) spectroscopy: capable of prob-
ing the molecular environment of catalytic centres.
However, due to the available isotopes this technique
is limited to Fe or Co catalytic centers.
Nuclear magnetic resonance ( N M R ) : although the
existence of intermediate species at a catalyst surface
can be demonstrated [l], quantification of these spe-
cies in terms of concentration is not yet possible.
Positron annihilation spectroscopy ( P A S ): capable of
probing electronic properties of surfaces, of localized
defects - pores, vacancies, shear planes, grain
boundaries - and of the solid bulk. Applications of
this technique to catalysis are fairly new and have
been reviewed in Ref. 2. A triad of detection tech-
niques may provide qualitative information on prob-
ing of inner catalytic sites, such as Brernsted acid
sites. In PAS the characteristics of positron anni-
hilation, such as positron life-time, momentum and
energy distribution, are utilized for the study of cat-
alytic materials. Here, the positrons are originating
from a suitable external positron-emitting source.
Consequently, this technique [2] is fundamentally
different from the application of positron emitters as
described in this review.
If data on reaction mechanisms and kinetics are re-
quired, isotopic labeling techniques may provide an
efficient means, as this is one of the few methods for
distinguishing two atomic species of the same element
in the same compound, in a similar chemical form (see
Section A.5.2.5). For hydrocarbon heterogeneous cat-
alysis the main elements of interest are isotopes of hy-
drogen, carbon, oxygen, and nitrogen and, to a lesser
degree, fluorine, phosphorus, and sulfur. Analytical
techniques which exploit physical characteristics of
isotopes are mass spectroscopy (nuclear mass, sampling
required), FTIR and FTRS (nuclear mass effects),
Mossbauer Spectroscopy (nuclear state), NMR (nuclear
spin) and radiotracer techniques (decay of unstable
nucleus). However, in special cases only the latter
technique can provide quantitative data from in situ
studies of heterogeneous catalysts.
Radiotracer techniques, exploiting radionuclides
such as 3H, 14C, 32P, and 35Sas tracers, may provide
quantitative information about reaction mechanisms
and rate limiting steps in reaction kinetics, but no in
situ information from the catalyst surface can be ob-
tained. However, due to the specific radiation charac-
teristics, radionuclides such as "C, 13N, I5O, and '*F
can be applied to quantitative, in situ studies of heter-
ogeneous catalysts. The latter radionuclides have had
application in biological and biochemical experiments
as early as 1936 [3]. The first applications to heteroge-
neous catalysis are more recent [4].
This article will briefly describes the radiation prop-
erties of /?-emitters, the production and synthesis of
/?-labeled compounds and the (in situ) detection of
/?-labels. Four different applications of llC, 13N and/
or, 150in heterogeneous catalysis are also summarized,
as are some developments areas and potential future
applications.
5.2.7.2 Characteristics of /I-Emitters
Any radioactive decay process, in which the mass
number A remains unchanged but the atomic number
2 changes, is classified as /?-decay.Unstable nuclei with
superabundant neutrons, such as 14C, emit a fast elec-
tron or /?--particle, thereby accomplishing a more
stable nuclear state, here 14N. Nuclei with insufficient
neutrons, such as "C, may also be unstable and emit a
fast positron (the antimatter particle of an electron) or
/?+-particle, thereby achieving a more stable nuclear
state, here B. The radiation characteristics of relevant
/?-emitters (2 < 18) which may be applied for hetero-
geneous catalysis studies are summarized in Table 1.
The SI unit of radioactivity is the becquerel (Bq),
which represents the number of nuclear disintegrations
per second. Once the molecular mass of the compound,
in which the radionuclide is embedded, and the half-life
of the radionuclide is known, the number of labeled
molecules giving rise to one nuclear disintegration per
second, i.e. the specific activity, can be computed. For
'normal' manageable quantities of radioactivity up to
about 109Bq (or lGBq), this activity corresponds to
about 1mmol 14C0 or 7 pmol "CO (Table l), empha-
sizing the sensitivity of "C-labeling experiments. For
quantification of the health, safety, and environment
(HSE) effects, the SI unit sievert (Sv), for equivalent
 5.2 Microkinetics 1025

Table 1. Radiation characteristics of relevant p-- and p+-emitters.

Radionuclide Half-life Decay mode Max. Energy Max. linear Max. specific
and yield (MeV) range in A1 activity
("/.) (mm) (Bq mol-I)

1'C 20.4 min p+ >99 0.96 1.4 3.4 x 1020

I3N 9.96 min p+ 100 1.19 1.9 7.0 x lozo
150 2.07 min p+ 99.9 1.72 3.0 3.4 x 1021
';F 110 min p+ 97 0.635 0.9 6.3 x 1019
H 12.4 y p- 100 0.0186 0.002 1.1 x 10'5
'4c 5730 y p- 100 0.155 0.1 2.3 x 10l2
32P 14.3 d p- 100 1.71 3.0 3.4 1017
35s 87.4 d p- 100 0.167 0.1 5.5 x 10'6

and effective does is used. This takes into account the
potential biological damage caused to the human body
by irradiation or intake of radionuclides. By taking
appropriate shielding measures these doses can by kept
as low as 1 pSv per pulse experiment (as compared to
the world-averaged natural background dose per indi-
vidual of about 2 mSv per annum).
The average number of 0-particles emitted per
nuclear disintegration is given by the P-yield which,
due to a competing decay process (electron capture),
may be lower than 100% for P+-particles. P-Particles
are emitted with a continuous energy distribution ex-
tending from zero to a maximum value. Coulomb in-
teractions with the (atomic) electrons of the matter
traversed will limit the maximum range of P-particles
(Table 1).
In cases where 3H, 14C, 32P,or 35S labeling studies
are applied to obtain quantitative information, the
limited range of p--particles requires sampling of the
outlet gases or liquids from a catalytic bed and sub-
sequent (destructive) analysis. After a fast P+-particle
has been retarded by Coulomb interaction, it will
associate with a "slow" electron (generally valence
electrons) thereby converting both the mass of the p+-
particle and the (valence) electron into electromagnetic
radiation. This so-called annihilation process is gov-
erned by the energy ( E = m x c2 where m is the mass of
the electron or positron and c is the speed of light
= 511 keV) and the momentum conservation law (two
y-photons emitted in the opposite direction). The re-
leased annihilation radiation is very penetrative and
can be detected external to catalytic beds and therefore
makes both in situ and quantitative studies of catalysts
feasible.
5.2.7.3 Production of Labeled Compounds
Though 3H and 14C are continuously generated in the
outer atmosphere by cosmic radiation, the commercial
production of 3H, 14C, 32P, or 35S is carried out by
irradiation of suitable target material in a neutron
irradiation facility, often a nuclear reactor. The radio-
labeled precursor material from the target has to be
radiochemically purified, after which this precursor
material has to be converted into the required com-
pound by chemical (micro)synthesis and, optionally,
subsequent chemical purification. As the shortest-lived
of these four radionuclides lives for more than 2 weeks,
there will be ample time to carry out the synthesis and
to deliver the labeled compound for experiments at a
radionuclide laboratory. As the range of P--particles is
limited, protection against radiation hazards is easily
achieved by shielding the radioactivity with layers of
perspex.
In general, "C, 13N, 150,and lSFare produced by
deuteron (8 MeV) or proton (16 MeV) bombardment
of suitable target materials, in which the labeled pre-
cursor molecules will be formed instantaneously. The
labeled precursor compound [5] is utilized either for
"direct" conversion into the required chemical com-
pound, or for conversion into other labeled precursor
compounds to be used as a starting compounds for
chemical (micr0)synthesis (Table 2). With respect to
the half-life time of the "C, I3N, 150,and 18Fradio-
nuclides, synthesis and purification steps should be fast,
and experiments with labeled compounds have to be
carried out in the near vicinity of the site where the
radionuclides were produced. Next to lead shielding
(hot cells and dedicated shielding) when handling Pt-
emitters, the synthesis and experimental procedures
should as far as possible be remotely controlled for
protection against radiation hazards.
Through the emergence of positron emission com-
puted tomography (PET) during the last decades as
a medical imaging tool [6] the general reputation of
I1C, 13N, 1 5 0 , and lSFhas become more wide spread.
For medical PET applications dedicated self-shielding
cyclotrons have been developed to allow medical re-
search and diagnostic studies [7]; however, existing
nuclear accelerator facilities at a university or nuclear
research institute generally may be used for "C, I3N,
References see page I032
1026 5 Elementary Steps and Mechanisms

Table 2. Precursor molecules available immediately, or with limited synthesis (<t112), after bombardment [ 5] .

p+-emitters Target Precursor Limited synthesis (<tip)

N2/02 (trace) "coz, "co Hzl'CO, "CH31, "CH3Li, C1211C0

NdH2 HWN
H2O 1 3 ~ 03 - , 1 3 ~ 20 - 1 3 ~ 01,3 ~ 0 213
, "0, H I ~ N O ~
CHI 13NH3
various 1500 c I 5 0 , c i 5 0 0 ,~ ~~~~~0
~ ~ 0 ,
'ONe or l80O 1 8 ~H I ~~ F , CH3I8F, cIO3lsF, N0I8F, Xe18FF, (C2H&NS18F
H2I80 18F-

Table 3. Comparison between 14C- and C-labeling for heterogeneous catalyst studies.

Step Radionuclide

14C(tip = 5730 years) "C (tip = 20.4 minutes)

Production of radio-
nuclide precursor
molecules
Synthesis of labeled
molecules
HSE aspects (synthesis)
Transport and storage of
labeled molecules
Measurement and
analysis
Potential application
area
HSE aspects (applica-
tion)
Relatively simple, cheap at nuclear irradiation
facility, e.g. nuclear reactor. Large amounts
can be produced, exceeding 10 GBq.
No time constraints. Complex compounds may
be synthesized by standard organic synthesis
routes, followed by purification. Costs vary
considerably, dependent on the molecule to be
labeled and available knowledge of synthesis
routes.
Handling of large amounts can easily be shielded.
Generation of radioactive waste.
Particularly during long time storage, compounds
with high specific activity may decompose
due to self-irradiation (radiolysis). Transport/
delivery over large distances is feasible.
Detection requires sampling. Sensitive techniques,
particularly in combination with accelerator
mass spectroscopy (AMS). Quantitative
molecular information only from sampled
(intermediate) reaction products.
Studies into reaction mechanisms, kinetics, effectiveness, and yields in biology, medicine, various types of
process chemistry (homogeneous/heterogeneous catalysis). Both these radionuclides can be substituted
for stable carbon ("C, I3C) in organic molecules without changing the chemical properties of these
molecules.
Radiotracer in oil/gas fields (mean residence
times > year). Natural I4C activity exploited
for carbon dating (AMS).
Handling of large amounts relatively simple.
Shielding can easily be achieved; potential for
contamination; generation of radioactive
waste.
Dedicated targets and target materials. Relatively
simple and at moderate costs from nuclear
accelerator facility, e.g. cyclotron, linear
accelerators; up to approximately 5 GBq.
Time constraints. Synthesis routes for industrially
interesting small molecules available. For
larger molecules further exploration required.
Costs vary, dependent on molecule to be
labeled and available knowledge of synthesis
routes.
Remotely controlled synthesis in lead-shielded
fume cupboards ("hot cells"). No generation of
radioactive waste (decay).
Prolonged storage not possible. Transport/
delivery only possible to nearby institutes (time
window between end of synthesis and start of
experiments, preferably shorter than 1 h).
Quantitative, external in situ detection of radio-
nuclide inside (transient, steady-state etc.)
reactor or body. Extremely sensitive. Quanti-
tative molecular information only from sam-
pled (intermediate) reaction products.
Only processes completed over time-scales shorter
than approximately 1 h.
Supply and temporary waste containers should
be lead-shielded; remote control of valves; no
generation of radioactive waste.

150,and/or 18Fproduction. The successful application
of llC, 13N and/or 150to, heterogeneous catalysis
requires close interdisciplinary cooperation between
radionuclide production experts, radiochemical syn-
thesis experts, radioanalytical experts and experts on
the discipline of application (Table 3).
5.2.7.4 Detection of p - and Annihilation Radiation
Before the emitted B--particles can be detected, the
outlet gases or liquids from catalytic beds have to be
sampled (see Section A5.2.7.2),after which the samples
may be separated into the constituting compounds

quantitatively analyzed in the presence of the radio-
label. Depending on the application, a variety of de-
tection techniques for b-tracers can be used [8], such
as ionization chambers, liquid scintillation counters,
pyrolysis combustion detection in combination with a
ionization chamber, and autoradiography (two-dimen-
sional b-tracer distributions).
The presence of p+-tracers can also be quantitatively
analyzed by external detection of the indirectly emitted
511 keV y-photons. This is achieved by, for example,
scintillation detectors consisting of relatively large vol-
ume, single crystals of Nal(T1) or small volume, single
crystals of Bi4(Ge04)3 (BGO). A detector is generally
contained in a lead collimator - comprising a slit which
defines the detector's field of view by stopping all 7 -
photons from unwanted directions - thereby also re-
ducing the signal due to natural background radiation.
In cases where only one of the emitted 51 1 keV y-
photons is used (single photon mode), the distance of
the detector to the p+-source contained in a reactor, the
dimensions of the detector and collimator slit, and the
settings of the ?-energy analyzer (pulse height analysis
or PHA) are the main factors determining the total
signal received and the attainable positional resolution
(order of magnitude, centimeters). Depending on the
effectiveness of the collimation and the PHA settings,
the signal due to natural background radiation may
vary from approximately 5 to 25 counts per second.
Two scintillation detectors on opposite sides of the
contained p+-source are required for simultaneous de-
tection of two 51 1 keV y-photons (coincidence mode).
Only limited absorptive collimation shielding is re-
quired, as the field of view of the pair of detectors is
now mainly determined by electronic collimation,
thereby rejecting 5 11 keV y-photons, which are not
registered within a preset time window (< 10 ns). An
improved positional resolution, which is more or less
determined by the range of the p+-particles in matter
(Table l), may be achieved by the coincident mode.
The total coincident background signal (accidental co-
incident events) will be very low, generally less than
0.01 coincident events per second, making "C-labeling
an extremely sensitive analytical technique for hetero-
geneous catalysis research.
5.2.7.5 Application to Heterogeneous Catalysis
The mechanism and kinetics of acetylene-propyne ad-
sorbate interactions on a Cr"' catalyst have been eval-
uated using "C-labeled acetylene and diacetylene [4,
91. These labeled molecules were directly produced
in the target during irradiation of appropriate hydro-
carbon gases. After chromatographic purification at
the start of each experiment about 0.2MBq H'ICCH
and 5KBq H"CCCCH was available. Together with
5.2 Microkinetics 1027
L l
A A
8
I I I I
C2H2 0 0.2 0.4 0.6 0.8 I .o
C,H, 1.0 0.8 0.6 0.4 0.2 0
M O L FRACTION A L K Y N E
Figure 1. CrV1catalyst: effect of carrier alkyne mole fraction on
the "C-labeled xylene product distribution.
propyne, the labeled molecules were condensed in
vacuo on the catalyst surface. Single annihilation pho-
tons from adsorbed C-labeled molecules were moni-
tored by a lead-collimated (pinhole 8 x 35mm), large
volume (76.2 x 76.2 mm) NaI(Tl).The products (ben-
zene, toluene, xylenes) desorbed from the catalyst sur-
face were analyzed using a (radio) gas chromatograph
(RGC) with thermal conductivity response detector
and a flow-through gas proportional counter.
The molecular orientation on CrVrof the adsorbates
with respect to product selectivity and the competitive
'
sorption of HI CCH and propyne could qualitatively
be described. The kinetics of HI'CCH displacement by
propyne from CrV1 was analyzed to be first order in
acetylene and second order in propyne concentration,
while the catalyst selectivity dependence modeled to the
relative HI 'CCH displacement rate could be quantified.
With this technique lop8 of a monolayer surface
coverage could be detected. Therefore it could be es-
tablished that the selectivity was altered at monolayer
alkyne coverage through variation of the acetylene-
propyne distributions. Calculation of modeled xylene
yields as a function of the acetylene displacement rate
resulted in distribution curves which were nearly iden-
tical with experimental data (Fig. 1).
The mechanism and kinetics of CO oxidation on a
platinum/ceria,$yalumina catalyst has been evaluated
using "C-labeled CO and C02, and 150-labeled CO,
C02, and 0 2 [lo-181 under the transient conditions
that apply during the cold start of a car. Large quanti-
ties of "C02 (up to 50 GBq) were extracted from a N2
References see page 1032
1028 5 Elementary Steps and Mechanisms

