NRO Re Rae
Aantal
eterogeneous ~
World Scientific| CONTENTS
| CONCEPTS INCATALYSIS. 5 sd
| L_References 12
CHAPTER 2:
THEORIES OF CHEMISORPTION,» ss
2.1_Introduction u
2.2 Theoretical Intermezzo 1. Background to the Extended-Hiickel
Method and other Semi-Empirical Methods. . . B
2.3 Elementary Theory of Chemisorption. One-Dimensional Models. __44
2.3.1 Elementary free-electron gas models. 46
2.3.2 Tight-binding calculations, all atoms equal; no hybridization 51
2.33 One-dimensional models, hybridization... _ 63
| 2.3.3.1 Physical background. 63
2.3.3.2 Elementary quantum chemistry of the Shockley surface state 68
| 2.4 Theoretical Intermezzo 2. Introduction to the Green’s Function Method,
Application to One-Dimensional Surfaces
Green’s function calculation of local density of states o
| open chain . 78
242 ‘The resolvent method. Application to Shodiley surface state 80
2.5 Elementary Quantum Chemistry of Chemisorption to Metal Surfaces.
76
Tamm Surface States, Surface Molecule Limit : . 85
2.5.1 __A model for three-dimensional chemisorption. Bethe lattice
approximation; the concept of group orbitals... _ 8
| 2.6 Theoretical Intermezzo 3. Formal Chemisorption Theory... 102
| 2.6.1 The expressions for the bond energy. 102
| 2.6.2 Relationship with HOMO-LUMO theory ‘and frobtier aia
| theory i - 113
| 2.6.3 Relationship with bond-order caleul ‘ions i 4s gS
| 2.1 The Effect of E -E I ;
| 211 ; icted HL Ri
approximation... ss ss
2.7.2 ‘The unrestricted Hartree-Fock solution of the Newns-Anderson
chemisorption model. 12
2.73 ‘The effective one-center electron-clectron repulsion integral
Time-dependent charge fluctuations of the chemisorbed state _124
2.7.3.1 The physical picture. sw 108% CONTENTS
2.7.3.2 Theoretical Intermezzo 4; tunnel time and electron m acoming
effects. Quantum chemistry of Uers - - 126
2.8 Application of Bethe Lattice Method to CO and H Chemisorption . _140
281 TI 7 f a 1 surh 140
2.8.2 Atop and bridge coordination of COandH . ws. 148
283 Coordination of CO to different £. on i 56
2.9 Free-Electron Approaches to Chemisorption. Some First-Principle
Calculations 0.
CHAPTER 3:
MOLECULAR BASIS OF METAL CATALYSIS Ci‘
3.1__Introduction 178
3.2 fees in Chemisorption _. 179
3.2. Adsorption energy as a Tunction of coverage; surface
reconstruction...
3.3 Quantum Chemistry of Surface Dissociation and Association
Reactions * 2 F 90
3.4 Shustorovich’s Empirical Approach to Surface Association and
Dissociation Reactions . . » + 204
3.4.1 | The Shustorovich Bond Order ‘Conservation ‘method . 204
3.4.2 Derivation of the formula for the heat of molecular sbceyhion: 206
3.4.3 Derivation of the formula representing the activation energy for
dissociation of a chemisorbed molecule se ee 208
3.5 Catalysis of Synthesis Gas Conversion . . 3 . . 3...) 21
3.5.1 General features. » 21
3.5.2 Quantum chemistry of CO chemisorption and dissociation 213,
3.5.3 CH, fragment chemisorption and reactivity... s 284
3.6 Hydrocarbon Conversion Catalysis... ess
3.6.1 General background sw sO
3.6.2 Theory of surface ligand effects. . 246
3.6.3 Quantum chemistry of cation-induced changes in chemisorption
to metal clusters . . . =. 0.
3.7 Selective Oxidation Catalysis hg) 87
3.7.1 General background . 267
3.72 Quantum chemistry of chemisorption of the O, molecule
tosilver . . 270
3.7.3 Epoxidation of ethylene; the electronic basis of the role of
subsurface oxygen...» . ss ss 7hconrents xi
3.8 References 0. . . ss
CHAPTER 4:
|
|
|
|
|
IDE SURFACE: 281
4.1 General Physicochemical Properties of the Tonogente Surface.
Pauling’s Valency Concept... ee
4.1.1 Surface relaxation effects __. -__281
4.1.2 _ The Madelung potential of the surface layer; Pauling’s valencies 285
4.1.2.1 Ton charge excess of the dense TiO, surface. _. 291
4.1.2.2 Elementary theory of surface Bronsted and Lewis acidity
and basicity. Implications for catalyst preparation and
acidcatalysis . ww we we es 04
2 ThE ic 8 f Transition-Metal Oxide Surt ans
4.2.1 Ligand-field theoretical considerations, the surface dangling
bond analog. . 303
j 7
| 4.22 Valence electron stmctore of the rutile surface a1
| 4.3 Quantum-Chemical Basis of Bronsted Acidity in Zeolites... __316
| 4.4 The Quantum-Chemical Basis of Heterogeneous Oxide Catalysis. 326
4.5 References 2. 2. . ee eee eee BBL
| 5.1 Introduction a
| 5.2__Gibbs’ Rule 336
| 5.2.1 Surface composition ofalloys . . =. =. =. =. —- ‘34
5.2.2 Statistical- mechanical theory ‘of surface enrichment in alloys 343
| 5.2.2.1 Regular solution model for a two-component ally .. . 343
5.2.3 Application of the cluster method “_ * 2 5 = wee
| 5.3 Lattice Stability of Zeolites =... . . . ss 852
5.3.1 Background. te
5.32 The lattice energy of zeolites... ~—~—=s=~SC~—ws~C“CS~Csi‘“‘<=‘“ 1-10 mbar). Hydrocarbon conversion catalysts may
become covered with hydrogen or a carbonaceous layer, ot may become converted
to a carbide as is the case for Fe. In view of this it is remarkable that for selected
reactions and conditions, kinetic data determined at vacuum conditions can be used
to predict the kinetic behaviour of the catalyst over a pressure gap of ten orders of
magnitude. This has been found for the ammonia synthesis reaction!) as well as
for the reaction of CO and NOM). Apparently the latter reactions can be classified
as condition invariant, whereas the former example of ethylene epoxidation is a
condition variant reaction.