target (Table 2) and subsequently trapped. "CO (up to

30 GBq) is obtained by reducing ' I COz and subsequent I T=130"C I
purification. The I50-labeled molecules were produced
on-line by a flow-through target, and appropriate traps
and convertors were applied to obtain C150, C i 5 0 0
or I5OO of required radiochemical purity. Synthetic
exhaust gas was passed through a catalyst bed, of
length approximately 15 cm, under the required con-
ditions for several hours to achieve steady state. Then,
approximately 1 MBq of the "C-labeled and 0.2 MBq
of the 150-labeled molecules were used to carry out 0 80 160 240 320 400
pulse experiments under steady state, transient con-
time. s
ditions. Gases from the catalyst bed were monitored
by a NaI(T1) scintillation detector (Fig. 2), then ana-
lyzed by mass spectrometry and a gas chromatograph
equipped with a conventional flame ionization detector
and an ionization chamber.
I II T= 140°C

Initially, the labeled molecules flowing through or

P
adsorbed on a catalyst bed [lo, 161 were monitored
both by one (single photon mode) and by two (co-
incident mode) large volume (76.2 x 76.2 mm) NaI(T1) -
d
E
3
detectors. Other than being shielded by lead rings, the 8
scintillation detectors were not collimated, resulting in
a sensitivity profile over the axis of the catalyst bed.
A computational model for the kinetics of CO oxi-
dation and C02-ceria interaction was set up. Using 0 80 160 240 320 400
literature data the measurements could be simulated time, s
reasonably well. A relatively high sticking probability
of 0 2 on the catalyst surface, caused by the promoting
effect of ceria, had to be used to simulate the mea-
surements. Additionally, the degree of occupation of I 1 T= 150°C
CO on a Pt surface and the CO oxidation rate under
different reaction conditions were determined. The P
adsorption-desorption equilibrium of C02 on ceria d
was quantified by establishing the equilibrium con- E
3
stant. However, accurate quantification of the kinetics 8
turned out to be complex using the experimental set-up
with two large volume Nal(T1) scintillation detectors,
because in a plug flow reactor steep gradients occur II . . .

while only the sum of the total reactor signal was reg-
istered. Also, the signal from the total catalyst bed was
subject to large inhomogeneous field-of-view effects.
More information on the presence of reactants and in-
termediates at the catalyst surface would considerably
increase the accuracy of the kinetic reaction parameters.
This need for more information required the devel-
opment of a linear array of opposing detectors. How-
ever, such an array of detectors was already present in
the PET apparatus [19] formerly used to explore in-
dustrial PET applications [20]. Now, the catalyst bed
could be monitored by two parallel arrays of 11 BGO
detectors resulting in 21 independent points (separa-
tion about 11 mm, spatial resolution per point about
lOmm), at which the "C- c.q. i50-concentration was
measured simultaneously; this measurement could be
repeated every 1.2 s. The recorded data were presented
0 80 160 240 320 400
time, s
Figure 2. Pt/ceria/y-alumina catalyst: (single photon) annihila-
tion signal as recorded by the Nal(T1) detector monitoring the
outlet gases of the catalyst bed during "CO pulse experiments at
130, 140 and 150 "C. "CO/"CO2 (product) distribution assessed
from RGC data points (indicated by open circles).
as a two-dimensional (x, t ) false-colour matrix, or so-
called reaction image, (Fig. 3) and subsequently ana-
lyzed. Although the banks of BGO detectors were an
integral part of a PET setup, this method of data re-
cording can at most be classified as one-dimensional
PET.
Some further experiments were carried out to
achieve an accuracy improvement with respect to the
 5.2 Microkinetics 1029

Figure 3. Ptlcerialy-alumina catalyst: reaction image (display of coincident data from 11 individual BGO detectors of two opposing
banks, simultaneously recorded approximately every second at 21 measuring points and subdivided over a total length of about 22cm) of a
I'CO pulse experiments, 75% conversion at 140°C. The vertical axis represents the time coordinate, the horizontal position and the grey
values are a measure of the I'C concentration. The I'C profile at 80 s, and the residence time distribution of the "C-label at -5 crn are
indicated. A computer simulation of the reaction image is shown in the top right comer.

kinetic data obtained with the NaI(T1). Although
labeled 13N0was not available, the effects of blocking
a significant amount of the available Pt reaction sites
could effectively be demonstrated by adding some sta-
ble NO to the synthetic exhaust gas. This was quanti-
fied by the lower CO desorption rate and a somewhat
higher activation energy for desorption. Rather un-
expectedly, the labeling of different atoms of the same
reactant molecule (e.g. " C O and C150), provided
much insight into the mechanism of the overall cata-
lyzed reaction. The rates of adsorption, desorption, and
surface reaction were determined by simulating the
quantitative experimental results by means of the
computational model. During low temperature CO
oxidation by oxygen, the catalyst surface is predom-
inantly covered by adsorbed CO molecules, while the
dissociative and irreversible adsorption of 0, at the
surface is the rate limiting step. Oxygen atoms of CO
remain much longer in the catalyst bed than the carbon
atoms, which could be ascribed to carbonate formation
at the ceria surface and exchange of the oxygen atoms
of these carbonate groups with the ceria lattice oxygen
atoms (Fig. 4).
The selective reduction of N O by NH3 over vanadial
titania catalysts at very low reactant concentrations has
been studied using 13N-labeled NO [21]. 13N-labeled
HNO3 (83%), NO (9"/0), and N2 (8%) were contin-
uously produced in the oxygen gas target during irra-
diation. By exploiting the appropriate convertors and
traps, either pure 13N0, about 50 Bq ml-' at an overall
References see page 1032
1030 5 Elementary Steps and Mechanisms

Figure 4. Pt/ceria/y-alumina catalyst: a collection of reaction images of different positron emitter labeled species at 140°C (75% con-
version). (A) I3NN over inert carrier material (noninteractive flow); (B) "CO (reaction and product interaction with ceria); (C) I'COz
(interaction with ceria; label hold-up, chromatographic effect); ( D)I5OO (25% immediately to outlet, reaction and I5O in product ex-
changes with ceria); (E) CI5O (reaction and I5O in product exchanges with ceria); (F) C 1 5 0 0(I5O exchanges with ceria).

flow rate of 580 ml min-' , or pure 13N02 was obtained.
Coincident photons were monitored by lead-collimated,
high purity Ge semiconductor detectors.
The selective catalytic reduction of NO by NH3 has,
been monitored with 5 x lop9ppm of 13N0,more than
11 orders of magnitude lower than the usually used
concentrations (Fig. 5 ) . These low concentrations re-
sult in very low conversion rates, enabling investigation
of the kinetics of the catalyst materials after pretreat-
ment with various reactive gases without adding these
gases during the conversion experiment. Catalysts pre-
treated with NH3 did reduce 13N0, whereas pretreat-
ment of H2 did not show any 13N0conversion. For the
NH3-pretreated catalyst the first-order reaction rate
for NO conversion and its activation energy could be
determined quantitatively. Using thermochromatog-
raphy, adsorption enthalpy for 13N0 and 13N02, and
hence NO and NO2 on various oxides, have also been
determined.
The heat of adsorption and surface coverage of n-
hexane on H-mordenite has been studied using "C-
labeled n-hexane [22]. C02 was batch-wise converted
into "CO using standard procedures (Table 2), after
which H3 l1 CCsHll was produced via a newly developed
route [23]. This is a two-step homologation reaction
in which "CO is pulsed at 350°C over a vanadium-
promoted Ru/SiO2 catalyst, after which the temper-
ature is rapidly reduced to about 100 "C and 1-pentene
is pulsed over the catalyst. Desorptive hydrogenation
leads to a mixture of alkanes in the c1-c6 range.
By optimizing the heat treatment protocol and the
subsequent chromatographic purification procedures
 5.2 Microkinetics 1031

PEP images - similar to the reaction images de-

scribed above - of adsorption/desorption phenomena
1500 on H-mordenite have been recorded between 150°C
A

a
v1 and 230°C at 10°C intervals by injecting H3"CCsH11
0
v into a hydrogen feed stream (150mlmin-'). The
x
.->
4-
1000 PEP images showed an enormous reduction in the
._
e H3 l 1 CCsH" retention time, when about 10 pL min-'
2 n-hexane was added to the hydrogen feed stream
500
(Fig. 6). From these studies a heat of adsorption for
n-hexane on H-mordenite was determined to be
0 66.7 k 1.3 kJ mol-' . In addition, the H-mordenite
0 50 100 150 200
surface coverage of n-hexane at various temperatures
Temperature ( " C ) and n-hexane partial pressures could be analyzed
quantitatively.
Figure 5. Vanadia/titania catalyst: NO conversion measured
as positron-emitting activities of I3NO (open circles) and I3NN
(solid squares) as a function of temperature. Gas hourly space
velocity was about 72 000 per hour with a feed composition of
5x ppm I3NO and 15% 0, in He.
5.2.7.6 Conclusions

The scope for "C, 13N, and 1 5 0 positron emitters in

''
towards the production of C-labeled hexane, about
2 MBq H3 CCsHl1 was available for experimental use.
A dedicated positron emission detector has been
designed and constructed, specifically for studies in
tubular reactors [24]. The detector, consisting of two
opposite arrays of nine independent (moveable) BGO
scintillation crystals, has been optimized to measure
the positron label concentration profile along the
longitudinal axis of the reactor (spatial resolution
3 mm, temporal resolution 0.5 s). This method of one-
dimensional data recording is called positron emission
profiling (PEP).
heterogeneous catalysis research has effectively been
demonstrated. These applications illustrate the benefits
of "C, 13N, and 150labelling techniques, for the ac-
quisition of time- and position-dependent data. The
high sensitivity of data acquisition and its quantitative
interpretation, in terms of reaction mechanisms, re-
action rates, concentration of intermediate species and
coverage of adsorption sites, activation energies and
transport phenomena, have been demonstrated.
Although the number of applications is as yet lim-
ited, the potential of this technique to investigale other
heterogeneous catalysis systems is clear. The recent de-

Figure 6. Positron Emission Profiling (PEP) images obtained on H-mordenite at 150°C at 150mL/min: u ) . I ' C H ~ C S H Iinjected
I into a
hydrogen feed stream, b ) , C H ~ C ~1Hinjection
I following presaturation using n-hexane (10 pL/min)/hydrogen.