Bonding between adsorbate and reactant and especially the rearrangement of
adsorbate molecules to chemisorbed reaction intermediates is usually accompanied
by a change in the formal valency of the surface atom to which the reaction inter-
mediate is coordinated. This will be discussed for the metal catalysts in chapter 3
and for the oxidic catalysts in chapter 4. Solid acids such as zeolites and clays can
be synthesized as Brgnsted acids, stable at temperatures up to 700 or 800 °C. They
catalyze reactions of the Bronsted acidity type that are known from classical carbo-
nium or carbenium ion organic chemistry. The presence of micropores in the zeolite
may change the selectivity of reaction due to spatial restraints. Also Lewis acid cat-
alysts are known. A familiar example is the epoxidation reaction of an alkene with
hydrogen peroxide or a hydroperoxide, catalyzed by silica or silica-zeolites contain-
ing small amounts of Ti. The large Ti ions substitute Si‘* in tetrahedral positions.
Exposed to the peroxide, the Ti‘* polarizes the electrons in the -O-OH group, en-
hancing the electrophilic nature of the OH group that is inserted into the alkene
bord.CONCEPTS IN CATALYSIS
Table 1.1. Examples of important catalytic materials and reactions.
~ SiO2/TiO2
(Lewis acids)
Sulfides
- NiS/MoS2, CoS/MoS»
Reduci i
~ Bi03/MoOs3
V203/TiO2
Chlorides
~ TiCls
Catalyst Reactions
Group-VIII/IB metals
~ Pt, Ni dehydrogenation, isomerization
Pt, Co, Ag oxidation
- Fe, Rh, Co synthesis gas conversion
NO reduction
Solid acids
~ zeolites, clays hydrocarbon cracking
(Bronsted acids) methanol to aromatics
alkylation
isomerization
shape-selective reactions
epoxidation
hydrodesulfurization
hydrodenitrogenation
selective oxidation
propylene to acrolein
NO reduction
polymerization
5
In the presence of a vapor phase with sulfur-containing molecules, metal surfaces
become sulfidized. A familiar class of sulfidic catalysts are MoS; and WS. They are
usually promoted with NiS or CoS. Such catalysts are very active in desulfurization
and hydrodenitrogenation catalysis.6 CHAPTER
However, other applications are also known. MoS catalysts can be used to pro-
duce higher alcohols from synthesis gas and for the water-gas shift reaction. The
elementary reaction steps on the sulfide metal center proceed in a way analogous to
those in organometallic complexes. On the edges of the sulfide the metal ion ligand
positions are partially occupied by reacting molecules. Catalysis occurs by oxidative
addition, insertion or reductive elimination steps. The electronic features controlling
these events are very similar to those known from coordination chemistry]. The
situation is more complex for the reducible oxides, represented by the BiyO3/MoOs
compounds. They are used for the selective oxidation of propylene to acrolein or as
ammoxidation catalyst. According to the Mars-van Krevelenl® mechanism the Mo
ion is the cation interacting with the organic molecule and Bi2O3 mainly serves to
conduct oxygen atoms to the catalytically active site. A recent development is the
use of mixed oxides of V203/TiO2, notably for the conversion of NO with NH to
No in stack-gas. Another important catalyst is based on the metal chorides. TiCls
dispersed on MgCl, is used as an olefin polymerization catalyst. In similar fashion
to the sulfide case, catalysis occurs by a coordination-type mechanism. Ethylene or
propylene coordinates to an unoccupied ligand position of Ti and inserts into the
growing hydrocarbon chain positioned on another ligand position of TiCls. MgClz
acts as support to TiCl3. There exist many other examples of coordination-type
mechanisms occurring at the catalytically active sites of heterogeneous catalysts.
We will just mention the Cr203/SiOz catalyst used for ethylene polymerization,
and the MoO03/SiO2 or Re2Os/SiO> catalyst used for the olefin metathesis reaction.
Most of the catalytic systems discussed result from accidental discovery or pur-
poseful development. Initially, both synthesis chemistry and reaction mechanism
used to be little understood and optimization studies were often highly empirical.
At present a considerable change is taking place in the toolkit of the practical cat-
alytic chemist or chemical engineer. In large industrial laboratories the use of X-ray
diffraction, Solid State NMR, Infrared Spectroscopy and X-ray spectroscopy is be-
coming state of the art. This adds to the more ”classical” analytical tools such as
Electron Microscopy and structural characterization techniques based on adsorp-
tion and desorption of gases. Such changes illustrate the impact of spectroscopic
techniques and more indirectly of theoretical chemistry on practical catalysis.
Catalysts are usually multicomponent systems in view of the need to stabilize
the active phase, to suppress nonselective sites and to optimize surface area. For
instance, metal particles are usually dispersed on inert high-surface oxidic supports.
Promotors or additives may be added to stabilize the particles or to poison nonselec-
tive sites (e.g. Sin catalytic reforming). There is not a single technique that enables
a complete characterization of such a system, but a combination of several spectro-
scopic techniques") may be very succesful to identify the different compound phases
present on a catalyst. Apart from developments with a direct impact on practical
catalysis, advances in spectroscopy as well as theory have significantly benefited
catalysis science, mostly through the study of well-defined model systems. The sur-
face science approach to catalysis as pioneered by Somorjail') and Ertl!) highlights
these efforts. The use of metal single crystal faces in combination with Ultra HighCONCEPTS IN CATALYSIS. T
Vacuum techniques makes the study of well-defined surfaces possible. In addition,
surface spectroscopic techniques can be used to identify chemisorbed species and
their changes as a function of temperature or upon contact with another reactant
gas. This approach has been very fruitful and has generated a significant amount
of detailed information on the behaviour of chemisorbed molecules indispensable for
the theoretical development of catalysis.
The study of chemisorbed molecules is of the essence for catalysis. Catalytic
events consist of elementary reactions on the catalyst surface. Chemical bonds are
formed between the surface atoms and the adsorbing molecule. These interactions
cause rupture of chemical bonds within the adsorbing molecule and formation of
new chemical bonds between the fragments.
Related to the Surface Science approach to catalysis is the use of molecular beam
techniques'®l, With Mass Spectroscopy and other detection techniques the reactivity
of small metal particles effusing into a vacuum is studied. Again reaction conditions
are well defined, but the temperature is close to zero Kelvin. This approach enables
the study of chemical reactivity as a function of particle size. Knowledge on the
reactivity of isolated particles is relevant especially for theoretical studies since these
data refer to well-defined particles without interaction with a support. Theoretical
studies can be conveniently done with isolated small clusters. As with surface science
studies, molecular beam experiments are fundamental to the study of elementary
reactions as a function of surface or particle structure. We will frequently refer to
data collected by these approaches.
Zeolites form another class of materials useful for fundamental studies(!“. As
mentioned earlier, zeolites are microporous silica-aluminates with micropores of di-
mensions comparable to organic molecules. The materials are unique, because these
micropores are determined by the three-dimensional crystallographic structure of
the material and catalytic events occur at the interphase of zeolite micropore and
zeolite lattice. As a result the catalytically active sites are well defined. Zeolites are
used in practice in the acidic form or promoted with metal or sulfide particles. High
Resolution Electron Microscopy, Neutron Diffraction and Solid State NMR are tech-
niques that are applied for structural characterization and to study the behaviour
of chemisorbed molecules.
Theoretical methods are required to derive structural information from spectro-
scopic data, which usually concern measurements of electronic features. Because of
the availability of large and efficient computer power and the current state of the
art of theoretical chemistry, electronic structure calculations on model systems of
relevance to experimental studies can be made. In addition, the catalytic chemist
needs insight into the factors that determine the transition-state potential energy
surface of reacting molecules. Also methods are needed to predict the geometry of
the adsorption site as a function of metal surface composition or charge distribution
in the zeolite. These methods will be extensively discussed in the next chapters.
For mechanistic and structural studies of practical catalysts, the possibility to8 CHAPTER
do in-situ spectroscopy of the working catalyst is important. Catalytic reaction
intermediates as well as active catalyst phases are often unstable with respect to
the noncatalytic condition. Techniques such as Infrared Spectroscopy using electro-
magnetic radiation are especially valuable so that high pressure-gas conditions can
be studied. For mechanistic studies pulse-type experiments with isotope-labelled
compounds are often very succesfully employed.
A recent interesting technique is the TAPl!1] (Temporal Analysis of Products)
reactor system that enables massspectrometric identification of reaction interme-
diates at a milliseconds timescale. The mechanistic knowledge generated by such
techniques is used not only to unravel details of elementary reaction steps, but also
to study catalyst synthesis ”in-situ” and as a function of catalyst activation. Usu-
ally an active catalyst is made by reaction of a support with an active catalytic
site precursor and activation procedures are applied to produce a catalyst with the
required performance. Two developments are of importance that enable the ”molec-
ular catalyst engineering approach” to be realized. Firstly there are the state of the
art and the rapid extension of organometallic chemistry. The arsenal of synthetic
methods can be used to produce a rich variety of complexes involving similar metal
centers. Clusters can be synthesized containing also oxygen or sulfur. Secondly we
have the molecularization of colloid chemistry. As a result, the properties of OH
groups adsorbed on a support surface, its basic or Bronsted acid properties, aad
also Lewis acid or basic sites become understood on a molecular level. By proper
variation of the conditions for catalyst support preparation, its reactivity with re-
spect to active catalytic site precursors may be controlled. This enables production
of more uniform and more selective catalysts.
Four major concepts appear to be important to molecular heterogeneous catal-
-lysis, They are:
1. Chemisorption and catalytically active site
2. Structure and reactivity relationships
3. Long-range versus short-range electronic and electrostatic effects
4. Surface and phase stability
ad 1, CHEMISORPTION
Earlier we described the catalytic reaction as a series of consecutive steps at the
surface, in which adsorbate and adsorbate-surface bonds are formed and/or broken
on the reaction path towards the product molecule, The forces between surface
atoms and adsorbate atoms responsible for rearrangement of the chemical bond
are similar to those responsible for strong adsorption (Egg > 10 kcal/mol). The
adsorption process dominated by such interaction is called chemisorption. Even
on a single crystal metal surface, several adsorption modes are conceivable and for
dissociation of a diatomic molecule many different reaction paths can be envisioned.
However, usually only one particular surface atom configuration is preferred to lead
to the idea of catalytic active site. If catalysis of a molecule is studied that has
several reaction possibilities, some desirable and others not, a selective reaction
usually requires a particular surface atom composition and rearrangement.CONCEPTS IN CATALYSIS. 9
ad 2. REACTIVITY RELATIONSHIP
The concept of chemisorption as a function of surface atom rearrangement implies
a relation between substrate reactivity and catalytic site rearrangement. Especially
surface-science-type experiments have demonstrated the dependence of catalytic ac-
tivity on single crystal surface exposed. Usually surfaces containing atoms of increas-
ing coordinative unsaturation appear more reactive. Molecular beam experiments
have shown dramatic changes in chemical reactivity as a function of particle size-An
optimum in reactivity is generally found for particles containing 6-10 metal atoms.