References see page 1032

1032 5 Elementary Steps and Mechanisms
velopment of a "C02 generator [25] makes feasible the
in situ, on-site application of "C02 or (in combination
with an appropriate catalytic convertor) "CO in
operational pilot plants. Storage of the "C-label in a
carbonate solution both facilitates the transport of the
C0:- supply vessel and the controlled release of
"co*.
However, there are some factors which may limit the
widespread application of this technique. These are
an awareness of the potential of this type of label,
the required close interdisciplinary cooperation, and the
number of available synthesis routes for labeling the
required molecular species. The first two factors may
be easily overcome, but the third requires some pre-
planning [22-241.
Although routes to some interesting molecules have
already been reported [5] or have recently been devel-
oped [23, 261, greater variety will extend the areas of
potential application of this very promising labeling
technique.
References
1. G. J. Nesbitt, H. H. Mooiweer, A. K. Novak, T. H. L.
Maessen, D. Schuize, A. Mizee, W. Kuhn, K. Mehr, P.
Dejon, S. Hafner, unpublished results.
2. R. Miranda, R. Ochoa, W.-F. Huang, J. Mol. Cat. 1993, 78,
67-90.
3. J. M. Buchanan, A. B. Hastings, Am. Physiol. Rev. 1946,26,
120-155, and references therein.
4. R. A. Ferrieri, A. P. Wolf, J. Phys. Chem. 1984, 88, 2256-
2263.
5. J. S . Fowler and A. P. Wolf, Positron Emitter Labelled
Compounds: Priorities and Problems, P. 391-450, in Positron
Emission Tomography and Autoradiography: Principles and
Applicationsfor the Brain and the Heart, eds. M. E. Phelps,
J. C. Mazziotta and H. R. Schelbert, Raven Press, New
York, 1986.
6. S. Webb (ed.), The Physics of Medical Imaging, Adam Hiiger,
Bristol, 1988.
7. M. E. Phelps, J. C. Mazziotta, H. R. Schelbert (eds), Positron
Emission Tomography and Autoradiography: Principles and
Applicationsfor the Brain and the Heart, Raven Press, New
York, 1986.
8. G. F. Knoll, Radiation Detection and Measurement, Wiley
and Sons, New York 1989.
9. R. A. Ferrieri, A. P. Wolf, J. Phys. Chem. 1984, 88, 5456-
5458.
10. K. A. Vonkeman, PhD thesis, Technical University Eind-
hoven, 1990.
11. K. A. Vonkeman, G. Jonkers and R. A. van Santen, Studies
to the functioning of Automotive Exhaust Catalysts using in-
situ Positron Emission Tomography, SAE technical paper
#910843 SOC.Automobile Eng., New York, 1991.
12. K. A. Vonkeman, G. Jonkers and R. A. van Santen, Studies
to the Functioning of Automotive Exhaust Catalysts using In-
situ Posftron Emission Tomography, p. 239-252 in Cataiy-
sis and Automotive Pollution Control II, (Ed.: A. Crucq),
Elsevier Science, Amsterdam, 1991.
13 G. Jonkers, K. A. Vonkeman, S . W. A. van der Wal, R. A.
van Santen, Nature 1992,355, 63-66.
14. K. A. Vonkeman, G. Jonkers, S . W. A. van der Wal, R. A.
van Santen, Ber. Bunsenges. Phys. Chem. 1993,97, 333-339.
15. G. Jonkers, K. A. Vonkeman, S. W. A. van der Wal, in Pre-
cision Process Technology, (Eds. M. P. C. Weijnen and
A. A. H. Drinkenburg), Kluwer, Deventer, 1993.
16. K. A. Vonkeman, G. Jonkers, S . W. A. van der Wal, J. de
Jong, H. Oosterbeek, R. A. van Santen, unpublished results.
17. K. A. Vonkeman, G. Jonkers, J. de Jong, H. Oosterbeek,
R. A. van Santen, unpublished results.
18. K. A. Vonkeman, G. Jonkers, P. Goethals and K. Strijck-
mans, R. A. van Santen, unpublished results.
19. E. J. Hoffman, M. E. Phelps, S.-C. Huang, J. Nucl. Med.
1983,24,245-257.
20. E. A. van den Bergen, G. Jonkers, K. Strijckmans, P. Goe-
thals, Int. J. Radiat. Appl. E. 1989, 3, 407-418.
21. U. Baltensperger, M. Ammann, U. K. Bochert, B. Eichier,
H. W. Gjggeler, D. T. Jost, J. A. Kovacs, A. Tijder, U. W.
Sherer, A. Baiker, J. Phys. Chem. 1993,97, 12 325-12 330.
22. R. A. van Santen, B. G. Anderson, R. H. Cunningham, A. V.
G. Mangnus, J. van Grondelle, L. J. van IJzendoorn, paper
submitted to the 11th Int. Congress on Catalysis, 1996.
23. R. H. Cunningham, R. A. van Santen, J. van Grondelle,
A. V. G. Mangnus, L. J. van IJzendoorn, J. Chem. SOC.
Comm., 1994, 1231.
24. A. V. G. Mangnus, L. J. van IJzendoorn, J. J. M. de Goeij,
R. H. Cunningham, R. A. van Santen, M. J. A. de Voigt,
Nuclear Instruments and Methods B 99, 1995, 649.
25. P. S. Kruijer, J. D. M. Herscheid, Appl. Radiat. Isot. 1995,
46, 337-338.
26. P. S. Kruijer, J. D. M. Herscheid 1995 personal communica-
tion ("C-labelled n-heptane synthesis).
5.2.8 Nonlinear Dynamics: Oscillatory
Kinetics and Spatio-Temporal
Pattern Formation
G. ERTL
5.2.8.1 Introduction
The rate of a catalytic reaction is governed by the un-
derlying mechanism, i.e. the sequence of elementary
reaction steps, and depends on the external control pa-
rameters such as the concentrations (= partial pres-
sures) p, of educt and product species and temperature
T , if transport processes are excluded in this context.
More specifically, usually the adsorbed species are
assumed to be randomly distributed over the surface
and their concentrations (= coverages) Bi are described
within a mean-field approximation so that the reaction
rate (=number of molecules r formed per unit time,
dn,/dt) is given by
whereby the coverages 8i are in turn determined by a
set of differential equations (modeling the individual

steps of adsorption, desorption and surface reaction) of
the type
Proper treatment then requires that additional con-
straints given by heat and mass balance are taken into
consideration.
Under flow conditions with the external control pa-
rameters being kept fixed, usually the reaction proceeds
at steady state with constant rate, i.e. dn,/dt =constant
and dOi/dt = 0. Due to the nonlinear character of the
quoted differential equations, this has, however, not
necessarily to be the case; the kinetics may - for certain
ranges of parameters - become oscillatory or even
irregular (chaotic). Such systems are typically far away
from equilibrium and as a consequence “dissipative
structures” [l] may emerge. The coverages Bi may
vary both in time and space and may give rise to phe-
nomena of spatio-temporal self-organization as being
treated in the general area of nonlinear dynamics [2].
Experimental observations on oscillatory kinetics
have been made with quite different systems, first with
electrochemical reactions [3] and later quite extensively
in homogeneous solution with the famous Belousov-
Zhabotinsky (BZ) reaction and related systems [4,51.
In heterogeneous catalysis, rate oscillations were first
reported about 25 years ago by Wicke and co-workers
for the oxidation of carbon monoxide on supported
platinum catalysts [6, 71. Since then numerous catalytic
reaction systems have been investigated in more or less
detail, and this subject has also been reviewed quite
extensively [8-161. Oscillatory kinetics have so far been
found with more than a dozen catalytic reactions, in
particular with CO oxidation, but also with oxidation
of H2, NH3, or hydrocarbons, with reduction of NO by
CO, H2, or NH3, as well as with hydrogenation re-
actions and even with the endothermic decomposition
of methyl amine. Further distinction is made with re-
spect to the type of catalyst (e.g. single crystal or poly-
crystalline material) as well as the pressure range stud-
ied. This classification is reasonable for the following
reasons:
(i) By far the most detailed insights into the under-
lying mechanisms and the general phenomenology
of spatio-temporal self-organization were obtained
in studies with well-defined single crystal surfaces
by applying the arsenal of modern surface physical
methods.
(ii) Single crystal studies are usually conducted with
bulk samples under low pressure conditions
(I lop4mbar) where the temperature changes as-
sociated with varying reaction rate are negligible.
In addition, the mean free path of gaseous mole-
cules is comparable or even larger than the di-
5.2 Microkinetics 1033
mensions of the reaction vessel which is operated
as a continuous flow reactor. Hence, concentration
gradients in the gas phase are practically instanta-
neously (I lop3s) transmitted. Analysis of such
experiments is thus appreciably simplified.
(iii) Experiments with polycrystalline, i.e. “real”, cata-
lysts are usually conducted at elevated partial
pressures ( 21 mbar) under which the heat released
by the reaction may lead to noticeable temper-
ature variations. The resulting thermokinetic effects
may become dominating, so that even the non-
uniformity of the surface chemistry is masked, and
quite novel phenomena may arise. Ideally, the
reaction is conducted in a way which may be
described by a continuously stirred tank reactor
(CSTR), i.e. perfect mixing ensures that the con-
centrations are everywhere identical. However,
frequently a concentration profile exists along the
reactor which is hence rather of the plug-flow
type. It is thus evident that heat and mass transfer
limitations may play important roles with such
systems.
Generally, an extended system such as a single crystal
surface, or even more a supported catalyst, exhibiting
temporal variations of the integral reaction rate has to
be subject to a coupling mechanism between various
parts. Otherwise, superposition of the uncorrelated
contributions from different regions would cause aver-
aging so that constant steady-state behavior would re-
sult. As a consequence, the coverages Bi as introduced
in eqs 1 and 2 will in general not only depend on time
but also on spatial coordinates, and proper mathemat-
ical modeling will be achieved in terms of partial dif-
ferential equations (PDE) as will be outlined further
below.
The following coupling mechanisms may be operat-
ing:
(i) Surface diffusion Local differences of the coverages
will lead to surface diffusion of the adsorbates, and
inclusion of this effect into mathematical modeling
extends eq 2 to reaction-diffusion (RD) equations
of the type
(3)
where Di is the diffusion coefficient of the respec-
tive adsorbed species. Diffusion lengths of the sys-
tems under consideration here are typically of
the order of about 1 pm. Concentration patterns
will hence be formed on extended single crystal
surfaces, but also on monolithic polycrystalline
material where the individual grains have typically
References see page 1049
1034 5 Elementary Steps and Mechanisms
much larger diameters [17]. However, other types
of RD phenomena will develop on the small
( I10 nm) crystal planes of supported catalyst
particles as modeled in experiments with field
emission tips [18, 191.
(ii) Gas-phase coupling Varying reactivity will also af-
fect the partial pressures of the reactants, even in a
flow system. As outlined above, at low pressures
these changes will propagate practically instanta-
neously and may offer a rather effective “global”
coupling mechanism which synchronizes different
parts of a single crystal surface or even of sepa-
rated samples [20]. Experiments with external
periodic modulation of one of the partial pressures
revealed that indeed amplitudes far below 1% may
suffice to cause such synchronization effects [21].
At elevated pressures, however, this coupling
mechanism will be of minor importance.
(iii) Heat conductance Local variations of reactivity
will be accompanied by temperature differences
due to the finite reaction enthalpies. Coupling be-
tween different regions may thus occur through
heat flux counterbalancing the temperature gra-
dients. This mechanism generally prevails at higher
pressures and also with supported catalysts where
temperature changes of up to order of 100 K may
arise. The characteristic “diffusion” length is in
this case of the order of about lmm, which is
much larger than the mean separation between
catalyst particles or even the diameters of their in-
dividual crystal planes. As a consequence, such a
system may again be regarded as being uniform
(on the relevant length scale), but the observed
phenomena will be strongly affected by these heat
phenomena rather than by the details of the sur-
face chemistry, and external constraints (such as
a constant average temperature) may provide effi-
cient global coupling.
In the following, after a brief outline of the theoretical
background, rather than to attempt to present an ex-
tended review of the field, the underlying principles will
be illustrated by selected examples with progressing in-
herent complexity.
5.2.8.2 Overview of the Theoretical Background
Sets of coupled nonlinear ordinary differential equa-
tions (ODES) of the type of eq 2 may be subject to
systematic analysis of possible solutions in the frame-
work of bifurcation theory [22, 231. Bifurcation simply
means a qualitative change of the dynamic behavior of
a system upon variation of one of its control parame-
ters, e.g. a transition from stationary to oscillatory be-
havior. The temporal behavior of a system described
by two variables can only be stationary or harmonic
oscillatory. These two situations are corresponding to
fixed points or limit cycles, respectively, in a phase-
space representation of the mutual dependencies of
the two variables where time is eliminated. With three
or even more variables the situation becomes much
more complex and may involve so-called mixed-mode
oscillations or chaotic behavior, the latter being char-
acterized by “strange attractors” in the phase-space
representation.
The specific nature of the differential equations de-
scribing the kinetics depends, of course, on the mecha-
nism of the underlying elementary steps. However, it
has been demonstrated that any exothermic reaction in
an open system can give rise to bistable, excitable or
oscillatory behavior for which the principle features
may be obtained by using just a single unimolecular
step [24, 251. Various mathematical models have been
proposed to describe oscillatory catalytic reactions,
either as general models or adapted to specific experi-
mental situations. The latter may quite often be traced
back to one of the various types of general models
which, e.g. may contain a slow buffer step [26], com-
prise coverage-dependent activation energies [27-3 11 or
involve the creation of vacant sites [32, 331.
The decision as to whether a proposed scheme yields
oscillatory solutions can, of course, be reached by in-
tegrating the differential equations or, even more ele-
gantly, by applying a more general method denoted as
stoichiometric network analysis (SNA) [24, 251.
As outlined in the introductory section, with any ex-
tended system the effects of spatial coupling between
different parts play an important part. If coupling by
diffusion is dominant, proper theoretical description
has to be based on partial differential equations
(PDEs) of the type of eq 3. The general nature of pos-
sible resulting concentration patterns has been explored
quite extensively in connection with the analysis of
chemical reactions in homogeneous solution [2, 5,
34-36]. The basic feature is that of a propagating
“chemical” wave, as recognized already in 1906
[37], which propagates with a velocity determined
by d m where D is the diffusion coefficient and k an
effective, first-order rate constant. Apart from prop-
agating waves, stationary so-called Turing patterns [38]
may be formed with which phenomenon, for example,
the periodic faceting of a Pt(ll0) surface in the course
of CO oxidation was identified [39, 401.
The different types of chemical waves may be clas-
sified with regard to whether the medium is bistable,
excitable, or oscillatory [2]. In a bistable medium a
transition from one state to another (differing in con-
centration) may proceed via a propagating reaction
front. In excitable media an external perturbation of
sufficient strength may cause a transient excursion of
the system to another state from where it returns to the

initial state. After some refractory time it may then be
excited again. Propagating pulses and spiral waves are
the characteristic features of such a situation. In con-
trast, oscillatory media do not need external perturba-
tion but perform autonomous periodic changes, asso-
ciated, for example, with the formation of so-called
target patterns, unless global coupling mechanisms are
operating. Finally, irregular and rapidly changing pat-
terns known as “chemical turbulence” may be formed
which represent the spatio-temporal counterpart of
temporal chaotic behavior.
In a bistable system both states may coexist just
at one specific value of the varied control parameter,
while otherwise the less stable state is pushed away by a
moving interface. If the front is not flat but exhibits a
curvature with radius R,it propagates with a modified
velocity c = co D/R [2], where D is the diffusion co-
efficient and co the velocity of the plane wave. Evi-
dently a critical radius, Rc = D / C Oexists
, below which
c < 0, i.e. a front does not expand but shrinks. That
means that Rc represents the critical radius for nucle-
ation of a concentration wave in a defect region which
is typically of the order of 1 pm, as determined for CO
oxidation on Pt [41].
Spiral waves are ubiquitous in reaction-diffusion
systems and characteristic for excitable media, whereby
the excitation not necessarily requires an external per-
turbation, but quite frequently may be identified with
an inherent local variation of kinetic parameters such
as found with a defect zone on a surface. The general
properties of spiral waves may be rationalized in the
framework of the so-called kinematic approximation
[2, 42, 431. The core of a spiral on a catalytic surface is
often formed by a defect to which the spiral is ‘pinned’
and which determines its characteristic properties, such
as rotation period and wavelength [42].
Global coupling is often decisive for the formation of
patterns in oscillatory media. Depending on the kind
of feedback, global coupling may either stabilize or
destabilize the uniformly oscillating situation through
symmetry breaking [44]. Systematic analysis of the
kinds of pattern formation in an oscillatory medium
with global coupling was performed with a model sys-
tem, the modified Ginzburg-Landau equation whose
characteristic features are, however, of general validity
[45]. It was shown that global coupling may, among
others, modify or even suppress turbulent behavior.
With realistic systems, such as nonuniform surfaces,
the observed spatio-temporal patterns may become
quite complex due to superposition and coupling of
differing contributions. Analysis may then become
rather difficult. One is then interested in methods that
allow extraction of the relevant features and sim-
plification of the dynamics. A suitable technique is the
Karhunen-Loeve decomposition into an optimal set
of eigenfunctions, also known as proper orthogonal
5.2 Microkinetics 1035
decomposition [46] which has already been successfully
applied to catalytic reactions in a few cases [47, 481.
5.2.8.3 CO Oxidation on Pt(ll0): A Case Study of a
Uniform Isothermal System
As mentioned above, kinetic oscillations associated
with a catalytic reaction were first discovered with the
oxidation of CO on supported platinum catalysts [6, 71,
and it was also this reaction which was first inves-
tigated using the “surface science” approach, i.e. with
well-defined single-crystal surfaces under low pressure
conditions [49]. It is still the system studied in most
detail, among which the Pt( 110) single-crystal surface
exhibits the richest variety of phenomena and is un-
derstood best.
The mechanism of this reaction is well established
[50] and proceeds schematically along the following
steps:
co + * * toad
0 2 + 2* + 20ad
oad + toad + c02 + 2*
Here, * denotes a free adsorption site; this has a dif-
ferent meaning for the two adsorbates. Whereas dis-
sociative oxygen adsorption is strongly inhibited by the
presence of preadsorbed CO, the chemisorbed 0 atoms
form rather open adlayers which do not significantly
affect the additional uptake of CO (asymmetric in-
hibition). Under steady-state flow conditions the tem-
perature has to be high enough ( 2400 K) to enable
continuous desorption of CO and creation of free ad-
sorption sites, otherwise complete blocking of the sur-
face by adsorbed CO would inhibit oxygen adsorption
and, consequently, the catalytic surface reaction. Fig-
ure 1 shows the steady-state rate of C02 formation on
a Pt(ll0) surface, which was either flat or periodically
stepped (faceted), as a function of CO partial pressure,
while the two other control parameters (PO, and T )
were kept fixed. At low pco, the surface is largely cov-
ered by adsorbed oxygen, and the reaction rate is gov-
erned by the supply of CO and therefore rises propor-
tionally to pco. Eventually, the stationary CO coverage
becomes so high that it starts to inhibit oxygen ad-
sorption noticeably, and further increase of pco causes
a decrease of the reaction rate which becomes now
limited by oxygen adsorption. Since the faceted surface
exhibits a higher oxygen sticking coefficient [51], for
identical control parameters the rate also becomes
larger. No such difference is observed at low pco where
the rate is limited by CO adsorption whose sticking
References see page 1049
1036 5 Elementary Steps and Mechanisms

I I I I I I
Pl11101
Po2 = 2.0 x mbar 1 m - f a c e t t e d "LEED"
T :L 5 5 K ,,--,2 facetted "LEED"
-
6- / \
L /
' \ l d I 1x1
nv1 /'2
5 \\ I

-
/
\ '
\ ' -
\I\
10011
1x l -
-c
1x 2
\
\
mcil \ Figure 2. The two structural modifications of the Pt(ll0)
\ - surface.
\

I
I
2
I
3 L 5
'.
I..
6
PCO mbar

Figure 1. Rate of COz formation in catalytic CO oxidation at a

Pt(ll0) surface which is either flat (full line) or faceted (broken
line) as a function of CO partial pressure pco, with the oxygen
partial pressure PO,, and temperature T , kept fixed.

coefficient is not noticeably affected by the surface

structure.
The steady-state kinetics can readily be modeled by
solution of the corresponding differential equations for
the CO and 0 coverages, and the resulting data may
even be transferred to "real" catalyst systems [52]. The t [sl
decrease of the rate with increasing pco becomes Figure 3. Oscillatory behavior resulting from integration of the
steeper at lower temperature (due to the increasing three ordinary differential equations modeling the kinetics of
mean residence time of adsorbed CO on the surface), CO oxidation on a Pt(ll0) surface for a particular set of control
and the system eventually becomes bistable, exhibiting parameters: T = 540 K, PO, = 6.7 x lo-' mbar, pco = 3 . 0 ~
mbar. Temporal variation of the reaction rate, the 0-and CO-
branches of high and low reactivity, and a clockwise coverages as well as the fraction of the surface being in the 1 X 1
hysteresis upon variation of pco [53, 541 which can structure, respectively (ML = monolayer) [55].
readily be modeled with a two-variable system [55, 561.
While such a behavior is characteristic for the densely
packed Pt( 111) surface, novel effects-namely rate os- readily be rationalized. Starting with the 1 x 2 surface,
cillations-come into play with the more open crystal the po2:pco ratio will be such that sufficient CO is ad-
planes such as (100) [49, 571, (110) [58] or (210) [59, 601. sorbed to lift the reconstruction. On the 1 x 1 patches
Returning to Pt( 1lo), these oscillations occur within formed more oxygen will be adsorbed which then re-
a relatively narrow range of pco, marked by a bar in acts with adsorbed CO, so that the coverage of the
Fig. 1, and are associated with a periodic variation latter drops below the critical value for maintaining the
of the atomic structure of the surface. As indicated in 1 x 1 phase stable. The surface structure transforms
Fig. 2 the clean Pt(ll0) surface is reconstructed into a back to 1 x 2 where the 0 2 sticking coefficient is
1 x 2 "missing row" structure which is transformed smaller, so that the CO coverage may build up again,
into the normal 1 x 1 structure if the CO coverage and so on. In this way, the microscopic surface struc-
exceeds a critical value of about f3co = 0.2 [61, 621. ture acts as an atomic switch between states of high
This structural transformation is driven by the en- and low reactivity and is thus responsible for the oc-
ergetics of CO adsorption and is a rather local process currence of rate oscillations. Proper inclusion of the
which involves displacements of surface Pt atoms only surface state as a further variable leads to a set of three
over short distances as demonstrated by scanning tun- ODES whose solutions exhibit, apart from steady-state,
neling microscopy (STM) [63]. On the other hand, the bistable, or excitable behavior, oscillatory kinetics for
oxygen sticking coefficient is larger on the 1 x 1 phase certain ranges of control parameters such as those dis-
than on the 1 x 2 phase [64]. Since the kinetic oscil- played in Fig. 3 [55]. Note that the 0 and CO cover-
lations are observed under conditions for which oxygen ages are strictly anticorrelated, while the fraction of the
adsorption is rate-limiting, their occurrence may now surface being present as 1 x 1 phase oscillates with a
 5.2 Microkinetics 1037

1.60

1- 10 sec

Figure 4. Experimental time series of the rate of COz formation on a Pt(l10): T = 550 K, PO, = 4.0 x mbar, the CO pressure was
changed stepwise at the points marked.

phase shift and plays the role of a slow variable whose

time constant is determined by the kinetics of the
structural phase transition. The reaction rate strictly
parallels the 0 coverage, and since the latter is reflected
by the variation of the work function, Aq, this quantity
may serve as a convenient monitor for the rate oscil-
lations rather than direct recording of the C02 partial
pressure by means of a quadrupole mass spectrometer.
Figure 4 shows an experimental time series which
was recorded in this way, for which po2 and T were
kept strictly fixed, while pco was varied stepwise in
small increments. (The onset of oscillations at even
larger pco occurs via small amplitudes which con-
tinuously grow upon decreasing the CO pressure as
characteristic for a Hopf bifurcation [58].) The oscil-
lations in Fig. 4 are at first strictly periodic, until at a Figure 5. Phase portraits constructed by the delay method from
the experimental data of Fig. 4 for (a) p c o = 1.65 x mbar,
certain value of pco (1.64 x lop4mbar) a qualitative (b) p c o = 1.63 x mbar, (c) p c o = 1.60 x mbar, and (d)
change (i.e. bifurcation) occurs. Alternating small and pco = 1.59 x mbar.
large amplitudes are formed and the period is doubled.
Upon further decrease of pco the amplitude ratio con-
tinuously varies. until at 1.60 x lop4mbar another ponents became positive. This is the first example of
doubling of the period takes place. Further slight var- “hyperchaos” identified with a chemical reaction [66].
iation of pco to 1.59 x lop4mbar causes the appear- Although mathematical modeling, in terms of the
ance of irregular variations-a state known as deter- quoted system of ordinary differential equations for the
ministic chaos. The whole transition to chaos through laterally uniform surface concentrations, accurately re-
the sequence of period doublings follows the well- produces the stationary kinetics as well as the bifurca-
known Feigenbaum scenario and is demonstrated more tion to periodic (harmonic) oscillations, it fails to de-
clearly by the phase portraits in Fig. 5, which were scribe the period-doubling transition to temporal chaos
constructed from the corresponding time series [65]. [55]. This is because spatial coupling has been com-
Detailed analysis shows that for the chaotic state asso- pletely neglected; it was tacitly assumed that the whole
ciated with a “strange attractor” one of the Lyapounov macroscopic surface is always strongly synchronized,
exponents became positive so that initially nearby tra- so that the concentrations depend only on time. As
jectories in the phase space diverge. Analysis of an- outlined above, this will generally not be the case, and
other set of data recorded further inside the chaotic
window revealed that then even two Lyapounov ex- References see page 1049
1038 5 Elementary Steps and Mechanisms
Figure 6 . Spatio-temporal concentration patterns on a Pt( 1 10) surface during CO oxidation recorded by photoemission electron micro-
scopy (PEEM) for conditions of dynamic bistability: T = 443 K, PO, = 4 x
as a consequence the formation of spatio-temporal
concentration patterns is to be expected.
If nonisothermal effects are neglected, coupling by
reaction-diffusion and global coupling through the
gas phase have to be taken into consideration, and a
proper theoretical description is achieved by inclusion
of surface diffusion, so that the equations are extended
into a set of partial differential equations (PDEs) of
the type of eq 3 [41, 67, 681. The resulting theoretical
skeleton bifurcation diagram for the spatio-temporal
pattern formation in the present system contains two
excitable and an oscillatory region, as well as two bi-
stable regimes with one- or two-front solutions.
Before discussing in some detail the characteristic
features of these patterns, their experimental verifica-
tion will be briefly described.
Imaging of lateral distributions of adsorbed species
at low pressures was initially performed with scanning
techniques [57, 691 which, however, suffered from
limited spatial and temporal resolution. These dis-
advantages are circumvented by photoemission electron
microscopy (PEEM) [70] which was used to record all
the images to be shown below. Related techniques such
as low-energy electron microscopy (LEEM) and mir-
ror electron microscopy (MEM) offer even higher lat-
eral resolution but, to date, have found few applica-
tions [71, 721. Field emission and, in particular, field
ion microscopy (FEM and FIM) [ 18, 191 enable almost
atomic resolution to be achieved, but these techniques
are confined to very fine tips (<1 pm) where the ex-
tension of the individual crystal planes is far below the
mbar, pco = 3.6 x mbar [73].
typical length scale of the reaction-diffusion patterns
described.
The principle of PEEM is based on the differing di-
pole moments of adsorbate complexes which give rise
to modification of the local work function. The yield of
photoelectrons emitted from a surface under the influ-
ence of ultraviolet irradiation is thus determined by the
type and concentration of adsorbed species, and the
lateral intensity distribution of these photoelectrons is
imaged through a system of electrostatic lenses onto
a channel plate and a fluorescent screen. From there,
the PEEM images are recorded by means of a CCD
(charge coupled device) camera and stored on video
tape. Typical resolutions are 0.2 pm and 20 ms, respec-
tively, the latter being determined by the video fre-
quency. Since 0 atoms chemisorbed on Pt(ll0) cause a
larger increase of the work function than adsorbed CO,
regions predominantly covered by oxygen will appear
dark in the images, while those on which CO prevails
are grey.
Even for those conditions under which the integral
reaction rate is constant, the composition of the surface
is not necessarily uniform, but may exhibit-frequently
transient-propagating concentration patterns com-
monly known as chemical waves [16].
The CO + Oz/Pt( 110) system exhibits, for example,
parameter ranges in which both 0 and CO wave fronts
are found to coexist and to propagate, as shown in Fig.
6 [73]. This situation is called “double metastability” or
“dynamic bistability” and is also reproduced in model
calculations [68]. The elliptical, rather than circular,
 5.2 Microkinetics 1039

Figure 7. PEEM images from a Pt(ll0) surface during CO oxi- Figure 8. Theoretical modeling of the temporal evolution of a
dation exhibiting the formation of spirals: interval between con- turbulent state from spiral patterns [75].
secutive images 30 s, T = 448 K, PO, = 4 x mbar, pc- =
4.3 x 10-5mbar [73].