Increasing particle size at first results in a decrease in reactivity, which again slowly
increases for larger particle sizes. These differences relate to changes in electronic
structure, especially for the smaller metal particles, as well as changes in exposed
surface geometry. The latter becomes very apparent from studies involving alloys
of catalytically active and nonactive metal atoms or coadsorption with catalytically
inert adsorbate atoms. It appears that some reactions (e.g. dissociation) require
coordination of the reactive molecules to a surface site with a minimum number of
neighbouring reactive surface metal atoms in order for the reaction to proceed. Such
reactions become suppressed if the concentration of these reactive surface atom en-
sembles decreases by dilution with nonreactive atoms. The earlier discussed concept
of shape selectivity observed for zeolites also implies a structure-activity relation-
ship. In this particular case, the selectivity of a reaction is governed by the size of
product molecules and by the transition state. Geometric considerations related to
the size of the crystalline micropore channels and cavities in the zeolite here prevail.
ad 3. LONG RANGE INTERACTION
Long-range effects can be defined as changes in the environment of the catalytically
reactive site due to alterations in composition or geometry of atoms in the second
coordination shell with respect to the chemisorbed molecule. The particle size effects
discussed above for metal particles, if not due to changes in geometry of the catalyt-
ically active site, belong to the long-range effect category. Changes in the surface
dipole layer for different metal surfaces or on surface kinks and steps also belong to
this category. An example of a work function change that affects the topology of
the chemisorbed molecule is the effect of positively charged potassium ions on the
coordination of CO. Under the influence of the positive charge of potassium on the
Pt surface, CO molecules shift from atop to bridging coordination sites. In zeolites
also changes in acidity occur due to changes in composition at positions beyond
the first coordination shell tetrahedron with respect to the Bransted acid site. An
interesting unresolved question is the range of the long-range interactions, a subject
to which we will also refer in this book.
ad 4, SURFACE COMPOSITION
In alloy catalysis it has been discovered that the surface composition and bulk
composition of a material may significantly differ. As a result, catalysis has to be
studied in relation to the surface composition rather than the bulk composition. In
addition, if the reactivity of the surface atoms is different the surface composition
may change, favouring enrichment of surface atoms of highest reactivity. The driving10° CHAPTERI
force to this phenomenon is minimization of surface-energy. As a consequence of
the surface energy minimization principle, small metal particles will have particular
shapes which may depend on gas-phase composition. Ionogenic materials in vacuum
will only expose their electrostatically neutral surfaces. As discussed in chapter 4
the stability of microporous solids in contact with a liquid phase is also ruled by
this surface-energy minimization principle.
One may wonder whether there is a relation between kinetic data collected on
uniform model catalyst surfaces and practical catalysts with a nonuniform distribu-
tion of catalytically active sites. A thorough analysis of this question is found in
the book by Boudart and Djéga-Mariadassow!4l, based upon pioneering studies of
Temkin and other Russian kineticists on ammonia synthesis. Assuming an expo-
nential distribution of adsorption energies and a linear variation of the activation
energy, one finds that the form of the overall rate constant is very similar for uni-
form and nonuniform distribution of active sites, so that the same type of kinetic
equation applies. But the interpretation of the kinetic constants is different for the
two cases, Fundamental to a kinetic analysis is the concept of the most abundant
reaction intermediate (MARI) and the rate-limiting step in a reaction sequence. On
a working catalyst on which many surface species may be present, some not active
in the reaction but just as a spectator, one has to identify the species that partici-
pates in the catalytic reaction chain. The dominant species is the one that is most
difficult to convert. It is one of the challenges of in-situ spectroscopy to identify the
MARI or MARIs proposed in the derivation of an overall kinetic equation. With
the help of this information the theoretical basis for a reaction mechanism can be
developed. We will consider the overall kinetic equation for a heterogeneous reaction
and compare this with the expressions used in homogeneous or enzyme catalysis!!),
Let us consider the sequence of two reversible elementary steps:
A
Ait 5S; = Bit Sz (1.1)
ka
ka
Sz+ Az = Bot+Si (1.2)
kaa
A, + Az = Bi + Bo
S; and Sz denote empty and occupied surface site positions respectively, with
L=54+5.
As is shown by Boudart, the turnover rate becomes:
kika[Ai}[Aa] — k-1k—2 [Bi)[B2)
= BIA + kalBil + falda] + bBA] 8)
<
andCONCEPTS IN CATALYSIS. 11
(Si) _ kali) + bolAo]
[Se] Ai[Ai] + &-2[Ba]
Let us assume that step k; is rate determining. Define @ as the fraction of occupied
sites:
(1.4)
IS) _ 1
Sa] 7 a)
The expression for the turnover rate becomes:
al Aa]@ = ky[Ao] K fy
= 1.6
= 1+ x fy a8)
with k
AL
=e (1.7)
The turnover rate becomes proportional to the surface coverage 9. Equation (1.6) is
the Langmuir- Hinshelwood expression and is very similar to the Michaelis - Menten
expression used in enzyme catalysis. Consider the following mechanism describing
enzyme catalysis:
p+sixEeyp (1.8)
ki ke
where E is the enzymatic site, S the substrate, X the enzyme substrate complex and
P the product of the reaction. Using the steady-state condition and assuming step
ky to be rate limiting, one derives for the turnover rate for product formation:
&:
1 dy _ 9, = Ease! __bKCs
ie x = 1.9)
Cay’ a T+ pbges 14K (1.9)
with 0, the steady state fraction of enzyme sites complexed with S4. Cpo is enzyme
site concentration, Cs is substrate concentration. One notes the great similarity of
Eq.(1.9), the Michaelis-Menten rate equation of enzyme catalysis and Eq.(1.6), the
equation for the surface reaction rate. The similarity between the expressions stems
from the conservation of surface sites or enzyme molecules available to the overall
reaction. Essential factors here are the complexation constant f- (Eq.(1.9)) or
adsorption equilibrium constant K (Eq.(1.6)) and the rate of decomposition of the
adsorption complex. Since at low reactant concentration the activation energy is
determined by the relation:
din(ky.K)
dh) (1.10)
(Hits Bact = ~Ro12° cuarreni
but at high reactant concentration the activation energy becomes:
, __p dint)
[Hing Pact = Pa “acy
one derives that relevance of kz or K with respect to the overall reaction rate depends
on the occupation of the catalytically active sites 0. In Eq.(1.10) and Eq.(1.11) Rg
is the gas constant. If the surface coverage or site occupation @ increases from 0 to 1,
the overall reaction order changes accordingly from 1 to 0. The examples considered
concern reactions with only one product. Analogous expressions can be derived for
reactions with several products. The selectivity is found to depend on the ratios of
decomposition rates and complexation constants.