shape of the growing islands is a consequence of the

anisotropy of the diffusion coefficient for adsorbed CO,
which is larger along the (110) direction than along the
perpendicular (001) direction [74]. As long as two con-
secutive fronts propagate with significantly different
velocities, the faster eventually overruns the slower
one. However, if the velocities become comparable,
pulses with stable width are formed, which in a two-
dimensional medium frequently develop into spirals.
Such a situation is shown in Fig. 7. Remarkably,
under identical external conditions, not all spirals exhibit
the same periods and wavelengths, but rather continuous
distributions. This has to be attributed to pinning of some
of the spiral cores to surface defects of varying size and
kinetic properties as reproduced by theoretical simu-
lations [68]. Extended ( 210 pm)defects may even form
the cores of multiarmed spirals [73].
Detailed theoretical modeling revealed that with the Figure 9. Experimental verification of chemical turbulence with
present system under certain conditions the spirals are the CO oxidation at a Pt(ll0) surface [76].
migrating (meandering). Eventually, the Doppler effect
becomes so pronounced that the spiral arms break up
and cause a transition to irregular patterns such as namely the formation of solitary pulses which prop-
shown in Fig. 8 [75] (chemical turbulence), quite sim- agate with constant velocity of about 3 pm/s along the
ilar to experimental observations (Fig. 9) [76]. Al- (001) direction as reproduced in Fig. 10 [77]. Upon
though not yet explored in detail, it is very likely that collision of two pulses these usually annihilate each
the situations of temporal chaos and hyperchaos de- other, as expected for reaction-diffusion systems, but
scribed above are to be attributed to interacting spatial sometimes one or even both pulses reemerged from the
and temporal instabilities. collision. The latter effect is denoted as soliton-like be-
The anisotropy of the surface diffusion with the
present system also gives rise to a novel phenomenon, References see page 1049
1040 5 Elementarv Stem and Mechanisms
Figure 10. PEEM images, recorded at an interval of 3s, from
a Pt(ll0) surface during CO oxidation at T = 485 K. PO, =
3.5 x 10-4mbar, pco = 1 x 10-4mbar. Dark pulses with in-
creased oxygen coverage propagate as solitary waves along the
(001) direction as indicated by arrows [77].
havior. Moreover, occasionally a pulse was observed
to emit another one in the opposite direction (wave
splitting). Since the external parameters were identical
everywhere, these unexpected effects have again to be
attributed to the local defect structure of the surface.
Under the assumption that the local 1 x 1 + 1 x 2
reconstruction is not perfect everywhere (and conse-
quently the oxygen sticking coefficient also varies),
Figure 11. Theoretical (one-dimensional) profiles of solitary waves propagating along the I direction with time where L> denotes the con-
centration of 0 atoms causing the dark pulses in the PEEM images of Fig. 10: (a) collision of two pulses propagating in opposite directions
in a defect region causes soliton-like behavior; (b) if a single pulse enters an appropriate defect zone, two pulses propagating in opposite
directions are created (wave splitting) [78].
all these phenomena could readily be reproduced by
theory, as for example shown in Fig. 11 for the effects
of soliton and wave splitting [78].
Parameter ranges for which oscillations of the in-
tegral reaction rate occur are characterized by global
coupling of different parts via the gas phase [79],
mainly by small variations of the CO pressure, since
oxygen is present in large excess under oscillatory con-
ditions. As a consequence of the asymmetric inhibition
of adsorption, a slight decrease ofpco (due to enhanced
reactivity) increases the probability for oxygen adsorp-
tion and hence causes a positive feedback loop and
favors synchronization. Frequently, the whole surface
oscillates uniformly in phase. At lower temperatures
(around 430 K) elliptical target patterns on an oscilla-
tory background are also observed (Fig. 12) [76], which
could be modeled theoretically by assuming defects act-
ing as pacemakers [80]. At high temperatures ( > 540 K),
standing waves are also formed which take the form of
stripes of rhombic cells [76], which again could be mod-
eled by taking gas-phase coupling into account [8 11.
In summary, the rich variety of phenomena asso-
ciated with catalytic CO oxidation on Pt( 110) single-
crystal surfaces under isothermal, low-pressure con-
ditions has been widely explored experimentally and
largely understood theoretically.
5.2.8.4 Oxidation of Carbon Monoxide on Other
Surfaces
Carbon monoxide oxidation has also been extensively
studied with other catalysts. While the Langmuir-

Figure 12. A sequence of PEEM images (200 x 300pm2) re-
corded at intervals of 4.1 s (30 s between the last two images) from
a Pt(l10) surface during CO oxidation showing the evolution of
so-called target patterns: T = 427 K, PO, = 3.2 x mbar,
pco = 3.0 x 10-5mbar [76].
Hinshelwood reaction scheme formulated above is
generally valid, there exist quite some differences both
with respect to the observed phenomena as well as
with the microscopic mechanisms causing kinetic
instabilities.
Prior to Pt(1 lo), the behavior of the Pt(100) single-
crystal surface was investigated in quite some detail
[49, 57, 82-85]. Again, a reconstruction mechanism
similar to that of Pt(ll0) is in operation. The clean
Pt( 100) surface exhibits a quasihexagonal (hex) config-
uration of surface atoms which is transformed under
the influence of adsorbed CO into the normal 1 x 1
configuration [86-901. However, in contrast to Pt(1 lo),
there is now a much larger difference in the oxygen
sticking coefficient, namely about 0.3 on the 1 x 1
phase and < l o p 3 on the hex surface [91, 921. As a
consequence, the parameter range for the occurrence of
kinetic oscillations is much wider than with Pt(ll0)
[79], and the global coupling mechanism through var-
iations in the partial pressures is less efficient. This
effect of imperfect lateral coupling causes the rate
oscillations as well as the associated spatio-temporal
concentration patterns to become much less regular
than with Pt(1 lo).
Even on the latter surface, however, the situation
may become even more complicated. Under certain
conditions, the structure of the flat surface transforms
into a sequence of steps and terraces (facets), which
process is associated with a continuous variation of the
reactivity of the type as displayed in Fig. 1 [93]. In
5.2 Microkinetics 1041
- 50 nm
Figure 13. Scanning tunneling microscopy (STM) from a
Pt(210) surface after prolonged CO oxidation at T = 480K,
PO, = 2.0 x mbar, pco = 6.7 x mbar, demonstrating
the formation of facets preferentially oriented along the (001)
direction with average mutual spacings around 15 nm [95].
the absence of a catalytic reaction, thermally activated
reordering takes place, and the initial situation is re-
stored. The increase in reactivity has to be attributed
to an enhancement of the oxygen sticking coefficient,
as verified in experiments with a cylindrical Pt single-
crystal sample exhibiting all surface orientations of the
(001) zone. There the (210) plane was found to exhibit
the highest value of the 0 2 sticking coefficient [94]. In-
deed, the latter plane was also found to exhibit kinetic
oscillations [59, 941. Since the clean Pt(210) surface is
not reconstructed, the simple reconstruction model
at first seemed to fail to account for the oscillatory
kinetics. However, it was found that this plane indeed
facets under reaction conditions, preferably into (1 10)
and (310) orientations. On the (110) (and related) mi-
crofacets then, the same mechanism is operating as
with the respective extended single-crystal surfaces [60].
Figure 13 displays an STM image from a Pt(210) sur-
face after prolonged reaction where the facet formation
is directly discernible [95].
Another, principally different mechanism comes
into play, however, with CO oxidation on palladium
at low pressures, and even occurs with Pt catalysts at
higher pressures. Kinetic oscillations have been found
with Pd(ll0) and (1 11) single-crystal surfaces at total
pressures exceeding about mbar with a large ex-
cess of oxygen in the gas phase [96-991. As outlined
References see page 1049
1042 5 Elementary Steps and Mechanisms
above, with Pt the normal Langmuir-Hinshelwood
mechanism causes the appearance of a clockwise (cw)
hysteresis in the CO2 production rate if the CO pres-
sure is continuously increased and decreased again.
With Pd such a cw hysteresis loop is also observed at
low po2, but it changes into counter-clockwise (ccw)
upon increasing the 0 2 pressure [97]. One therefore
obtains a cross-shaped stability diagram by plotting the
transition points of the cw/ccw hysteresis on a poZ/pco
diagram [99], and the crossing point marks the lower
limit of poZ for the occurrence of oscillations. This
mechanism has to be attributed to the participation of
another species, namely oxygen atoms which may be
dissolved below the surface (i.e. subsurface oxygen)
[97]. If the surface is largely covered with oxygen, 0
atoms will start to penetrate below the surface whereby
the surface itself becomes less active, i.e. the oxygen
sticking coefficient is reduced. As a consequence, the
surface becomes predominantly covered by CO, and
now subsurface 0 atoms are segregating back to the
surface from their reservoir until the initial situation is
reestablished and one oscillation cycle is completed.
Inclusion of such a subsurface species into the reaction
model yielded satisfactory theoretical description of the
observed kinetic phenomena [loo, 1011.
The existence of subsurface oxygen at Pd surfaces
manifests itself mainly through the inverse dipole mo-
ment which leads to a lowering of the work function
Ap, instead of the increase usually caused by negatively
charged 0 atoms on the surface [102, 1031.
A similar subsurface species may also be formed
on Pt(100) and (110) surfaces under low pressure con-
ditions where the work function may become even
lower (by up to l e v ) than that of the clean surface
which effect causes the appearance of very bright spots
in the PEEM images [104, 1051. However, it does not
play a major role with the kinetic instabilities at low
pressures and does, hence, not affect the validity of the
reconstruction model outlined above.
For platinum, the situation may also become quite
different at higher oxygen pressures. It is known that
for poZ 2 1 mbar Pt may form oxides, and the time-
scale on which such oxides are reduced by CO were
found to be similar to the period of rate oscillations
[106]. Therefore, an oxide model was proposed [lo71
which assumes that part of the active surface covered
by chemisorbed oxygen atoms is transformed into an
inactive oxide state. This oxide may then be slowly re-
duced by CO whereby the initial active state of metallic
Pt surface is restored. Experimental verification of this
mechanism was sought in Fourier transform infrared
spectroscopy (FTIR) experiments which were, how-
ever, not very conclusive [108, 1091. Indirect proof was
obtained by solid-state potentiometry which experi-
ments demonstrated that the conditions under which
rate oscillations occur coincide with those for the for-
mation of platinum oxides [110]. Quite recently, the
application of in-situ X-ray diffraction with supported
Pt catalysts revealed even more direct evidence for the
operation of the oxide mechanism. From analysis of
angular diffraction profiles it was concluded that the
rate oscillations are associated with periodic oxidation
and reduction of PtO and Pt304, reaching a maximum
degree of oxidation of about 20-30% [lll]. It is felt
that the oxide mechanism is prevailing in CO oxidation
on Pt catalysts, whenever the experiments are per-
formed near atmospheric pressure, even if single-crystal
samples are used [112].
Still another mechanism, the carbon model, was
proposed in order to explain kinetic instabilities in CO
oxidation [113, 1141 in which it was assumed that de-
activation of the surface by buildup of C atoms is fol-
lowed by their oxidative removal. So far, however, no
convincing evidence for the operation of this mecha-
nism can be offered.
It should be mentioned in this context, that with CO
oxidation on polycrystalline Pt catalysts near atmo-
spheric pressure conditions, not only more or less
regular periodic rate oscillations were found, but that
chaotic temporal behavior was also identified in a
number of studies [115-1181.
5.2.8.5 Other Isothermal Systems with Oscillatory
Kinetics
Apart from CO oxidation, quite a number of other
catalytic systems was found to exhibit rate oscillations
and other kinetic instabilities which have not primarily
to be attributed to thermokinetic effects but to the un-
derlying surface chemistry. Among these, reactions of
NO with either CO, H2, or NH3 on a Pt(100) single-
crystal surface have been explored in most detail and
will hence be described here to some extent. These re-
actions have a basic feature in common, namely an
autocatalytic step associated with the dissociation of
adsorbed NO.
+
With the reaction NO CO --f 4N2 + CO, kinetic
oscillations were observed with a polycrystalline Pt
ribbon in the low4mbar pressure range [119], following
a short report on such effects with Pt(100) at extremely
low pressures ( x mbar) [120]. Subsequently, this
reaction became subject of extended investigations with
Pt(100) in the lop7 to lop5mbar pressure regime [121-
1271. No oscillations were observed on Pt(ll1) and
Pt(llO), which planes were found to be much less re-
active in dissociating NO, in accordance with findings
made with the cylindrical Pt single-crystal sample [ 1251.
With Pt(100) kinetic oscillations occur over a wide
range of the p ~ 0 : p c oratio within two temperature
windows. In the low temperature (1430 K) range the
oscillations occur on a pure 1 x 1 substrate. They are

3.0
I ,
T [KI
Figure 14. Bifurcation diagram for the NO + CO reaction on
Pt(100) for fixed pco NO = 4 x mbar as a function of
temperature T in the upper temperature window. Open squares
denote the amplitudes in the oscillatory regime, while full squares
represent stationary rates. The inset shows the Feigenbaum sce-
nario with the sequence of period doublings ( P I to p ~ and
) tran-
sition to chaos (c) [126].
not sustained, but may be initiated by a small temper-
ature jump after which they decay within a few cycles
[123]. This effect is due to insufficient lateral coupling
as was verified by imaging spatio-temporal concen-
tration patterns by PEEM [127]. Periodic modulation
of the temperature with small amplitude ( x 1-3 K)
suffices to create sustained oscillations [124]. In the
upper temperature range ( 2450 K) the oscillations are
not damped and occur on a laterally uniform surface
whose structure switches periodically between 1 x 1
and hex, similar to the CO oxidation reaction. The
transitions to steady-state rates at both limits of the
high temperature range are illustrated by the bifurca-
tion diagram reproduced in Fig. 14. At the upper limit,
oscillations develop via a Feigenbaum scenario from
small amplitude chaotic oscillations, while at the lower
limit the transition to stationary kinetics is discontin-
uous and associated with the onset of spatio-temporal
pattern formation [126, 1271.
+
The mechanism of the NO CO reaction (with for-
mation of N20 in a side reaction being neglected) in-
volves dissociation of adsorbed NO as rate-limiting
Step [123]. This process, NOad + * + N,d Oad, re- + terns and spiral waves. This effect has been attributed
quires (formally) an additional vacant adsorption site
(*). In the subsequent steps of product formation,
+
namely 2N,d + N2 2* and toad o a d + C02 23, +
more free sites are created than are consumed by the
initiating process of NO dissociation. Hence, autoca-
talysis with respect to the formation of free sites is the
consequence which explains the occurrence of a surface
explosion, i.e. of extremely narrow peaks, in thermal
desorption spectroscopy experiments with NO CO
coadsorbed on Pt(100) [128, 1291. This autocatalytic
process is also responsible for the occurrence of kinetic
oscillations without involving a structural transfor-
mation of the surface as demonstrated by proper theo-
5.2 Microkinetics 1043
retical modeling [123, 1301. It therefore accounts for
the experimental observations in the low-temperature
window.
The oscillations in the high-temperature range are
associated with the adsorbate-driven hex .e 1 x 1
transformation of the surface structure. A qualitative
theoretical description could be achieved by including
I this step in the model [123], which was recently ex-
tended further by incorporating more accurately the
growth kinetics of the 1 x 1 phase triggered by lifting
the hex reconstruction [ 1311. Even then, however, the
agreement between theory and experiment is not com-
pletely satisfactory, most probably due to the presence
of surface defects which play a significant role in the
dissociation of NO [123] and may be modified by the
reaction [132].
Reduction of NO by either H2 or NH3 does not
proceed along single pathways but through branching
channels whereby the aspect of selectivity also comes
+
into play. In the NO H2 reaction on Pt(100) both N2
and NH3 are formed (apart from a minor yield of
N20) [133, 1341 whereby chaotic kinetics were also
identified [135]. In the NO + NH3 reaction the main
products are N2 and N20, and oscillatory kinetics were
found both with polycrystalline platinum at higher
pressures [136, 1371 as well as in the lo-’ to lO-’mbar
pressure range with Pt(100) [138, 1391. The dynamic
behavior of both systems is nearly identical, and the
mechanism is obviously rather similar to that in the
NO + CO reaction. Consequently, successful mathe-
matical modeling could be achieved on the basis of
an analogous reaction scheme [140], and the participa-
tion of the hex + 1 x 1 structural transformation was
demonstrated experimentally [138, 1411.
NO is much more readily dissociated on Rh than on
Pt, and oscillatory kinetics with this catalyst material
+
were again found with the NO H2 and NO NH3 +
reactions with field electron microscopy [18, 1421 as
well as with a Rh(l10) single-crystal surface [143]. In
the latter study, by applying PEEM, quite unusual
spatio-temporal patterns of the type shown in Fig. 15
were detected, namely rectangularly shaped target pat-
to an anisotropy of the surface diffusion which, in ad-
dition, is influenced by adsorbate-induced reconstruc-
+ tion effects [144].
The H2 + O2 reaction on Pt deserves brief mention-
ing in this connection, since kinetic oscillations at high
pressures had been reported quite early for this system
[9] and were later investigated again in quite some de-
tail [145]. Quite recently, the heat generated with this
+ reaction at a pressure of 10-2mbar was directly de-
termined experimentally [146]. Obviously, the quoted
studies were not conducted under isothermal condi-
References see page 1049
1044 5 Elementary Steps and Mechanisms

Figure 15. Concentration patterns associated with the NO + H2 reaction on a Rh(l10) surface: (a) elliptical target patterns, T = 427 K,
PNO = 1.6 x 10-6mbar, PH> = 1.8 x mbar; (b) nucleation of square-shaped target patterns, t = 595 K, PNO = 1.6 x 10-6mbar,
pH2= 5 x mbar; (c) same conditions as (b) but 60 s later [143].

tions so that heat transfer comes into play which effect

does, however, not account for the observed time scales
of oscillations. However, the experimental findings
could be satisfactorily modeled on the basis of an oxi-
dation-reduction mechanism similar to that proposed
for CO oxidation [147].
No kinetic oscillations could be observed with this
reaction on an extended Pt( 100) single-crystal surface
at low pressures, but merely simple bistable behavior.
Interestingly, if a field emitter tip was used instead on
which the (100) plane of only about 50nm diameter
was adjacent to other crystal planes, oscillatory phe-
nomena occurred [148]. It is felt that this effect is of
more general relevance for the catalytic properties
of small particles and it is therefore discussed further
below.
Figure 16. Space-time diagram of the temperature profile of a
5.2.8.6 Thermokinetic Phenomena 1 cm long segment of a Pt ribbon during oxidation of propylene.
While the average temperature is kept constant, a local temper-
ature front moves back and forth [162].
Due to the pronounced temperature dependence of
chemical rate processes, even small deviations from
isothermal conditios are expected to be of strong influ-
ence on the dynamic behavior. Such effects are very which was solely heated by the enthalpy released by the
likely to occur in any studies with “real” (either mon- reaction. It turned out that the oscillatory behavior
olithic or supported) catalysts conducted near atmo- changed substantially if the wire loop was cut into
spheric pressure, where frequently temperature changes two parts, again demonstrating the importance of heat
of several tens of degrees or even more are found. transport in such systems [157].
Local variations of temperature can most con- Systematic studies on wave propagation in electri-
veniently be recorded by infraed thermography [ 1491, cally heated wires have been reported for various re-
and numerous reports on such phenomena associated actions, including oxidation of ammonia and hydro
with catalytic CO oxidation on supported catalysts can carbons [149, 158-1641 as well as the endothermic de-
be found in the literature [150-155]. Among others, composition of CH3NH2 on a Pt wire [165]. Stationary
these experiments demonstrated the importance of (Turing) patterns as well as moving fronts were ob-
thermal coupling between different catalyst pellets served, the latter being associated with macroscopic
whose variation affects the degree of synchronization rate oscillations. As an example, Fig. 16 shows the
of the rate oscillations [154, 1551. By using a thin irregular motion of fronts associated with oxidation of
evaporated Pt film, the propagation of reaction waves propylene on a Pt ribbon which was heated electrically
during CO oxidation could be studied [156]. In another and maintained at constant total resistance whereby
experiment with the same reaction, a Pt wire was used the integral kinetics became chaotic (Fig. 17) [162].

.22 1 -I
1 I I , I , I I
0 500 1000 1500 2000 2500
t [secl
Figure 17. Chaotic oscillations of the total heat generated by
propylene oxidation on an electrically heated platinum ribbon
while the electrical resistance is kept constant [162].
Instead of keeping the resistance (= average temper-
ature) constant, similar experiments were performed
under different constraints, in which either the voltage
or the heating current [161] was kept constant, and
the resulting dynamic properties were found to differ
significantly. Quite generally, such conditions represent
global constraints and may lead to spatio-temporal
pattern formation as a consequence of symmetry
breaking [ 166, 1671.
However, similar types of patterns may also be
formed without external constraints, such as those ob-
served with the oxidation of hydrogen on a Ni ring
without additional electrical heating [ 168, 1691. Under
certain conditions, a pulse rotating around the Ni ring
with a period of about 10min was observed. The in-
tegral behavior exhibited oscillations with three max-
ima and minima per rotation period. This effect dem-
onstrates that the rotating pulse changes its shape
periodically which is most likely due to nonuniformities
of the catalyst.
If two-dimensional (e.g. foils) rather than one-
dimensional (rings or wires) catalysts are used, the re-
sulting spatio-temporal patterns are usually still more
complex, as demonstrated with the HZ 0 2 reaction +
on Pt [170, 1711 and Ni [160, 1721.
Various approaches can be found in the literature to
implement nonisothermal effects in the theoretical
modeling of the (integral) kinetics. For example, a
simple mechanism for thermokinetic oscillations is
possible in which catalytic sites are blocked at low
temperature and reactivated at higher temperature.
Such schemes have been analyzed in general form [ 173,
1741 and applied to various oscillatory reactions such
as NO + CO on Pd [175], H2 + 0 2 on Pt [53] and, in
particular, CO + 0 2 on Pt [176-1781 as well as with
the endothermic decomposition of CH3NH2 mentioned
5.2 Microkinetics 1045
above [165]. The basic mechanism of the kinetic oscil-
lations in this latter system (exhibiting temperature
amplitudes of up to 500K) is in fact rather simple. In
the reactive state the temperature drops due to the
endothermicity, and at low enough temperature the
surface is then blocked by an inhibiting species. As a
consequence, the continuing electrical heating causes
the temperature to rise again until the inhibiting species
desorbs and the high reactivity is restored.
Although with the systems discussed in this section
the details of the reaction mechanism and of the ki-
netics of the elementary steps is much less understood
than for the well-defined single-crystal systems de-
scribed in the preceding sections, in these cases the
basic features of spatio-temporal pattern formation
could also be modeled successfully. This is because the
decisive effects are thermokinetic in nature and can be
approximated by a heat-balance equation in which the
chemistry of the reaction is reduced to a single variable
and the surface diffusion of the adsorbates is neglected
[ 179-1 841. Coupling through the gas phase may also be
mostly neglected, although its inclusion may give rise
to quantitative changes [ 1791. The basic equations take
the general (dimensionless) form
(5)
The termf( T ,8, I ) denotes the heat balance, QR is the
heat generated by the reaction, Q E ~is the heat ex-
change with the surroundings, and QH(I) is the energy
input by electric current I , if applied. The term g(8. T )
represents the surface chemistry of the catalytic re-
action, with 8 being a characteristic concentration
variable. Typically, a species with concentration 8 is
consumed by the decisive reaction step and is produced
in another process.
Various types of solutions were extensively discussed
in the quoted literature to which the interested reader is
referred to. It is just mentioned that for certain param-
eter ranges many stable solutions were found to coex-
ist, depending on the initial conditions. That means
that even for identical steady-state control parameters
the dynamic behavior may differ considerably, which
renders this field rather complex.
5.2.8.7 Some Consequences and Future Prospects
The phenomena associated with the nonlinear dynam-
ics of catalytic reactions demonstrated that the true
situation may be far more complex than expected on
References see page 1049
1046 5 Elementary Steps and Mechanisms

the basis of conventional kinetic considerations. This

may also have implications for other areas of catalysis,
a few of which will be outlined by means of recent
experimental developments.

A Periodic Forcing
Instead of adjusting the control parameters in a way
that sustained rate oscillations develop, a system may
also be subject to periodic modulation of one of these
external parameters. The goal of increasing the yield 1

and (what would be even more important) the selec- 100 200
t Is1
tivity in this way has been achieved in several reports
on “real” catalytic systems [13, 185-1891. The response
of a system exhibiting autonomous oscillations with
period TOto external forcing with period T,, and am-
plitude A may be classified in the following way [190- - 50
>
1921. If the response of the system with period T, oc- LE 40
curs with fixed phase relation to the modulation, the 2 30
system is entrained, and the ratio between T, and T,, 20
may be expressed as that between two small integer
10
numbers, T,/Tex= k / l . With k / l = 1, the entrainment
is harmonic, for k / l > 1 it is superharmonic, with
k / l < 1 it is subharmonic. If the phase difference be- tlsl
tween response and modulation is continuously vary-
4.05
ing, the oscillations are denoted as quasiperiodic. The N

a
0

behavior of such a system may be rationalized by a 3.95

“dynamic phase diagram” in which the regions of
entrainment and quasiperiodicity are marked as a
function of TeX/Toand A . The structure of this dia-
gram does not depend on the specific features of the
reaction system, but only on its type of bifurcation
around which the perturbation occurs [ 191, 1921.
Single-crystal experiments on periodic forcing have
been performed with the CO oxidation on Pt( 100) [ 1931 50 100 150 200
and Pt(ll0) [21]. As an example, Fig. 18 shows a se- t 151
quence of time series for entrained and quasiperiodic
oscillations recorded with the latter system. The result-
ing experimental phase diagram (Fig. 19(a)) is in good
agreement with the results of a theoretical treatment
of the system on the basis of the mechanism outlined
above [194] (Fig. 19(b)).

B Uniform Modification of the Surface Composition

It is common practice to alter the properties of a cata-
lyst by uniform modification of its chemical com-
position, e.g. by alloying. The resulting effects on the
catalytic properties are usually discussed in terms of
local (i.e. atomistic) concepts, such as the “ligand” or 1
100 200 300 400 500
“ensemble” effects [195] (see this Handbook Part A, t Is1
Chapter 5.3.3) by which the modification of the elec-
Figure 18. Time series for periodically forced oscillations in the
tronic properties of a surface atom by its neighbors or CO oxidation on a Pt(ll0) surface exhibiting sustained oscil-
the specific configuration of surface atoms forming the lations with frequency vo. Modulation of PO, with frequency vp
active site are characterized. Quite different phenom- and amplitude A (as percentage of the basic value): (a) 1 : 2 sub-
ena on mesoscopic length scales may come into play harmonic entrainment, YO = 0.09s-l, vp = 0.18s-’, A = 1.2%; (b)
with systems exhibiting characteristics of nonlinear dy- 2: 1 superharmonic entrainment, vo = 0.25 s-’ , vp = 0.1 1 s-’ ,
A = 1.2%; (c) 7 : 2 superharmonic entrainment, vo = 0.185 S K I ,
namics such as spatio-temporal pattern formation. vp = 0.047s-’, A = 1.2%; (d) quasiperiodic behavior, vo =
These features are governed by kinetic parameters 0 . 2 5 ~ - ’ vp
, = 0.016~-’,A = 1.2% [21].

ns
At% -..-pd ,
I
I
0
112 2/3 1 L / 3 3/2 513 2
Figure 19. Dynamic phase diagram for periodically forced os-
cillations in the CO oxidation on a Pt(ll0) surface. Existence
range for entrained and quasiperiodic oscillations as a function of
the period ratio of modulation and autonomous oscillations,
i“, : To, and of the amplitude (as percentage of the base pressure)
of the modulated 0 2 partial pressure. (a) Experimental results,
where shaded areas indicate conditions for quasiperiodic behavior
between the sub- and superharmonic entrainment bands [21]. (b)
Theoretical result, obtained with the quoted system of differential
equations modeling the kinetics. Types of bifurcation: ns = Nei-
mark-Sacker, pd = period doubling, snp = saddle node [194].
(such as rate constants, diffusion coefficients etc.) uuer-
aged over length scales typical for the resulting con-
centration patterns, i.e. of the order of about 1 pm, and
may already be substantially altered by rather small
amounts of foreign atoms. As an example, Fig. 20
reproduces a PEEM image reflecting spatio-temporal
pattern formation on a Pt(ll0) sample subject to
steady-state CO oxidation where a part of the surface
was uniformly covered by about 5% of a monolayer
of Au atoms, while the other part was pure platinum
[196]. Both the phenomenology as well as the dynamics
(velocity of wave propagation) are obviously signif-
icantly affected. The resulting effects on the overall
kinetic properties have still to be explored, but it will
certainly not be possible to describe them properly in
terms of the quoted atomistic concepts.
C Surfaces with Preformed Mesoscopic Structures
Instead of modifying a catalytic surface homogene-
ously, e.g. by uniform deposition of another material,
5.2 Microkinetics 1047
,? b
, I
I
,
/
Figure 20. PEEM image from a Pt(l10) surface on which the left
part was covered by 5% of a monolayer of Au during CO oxida-
tion: T = 460 K, po2 = 4 x mbar, pco = 4.6 x mbar
[196].
the next step consists of prefabricating heterogeneous
structures with appropriate dimensions. It is generally
known that formation and propagation of nonlinear
waves may be markedly affected by the spatial bound-
ary conditions in that, for example, certain modes are
selected, or the propagation through narrow channels
is suppressed etc. [197-1991. Photolithographic tech-
niques, as developed for microelectronics, offer a con-
venient way to prepare surfaces with such micro-
structures. The first experiments of this type were
performed with a Pt(ll0) surface, onto which a tita-
nium mask was deposited. The latter oxidizes and is
then completely inert in the CO oxidation reaction
which is hence restricted to the bare Pt areas [200]. A
PEEM image from such a surface exhibiting formation
of concentration patterns is displayed in Fig. 21.
Clearly in this way novel dynamic effects develop
which would not exist with uniform surfaces.
D Phenomena on the Atomic Scale
The small particles of a “real” supported catalyst usu-
ally have dimensions of only a few nm, much smaller
than the typical patterns found with isothermal re-
action-diffusion systems on extended uniform surfaces.
Typical nonlinear effects with such systems are, how-
References see page 1049
1048 5 Elementary Steps and Mechanisms
Figure 21. Pattern formation on a microstructured Pt(ll0) sur-
face during CO oxidation. Two of the patches are in uniform states
while on the H spirals and propagating pulses develop [200].
ever, of thermokinetic origin, for which only the prop-
erties averaged over macroscopic (= 1 mm) length scale
are of significance, as outlined above.
Quite different problems come into play, however, if
processes occurring near the atomic scale become sub-
ject of closer inspection. On the theoretical side, for-
mulation of kinetics in the framework of a mean-field
description in terms of differential equations becomes
questionable. Such an approximation is per se only
justified if the mobility of the adparticles is sufficiently
high and if interactions between them may be ne-
glected. An alternative approach consists of performing
lattice-gas-type computer simulations [201], among
which models for CO oxidation have attracted par-
ticular attention [202-2041. Further progress in the de-
Figure 22. Series of field ion microscopy (FIM) images from a Pt tip during the Hz + 02 reaction at 3 0 0 K : p ~ ,= 6 x 10-4mbar,
po2 = 5 x mbar. Imaging gases are 0 2 and H20 formed by the reaction [148].
velopment of kinetic lattice-gas models, applicable also
to conditions far from equilibrium, is to be expected in
the near future.
The experimental situation has been significantly
affected by the possibility of observing nonlinear phe-
nomena with near-atomic resolution by the application
of field emission and field ion microscopic techniques
(FEM and FIM) [18, 19, 205, 2061 (see this Handbook
Part A, Chapter 5.3.5). Even if well-developed patterns
such as spirals may not develop for geometric reasons
on the small tips used in these experiments, propagat-
ing reaction-diffusion waves may become clearly dis-
cernible. Usually, their interfaces are much sharper
than expected for reaction-diffusion systems of non-
interacting particles, and this effect is most likely at-
tributed to , the operation of attractive interactions
by which mean-field descriptions become invalid, and
which may indeed lead to the formation of very sharp
interfaces and even to “uphill” diffusion opposite to an
existing concentration gradient [207].
A quite interesting effect was observed with the
H2 + 0 2 reaction on Pt [148]. As outlined in Section
5.2.8.4, on an extended Pt(100) surface only bistable
behavior is found, whereas if the same plane is exposed
as small area on a field emitter tip under certain con-
ditions oscillatory behavior is observed. This is illus-
trated by the sequence of images reproduced in Fig. 22.
The central Pt( 100) region becomes periodically acti-
vated by reaction-diffusion waves triggered from ad-
jacent higher index planes, and as a consequence also
the time-averaged reactivity of this plane is much
higher than if existing isolated. It is important to dis-
tinguish this effect from the phenomenon of “spill-
over”, well-known in catalysis (see this Handbook Part