(1.11)
Clearly the kinetics of homogeneous and enzyme catalysis as well as heteroge-
neous catalysis are closely related. They are governed by the geometric and elec-
tronic details of the catalytically active site. The molecularization of heterogeneous
catalysis implies an increasing aptitude to manipulate catalyst preparation, so as to
direct catalytic reactivity towards required selectivity. At the root of this molecu-
lar knowledge is the increasing sophistication of spectroscopy that can be usefully
applied to complex catalytic systems. Spectroscopic as well as mechanistic interpre-
tation is founded on advances in theoretical chemical techniques currently able to
simulate systems of catalytic relevance. It is to the description of those endeavours
that this book is committed.
1.1 References
la. M. Boudart, G. Djéga - Mariadassou, "Kinetics of Heterogeneous Reactions”,
Princeton University Press, Princeton, N.J. (1984).
1b. J.M. Thomas, W.J. Thomas, "Introduction to Heterogeneous Catalysis”,
Academic Press (1967).
2. P.Stoltze, J.K. Norskov, Phys. Rev. Lett, 55 (1985) 2502;
P. Stoltze, Physica Scripta 36 (1987) 824.
3. S.H. Oh, G.B. Fischer, J.E. Carpenter, D.W. Goodman, J. Catal. 100 (1986)
360.
R. Hoffmann, Science 211 (1981) 995.
5. G.A.Somorjai,” Chemistry in Two Dimensions: Surfaces”, Cornell University
Press (1981).
6. P. Mars, D.W. van Krevelen, Chem. Eng. Sci. Suppl. 3 (1954) 41.
7. J.M. Thomas, R.M. Lambert, "Characterization of Catalysts”, J. Wiley
(1980).
8. G. Ertl in Catalysis 4, chapter 3, Ed. J.R. Anderson, M. Boudart, Springer
(1983).
9. Kaldor, R.L. Whetten, D.M. Cox, D.J. Trevor, A Kaldor, Phys. Rev. Lett.
54 (1985) 1494.
10. J.M. Thomas, Ang. Chem. Int. Ed. Engl. 27 (1988) 1673.
*CONCEPTS IN CATALYSIS 13
11. J.T. Gleaves, J.R. Ebner, T.C. Kuechler, Catal. Rev. - Sci. Eng. 30 (1)
(1988) 49.
12. R.A. van Santen, W.M.H. Sachtler, Adv. in Catal. 26 (1977) 69.CHAPTER 2
‘THEORIES OF CHEMISORPTION
2.1 Introduction.
In order to appreciate the different theoretical approaches of use in chemisorption
theory, it is necessary to know some of the fundamental observations on chemisorp-
tion derived from surface science studies. Each theoretical method has its limita-
tions, so it is also useful to have an idea of the current status of theoretical chemistry.
In this section these two topics will be highlighted and an outline of the material to
be presented in this chapter will be given.
Before the era of modern surface science in metal catalysis alloys were inves-
tigated to solve the question of the so-called ’electronic factor’ in heterogeneous
catalysis. Experiments were aimed at gaining an insight into the electronic basis
for the unique position of the group VIII metals as hydrogenation catalysts. The
group VIII metals, i.e. Fe, Co, Ni; Ru, Rh, Pd; Os, Ir, Pt, have d-valence electrons
distributed over a narrow nearly filled d-valence electron band and a broader s,p-
valence electron band. The d-valence electron-band is completely filled for the IB
metals, Cu, Ag and Au, that have a high activation energy as hydrogenation cata-
lysts and are little active as hydrocarbon conversion catalysts. For this reason one
of the propositions of that time was that the catalytic activity relates to the number
of d-valence electron holes. This seemed reasonable since the number of s, p-valence
electrons varies little. It is of the order of one electron per atom. The distribu-
tion of valence electrons over d- and s,p-electron bands is sketched in Fig.(2.1).
Information on the distribution of valence electrons is derived from photoemission
measurements, in which the kinetic energy of emitted electrons is measured as a
function of the frequency of irradiating light. The kinetic energy of the emitted
electrons (Exjn) is related to the valence electron energy of the electrons (Fyq)) via
the Einstein relation:
Ein = hy — Evat (2.1)
In Eq.(2.1) v is the frequency of light and 4 Planck’s constant. Before the develop-
ment of photoemission techniques one could measure the electronic structure only
indirectly, e.g. by measuring the thermal or electric conductance. These properties
are mainly governed by the energy density of the electrons in the highest occupied
orbitals of the metal p(Er). The energy of the highest occupied molecular orbital
and the lowest unoccupied orbital is the same in a conductor and is defined as the
Fermi level Ep. In the fifties, catalytic chemists tried to establish relations between
catalytic activity and p(Ep). Alloying may change (Ep) considerably. Sometimes
these changes correlate with catalytic activity, sometimes they do not.THEORIES OF CHEMISORPTION 15
fe “(itt
EF
Figure 2.1. Distribution of valence electrons over d- and s, pelectron bands for a
transition metal (schematic). a: group VIII metal; b: group IB metal.
E kcal
9
O
20 Au
Pd 80 4
6
Figure 2.2. Activation energy of H2/D2 by Pd/Au alloys as a function of conversion.
‘The broken line x denotes the paramagnetic susceptibility in arbitrary units!"),
‘An example is provided by the work of Couper and Eley!!!, who demonstrated that
the rate of the reaction:
Hy + Dz —+ 2HD
decreases dramatically if palladium is alloyed with gold. As shown in Fig.(2.2)
this decrease in catalytic activity correlates with a dramatic decrease in magnetic
susceptibility, i.e. p(Ep). We will return later to the question which features of the16 CHAPTER?
mA
°
SNS AS"
Cc
\ | i
i i i
} i }
i i |
: : :
METAL,
Ho Ho
Hc c c CH
\ on NY \ oe /
'
|
i
|
s ‘ .
ALLOY
Figure 2.3. Primary ensemble effect.
oO
c
i
: i
i
’ mw. s
site blocking
Figure 2.4. Secondary ensemble effect.‘THEORIES OF CHEMISORPTION 17
electronic structure of this alloy are responsible for the decrease of p(Ep) with Au
composition and whether one can theoretically justify a relation between chemical
reactivity and p(E£p).
Progress in the development of surface techniques at present not only makes it
possible to distinguish between surface and bulk electronic structure, it also pro-
vides the possibility to distinguish between surface and bulk atoms. Electron emis-
sion techniques can be employed exploiting the difference in electron freepath as a
function of electron energy. Other techniques provide direct information on the sur-
face composition; e.g. Low Energy Ion Scattering (LEIS) relates the kinetic energy
of scattered noble atom ions directly to the mass of the surface atom scatterers. A
classical method to probe the surface composition of an alloy is the use of chemisorp-
tive titration. This can be applied if alloys are studied that consist of a mixture of
elements such that the atoms of one element strongly interact with the gas atoms
used for chemisorptive titration (e.g. Pd adsorbs CO or H2), but the other elements
only very weakly (e.g. Au does not adsorb CO or Hz). One finds that for alloys the
surface composition is usually different from the bulk. In chapter 5 we will return to
theories useful to predict surface enrichment of alloys. Catalytic properties correlate
well with the surface composition of alloys. The existence of a surface phase explains
why correlations with p(E{p) measured in the bulk sometimes are and sometimes are
not found. A correlation has to be established with a property that characterizes
the surface electronic structure and not the bulk electronic structure.
Alloys can also change the selectivity of reactions. Alloying of Ni with Cu changes
hydrocarbon conversion by Ni in the following way:
a. The rates of hydrogenolysis, the reaction that decreases the size of the molecule
by breaking C-C bonds, become suppressed.
b. The rates of isomerization reactions, in which the number of carbon atoms in
the molecule remains the same, decreases much less, whereas hydrogenation or
dehydrogenation reactions change little. The rates are expressed as molecules
reacted per unit time per Ni atom exposed. One finds in general that association
and dissociation reactions are affected most if one studies the effect of an inactive
component such as Cu or Au on the catalytic properties of an active component
Ni.
Changes in selectivity can be explained on the basis of two effects:
a, geometric
b. electronic
Geometric effects can be subdivided into so-called primary and secondary ensemble
effects. According to the primary ensemble effect a change in surface reactivity
occurs, if a molecule adsorbs with several atoms to the metal surface (see Fig.(2.3)).
Molecules adsorbed in this way, when partially stripped of hydrogen atoms, will
hydrogenolyse. If the surface is diluted with metal atoms that are catalytically much
less active (e.g. Cu), the probability decreases that active metal atoms (e.g. Ni) keep
active metal neighbors. As a result the probability for multisite adsorption is reduced18 CHAPTER?
and hence the rate of the hydrogenolysis reaction. The secondary ensemble effect
decreases the strength of the metal surface-adsorbate atom bonds, by diminishing
the surface metal atom coordination number of the adsorbate atom in contact with
the metal surface Fig.(2.4). An example is the adsorption of CO to the Pd surface.
CO coordinated to Pd surfaces may chemisorb atop to a single Pd atom or bridge
between two Pd atoms. The CO frequency in the atop configuration is 2050 cm™!,
in the bridging configuration 1980 cm~1. In a subsequent section we will discuss
the reasons for this frequency difference. While on nonalloyed Pd predominantly
CO in bridging coordination occurs, one observes that upon alloying with silver
the probability increases for CO to chemisorb atop”) Fig.(2.5). A simple geometric
explanation for this phenomenon can be given. The interaction of CO with Ag is
much weaker than that with Pd and the probability for Pd neighbor atoms decreases.
Ina later section we will discuss the clectronic basis of this phenomenon.
Modern surface science as well as molecular beam techniques indicate very large
variations in surface reactivity if one compares different metal surfaces (semi-infi-
nite systems) or metal particles of different size. As an example in Fig.(2.6) for
different Fe and Re single crystal surfaces the turnover numbers (rates expressed as
molecules converted per unit time per surface atom exposed) are given for the con-
version of Nz and H2 to ammonia. Both the geometric arrangement of surface atoms
and their average coordination number change. The larger the average coordinative
unsaturation of the surface atoms, the higher the turn over number for ammonia
synthesis. One of the primary aimes of the theoretical discussion presented in this
chapter is to study the reactivity of surface atoms as a function of coordination as
well as valence electron occupation. The large differences in reactivity as a function
of particle size Fig.(2.7), found for very small particles in molecular beams, are espe-
cially of interest since current quantum-chemical ab-initio methods are best suited
to study the reactivity of small metal particles. Metal catalysts used in hydrocarbon
reforming processes contain metal particles of very small size (<2 nm). This is also
the case for catalysts with metal particles distributed in the micropores of zeolites.
The experimental results indicate that the behavior of small metal particles may be
very different from that of large surfaces. As a matter of interest this is also found
theoretically. The differences in covalent bonding that are the result of changes in
electron density relate to some extent with differences in ionization potential. This
is nicely illustrated by the molecular beam results presented in Fig.(2.8).
Changes in ionization potential do also play a role on metalsurfaces. Because of
the electric double layer caused by spill-over of surface electrons to the vacuum,
the ionization potential of a metal surface (the workfunction) decreases if one com-
pares the workfunction of a close-packed surface containing surface atoms of high
coordination, with a more open surface, containing surface atoms with a lower co-
ordination. As we will discuss, not only changes of the surface-dipole layer occur
if one compares different surfaces, but also metal-metal atom distances may change
(they usually decrease if one compares dense surfaces with more open surfaces).
In this chapter first the current status of quantum chemistry will be summarized
with respect to the problem of surface reactivity and a start will be made with the‘THEORIES OF CHEMISORPTION 19
2100 2000 1900 1800 1700 cm~1
Pd/Ag= 9.75/0.25
9.5/0.5
9/1
3.5/6.5
Figure 2.5. Infrared spectra of CO adsorbed to Pd-Ag alloys.