A, Chapter 5.3.2) which is just due to diffusion of
reacting species from one part of the surface to the
other [208]. In the present case, it is the coupZing of re-
action and diffusion which causes the propagation of
nonlinear concentration waves under conditions far
from equilibrium and thus affects the overall reactivity.
It is felt that such phenomena are of general im-
portance for the reactive properties of small supported
catalyst particles whose dimensions are comparable to
those of a field emitter tip. Their activity is commonly
considered to be simply composed by a superposition
of the contributions from the different crystal planes
and other structural elements such as steps and edges.
The described experiments demonstrated that this lin-
ear approach may have to be corrected fundamentally.
Certainly, effects of nonlinear dynamics will govern
widely the kinetics of catalysis, even if there exists no
macroscopic manifestation such as oscillations of the
global reaction rate.
References
1. G. Nicolis, I. Prigogine, Self-Organization in Non-Equilib-
rium Systems, Wiley, New York, 1977.
2. A. S. Mikhailov, Foundations of Synergetics, Springer, Ber-
lin. 1991. Vols I and 11.
3. U.’F. Franck, Ber. Bunsenges. Phys. Chem. 1980,84, 334.
4. A. M. Zhabotinsky, Ber. Bunsenges. Phys. Chem. 1980, 84,
303.
5. R. J. Field, M. Burger (Eds), Oscillations and Travelling
Waves in Chemical Systems, Wiley, New York, 1985.
6. P. Hugo, Ber. Bunsenges. Phys. Chem. 1970, 74, 121.
7. H. Beusch, D. Fieguth, E. Wicke, Chem. Zng. Techn. 1972,
44, 445.
8. M. Sheintuch, R. A. Schmitz, Catal. Rev.-Sci. Eng. 1977,
15, 107.
9. M. G. Slinko, M. M. Slinko, Catal. Rev.-Sci. Eng. 1978, 17,
119.
10. L. F. Razon, R. A. Schmitz, Catal. Rev.-Sci. Eng. 1986,28,
89.
11. G. Ertl, Ado. Catal. 1990, 37, 213.
12. R. Imbihl, Progr. Surf: Sci. 1993, 44, 185.
13. F. Schiith, B. E. Henry, L. D. Schmidt, Adv. Catal. 1993,
39, 51.
14. M. M. Slinko, N. Jaeger, Oscillating Heterogeneous Cata-
lytic Systems, Elsevier, Amsterdam, 1994.
15. R. Imbihl, G. Ertl, Chem. Rev., 1995, 95, 697.
16. M. Eiswirth, G. Ertl in Chemical Waves and Patterns (Eds:
R. Kapral, K. Showalter), Kluwer, Dordrecht, 1994, 447.
17. J. Lauterbach, G. Haas, H. H. Rotermund, G. Ertl, Surf:
Sci. 1993,294, 116.
18. M. F. H. van Tol, A. Gielbert, B. E. Nieuwenhuys, Catal.
Lett. 1992, 16, 297.
19. V. Gorodetskii, J. H. Block, W. Drachsel, M. Ehsasi, Appl.
Surf:Sci. 1993, 67, 198.
20. M. Ehsasi, 0. Frank, J. H. Block, K. Christmann, Chem.
Phys. Lett. 1990, 165, 115.
21. M. Eiswirth, G. Ertl, Phys. Rev. Lett. 1988, 60, 1526.
22. J. Guckenheimer, P. Holmes, Nonlinear Dynamics, Dynam-
ical Systems, and Bifurcations of Vector Fields, Springer,
Berlin, 1986.
5.2 Microkinetics 1049
23. J. M. T. Thompson, H. B. Stewart, Nonlinear Dynamics and
Chaos, Wiley, New York, 1987.
24. M. Eiswirth, A. Freund, J. Ross, J. Phys. Chem. 1991, 95,
1294.
25. M. Eiswirth, A. Freund, J. Ross, Adv. Chem. Phys. 1991,
80, 127.
26. G. Eigenberger, Chem. Eng. Sci. 1978,33, 1263.
27. V. D. Belyav, M. M. Slinko, V. L. Timoshenko, M. G.
Slinko, Kinet. Katal. 1973, 14, 810.
28. C. A. Pikio, D. Luss, Chem. Eng. Sci. 1977, 32, 191.
29. I. Kevrekidis, L. D. Schmidt, R. Aris, Surf: Sci. 1984, 137,
151.
30. S. S. Tambe, V. R. Kumar, K. N. Ponnani, B. D. Kulkarni,
Chem. Eng. J . 1987,34, 143.
31. D. G. Vlachos, L. D. Schmidt, R . Aris, J. Chem. Phys.
1990, 93, 8306.
32. C. G. Takoudis, L. D. Schmidt, R. Aris, Surf: Sci. 1981,
105, 325.
33. M. A. McKamin, R. Aris, L. D. Schmidt, Proc. Roy. SOC.A
1988,415, 363.
34. A. Winfree, When Time Breaks Down, Princeton University
Press, Princeton, 1987.
35. H. Swinney, V. I. Krinsky (Eds), Wave Patterns in Chemical
and Biological Systems, Physica D 1991, 49.
36. A. V. Holden, M. Markus, H. G. Othmer (Eds), Nonlinear
Wave Processes in Excitable Media, Plenum Press, New
York, 1991.
37. R. Luther, Z. Elektrochem. 1906, 12, 596.
38. A. M. Turing, Phil. Trans. Roy. SOC.B 1952, 237, 37.
39. J. Falta, R. Imbihl, M. Henzler, Phys. Reo. Lett. 1990, 64,
1409.
40. R. Imbihl, A. E. Reynolds, D. Kaletta, Phys. Rev. Lett.
1991, 67, 275.
41. M. Bar, C. Ziilicke, M. Eiswirth, G. Ertl, J. Chem. Phys.
1992, 96, 8595.
42. J. P. Keener, J. J. Tyson, Physica D 1986,21, 300.
43. A. S. Mikhailov, V. S. Zykov, Physica D 1991, 52, 379.
44. M. Sheintuch, Chem. Eng. Sci. 1981,36, 893.
45. F. Mertens, R. Imbihl, A. S. Mikhailov, J. Chem. Phys.
1993,99, 8668.
46. J. L. Lumley, Stochastic Tools in Turbulence, Academic
Press, New York, 1970.
47. M. D. Graham, S. L. Lane, D. Luss, J. Phys. Chem. 1993,
97, 889.
48. K. Krischer, R. Rico-Martinez, I. Kevrekidis, H. H. Roter-
mund, G. Ertl, J. L. Hudson, AZChEJ . 1993,39, 89.
49. G. Ertl, P. R. Norton, J. Riistig, Phys. Rev. Lett. 1982, 49,
177.
50. T. Engel, G. Ertl, Adv. Catal. 1979, 28, I .
51. R. Imbihl, M. Sander, G. Ertl, Surf: Sci. 1988,204, L 701.
52. S. Oh, G. B. Fisher, J. E. Carpenter, D. W. Goodman, J.
Catal. 1986, 100, 360.
53. E. Wicke, P. Kummann, W. Keil, J. Schiefler, Ber. Bun-
senges. Phys. Chem. 1980,84, 315.
54. M. Ehsasi, M. Matloch, 0. Frank, J. H. Block, K. Christ-
mann, F. Rys, W. Hirschwald, J. Chem. Phys. 1989,91,4949.
55. K. Krischer, M. Eiswirth, G. Ertl, J. Chem. Phys. 1992, 96,
9161.
56. V. P. Zhdanov, B. Kasemo, Surf: Sci. Rep. 1994, 20, 111.
57. M. P. Cox, G. Ertl, R. Imbihl, Phys. Reo. Lett. 1985, 54,
1725.
58 M. Eiswirth, G. Ertl, Surf: Sci. 1986, 177, 90.
59 M. Ehsasi, S. Rezaie-Serej, J. H. Block, K. Christmann, J.
Chem. Phys. 1990, 92, 7596.
60. M. Sander, R. Imbihl, G. Ertl, J. Chem. Phys. 1991, 95,
6162.
61. S. R. Bare, P. Hoffmann, D. A. King, Surf: Sci. 1984, 144,
347.
1050 5 Elementary Steps and Mechanisms
62. T. E. Jackman, J. A. Davies, D. P. Jackson, W. N. Unertl,
P. R. Norton, Surf: Sci. 1982, 120, 389.
63. T. Gritsch, D. Coulman, R. J. Behm, G. Ertl, Phys. Rev.
Lett. 1989, 63, 1086.
64. N. Freyer, M. Kiskinova, G. Pirug, H. P. Bonzel, Surf: Sci.
1986, 166, 206.
65. M. Eiswirth, K. Krischer, G. Ertl, Surf: Sci. 1988,202, 565.
66. M. Eiswirth, T. M. Kruel, G. Ertl, F. W. Schneider, Chem.
Phys. Lett. 1992, 193, 305.
67. M. Falcke, M. Bar, H. Engel, M. Eiswirth, J. Chem. Phys.
1992,97,4555.
68. M. Bar, N. Gottschalk, M. Eiswirth, G. Ertl, J. Chem. Phys.
1994,100, 1202.
69. H. H. Rotermund, S . Jakubith, A. von Oertzen, S . Kubala,
G. Ertl, J. Chem. Phys. 1989, 91, 4942.
70. W. Engel, M. Kordesch, H. H. Rotermund, S. Kubala,
A. von Oertzen, Ultramicroscopy 1991, 95, 6162.
71. W. Swiech, B. Rausenberger, W. Engel, A. M. Bradshaw,
E. Zeitler, Surf: Sci. 1993, 294, 297.
72. W. Swiech, C. S . Rastomjee, R. Imbihl, J. W. Evans,
B. Rausenberger, W. Engel, A. K. Schmidt, A. M. Brad-
shaw, E. Zeitler, Surf: Sci. 1994, 307, 138.
73. S . Nettesheim, A. von Oertzen, H. H. Rotermund, G. Ertl,
J. Chem. Phys. 1993, 98, 9977.
74. A. von Oertzen, H. H. Rotermund, S . Nettesheim, Surf: Sci.
1994,31I , 322.
75. M. Bar, M. Eiswirth, Phys. Rev. E 1993, 48, R 1635.
76. S . Jakubith, H. H. Rotermund, W. Engel, A. von Oertzen,
G. Ertl, Phys. Rev. Lett. 1990, 65, 3013.
77. H. H. Rotermund, S . Jakubith, A. von Oertzen, G. Ertl,
Phys. Reu. Lett. 1991, 66, 3083.
78. M. Bar, M. Eiswirth, H. H. Rotermund, G. Ertl, Phys. Rev.
Lett. 1992, 69, 945.
79. M. Eiswirth, P. Moller, K. Wetzl, R. Imbihl, G. Ertl, J.
Chem. Phys. 1988, 90, 510.
80. M. Falcke, H. Engel in Spatio-Temporal Organization in
Nonequilibrium Systems (Eds: S . C. Muller, T. Plesser),
Projektverlag, Dortmund, 1992.
81. H. Levine, X. Zhou, Phys. Rev. E 1993, 48, 50.
82. P. R. Norton, P. E. Bindner, K. Griffiths, T. E. Jackman,
J. A. Davies, J. Riistig, J. Chem. Phys. 1984, 80, 3859.
83. R. Imbihl, M. P. Cox, G. Ertl, J. Chem. Phys. 1986,84, 3519.
84. R. Imbihl, M. P. Cox, G. Ertl, H. Miiller, W. Brenig, J.
Chem. Phys. 1985,83, 1578.
85. P. Moller, K. Wetzl, M. Eiswirth, G. Ertl, J. Chem. Phys.
1986, 85, 5328.
86. R . J. Behm, P. A. Thiel, P. R. Norton, G. Ertl, J. Chem.
Phys. 1983, 78, 7438; 7448.
87. K. Heinz, E. Lang, K. Straws, K. Miiller, Appl. Surf: Sci.
1982, 11/12, 611.
88. P. R. Norton, J. A. Davies, D. K. Geber, C. W. Sitter, T. E.
Jackman, Surf: Sci. 1981,108,205.
89. A. Crossley, D. A. King, Surf: Sci. 1980, 95, 131.
90. A. Hopkinson, J. M. Bradley, X. C. Guo, D. A. King, Phys.
Rev. Lett. 1993, 71, 1597.
91. P. R. Norton, K. Griffiths, P. E. Bindner, SurJ Sci. 1984,
138, 125.
92. X. C. Guo, A. T. Pasteur, D. A. King, Surf: Sci. in press.
93. S . Ladas, R. Imbihl, G. Ertl, Surf: Sci. 1988, 197, 1153; 198,
42.
94. M. Sander, R. Imbihl, G. Ertl, J. Chem. Phys. 1992, 97,
5193.
95. M. Sander, R. Imbihl, R. Schuster, J. V. Barth, G. Ertl,
Surf: Sci. 1992, 281, 159.
96. M. Ehsasi, C. Seidel, H. Ruppender, W. Drachsel, J. H.
Block, K. Christmann, Surf: Sci. 1989,210, L 198.
97. S . Ladas, R. Imbihl, G. Ertl, Surf: Sci. 1989, 219, 88.
98. M. Ehsasi, 0. Frank, J. H. Block, K. Christmann, Chem.
Phys. Lett. 1990, 165, 115.
99. M. Ehsasi, M. Berdau, T. Rebitzki, K. P. Charle, K.
Christmann, J. H. Block, J. Chem. Phys. 1993, 98, 9177.
100. M. R. Bassett, R. Imbihl, J. Chem. Phys. 1990, 93, 811.
101. M. Berdau, M. Ehsasi, A. Karpowicz, W. Engel, K. Christ-
mann, J. H. Block, Vacuum 1994,45, 271.
102. J. Witte, U. Memmert, K. Griffiths, P. R. Norton, J. Chem.
Phys. 1989, 90, 5082; 5088.
103. M. Milun, P. Pervan, M. Vajic, K. Wandelt, Surf: Sci. 1989,
2111212, 887.
104. H. H. Rotermund, J. Lauterbach, G. Haas, Appl. Phys.
1993, A 57, 507.
105. J. Lauterbach, K. Asakura, H. H. Rotermund, Surf: Sci.
1994, 313, 52.
106. J. E. Turner, M. B. Maple, Surf: Sci. 1984, 147, 647.
107. B. C. Sales, J. E. Turner, M. B. Maple, Surf: Sci. 1982, 114,
381.
108. D. J. Kaul, E. E. Wolf, J. Catal. 1985, 91, 216.
109. T. Lindstrom, T. T. Tsotsis, Surf: Sci. 1985, 150, 487.
110. I. V. Yentakakis, C. G. Vayenas, J. Catal. 1989, I l l , 170.
111. N. Hartmann, R. Imbihl, W. Vogel, Catal. Lett. 1994, 28,
373.
112. R. C. Yeates, J. E. Turner, A. J. Gellmann, G. A. Somorjai,
Surf: Sci. 1985, 149, 175.
113. V. A. Burrows, S. Sundaresan, Y. J. Chabal, S . B. Christ-
mann, Surf: Sci. 1987, 180, 110.
114. V. A. Burrows, S . Sundaresan, Y. J. Chabal, S . B. Christ-
mann, Surf: Sci. 1985,160, 122.
115. H. U. Onken, E. Wicke, Ber. Bunsenges. Phys. Chem. 1986,
90, 976.
116. M. M. Slinko, N. I. Jaeger, P. Svensson, J. Catal. 1989,118,
349.
117. L. F. Razon, S . M. Chang, R. A. Schmitz, Chem. Eng. Sci.
1986,41, 1561.
118. J. Kapicka, M. Marek, Surf: Sci. 1989, 222, L 885.
119. W. Adlhoch, H. G. Lintz, T. Weisker, Surf: Sci. 1981, 103,
576.
120. S. P. Singh-Boparai, D. A. King, Proc. 4th Int. Congr. Surf:
Sci., Cannes, 1980, p. 403.
121. S . B. Schwartz, L. D. Schmidt, Surf: Sci. 1987,183, L 269.
122. S . B. Schwartz, L. D. Schmidt, Surf: Sci. 1988,206, 169.
123. T. Fink. J. P. Dath, R. Imbihl, G. Ertl. J. Chem. Phys. 1991,
95, 2109.
124. J. P. Dath, T. Fink, R. Imbihl, G. Ertl, J. Chem. Phys. 1992,
96, 1582.
125. G. Veser, R. Imbihl, J. Chem. Phys. 1992, 96, 7155.
126. G. Veser, F. Mertens, A. S . Mikhailov, R. Imbihl, Phys.
Rev. Lett. 1993, 71, 935.
127. G. Veser, R. Imbihl, J. Chem. Phys. 1994, 100, 8483; 8492.
128. M. W. Lesley, L. D. Schmidt, Surf: Sci. 1985,155, 215.
129. T. Fink, J. P. Dath, M. R. Bassett, R. Imbihl, G. Ertl, Surf:
Sci. 1991, 245, 96.
130. R. Imbihl, T. Fink, K. Krischer, J. Chem. Phys. 1992, 96,
6236.
131. A. Hopkinson, D. A. King, Chem. Phys. 1994,177.
132. Y. Uchida, R. Imbihl, G. Lehmpfuhl, Surf: Sci. 1992, 275,
253.
133. J. Siera, P. Cobden, K. Tanaka, B. E. Nieuwenhuys, Cutul.
Lett. 1991, 10, 335.
134. M. Slinko, T. Fink, T. Loher, H. H. Madden, S. J. Lom-
bardo, R. Imbihl, G. Ertl, Surf: Sci. 1992,264, 157.
135. P. D. Cobden, J. Siera, B. E. Nieuwenhuys, J. Vac. Sci. A
Technol. 1992,10, 2487.
136. C. G. Takoudis, L. D. Schmidt, J. Phys. Chem. 1983, 87,
958; 964.
137. T. Katona, G. A. Somorjai, J. Phys. Chem. 1992, 96, 5465.
138. S. J. Lombardo, F. Esch, R. Imbihl, Surf: Sci. 1992, 271,
L 367.
139. M. F. H. van Tol, J. Siera, P. D. Cobden, B. E. Nieu-
wenhuys, Surf: Sci. 1992,274, 63.