—Pco = 0.01 Torr; -— Poo = 0.5 Torr!’
theoretical treatment of chemisorption. We begin with semi-empirical approaches
such as the Extended-Hiickel method or the tight-binding method, as it is called by
solid-state physicists, because they are convenient to analyze. The concepts devel-
coped by applications of these methods to the surface reactivity problem will be used
to analyze also the results of more extensive and accurate calculations. At present
surface chemistry is in a comparable situation to that of organic chemistry in the
| sixties. Using elegant approaches and a simple method such as the Extended-Hiickel
method, Woodward and Hoffmann] developed important concepts extremely fruit-
ful to organic chemists. An enormous body of theoretical work evolved to test the
concepts with the aid of highly accurate quantitative calculations. Such calculations
have been rendered possible by the advent of supercomputers, which are steadily
improving in computation power. At present they are applied also to large-scale
quantum-chemical calculations on small metal clusters and have even become avail-
| able for use on metal surfaces.
Again especially due to efforts of Hoffmann!4l, but also many others("), the Extended-20 CHAPTER?
Fe(111) Fe(100) Fe(110)
Cg (solid)
Cg (broken)
Sh, SB
C5 (solid) Cq(solid)
C4o (broken)
Figure 2.6. Top figure: Surface structure sensitivity of iron-catalyzed ammonia syn-
thesis.
Bottom figure: Strong surface structure sensetivity of ammonia synthesis on rhenium
single crystals. ‘The symbols C, indicate n-coordinated sites. Yield is expressed in
arbitrary unite),‘THEORIES OF CHEMISORPTION 21
2 4 6 8 10 12 14 16 18
CLUSTER SIZE
Figure 2.7. Reactivity of metal clusters with Dz as a function of clustersize. Triangles
denote metal cluster responses in the absence of D2. Drawn peaks denote cluster
responses in the presence of D2. Loss in intensity implies a large cluster reactivity'®),
Hiickel method has been applied to small metal clusters and to the study of elemen-
tary surface reactions], Important concepts have evolved, which will be discussed
and can be verified with ab-initio calculations. Earlier, physicists had been oc-
cupied with the question of the relation between chemisorption and surface states
using related tight-binding techniques. Their adaptation of those techniques to the
surface problem is of especial interest for the development of embedding methods,
which study changes in chemisorption if a cluster becomes embedded in a semi-
infinite surface. Currently this work is being extended to systems including explicit
treatment of electron-electron interactions using extensions of local density approx-
imation (LDA) solid-state methods. Interestingly the Hartree-Fock-Slater method,
a quantum-chemical version of the LDA method developed by quantum chemists to
study organometallic complexes, is being extended to the study of transition-metal
surfaces(!®),
In the next section we will start with a short theoretical intermezzo to show the
approximations involved in the use of the Extended-Hiickel and related methods and
introduce some basic theoretical analytical tools. Then the tight-binding method
will be applied to study changes in electronic structure in a one-dimensional system22 cuapren?
lectron Binding Energy (eV)
Relative Recctivity
5 10 15, 20 25
Cluster Size (No. of Atoms)
Figure 2.8. Cluster ionization potential as a function of cluster size. For cluster sizes
larger than 10 an apperent relationship with reactivity is found"®),
simulating surface formation. In the next section this approach will be extended
to the study of three-dimensional systems, where chemisorption will be related to
surface states. Then the basic concepts of LDA theory will be introduced and their
applications to chemisorption discussed. Since electron-electron interactions can now
be accounted for, the physics of the image potential in relation with chemisorption
can be introduced. The results of first-principle calculations on chemisorption to
metal surfaces will be presented and interpreted. For chemical applications, analysis
in terms of formal chemisorption theory is useful. This is illustrated by analyzing
Hand CO chemisorption to transition-metal surfaces. The chapter will be finished
with a discussion of the use of free-electron theories. Here again simplifications are
possible and it will appear that such theories are complementary to theories based
on molecular orbital approaches.‘THEORIES OF CHEMISORPTION 23
2.2 Theoretical Intermezzo 1.
Background to the Extended-Hiickel Method and other Semi-Empirical
Methods.
‘The classical energy of a system can be written as:
E = Exin + Epot (2.2)
According to quantum mechanics the motion of the electrons is to be calculated
from wave functions (r). The chance density to find an electron at position r is
given by:
AT) = b'(F).d(F) (2.3)
Since the total chance to find an electron in space eqals 1, it follows that:
[are @).0@) = [ao =1 (2.4)
dr denotes an infinitesimally small volume element dr. Relation (2.4) defines the
normalization of p(r). The wave functions (r) are the solutions of the Schrddinger
equation:
[P+ V7) vir) = Bd) (25a)
AF ).vi(F) = Ev(F) (2.56)
7(*) and V(#) are kinetic and potential energy operators. The operator H{(7) is
called the Hamiltonian operator. For a particle with only one degree of freedom
along the z-axis, the kinetic energy operator 7'(z) equals:
P
“2m de®
If the potential V(r) only depends on position, as presupposed in relation (2.5a),
the potential energy operator equals the classical potential. (2) is the solution of:
a yey + )o(z) = Ed(z) (2.7)
2m dz ave) = i .
Because of the normalization condition Eq.(2.4) this equation has a solution for only
a limited number of values Ej. These values are the eigenvalues of the Schrédinger
equation. The corresponding solutions for (z) are called the eigenfunctions. We
shall focus the discussion on the approximations that lead to the Extended-Hiickel
method. In the course of its derivation we will also have an opportunity to refer
to a few other methods such as the Local Density Approximation (LDA) and the
Xa method, A major assumption involved is that the electrons move in the average
potential of the other electrons. This implies that the for n-electrons the electron
T(z) = (2.6)24 CHAPTER?
distribution p(71,......;7n) is equal to the product of the probability distributions
of the single atoms:
OF sonFn) = TL oC) (28)
This is the equivalent of the mean field assumption in statistical mechanics.