140. S. J. Lombardo, T. Fink, R. Imbihl, J. Chem. Phys. 1993,
98, 5526.
141. G. Veser, F. Esch, R. Imbihl, Catal. Lett. 1992, 13, 371.
142. M. F. H. van Tol, J. de Maaijoe-Gielbert, B. E. Nieu-
wenhuys, Chem. Phys. Lett. 1993,205,207.
143. F. Mertens, R . Imbihl, Nature 1994, 370, 124.
144. N. Gottschalk, F. Mertens, M. Bar, M. Eiswirth, R. Imbihl,
Phys. Rev. Lett. 1994, in press.
145. B. Hellsing, B. Kasemo, V. P. Zhdanov, J. Catal. 1991, 132,
210.
146. J. K. Gimzewskii, C. Gerber, E. Meyer , R. R. Schiller,
Chem. Phys. Lett. 1994,217, 289.
147. V. P. Zhdanov, Surf: Sci. 1993,296,261.
148. V. Gorodetskii, J. Lauterbach, H. H. Rotermund, J. H.
Block, G. Ertl, Nature 1994, 370, 277.
149. V. Barelko, I. I. Kurachka, A. G. Merzhanao, K. G. Shka-
dinskii, Chem. Enq. Sci. 1978, 33, 805.
150. J. C. Kellow, E. E. Wolf, Chem. Eng. Sci. 1990, 45, 2597.
151. F. Quin, E. E. Wolf, A. C. Chang, Phys. Rev. Lett. 1994, 72,
1459.
152. R. Sant, E. E. Wolf, J. Catal. 1988, 110, 210.
153. C. C. Chan, E. E. Wolf, H. C. Chang, J. Phys. Chem. 1993,
97, 1055.
154. H. U. Onken, E. Wicke, Ber. Bunsenges. Phys. Chem. 1986,
90, 976.
155. H. U. Onken, E. Wicke, 2.Phys. Chem. NF 1989,165, 23.
156. J. P. Dath, J. P. Dauchot, J. Catal. 1989, 115, 593.
157. P. K. Tsai, M. B. Maple, R. K. Herz, J. Catal. 1988,112,453.
158. S. A. Zhukov, V. Barelko, Sou. J. Chem. Phys. 1982, I , 883.
159. Yu. E. Volodin, V. Barelko, A. G. Merzhanov, Sou. J.
Chem. Phys. 1982, 1, 1146.
160. L. Lobban, D. Luss, J. Phys. Chem. 1989, 93, 6530.
161. L. Lobban, G. Philippou, D. Luss, J. Phys. Chem. 1989, 93,
733.
162. G. Philippou, F. Schultz, D. Luss, J. Phys. Chem. 1991, 95,
3224.
163. G. Philippou, D. Luss, J. Phys. Chem. 1992, 96, 6651.
164. G. Philippou, D. Luss, Chem. Eng. Sci. 1993, 48, 2313.
165. G. A. Cordonnier, F. Schuth, L. D. Schmidt, J. Chem.
Phys. 1989, 91, 5374.
166. M. Sheintuch, Chem. Eng. Sci. 1989,44, 1081.
167. M. Sheintuch, J. Phys. Chem. 1990, 94, 5889.
168. S. L. Lane, D. Luss, Phys. Rev. Lett. 1993, 70, 830.
169. M. D. Graham, S. L. Lane, D. Luss, J. Chem. Phys. 1993,
97, 7564.
170. R. A. Schmitz, G. A. D’Netto, L. F. Razon, J. R. Brown
es (Eds.: G. Nicolis, F. Baras), Dor-
drecht, Amsterdam, 1989.
171. P. C. Pawlicki, R. A. Schmitz, Chem. Eng. Progr. 1987,83,40.
172. M. D. Graham, S . L. Lane, D. Luss, J. Chem. Phys. 1993,
97, 889.
173. M. Sheintuch, J. Catal. 1985, 96, 326.
174. R. Dagonnier, M. Dumont, J. Nuyts, J. Catal. 1980,66, 130.
175. F. Schiith, E. Wicke, Ber. Bunsenges. Phys. Chem. 1989. 93,
491.
176. H. C. Chang, M. Aluko, Chem. Eng. Sci. 1984,39, 36; 51.
177. M. Aluko, H. C. Chang, Chem. Eng. Sci. 1986,41, 317.
178. M. P. Harold, M. Sheintuch, D. Luss, I & EC Research
1987,26,786; 795; 1616.
179. S. A. Zhukov, V. Barelko, Dokl. Akad. Nauk SSSR 1978,
238, 135.
180. Yu. E. Volodin, V. N. Zuyagin, A. N. Ivanova, V. Barelko,
Ado. Chem. Phys. 1990, 77, 551.
181. 0. Middaya, M. D. Graham, D. Luss, M. Sheintuch, J
Chem. Phys. 1993, 98, 2823.
182 U. Middya, M. Sheintuch, M. D . Graham, D. Luss, Physica
D 1993, 63, 393.
183 U. Middya, D. Luss, M. Sheintuch, J. Chem. Phys. 1994,
100, 3568.
5.3 Factors Influencing Catalytic Action 105 1
184. V. Barelko, E. C. Pechatnikov, A. N. Ivanova, Sou. J.
Chem. Phys. 1991, 7, 1153.
185. Y. S. Matros in Unsteady State Processes in Catalysis (Ed.:
J. S . Matras), VSP, Utrecht, 1990, p. 131.
186. L. L. Hegedus, C. C. Chang, D. J. McEwen, E. M. Sloan,
Ind. Eng. Chem. Fundam. 1980,19, 367.
187. M. B. Cutlip, C. J. Hawkins, D. Mukesh, W. Morton, C. N.
Kenney, Chem. Eng. Commun. 1983,22, 329.
188. G. Vaporciyan, A. Annapragada, E. Gulari, Chem. Eng.
Sci. 1988, 43, 2457.
189. M. Marek in Temporal Order (Eds: L. Rensing, N. I.
Jaeger), Springer, Heidelberg, 1985.
190. W. Vance, G. E. Tsahouras, J. Ross, Progr. Theor. Phys.
1989, 99, 331.
191. G. E. Tsahouras, J. Ross, J. Chem. Phys. 1987, 87, 6538;
1989,89, 5715.
192. G. E. Tsahours, J. Ross, J. Phys. Chem. 1989, 93, 2833.
193. R. Schwankner, M. Eiswirth, P. Moller, K. Wetzl, G. Ertl,
J. Chem. Phys. 1987, 87, 742.
194. K. Krischer, M. Eiswirth, G. Ertl, J. Chem. Phys. 1992, 97,
307.
195. W. M. H. Sachtler, R. A. van Santen, Adu. Catal. 1977,26,
69.
196. K. Asakura, J. Lauterbach, H. H. Rotermund, G. Ertl,
Phys. Rev. B 1994, 50, 8043.
197. P. Manneville, Dissipative Structures and Weak Turbulence,
Academic Press, London, 1990.
198. A. Babloyantz, J. A. Sepulchre, Physica D 1991, 49, 52.
199. J. Enderlein, Phys. Lett. 1991, 156, 429.
200. M. D. Graham, I. G. Kevrekidis, K. Asakura, J. Lau-
terbach, K. Krischer, H. H. Rotermund, G. Ertl, Science
1994, 264, 80.
201. J. W. Evans, Langmuir 1991, 7, 2514.
202. R. M. Ziff, E. Gulari, Y. Barshed, Phys. Rev. Lett. 1986,56,
2553.
203. P. Moller, K. Wetzl, M. Eiswirth, G. Ertl, J. Chem. Phys.
1986,85, 5328.
204. X. G. Wu, R. Kapral, Physica A 1992, 188, 284.
205. V. Gorodetskii, W. Drachsel, J. H. Block, Catal. Lett. 1993,
19, 223.
206. N. Kruse, personal communication.
207. A. S. Mikhailov, G. Ertl, Chem. Phys. Lett. 1995,238, 104.
208. W. C. Conner, G. M. Pajonk, S. J. Teichner, Adv. Catal.
1986, 34, 1.
5.3 Factors Influencing Catalytic
Action
5.3.1 Substituent Effects
M. KRAUS
5.3.1.1 Substituent, Reaction Center, and Surface
Reaction Complex
In the nomenclature of physical organic chemistry, the
grouping of atoms within a molecule, whose bonds are
rearranged during its transformation, is called the re-
References see page 1063