The n-electron wave function can be written as the antisymmetrized product of
one-electron functions ¥;(T;,0;). 0; denotes the spin of the electron:
UF 1. Fnyonon) = AT] ay (F 1501)-ag(F2502).-Yan(F ny on) (2.9)
Ais the antisymmetrization operator. One can improve on this one-electron
approximation if one includes electron correlation by assuming the n-electron state
function (71... 7 n,01...n) to be a linear combination of functions
(7) ... Tn,o1---on) with the electrons distributed over different one-electron
functions ¥a;(7,0). This leads to so-called configuration interaction methods and
the n-electron density cannot be written any more in the form of Eq.(2.8) because
now the electron motions become correlated. If one deals with weak interactions such
as the van der Waals interaction, that depends on the polarization of the electrons
in one fragment due to the fluctuating dipole of the electrons in the other fragment,
the one-electron approximation does not satisfy and configuration interaction has
to be considered explicitly.
‘The wave functions can be written as a product of a spatial function and spin
function if electromagnetic interactions are ignored:
vilF 0) = vir) (2.10)
Different methods exist to compute wave functions (7) that are the solutions
of equations of the type Eq.(2.5.) If the potential V(r) is small compared with
the kinetic energy, it is useful to consider ¥(7) approximately as a sum of plane
waves, an approximation often used in solid physics. If ¥(#) is approximated by
a single plane wave the method is called the free-electron method. Such solutions
are discussed later. We will first consider the wave function ¥;(7) to be a linear
combination of atomic orbitals centered on the lattice atoms.
vilF) = DO Ges) (2.11)
d
This method is called the LCAO (Linear Combination of Atomic Orbitals)
method. The electron density p(7) becomes:
AF) =D mpi) = Do mali)? (2.124)‘THEORIES OF CHEMISORPTION 25
= x [zt IGE les(F)P + 2x Reef cf etres(7)} ou (2.128)
nj is the orbital occupation. According to the Pauli principle the orbitals of a
molecule are occupied such that only a maximum of two electrons per molecular
orbital ¥;(r) are allowed. The contribution due to the first term in Eq.(2.12b) can
be considered as the sum of the electron densities on the atoms. The term:
Ee Rect & oi (F)9j(F me (2.13)
igg
due to the interference of atomic wave functions y;(7) and y;(F) is responsible for
covalent bonding. It relates to the interatomic electron density. If the electrons are
considered to move independently, Eq.(2.8) is satisfied and the total Hamiltonian
can be written as:
AF... Fn) = DAG +O Zrbye (2.14)
yey | Ray |
f(#;) is the one-electron operator that operates on a single electron and has a form
similar to Eq.(2.5a). The second term in Eq.(2.14) is the nuclear repulsion term.
Ryy represents the nuclear distance and Zy the nuclear charge. The general form
of H(#) is:
Weg
A) = P)- Tes far Oe + Blezch,7) (2.15)
Ry-F | [r-r]
‘The second term in Eq.(2.15) is the nuclear attraction between nuclei and electrons,
the third the repulsion an electron experiences due to the presence of other electrons
and the last operator /3(ezch, 7) is the contribution due to the so-called exchange
energy. It arises from the Pauli requirement that the n-electron wave function
has to be antisymmetric. H(exch, 7) partly compensates for the interaction of the
electron with itself included in the electron repulsion term.
Several approximations are used to compute H(ezch,7)¥,(7). In the Hartree-
-Fock approximation it follows directly from substitution of Eq.(2.9) in the exact
n-electron Hamiltonian and depends on the molecular orbital considered. For molec-
ular orbital j the result becomes:
EPP (exch, 7 )bj(r) = — me jf ar! ber) bj(r! oe(F Sopa; (2-16)
| 7 a
In the Local Density mene one uses the property that
E(ezch. 7) is a function of the total density p(7). Often it is approximated by the
functional dependence one derives for the exchange energy of a free-electron gas:
B(exch,#) ~ p(r)$ (2.17)26 CHAPTER?
In the Xq method as originally introduced, in addition, assumptions are made con-
cerning the form of Dy —“*—. It is computed within the muffin tin approxima-
tion. Within a sphere radius ay the potential is assumed to be equal to a particular
value assigned to atom 7, outside the sphere the potential is assumed to be equal
to zero. The advantage of these assumptions is that scattering techniques can be
used to find the eigenvalues of the electron distribution. These methods are efficient
and can be used for very large systems. We will return to this approximation later.
Usually its accuracy is insufficient to study the interaction energy of molecules. In
the Hartree-Fock-Slater (HFS)-LCAO method the E(exch, 7) term is approximated
bya term similar to Eq.(2-17) but further no assumptions on the potential are made.
In the Extended-Hiickel method the interaction the electron experiences from elec-
trons on neighboring atoms is supposed to be canceled by the attraction with nuclei.
In addition the electron-electron interaction on one atom is averaged or can be sim-
ulated by an effective one-center repulsion term. This implies that E(each,7) is
independent of the molecular orbital j, The Extended-Hiickel method therefore is
to be considered an approximation to the LDA. The approximation involved corre-
sponds with the situation that the atoms are assumed to be neutral with localized
electron distributions. Since the eigenvalues Ej are the solution of equations of type
Eq.(2.5a), they are given by:
By = [dr ve) Aw) (2.18)
The values B; follow from the set of secular equations generated from Eq.(2.5) and
Eq.(2.11):
HF oF) = Bedi(7) (2.19)
Multiplication at the left by the atomic orbital ¢$(7), substitution of Eq.(2.11) and
integration yield:
ry [rgd ol) = BIE Sy (2.20)
k
Defining:
Sie = f ar e(rdex(r) (2.21a)
Hyp = [erg mer) (2.218)
one finds as secular equation:
Dj — 215i) =0 (2.22)
a‘THEORIES OF CHEMISORPTION 27
Normalization gives as additional condition:
1= fdrvt@yiF) = Dial +20 Rech Se (2.28)
k i